The Impact of Microbiome Alterations in the NOD …...treated NOD mice showed quantifiable, durable...
Transcript of The Impact of Microbiome Alterations in the NOD …...treated NOD mice showed quantifiable, durable...
The Impact of Microbiome Alterations in the NOD Mouse Model of Type 1 Diabetes
by
Christopher Yick Lam Yau
A thesis submitted in conformity with the requirements for the degree of MSc.
Graduate Department of Immunology University of Toronto
© Copyright by Christopher Yick Lam Yau 2014
ii
The Impact of Microbiome Alterations in the NOD Mouse Model of
Type 1 Diabetes
Christopher Yick Lam Yau
MSc.
Graduate Department of Immunology University of Toronto
2014
Abstract
Epidemiological and experimental data suggest Type 1 Diabetes (T1D) and gut inflammatory
disorders share genetic and environmental risk factors. Genetics and microbiome composition
differentiate T1D-prone NOD mice and T1D-resistant NOR mice. In the first part of this study,
we examined the possible protective effects of NOR-derived microbiome transplant in NOD
mice. We report, despite quantifiable, durable changes in the gut microbiome of NOD recipients
of NOR-derived microbiome transplant compared to NOD and NOR controls, these mice were
not protected from T1D.
Studies have also suggested that probiotic bacteria can have protective effects in T1D and gut
inflammatory disorders. In the second part of this study, we examined the possible protective
effects of treatment with probiotic microbes in the NOD mice. We report Protecflor probiotic-
treated NOD mice showed quantifiable, durable changes in gut microbiome composition, and
were protected from autoimmune infiltration of the islets compared to vehicle-treated controls.
iii
Acknowledgments
I would like to thank my supervisor, Jayne Danska, first for her patience, for her insistence on
scientific rigor, and her dedication towards helping her students develop as scientists. To my
committee members Philippe Poussier, Dana Philpott and Shannon Dunn: thank you for
providing helpful input over the past years, helping with navigating personal challenges and
asking the right questions to drive my thinking forward. I would like to thank all members of the
Danska and Guidos Laboratory, with special thanks to Andrea Wong and Janet Markle, for their
mentorship and guidance.
The NOD/NOR microbiome project benefitted from the invaluable assistance of Steve Mortin-
Toth and keen scientific advice and previous work provided by Janet Markle. Dan Frank
provided useful advice on primer design strategies and qPCR standards.
The Protecflor project was contributed to hugely by the excellent technical skills and advice from
Steve Mortin-Toth. Thanks to Dan Frank for providing 16S sequencing and insightful
discussions on analysis methodology. We thank Thomas Tompkins and Mathieu Olivier at
Lallemand for providing Protecflor probiotic and information on the organisms.
Lastly, I would like to thank my friends, family, and my wife Jenn for their boundless emotional
support and encouragement. Without their support, this would not have been possible.
iv
Table of Contents
Acknowledgments ........................................................................................................................... ii
List of Tables ............................................................................................................................... viii
List of Figures ................................................................................................................................ ix
Chapter 1 Introduction .................................................................................................................... 1
1 Introduction ................................................................................................................................ 2
1.1 Introduction to T1D ............................................................................................................ 2
1.2 Genetic Risk Factors ........................................................................................................... 2
1.2.1 Genetic link to inflammatory diseases of the gut .................................................... 3
1.3 Environmental modifiers .................................................................................................... 3
1.4 NOD Model of T1D Genetics & Environment ................................................................... 4
1.4.1 Environmental modifiers in NOD ........................................................................... 5
1.5 The Hygiene Hypothesis ..................................................................................................... 6
1.6 The Gut Microbiome ........................................................................................................... 7
1.7 Tools and methods for investigating microbiome composition .......................................... 8
1.8 Previous Work .................................................................................................................. 10
1.9 Project Rationale ............................................................................................................... 10
1.9.1 The effect of NOR-derived microbiome transfer into NOD on microbiome composition and T1D incidence ........................................................................... 10
1.9.2 The effect of Protecflor probiotic treatment on NOD microbiome composition and insulitis severity ............................................................................................. 11
Chapter 2 Materials and Methods ................................................................................................. 13
2 Materials and Methods ............................................................................................................. 14
2.1 Mice .................................................................................................................................. 14
2.2 Design and Validation of Taxa-Specific qPCR ................................................................ 14
2.3 PCR ................................................................................................................................... 15
v
2.4 qPCR ................................................................................................................................. 15
2.5 Oral gavage transfer of the gut microbiome ..................................................................... 16
2.6 T1D assessment ................................................................................................................ 16
2.7 Treatment protocol for Protecflor probiotics .................................................................... 16
2.8 Insulitis assessment ........................................................................................................... 17
2.9 Bacterial 16s rDNA PCR library construction .................................................................. 17
2.10 454 pyrosequencing of NOD and NOR microbiome samples .......................................... 18
2.11 Illumina paired-end sequencing and analysis ................................................................... 18
2.12 Multivariate data analysis ................................................................................................. 19
2.12.1 PCA ....................................................................................................................... 19
2.12.2 Clustering Analysis ............................................................................................... 19
2.12.3 Statistical Analysis ................................................................................................ 19
Chapter 3 The effect of NOR-derived microbiome transfer into NOD on microbiome composition and T1D incidence ............................................................................................... 20
3 The effect of NOR-derived microbiome transfer into NOD on microbiome composition and T1D incidence ................................................................................................................... 21
3.1 Background and Rationale ................................................................................................ 21
3.2 Results ............................................................................................................................... 22
3.2.1 Design and Validation of Taxa-Specific qPCR .................................................... 22
3.2.2 Comparison of Abundance of Caecal Bacterial Taxa in NOD Mice from Weaning to Adulthood .......................................................................................... 23
3.2.3 Comparison of abundance of Caecal Bacterial Taxa in NOD and NOR Mice ..... 24
3.2.4 Effects of NOR-derived Caecal microbiota on NOD recipients ........................... 25
3.3 Summary ........................................................................................................................... 26
Chapter 4 The effect of Protecflor probiotic treatment on NOD microbiome composition and insulitis severity ....................................................................................................................... 39
4 The effect of Protecflor probiotic treatment on NOD microbiome composition and insulitis severity ....................................................................................................................... 40
vi
4.1 Background and Rationale ................................................................................................ 40
4.2 Results ............................................................................................................................... 41
4.2.1 Detection of Protecflor Probiotic organisms by qPCR ......................................... 41
4.2.2 Gastrointestinal Transit times of Protecflor Probiotic organisms in the NOD model ..................................................................................................................... 42
4.2.3 Evaluating the Effects of Protecflor Probiotic on Insulitis Severity in NOD model ..................................................................................................................... 43
4.2.4 The Effects of Protecflor Probiotic treatment on the gut microbiome in NOD mice ....................................................................................................................... 44
4.2.4.1 Differential abundance of bacterial taxa in Protecflor and vehicle treated NOD mice ................................................................................... 45
4.2.4.2 β-diversity analysis of 16S rDNA sequencing data ................................ 46
4.2.4.3 Principal components analysis of 16S rDNA sequencing data .............. 46
4.2.4.4 Correlation of principal components analysis with insulitis .................. 47
4.2.4.5 Correlation of caecal bacterial community richness with principal components analysis and insulitis .......................................................... 48
4.2.4.6 Correlation of caecal bacterial diversity with principal components analysis and insulitis ............................................................................... 49
4.3 Summary ........................................................................................................................... 50
Chapter 5 Discussion .................................................................................................................... 78
5 Discussion ................................................................................................................................ 79
5.1 The effect of oral gavage transfer of female NOR caecal microbiota into female NOD recipients on microbiome composition and T1D incidence ............................................. 79
5.1.1 Summary ............................................................................................................... 79
5.1.2 Use of “indicator” taxa .......................................................................................... 79
5.1.3 Comparison of abundance of caecal bacterial taxa in NOD mice from weaning to adulthood .......................................................................................................... 80
5.1.4 Comparison of abundance of caecal bacterial taxa in NOD, NOR and NOR-gavaged NOD Mice .............................................................................................. 81
5.1.5 Interstrain microbiome variation ........................................................................... 82
vii
5.2 The effect of Protecflor probiotic treatment on NOD microbiome composition and insulitis severity ................................................................................................................ 83
5.2.1 Summary ............................................................................................................... 83
5.2.2 Gastrointestinal Transit times of Protecflor Probiotic organisms in the NOD model ..................................................................................................................... 84
5.2.3 Evaluating the Effects of Protecflor Probiotic on Insulitis Severity in NOD model ..................................................................................................................... 85
5.2.4 The Effects of Protecflor Probiotic treatment on the gut microbiome in NOD mice ....................................................................................................................... 86
5.2.4.1 Taxa enriched in Protecflor treated mice ................................................ 86
5.2.4.2 Taxa depleted in Protecflor treated mice ................................................ 87
5.2.5 Principal components analysis of Protecflor and vehicle-treated caecal samples .................................................................................................................. 88
5.2.5.1 Bacterial taxa associated with T1D in NOD mice and humans ............. 89
5.2.5.2 Proteolytic Bacteria ................................................................................ 90
5.2.5.3 Short Chain Fatty Acid Producers .......................................................... 90
5.2.5.4 Enterotypes in NOD and T1D ................................................................ 91
5.2.6 Species richness and diversity in Protecflor and vehicle-treated animals ............ 92
References ..................................................................................................................................... 93
viii
List of Tables
Table 3-1: Taxonomic groups/OTUs with differential abundance selected from NOD vs. NOR 16S HTS
data
Table 3-2: Bacterial 16S taxa specific qPCR primers.
Table 3-3: qPCR-quantification of bacterial taxa
Table 4-1: Oligonucleotide primers specific for Protecflor organisms.
Table 4-2: Number of 16S rDNA sequencing reads per sample.
Table 4-3: PC1 OTU object loadings on Protecflor treated mouse caecal samples.
Table 4-4: PC2 OTU object loadings on Protecflor treated mouse caecal samples.
ix
List of Figures
Figure 3-1: 16S rDNA sequencing of female NOD and NOR caecal flush microbiome.
Figure 3-2: Quality analysis and validation for bacterial taxa-specific qPCR assays.
Figure 3-3: qPCR quantification of selected bacterial taxa in female NOD mice, ages 25-55.
Figure 3-4: qPCR quantification of selected bacterial taxa in female adult NOD and NOR mice.
Figure 3-5: qPCR quantification of selected bacterial taxa in female adult NOD, NOR-gavaged NOD and
NOR mice.
Figure 3-6: Gavage of female NOD pups with female NOR-derived intestinal microbiome.
Figure 4-1: Quality analysis and validation for Protecflor-specific qPCR assays.
Figure 4-2: In vivo transit time for Protecflor probiotic strains
Figure 4-3: Insulitis outcomes in female NOD pups with Protecflor probiotic.
Figure 4-4: Quality analysis of Illumina 16S rDNA sequencing of Protecflor and vehicle-treated mice.
Figure 4-5: Top 20 caecal and fecal microbial taxa in Protecflor and vehicle-treated mice.
Figure 4-6: Manhattan plot comparing enriched or depleted OTUs in Protecflor-treated vs vehicle-
treated animals.
Figure 4-7: Morisita-Horn distance heatmap of Protecflor-treated vs vehicle-treated caecal samples.
Figure 4-8: Scree plots of PCA comparing all bacterial OTUs across caecal flush samples from Protecflor
and vehicle-treated animals.
Figure 4-9: PCA comparing all bacterial OTUs across caecal flush samples from Protecflor and vehicle-
treated animals.
Figure 4-10: Unsupervised clustering analysis of caecal flush samples from Protecflor and vehicle-
treated animals.
Figure 4-11: Correlation of principal components with insulitis protection in Protecflor and vehicle-
treated mice.
Figure 4-12: Correlation of principal components with bacterial community richness.
Figure 4-13: Correlation of bacterial community richness with insulitis protection in Protecflor-treated
mice.
Figure 4-14: Correlation of bacterial community diversity with principal components in Protecflor-
treated mice.
Figure 4-15: Correlation of bacterial community diversity with insulitis protection in Protecflor-treated
mice.
1
Chapter 1 Introduction
2
1 Introduction
1.1 Introduction to T1D
Type 1 Diabetes (T1D) is an autoimmune disease which is characterized by the T cell mediated
destruction of insulin producing β cells in the pancreas, leading to loss of blood glucose
control(reviewed in [1]). The current treatment for T1D is exogenous administration of insulin
and frequent monitoring of blood glucose levels[1]. Despite treatment, T1D patients suffer
chronic hyperglycemic fluctuations resulting in significant comorbidities, including increased
risk of cardiovascular disease, retinopathy, nephropathy, and neuropathy[2]. Due to these
comorbidities, the global health burden of T1D is high, and is on the rise.
Global incidence of T1D has been rising over past 50 years[3]. Recent estimates in Canada
suggest an annual rate of 20 to 26 new cases per 100,000 children aged 0 to 14 years, among the
highest reported in the world[4]. The global rate is reported to be increasing by 232% per
year[5]. Clinical management of T1D has not changed significantly over the last several decades.
This contrast highlights the need for further research into the genetic/environmental interactions
which underlie the etiology of this disease.
1.2 Genetic Risk Factors
T1D is a complex disease involving multiple genes and epigenetic environmental modifiers.
Although ~90% of new T1D cases occur in individuals without family history of T1D, the
disease is known to be hereditary[6]. Family/pedigree studies have demonstrated that first degree
relatives of T1D patients have a 5% to 7% lifetime risk of developing T1D, as compared to 0.4%
risk in the general population[7]. The higher concordance rates in monozygotic twins than in
dizygotic twins further suggest genetic factors play a role in this disease[8]. Over the past
decades, multiple approaches have been used to identify loci associated with risk of T1D. Over
40 T1D-associated loci have been identified, and causal variants have been identified in a
number of them[9].
The strongest genetic contribution to risk is MHC class II haplotype (HLA). high-risk alleles at
this locus account for approximately half of the increased genetic risk relative to the general
population[10]. Other putatively causal single nucleotide polymorphisms (SNPs) have been
found in genes with immunological functions: PTPN22[11], IL2RA[12], CTLA4[13], IFIH1[14].
3
Despite the success of genome-wide association studies (GWAS) in identifying T1D
susceptibility loci, most of these SNPs display low effect sizes on the disease phenotype. It has
been estimated that all non-HLA SNPs identified to date to account for only 10.9% of T1D
variance observed[15]. This gap between the observed heritability of T1D and explanatory
power of known genetic factors is known as the “missing heritability.” This missing heritability
may be explained by multiple common alleles of weak effect, gene-gene, and/or gene-
environment interactions. Recent studies have suggested that the rare allele hypothesis is
unsupported for T1D[16].
1.2.1 Genetic link to inflammatory diseases of the gut
A number of T1D associated genetic loci are shared with inflammatory diseases of the gut, in
particular Crohn’s and celiac disease. ZNF365, MST1, ERBB3, and PTPN22 have been identified
as a common set of susceptibility genes in both Crohn’s disease and T1D[17]. T1D and celiac
disease share the some susceptible alleles of HLA-DQB1. Additionally, IL2, RGS1, TAGAP and
IL18RAP have been reported as susceptibility genes for both T1D and celiac disease[18]. T1D
patients are also at a higher risk of developing celiac disease compared to the general population.
Anti-tissue transglutaminase antibodies, a marker for celiac disease, are present in ~10% of
children and ~2% of adults with T1D[19]. This connection suggests underlying commonalities in
the genetic and environmental factors which cause T1D and autoimmune diseases of the gut.
1.3 Environmental modifiers
While there is strong evidence that genetics plays an important role in T1D, several pieces of
evidence point to the importance of environmental factors in T1D. Relatively weak concordance
between the incidence of T1D in monozygotic (“identical”) twins indicates involvement of
stochastic or environmental factors in the etiology of the disease. Studies have placed this
concordance rate for T1D between 30-80%[20]–[22]. Meta-analyses of T1D cohort studies over
the last 44 decades have also shown decreasing age at onset, as well as accelerated pathogenesis,
as measured by loss of beta cell mass[23]. Within a genetically stable population, rapid changes
in the development of the disease are not expected, and point to changing environmental factors.
The wide geographical distribution in T1D incidence also suggests gene-environment
interactions underlie the disease. Overall age-adjusted incidence of type 1 diabetes ranges from
4
0.1/100,000 per year in China and Venezuela to 36.8/100,000 per year in Sardinia and
36.5/100,000 in Finland[4]. Some studies have noted that this corresponds roughly to a “north-
south” gradient globally[4]. Other studies on geographic distribution have observed a correlation
between T1D incidence and latitude, both globally, and within countries; an Australian study,
found that T1D was three times prevalent in southerly region than in northerly regions[24]. Data
from migrant populations suggest that these geographical tendencies are not entirely dependent
on genetic factors. Immigrants tend to adopt the T1D incidence in their destination country and
have rates of T1D which approach those of the indigenous population[25], [26]. Furthermore,
adjacent countries with similar genetic backgrounds can often have divergent T1D incidences; a
six-fold difference in T1D incidence is observed at the eastern Finland-Russian border[27].
These data highlight the need to investigate environmental modifiers in T1D in a controlled
system.
1.4 NOD Model of T1D Genetics & Environment
Two rodent models of T1D, the Bio-Breeding (BB) Rat[28] and the Non-Obese Diabetic (NOD)
mouse model[29], have enabled studies of the genetics and gene-environment interactions that
affect T1D incidence and pathogenesis. The NOD mouse is an inbred mouse strain with a high
incidence of spontaneous T1D. Unlike in humans, T1D in the NOD displays a marked sex bias,
with ~80% of females and ~30% of males becoming diabetic at 6 months of age[30]. The disease
displays many of the known features of T1D in humans. Like in humans, T1D in the NOD
mouse is polygenic, with over 30 Insulin-Dependent Diabetes (Idd) loci identified[31].
Furthermore, the MHC haplotype contributes the highest genetic risk in both humans[10] and
NOD mice[32]. The NOD mouse model also shares a number of other susceptibility loci and
immune pathways with the human disease.T1D in NOD mice also displays much of the
immunopathology seen in human T1D: auto-antibody production[33] and APC-mediated
lymphocyte infiltration of the islets, termed insulitis[34].
To narrow the study of these genetic susceptibility loci in the NOD model, our lab has compared
NOD mice to the closely-related Non-Obese Resistant (NOR) mice, an inbred strain with genetic
material derived from NOD and bk/s strains[35]. NOR shares 88% of the NOD genome by
descent, and differs from NOD at only 4 of the 20 Idd loci identified; Idd4, Idd5, Idd9,
Idd13[36][37][38, p. 13][39]. However, NOR is completely resistant to T1D, although it exhibits
5
some of the inflammatory pre-clinical traits which precede T1D in NOD mice and humans.
Using NOD background congenic strains harboring NOR intervals, our lab has previously
mapped early T cell recruitment events in the initiation of insulitis to Idd5 and Idd13, and found
these loci to regulate a T cell-dependent progression from a benign to a destructive stage of
insulitis[40]. Work in our lab and others has also identified candidate genes in the four loci
which differ between NOD and NOR. In conclusion, these two strains provide a system to study
the genetics of T1D in a model through the use of inter-crossing and genetic mapping.
1.4.1 Environmental modifiers in NOD
Like in the human disease, there is evidence of environmental modifiers of T1D in NOD mice.
Meta-studies compiling natural history and other data from NOD colonies in multiple facilities
have noted a variable diabetes rate, ranging from 50%-80% T1D incidence in female NOD
mice[30]. It has been reported that T1D incidence is inversely correlated with reports of
infectious disease in colonies[30]. Furthermore, generations of NOD mice re-derived from
conventional conditions to SPF showed a twofold increase in T1D incidence[41]. These data
suggest a role for microbial exposure in modulating T1D incidence. This is further supported by
studies showing deliberate infection with multiple organisms, including Schistosoma
mansoni[42], Mycobacterium avium[43], Trichinella spiralis[44], Heligmosomoides
polygyrus[45] and Salmonella typhimurium[46], protects NOD mice from diabetes.
Interperitonial injection of zymosan, a fungal cell wall component, has also been found to be
protective in the NOD model[47]. The potential mechanisms responsible for these protective
effects include T helper cell type 2 (Th2) skewing, regulatory T cell (Treg) and natural killer T
(NKT) cell activity, and modulation of antigen presenting cells (APCs) towards a diabetes-
protective phenotype.
In addition to systemic antigen exposure, recent studies have also shown that the intestinal
microbiota and the innate immune system in NOD mice can influence the development of T1D.
NOD mice genetically deficient in MyD88 were protected from T1D and resulted in altered gut
microbiota[48]. Transfer of NOD.MyD88-/--derived gut microbiota into germ-free NOD mice
protected against T1D in these mice[48]. Oral exposure of NOD mice to a number of non-
infectious bacterial strains has also been shown to be protective. The probiotic mixture VSL#3; a
mixture of Bifidobacterium and Lactobacillus species and Streptococcus salivarius subsp.
6
thermophiles was found to be protective of SPF-colonized NOD mice if given by oral gavage
three times a week from 4 to 32 weeks of age[49].
Changes in the intestinal microbiota of mice have also been found to induce hormonal and
metabolic changes leading to protection against T1D. In the NOD model, T1D incidence exhibits
a distinct sex bias, with ~80% of females and ~30% of males becoming diabetic at 6 months of
age[30]. This bias is colonization dependent, as the sex bias is absent in germ-free NOD mice.
Work in our lab showed that the transfer of male NOD-derived microbiota into weanling female
NOD mice by oral gavage significantly protected them against T1D, and induced changes in
testosterone levels and the metabolomics profile of treated mice[50].
These data suggest that changes in the gut microbiota of NOD mice may be correlated with
protection, either by direct interaction with the immune system, or by modifying other
physiological factors with known influence on T1D incidence. These and other data lead to the
idea that microbial exposure may be an important environmental factor in auto-immune diseases
in both mice and humans.
1.5 The Hygiene Hypothesis
Epidemiological meta-studies on infectious and autoimmune diseases have shown a dramatic rise
in the incidence of T1D and other autoimmune diseases, and a dramatic decline of infectious
diseases[41]. This decline is mostly attributable to advances in medicine and sanitation;
antibiotics, improved hygiene have helped to manage infectious disease in the developed world.
Epidemiological evidence suggests that the increased incidence of autoimmune diseases is linked
to improved socio-economic status, but does not identify a specific environmental modifier. This
raises the hypothesis that there exists a connection between the rise in autoimmune diseases and
the decline in infectious diseases, called the hygiene hypothesis.
The hygiene hypothesis posits that advances of medicine and sanitation have altered the
microbial environment to which humans are exposed, and that this altered environment has led to
an increase in auto-immune and auto-inflammatory diseases[51]. Correlations have been found
between the increasing prevalence and severity of asthma among children in developed nations
and improved sanitation and widespread antibiotic use[52]. Recent studies suggest there is a
‘critical window’ in the first 6 months of life; during which immunological developments in
7
infancy may be influenced by the commensal microbial community might alter allergic
sensitization later in life[53]. It has been suggested that rising autoimmune disease incidence
may reflect immune responses which were evolved to survive infection, but are now maladaptive
in light of the reduced immunological challenge due improved public health. Alternatively,
alterations to our microbiota due to decreased exposure to microbes have resulted in the loss of
immunological beneficial microbiota. The hygiene hypothesis raises some interesting
possibilities, but more work is needed to identify mechanisms linking microbial exposures to
immune mediated disease.
1.6 The Gut Microbiome
The microbiome is the collection of microbes which live as a complex community/ecosystem in
or on a host organism. This collection of microbes consists of hundreds of species and strains,
each with their own complex metabolism and products. Commensal (non-pathogenic) microbes
on all colonized body surfaces out-number human cells 10:1, while the human gut alone harbors
roughly 1.5 kg of bacteria, with a metagenome of over 3 million genes[54].
The gut microbiome of an individual develops soon after birth. Before birth, the gut is an
essentially sterile environment. Soon after birth, the mammalian neonate’s gut is colonized by
mainly maternally-derived bacteria; vaginal bacteria for vaginally delivered babies or skin-
associated bacteria following cesarean section deliveries[55]. After a brief period of instability,
phylogenetic diversity increases over time. The microbiome of children begins to transition to
more similar to the adult state starting at 0.6-1 year of age[56]. In a stable gut microbiome,
~1,000 species are represented, over two thirds of which are members of two phyla,
Bacteroidetes and Firmicutes[57].
The microbiome influences multiple aspects of host physiology, including nutrition, metabolism
and immunity. The presence and composition of the gut microbiome has a profound influence on
the development and function of the immune system as a whole. The development of the gut-
associated immune system is altered in the absence of the microbiome; in germ-free mice, GALT
and Peyer’s patches are deficient in lymphocytes, and IgA production is essentially
abrogated[58].
8
Furthermore, a number of bacterial groups have been identified as immunomodulatory in mice.
Bacteroides fragilis, a human commensal, induced systemic Th1 responses in mice mediated by a
surface polysaccharide, polysaccharide-A (PSA)[59]. Segmented filamentous bacteria (SFB)
were found to be associated with the abundance of lamina propria Th17 cells in the gut of mice,
and were found to be sufficient to generate Th17 cells in germ-free mice mono-colonized with
SFB[60]. Although SFB have not been identified in humans, it may be that other bacterial strains
play similar roles in humans. Clostridium species belonging to clusters IV and XIVa were found
to induce differentiation of inducible Foxp3+ and IL-10+ Tregs in the mouse colonic lamina
propria[61]. 17 strains of human-derived Clostridium species belonging to clusters IV, XIVa,
and XVIII (recently reclassified as the class Erysipelotrichi) were found to induce Treg expansion
and functional potency by upregulation of IL-10 and ICOS in mono-colonized mice. It is clear
that some commensals can modulate immune cell subsets in the host, in both humans and mice.
Other bacterial strains have also been found to be associated with effects in models of
autoimmunity. SFB have been found to co-segregate with T1D protection in the NOD
mouse[62]. In another study, SFB was also found to drive autoimmune arthritis in germ-free
mice via promotion of the Th17 subset[63]. Faecalibacterium prausnitzii was found to be
associated with a lower risk of postoperative recurrence of ileal CD in Crohn’s patients and
reduced the severity of TNBS-induced colitis in a murine experimental model of colitis[64]. It is
clear that the gut microbiome and alterations in intestinal microbiota can influence incidence in
rodent models of T1D, and other models of autoimmune disease.
1.7 Tools and methods for investigating microbiome composition
The detection and quantification of microbes in the gut has been integral to our growing
understanding of the effects of both single bacterial strains and broad community structure of
host physiology. Current technologies for investigating microbiome composition are largely
dependent on sequence variation amongst bacteria in highly conserved genes. The 16S rRNA
gene is the most commonly used gene as it is conserved across most prokaryotes[65]. Pan-
prokaryotic multiple alignments of 16S rRNA gene sequences reveal regions of high interspecies
conservation, and regions of hypervariability[66]. Sequence variation in these hypervariable
regions is proportional to phylogenetic distance and therefore can be used to phylogenetically
classify bacterial sequences.
9
By designing degenerate oligonucleotide primers complementary to taxa-specific consensus
sequences within the hypervariable regions, PCR amplification of sequences only from templates
derived from specific bacterial taxa is possible. In a quantitative PCR (qPCR) reaction, this
specific amplification can be used to quantify 16S rRNA copy numbers from specific taxa,
providing taxa-specific bacterial quantification, even from a complex mixture.
Conversely, libraries of bacterial 16S rDNA representative of the community variation can be
generated using degenerate PCR primers complementary to conserved regions, amplifying into
hypervariable regions. Combined with high-throughput DNA sequencing, sequence variation in
these hypervariable regions can be used to quantify and phylogenetically classify bacteria in
complex community. Sequences are classified into operational taxonomical units (OTUs), which
represent the lowest level of taxa reliably assignable to each sequence. Data from high-
throughput 16S sequencing can be further analyzed using multivariate analysis techniques, which
are mainly aimed at reducing the complexity of large multidimensional datasets.
One of the main methods of dimensional reduction is Principal Components Analysis (PCA).
This is done by finding a set of orthogonal linear combinations of the original variables, called
principal components, which account for the greatest variance in a dataset. The number of
significant principal components is restricted to those which represent most of the variances,
thereby representing the microbial composition of a gut community with a greatly reduced
number of variables. By looking at the component variables and their weightings, called
loadings, which make up each principal component, factors which influence the composition of a
microbiome can be inferred. The distance in a PCA plot between two microbiomes can indicate
degree of compositional similarity, and clustering analysis can be used to further group
microbiomes by similarity.
Ecological community metrics can be used for further comparison of microbiome 16S
sequencing data. The two main metrics use are community richness and diversity. More
qualitatively similar communities often have similar richness and diversity. Richness is an
estimate of the true total number of OTUs in a sample, regardless of the distribution of sequences
across these OTUs. The Chao1 statistic is a measure of sample richness [67] which estimates the
true total number of OTUs in a sample based on the number of singletons (OTUs observed once
in the sample), and doubletons (OTUs observed twice in the sample) observed [67].
10
The second most common community metric used is diversity. This statistic depends on the
number of OTU observed in a sample, and the evenness of the distribution of individual samples
across OTU categories[68]. Shannon’s H statistic is a commonly used measure of the diversity or
the predictability of a given OTU in a sample[69]. A higher Shannon’s H value indicate higher
entropy or “surprise” in a sample; the next OTU to be called is harder to predict [68].
1.8 Previous Work
Previously in our lab, we performed 16S sequencing on female NOD and NOR mice (Figure
3-1). Our data shows there are quantifiable differences between the gut microbiota of NOD and
NOR mice, most notably in the 5 taxonomical groupings: Bacteroidetes, Clostridia IV,
Ruminococcus, Clostridia XIVa, and Helicobacter (Table 3-1). Given their shared
environmental exposures under controlled, laboratory conditions, these differences suggest that
there are genetic factors which influence the composition of the gut microbiota. Previously, our
lab also showed the transfer of male-derived microbiome into recipient NOD females resulted in
a durable alteration in the recipient microbiome, alterations in metabolomics profiles in serum,
and a protective effect on T1D in recipient females[50]. These results suggest differences in
microbiome composition can affect the risk of T1D and other autoimmune diseases in the host.
Together, these data lead to the hypothesis that these differentially abundant bacterial groups in
NOD and NOR mice may contribute to the protection of NOR mice from T1D, and that the
transfer of these organisms to NOD mice may result in durable changes to the gut microbiome
composition, and afford protection against T1D.
1.9 Project Rationale
1.9.1 The effect of NOR-derived microbiome transfer into NOD on microbiome composition and T1D incidence
The NOD mouse is an inbred mouse strain with a high incidence of spontaneous T1D. NOR
mouse is a strain closely related to the NOD, but is protected by the genetic contributions from
the C57BLKS/J strain. In the NOD model, T1D incidence exhibits a distinct sex bias, with ~80%
of females and ~30% of males becoming diabetic at 6 months of age[30]. This bias is
colonization dependent, as the sex bias is absent in germ-free NOD mice. Work in our lab
showed that the transfer of male NOD-derived microbiota into weanling female NOD mice by
11
oral gavage significantly protected them against T1D, and induced changes in the microbiome
composition, testosterone levels and metabolomics profile of treated mice[50].
We hypothesized that the T1D protection afforded by genetic contributions from the C57BLKS/J
strain in the NOR strain might also be linked to microbiome composition. Previous work in our
lab found 5 bacterial taxa (Bacteroidetes, Clostridia IV, Ruminococcus, Clostridia XIVa, and
helicobacter) that were differentially abundant between NOD and NOR animals. Other studies
have suggested that members of these taxa may have immunomodulatory effects. The first goal
of my work was to investigate the possible protective effects of NOR-derived microbiome
transplant in the NOD model of T1D.
1.9.2 The effect of Protecflor probiotic treatment on NOD microbiome composition and insulitis severity
T1D and gut inflammatory disorders have been found to share genetic risk factors, and can occur
as co-morbidities suggesting shared genetic and environmental factors in T1D and autoimmune
diseases of the gut. Studies in ulcerative colitis, inflammatory bowel syndrome and T1D have
suggested that changes in the gut microbiome associated with treatment with probiotic bacteria
such as Lactobacillus and Bifidobacterium can have protective effects in these diseases.
Probiotics are generally derived from fermented food products or the gut microbial communities
of ruminants. They are generally cultured organisms and relatively well-characterized compared
to most of the strains in complex microbial communities such as the gut.
Studies in ulcerative colitis, inflammatory bowel syndrome and T1D have suggested that specific
probiotic bacterial strains can have protective effects in these diseases. For example,
Lactobacillus helveticus R0052 and Lactobacillus rhamnosus R0011 have been shown to reduce
clinical symptoms in open label experiments on human adults with inflammatory bowel
syndrome[61], and to reduce epithelial injury due to Escherichia coli O157:H7- and O127:H6-
infection in an in vitro gut epithelial cell co-culture system [70]. Bifidobacterium longum R0175,
given in combination with inulin, a plant polysaccharide, was found to improve self-reported
quality of life, symptoms of blood and mucous in stool, and to reduce relapse rate in a small
pediatric study of ulcerative colitis (UC) [71].
12
We hypothesized that continuous treatment with Protecflor probiotic, a formulation containing
Lactobacillus helveticus R0052, Lactobacillus rhamnosus R0011, and Bifidobacterium longum
R0175 and the yeast Saccharomyces cerevisiae boulardii, would impact the gut microbial
composition of diabetes-prone NOD mice, and consequently, have an effect on the progression
of islet autoimmunity.
13
Chapter 2 Materials and Methods
14
2 Materials and Methods
2.1 Mice
NOD/Jsd (NOD) and NOR/Jsd (NOR) Specific Pathogen Free (SPF) mice used in this study
were maintained at The Hospital for Sick Children (Toronto), housed in sterilized static caging.
The animals received standard mouse diet (LabDiet #5015, PMI Nutrition International) and
autoclaved water. All staff wore autoclaved gowns, caps, masks, shoe covers, and sterile gloves.
Animal handling and cage changes were done under HEPA filtered air. The health status was
determined by weekly exposure of CD-1 sentinels to soiled bedding from the cages in the room.
Quarterly serological testing of sentinels confirmed the mice were negative for: Mouse Hepatitis
Virus, Minute Virus of Mice, Mouse Parvovirus, Murine Norovirus, Sendai Virus, Theiler's
Murine Encephalomyelitis, Reovirus and for endo- and ectoparasites. The results of sentinel
testing indicated presence of Murine Norovirus and Helicobacter sp. in the room. In addition,
live animals were subjected to additional, annual comprehensive necropsy, histopathology,
bacteriology and parasitology testing. All experiments were carried out in accordance with the
Institute animal utilization protocols and Canadian Council on Animal Care (CCAC) guidelines.
2.2 Design and Validation of Taxa-Specific qPCR
16S rRNA gene sequences belonging to the target taxa were retrieved from the RDP
database[72] and aligned in PRIMROSE[73]. Primer sequences were then found in consensus
regions, with a target theoretical Tm of 60C. These preliminary candidates were validated in
silico by performing ProbeMatch searches on the RDP database for spurious amplicons (matches
to 16S rDNA sequences classified out of the target taxa).
In vitro validation of primer candidates which passed in silico validation was performed by the
following pipeline: the candidate oligonucleotide primer sequences were sent for synthesis
(IDT). PCR (see below) was performed on template DNA (see below for bacterial DNA
extraction) derived from female NOD caecal contents. PCR products were run on 2% agarose gel
electrophoresis, and visualized using ethidium bromide staining and UV trans-illumination.
Approximate molecular weight/amplicon lengths were determined by comparison with a
concurrently run 100bp DNA molecular weight ladder (FroggaBio). Bands at the predicted
amplicon sizes (Table 3-1) were excised using a razor blade and PCR products were extracted
15
using the QIAquick Gel Extraction Kit (Qiagen) according to manufacturer’s instructions. PCR
amplicons were cloned into pCR2.1 plasmids using TOPO TA-cloning according to
manufacturer’s instructions (Life Technologies) and transformed into chemically-competent
DH5α E. coli for cloning and selection. Transformed bacteria were plated on LB agar containing
100 µg/mL ampicillin and 40 µg/mL x-Gal for blue-white selection. Individual colonies were
picked and grown in LB broth.
Plasmid DNA was extracted using QIAprep Spin Miniprep Kit (Qiagen) according to
manufacturer’s instructions, and sent for Sanger sequencing by TCAG using M13F and M13R
primers. At least 20 clones for each primer set were Sanger sequenced. Sequences were pre-
processed to remove pCR2.1 sequences in MEGA4[74] and taxonomically classified by the web-
interface RDP classifier[75] using RDP 16S rRNA training set 9 and a confidence threshold of
80% for family-level taxonomical assignment, and 50% for genus-level assignment[76] to
validate specificity of designed primers to the target taxa.
2.3 PCR
PCR amplification was conducted in polypropylene 12-cap strips on GeneAmp PCR system
9700 (Applied Biosystems). Each 25-µl reaction contained the following: 2.5 µL of ThermoPol
reaction buffer (New England bioLabs), 0.5 µL of 10 mM dNTPs, 0.5 µL of each primer (200
µg/mL; IDT), 0.125 µL Taq DNA polymerase (5000 U/mL), 10 ng template DNA and H2O to
make up the volume to 25 µL. PCR conditions were 30s at 95°C, followed by 35 cycles of 95°C
for 15s, 60 s at the annealing temperature, and 1 min at 68°C.
PCR products were run on 2% agarose gel electrophoresis, and visualized using ethidium
bromide staining and UV trans-illumination. Approximate molecular weight/amplicon lengths
were determined by comparison with a concurrently run DNA molecular weight ladder
(FroggaBio).
2.4 qPCR
qPCR assays were conducted in polypropylene 96-well plates on an ABI Prism 7900HT
(Applied Biosystems). Each 25-µl reaction contained the following: 12.5 µL of
PowerSYBRGreen PCR Master Mix (Applied Biosystems), 0.5 µL of each primer (200 µg/mL;
16
IDT), 10 ng template DNA and H2O to make up the volume to 25 µL. PCR conditions were 10
min at 95°C, followed by 40 cycles of 95°C for 15s, 60 s at the annealing temperature.
Each plate included triplicate reactions per DNA sample and the appropriate set of standards and
DNA-negative controls. Melting curve analysis of the PCR products was conducted following
each assay to confirm that the fluorescence signal originated from specific PCR products and not
from primer-dimers or other artifacts.
2.5 Oral gavage transfer of the gut microbiome
Weanling NOD females (aged 22-26 days) were gavaged with diluted caecal contents from adult
(10-12 weeks) female NOR mice, according to previously published methods[50] with slight
modifications. Briefly, the cecum of one adult microbiome NOR donor was dissected, opened
along its length, the contents transferred to a sterile tube, and resuspended in 50 volumes of
sterile water. 250 µL of this suspension was given to each recipient by oral gavage using a 24G
round tip gavage needle. Recipients were rested for 24 hours, and this procedure was repeated
once.
2.6 T1D assessment
Blood and/or urine glucose levels were measured in females biweekly. Animals were classified
as diabetic when blood glucose exceeded 16mmol/L or urine glucose exceeded 250mg/dL.
Statistical analysis on T1D life table data was performed using the Log-rank (Mantel-Cox) test,
on Prism 5.0b (GraphPad Software, San Diego California USA)
2.7 Treatment protocol for Protecflor probiotics
A commercially available probiotic mixture consisting of Lactobacillus rhamnosus strain R0011,
L. acidophilus strain R0052, Bifidobacterium longum R0175 and the yeast Saccharomyces
cerevisiae boulardii (Protecflor) and equivalent microbe-free vehicle was provided by Institut
Rosell-Lallemand. Weanling NOD females (aged 22-26 days) were gavaged with 1x109 CFU of
Protecflor probiotic suspended in water, or the equivalent weight of Protecflor vehicle according
to previously published methods[50] with slight modifications. Briefly, 0.017 g of Protecflor
probiotic or vehicle was suspended in 1 mL of distilled water. 250 µL of this suspension was
17
given to each recipient by oral gavage using a 24G round tip gavage needle. Recipients were
rested for 24 hours, and this procedure was repeated once.
To maintain treatment of probiotic recipients post-gavage, Protecflor probiotic or vehicle was
administered by suspension in drinking water. Briefly, 0.691g or 2 x 108 CFU/mL of probiotic
formulation or vehicle was suspended in 200 mL of sterile drinking water and provided as the
sole drinking water source for treated mice. Treated water was changed every 48 hours. Water
consumption was measured and recorded to calculate an approximate per day CFU dose, and to
ensure normal water consumption.
2.8 Insulitis assessment
Pancreata were dissected and immediately immersed in OCT media (Tissue-Tek, Torrance, CA),
frozen in -20°C 2-methylbutane, and stored at –70°C. Preparation of frozen sections was
performed with a Leica CM 3050 Cryostat (Leica Canada). To maximize analysis of independent
islet infiltrates, three 5-µm sections were cut at least 400 µm apart. Pancreatic sections were
stained with Mayer’s hematoxylin and eosin Y (H+E, Sigma) to visualize leukocyte infiltration.
Assessment of insulitis severity in pancreatic sections was performed as previously described by
our group [50][36][37]. Briefly, islets were graded according to the following criteria: 0, no
visible infiltrates; 1, peri-insulitis as indicated by peri-vascular and peri-islet infiltrates; 2, <25%
of the islet interior occluded by leukocytes displaying invasive infiltrates; 3, >25% but <50% of
the islet interior invaded by leukocytes; or 4, invasive insulitis involving 50%-100% of the islet
field. Statistical analysis on insulitis severity data was performed using the χ2 test, on Prism 5.0b
(GraphPad Software, San Diego California USA.
2.9 Bacterial 16s rDNA PCR library construction
Bacterial profiles were analyzed by broad-range PCR of 16S rRNA genes and phylogenetic
sequence analysis. DNA was extracted using the UltraClean fecal DNA kit (MoBio, Inc).
Amplicons of the 16S rRNA gene were generated via broad-range PCR (30 cycles) using
barcoded primers [77] modified with sequences required for either 454 pyrosequencing (~500
b.p. V1V3 region; primers 27FYM+3 and 515R) [78], or Illumina sequencing-by-synthesis
(~250 bp V4 region; primers 538F and 806R) [79]. PCR yields were normalized using a
SequalPrepTM kit (Invitrogen, Carlsbad, CA), pooled, lyophilized, and gel purified, as previously
18
described [80]. Pyrosequencing was performed at The Center for Applied Genomics at the
Hospital for Sick Children on a 454/Roche Life Sciences GS-FLX instrument using Titanium
chemistry (Roche Life Sciences, Indianapolis, IN). Illumina paired-end sequencing was
performed at the University of Colorado, Denver by collaborator Dr. Dan Frank’s laboratory on a
MiSeq sequencing platform using the 500-cycle MiSeq Reagent Kit version 2.
2.10 454 pyrosequencing of NOD and NOR microbiome samples
Pyrosequences were sorted into libraries by barcode and quality filtered using bartab[77]. All
pyrosequences were screened for nucleotide quality (bases at 5’ and 3’ ends with mean Q <20
over a 10 n.t. window were discarded), ambiguous bases (sequences with >1 N were discarded),
and minimum length (sequences <200 n.t. were discarded). Infernal [81] and ChimeraSlayer [82]
were used to screen for bacterial 16S rRNA sequences as described previously[83]. Sequences
were classified by parsimony insertion into the Silva108 guide tree after alignment with SINA
using the ARB software package[84][85][86]. Sequences were then assigned the taxonomies of
the clades into which they were inserted, using the ARB export node display setup (NDS)
function. Operational taxonomic units (OTUs) were determined by clustering sequences with
identical taxonomic assignments.
2.11 Illumina paired-end sequencing and analysis
Illumina Miseq paired-end sequences were sorted by sample via barcodes in the paired reads
with a python script. The sorted paired reads were assembled using phrap[87][88][89].
Assembled sequence ends were trimmed over a moving window of 5 nucleotides until average
quality met or exceeded 20. Trimmed sequences with more than 2 ambiguities or shorter than
200 nt were discarded. Potential chimeras identified with Uchime [90] using the Schloss Silva
reference sequences [91] were removed from subsequent analyses. The remaining 4,578,881
sequences were aligned and classified with SINA (1.2.11)(47) using the 244,077 bacterial
sequences in Silva 111NR (55) as reference configured to yield the Silva taxonomy. Sequences
with >97% similarity (ie., genus level) were grouped into OTUs. Before analysis, OTUs
representing less than 0.005% of all reads (228 reads) were discarded.
Ecological statistics (such as taxa observed, Good's Coverage estimator, Morisita-Horn
index[92], Chao1 statistic[67], Shannon’s H statistic[69] and Wilcoxon signed-rank test),
19
Manhattan plots and heatmaps were prepared with the Explicet software package[93]. Chao1
statistic, Shannon’s H statistic indices were estimated through bootstrap resampling (1000
replicates) of the OTU distributions obtained from each specimen. Collector’s curves were
prepared in Explicet. All ecological statistics were generated at a sample rarefaction of 40795
reads, representing the sample with the lowest reads sequenced. All 16S amplicon libraries were
sequenced to >95% coverage. Individual microbial groups that differed in prevalence or
abundance between treatment groups were identified by Wilcoxon rank-sum test. Because of the
exploratory nature of this study, p-values were not corrected for multiple tests.
2.12 Multivariate data analysis
2.12.1 PCA
Multivariate data analysis (principal components analysis and clustering analysis) was performed
using Ginkgo[94]. Principal components analysis was performed using Ginkgo [94]. Relative
abundance data were standardized by centered-log transformation, then PCA performed on the
resulting correlation matrix. The number of principal components to display as significant was
determined by the “inclusive elbow” method in which the significance cutoff is set at the next
component past the largest difference in variance explained. PCA displays were then generated.
2.12.2 Clustering Analysis
A K-medians/fuzzy C-medians clustering was performed using Ginkgo. K-medians/fuzzy C-
medians clustering was performed with randomized seeding and 2-5 starting clusters. Clustering
was evaluated by Dunn’s Coefficient; with 2 starting clusters yielding the best clustering
(Dunn’s Coefficient = 0.86180). A single linkage hierarchical agglomerative method using Bray-
Curtis dissimilarity matrix[95] was used independently used to corroborate the K-medians/fuzzy
C-medians clustering.
2.12.3 Statistical Analysis
Spearman rank correlation tests were performed in Prism5.0b to evaluate associations between
two variables. Fisher’s exact tests and χ2 tests were performed in Prism 5.0b for comparisons of
categorical data. A P-value of <0.05 was considered significant. False discovery rate corrections
were not performed and P-values are uncorrected.
20
Chapter 3 The effect of NOR-derived microbiome transfer into NOD on
microbiome composition and T1D incidence
21
3 The effect of NOR-derived microbiome transfer into NOD on microbiome composition and T1D incidence
3.1 Background and Rationale
Recent studies have shown that the intestinal microbiota and the innate immune system in NOD
mice can influence the development of T1D. NOD mice genetically deficient in MyD88 were
protected from T1D and 16S sequencing showed MyD88 deficiency resulted in altered gut
microbiota[1]. Transfer of NOD.MyD88-/--derived gut microbiota into germ-free NOD mice
protected against T1D in these mice[1]. Changes in the intestinal microbiota of mice have also
been found to induce hormonal and metabolic changes leading to protection against T1D.
Previous work in our lab by Markle et al. [2] showed that the transfer of male NOD-derived
microbiota into weanling female NOD mice by oral gavage significantly protected them against
T1D, and induced changes in testosterone levels and the metabolomics profile of treated mice.
These data suggest that changes in the gut microbiota of NOD mice may be correlated with
protection, either by direct interaction with the immune system, or by modifying other
physiological factors with known influence on T1D incidence. These and other data lead to the
idea that microbial exposure may be an important environmental factor in auto-immune diseases
in both mice and humans.
Previously in our lab, we performed 16S sequencing on female NOD and NOR mice (Figure
3-1). Our data shows there are quantifiable differences between the gut microbiota of NOD and
NOR mice, most notably in the 5 taxonomical groupings: Bacteroidetes, Clostridia IV,
Ruminococcus, Clostridia XIVa, and Helicobacter (Table 3-1). Given their shared
environmental exposures under controlled, laboratory conditions, these differences suggest that
there are genetic factors which influence the composition of the gut microbiota. These results,
suggest that these differentially abundant groups in NOD and NOR mice may contribute to the
protection of NOR mice from T1D, and that the transfer of these organisms to NOD mice may
afford protection.
These preliminary data suggest that, similar to our observations in NOD males vs. females, the
NOR female caecal microbiota may also transfer protection to NOD females. Therefore, we
established a system to test whether transfer of caecal contents from NOR into NOD animals
22
provokes a change in the recipient’s microbiota and whether such changes were associated with a
T1D-protective effect on recipient NOD mice.
3.2 Results
3.2.1 Design and Validation of Taxa-Specific qPCR
To assess the impact of these caecal microbe transfers, we developed an assay to assess
alterations to the recipient microbiome. We expected the changes in the recipient microbiomes to
include differentially abundant taxa we had detected in the NOD vs. NOR female 16S rDNA
HTS analysis. Therefore, taxa-specific qPCR was chosen to quantify these bacterial groups as it
allowed greater sample throughput, and was cheaper than high-throughput sequencing.
qPCR amplification assays were designed to provide specificity for the targeted taxa
differentially abundant in NOD vs. NOR females. Related bacterial species display sequence
similarity in some regions of the 16S rRNA gene[3]. We selected 5 bacterial taxa (Bacteroidetes,
Clostridia IV, Ruminococcus, Clostridia XIVa, and Helicobacter) based on their differential
abundance in NOD and NOR mice (Table 3-1). We designed primers by aligning of the 16S
rRNA gene from sequenced members of the target taxa (Table 3-1) as assigned in the RDP
database[4], then locating regions of the 16S rRNA sequences that were conserved only within
targeted taxa using PRIMROSE[5]. These taxa-specific qPCR primers were then tested in silico
by performing Probe Match searches using the RDP web interface against RDP 16S rDNA
database entries [4]. Primers which matched sequences in RDP from non-target taxa were
redesigned.
Once a set of primers which passed in silico testing were found, we used these primers to
produce amplicons from NOD and NOR caecal extract DNA by PCR. we then cloned and
separately sequenced 20 amplicons by Sanger sequencing and identified the amplified sequences
BLASTing against available reference bacterial genome sequences. Any primers which produced
amplicons identified as members of a non-target taxa were discarded and redesigned. The final
primer sets, which passed both stages of specificity testing are shown in Table 3-2.
The next requirement for the qPCR assay was to ensure consistent quantification and
normalization between assay runs. The selected primers were used in qPCR reactions with NOD
caecum-derived bacterial DNA as template. Serial dilutions of NOD caecal flush-derived DNA
23
were quantified by qPCR using each of the primer sets. The copy number of each template type,
determined by a standard curve, was plotted against the dilution of the template DNA (Figure
3-2a). All primer sets showed linearity between 25ng/µL and 25pg/µL of template DNA (R2 >
0.99). To test their reproducibility, these assays were then repeated over 3 separate days, and
found to display no significant day-effects (p=0.48). These results demonstrated that the qPCR
assays provided consistent quantification over multiple experimental replications. PCR amplicon
copy numbers, determined by a standard curves, were plotted against average threshold cycle
(Ct) values to compare amplification efficiencies (Figure 3-2b). The amplification efficiency in
qPCR assays for each of the 5 taxa varied by ≤5%. The consistency in amplification efficiencies
suggested that quantification between assays would be comparable without inter-assay
normalization.
3.2.2 Comparison of Abundance of Caecal Bacterial Taxa in NOD Mice from Weaning to Adulthood
The gut microbiome is in a dynamic state of flux between weaning and adulthood in both mice
and humans[6][7]. In humans, multiple successions of dominant microbial groups occur until the
gut microbiome stabilizes at ~1-2 years of age [8]. Schloss et al. found that in C57BL/6 mice, the
microbiome was found to from a less stable state 10 days post-weaning to a more stable state 15
days post-weaning[9]. We reasoned that developmental changes in the mouse microbiome would
provide a source of variation to test our taxa-specific qPCR assays.
To test if the taxa-specific qPCR assay panel could detect changes in the microbiome from
weaning to adulthood, DNA from caecal contents from female NOD mice ages 24-55d was
extracted and assayed by qPCR (Figure 3-3). We were able to detect fluctuations in the qPCR
profiles of the gut microbiome across this age range for the 5 target taxa. This result suggested
that the taxa-specific qPCR assays had the dynamic range and sensitivity to detect the variation
in the gut microbiome of mice, as represented by the changes between weaning and adulthood.
qPCR profiles of the gut microbiome showed the most fluctuation between 20-42 days of age,
but changes became less pronounced after 42 days. This observation suggests that by 42 days,
the gut microbiome begins to stabilize and resemble the meta-stable adult state[7].
24
3.2.3 Comparison of abundance of Caecal Bacterial Taxa in NOD and NOR Mice
Previous 16S sequencing experiments had identified differentially abundant bacterial taxa in the
caecum of NOD and NOR mice (Figure 3-1). We hypothesized that these taxonomic groups
might contribute to the T1D protection observed in NOR compared to NOD mice. Moreover, we
proposed that that transfer of caecal contents from female NOR into female NOD animals
(NOR→NOD) might impart a protective effect on the recipients. We expected that changes in
the microbiome of NOR→NOD mice would include those differentially abundant taxa detected
in of 16S rRNA HTS study (Figure 3-1 and Table 3-1). Therefore we used the qPCR assays to
quantify these differentially abundant taxa in DNA extracted from caecal samples from
unmanipulated NOD and NOR mice (Figure 3-4).
We observed differences between NOD and NOR in the Bacteroidetes, Clostridia IV, and
Helicobacter groups (p = 0.0015, 0.0019, 0.0288, respectively) (Figure 3-4a). We observed a
difference trending towards significance in the Ruminococcus group (p = 0.0606) (Figure 3-4a).
The relative abundance of these taxa as quantified by qPCR is consistent with the results from
the 16S sequencing experiment, although exact proportional difference between mouse strains
for each taxa differed between the two types of analysis platforms (Table 3-3). These results
suggested that taxa-specific qPCR could reliably differentiate between NOD and NOR mouse
gut microbiome samples.
We also observed some intra-mouse strain variation of bacterial taxa, for example in the
Clostridia IV and Ruminococcus groups (Figure 3-4b). We hypothesized that this variation
might reflect cage to cage variation in housing environment resulting in intra-strain variation in
the microbiome. To examine this potential source of variation, we segregated the qPCR results
by cage for these two taxa (Figure 3-4b). We observed the no clear clustering of cage/littermates
into groups of similar relative abundance. We observed that relative abundance of taxa was
actually highly variable within cages. Overall, these data suggests that residence in different
cages did not explain the intra-genotype variation detected in these taxonomic groups.
We noted that the variance in abundance in NOD and NOR samples seemed to differ in some of
the 5 taxa quantified. We hypothesized that this variation might reflect differences in intra-strain
homogeneity in the gut microbiome. To examine these potential differences in variance, we log-
25
transformed all the data then performed a Brown–Forsythe test[10][11] to compare the sample
variance in each taxa between NOD and NOR samples. We found that the variance of abundance
in Clostridia IV, Helicobacter, and Ruminococcus was higher between NOR animals compared
to NOD (p = 0.00077, 0.0045, 0.0063 respectively). These results suggest that there is greater
intra-strain differences in the NOR gut microbiome compared to NOD.
3.2.4 Effects of NOR-derived Caecal microbiota on NOD recipients
Prior experiments in our lab had shown that the transfer of caecal contents derived from adult
male into 21 day old female NOD mice altered the microbiome of the recipients and induced
protection against T1D[2]. Gavage transfer-induced changes in the recipient’s microbiome were
found to persist 100 days after gavage treatment. These results suggested that the transfer of
caecal contents derived from an adult donor mouse could alter the microbiome of the recipients.
We hypothesized that we would be able to detect persistent changes in female NOD recipients of
female NOR caecal contents using taxa-specific qPCR. The qPCR assay panel was used to
quantify selected taxa in DNA extracted from caecal samples from 60 day old un-manipulated
female NOD or NOR mice, and from 60 day old NOD females gavaged at weaning with adult
NOR female microbiota (Figure 3-5). These data showed differences between untreated NOD
and NOR→NOD mice in the Clostridia IV and Clostridia XIVa groups (p = 0.008 for both
groups) (Figure 3-5), suggesting that the gavage treatment produced changes in the recipient gut
microbiome detectable 40 days post-gavage, at 60 days of age. We also observed differences
between NOR and NOR-gavaged NOD mice in the Clostridia IV, Clostridia XIVa, and
Helicobacter groups (p = 0.0079, 0.0119, 0.0424 respectively). These data suggested that female
NOR→NOD caecal gavage altered the recipient’s microbiome and that these alterations
persisted for at least 40 days post-gavage.
Both the 16S rDNA HTS (Figure 3-1) and the qPCR assays documented differences in the
composition of the sex and age-matched NOD and NOR microbiota. We hypothesized that
altering microbiome composition by oral gavage of the T1D-resistant NOR in T1D-susceptible
NOD recipients might impact the incidence of T1D in the recipient NOD mice. To test this
hypothesis, a cohort (n=30) of weanling (21d) female NOD mice were gavaged with diluted
material from the caecum of adult female NOR mice as we described for the NOD male
→female transfers[2]. The NOD recipient mice were then monitored for the development of T1D
26
(Figure 3-6). We observed no difference in the average age of onset, or incidence of T1D in the
NOR→NOD vs. unmanipulated NOD females survival from T1D. This observation suggests
that, although the oral gavage introduction of NOR females microbiota caused detectable
alterations in the microbiome of the recipient NOD mice, these alterations did not have a
detectable effect on the development of T1D, in stark contrast to the female NOD recipients of
male microbiome transfer previously reported[2].
3.3 Summary
We present this set of experiments as our effort to validate qPCR as viable method for the
detection and quantification of members of specific bacterial taxa, and the application of such an
assay panel to probe the effect of the microbiome on T1D development in the NOD mouse
model.
By in silico and in vitro methods, we designed and validated the specificity of a set of taxa-
specific qPCR primers targeted to the quantification of 5 bacterial taxa found previously to be
differentially abundance between NOD and NOR mice. We applied our panel of assays to two
experimental settings to provide a proof of concept. We first applied our panel of assays to
tracking changes in the composition of the NOD gut microbiome from weaning to adulthood.
We found that changes in the gut microbiome from weaning to adulthood could be detected and
quantified by taxa-specific qPCR, and a major shift in the microbiome composition was observed
between 25d and 41d female NOD mice. Secondly, we applied the qPCR panel to quantifying
bacterial taxa in unmanipulated NOD and NOR mice. We observed differences in the
microbiome composition of NOD and NOR mice which corroborated previous results in our lab.
Bacteroidetes and Helicobacter were enriched in NOR animals and Clostridia IV, and
Ruminococcus were enriched in NOD animals.
Having shown that our panel of qPCR assays could quantify the targeted bacterial taxa in NOD
and NOR mice, and that these results were in agreement with previous sequencing results, we
then applied the qPCR panel to quantifying bacterial taxa in unmanipulated NOD, NOD
recipients of NOR caecal material by gavage (NOR-gavaged NOD), and NOR mice. We found
significant differences between untreated NOD, NOR-gavaged NOD mice, and NOR mice.
These differences suggest that the gavage treatment altered microbiomes of treated NOD mice,
27
and that the resulting microbiome of recipient mice is not identical to that of the donor, nor is it
clearly partially composed of NOD and NOR contributions.
To test the effect of transferring NOR-derived microbiota into recipient weanling female NOD
mice on T1D, we assessed T1D incidence in NOD mice treated with NOR-derived microbiota by
gavage. We found that NOR-derived microbiota treatment was not protective against T1D.
28
Figure 3-1: 16S rDNA sequencing of female NOD and NOR caecal flush microbiome. 16S rDNA
PCR libraries were generated from caecal flush extracted from 84 day female NOD and NOR mice.
Libraries were pyrosequenced and the ratio NOD/NOR for each OTU count was calculated. These data
were plotted as bars mapped onto a taxonomical tree to define phylogenetic relationships between OTUs.
The dotted line indicates a ratio of 1. Graphic and data analysis provided by Dan Frank.
29
A
B
30
Figure 3-2: Quality analysis and validation for bacterial taxa-specific qPCR assays. (A): Evaluating
linearity and technical consistency of qPCR assays over serial dilutions. DNA was extracted from
samples of caecal contents from adult female NOD and NOR mice. DNA concentration was determined
by Nanodrop (Thermo Fisher Scientific Inc), and a serial dilution (1 - 1/10000) 50% NOD-NOR
admixture template DNA was prepared. Taxa-specific 16S qPCR primers (see Table 3-2) were used to
amplify partial 16S sequences for quantification by SYBRGreen. Raw data were then normalized using a
parallel reaction quantifying total bacteria, and repeated using the same template on three separate days.
Results from 3 days are plotted adjacent. (B): Evaluating efficiency of qPCR assays over serial dilutions.
Template dilution was plotted against Cthreshold to determine amplification efficiency by slope. Template
DNA serial dilutions and qPCR was performed as in (A). Points on the graphs represent the average of
three technical replicates. Lines represent lines of best fit for each primer set.
31
Figure 3-3: qPCR quantification of selected bacterial taxa in female NOD mice, ages 25-55. Female
NOD mice were aged to 25, 33, 41 and 55d. qPCR assays were performed as described above. Points on
the graph represent the average of three biological replicates, each biological replicate represents an
average of three technical replicates.
32
B
A
33
Figure 3-4: qPCR quantification of selected bacterial taxa in female adult NOD and NOR mice.
(A): Comparison of selected bacterial taxa in NOD and NOR. Caecal DNA from adult female NOD and
NOR mice was evaluate by taxa-specific qPCR primers and quantified by SYBRGreen. Resulting taxa-
specific bacterial counts were normalized to qPCR reactions quantifying total bacterial counts. Filled
circles indicate NOD samples, empty circles indicate NOR samples. The Mann-Whitney U test was used
to calculate the P-values presented between NOD and NOR samples. (B): Segregation of NOD and NOR
data from (A) by cage in Clostridia IV and Ruminococcus. Each cage is represented by a color; symbols
of the same color within a strain represent cagemates; data shown are the same as that displayed in (A).
Points on the graph represent the average of three technical replicates; the data presented represent results
from 18 mice.
34
Figure 3-5: qPCR quantification of selected bacterial taxa in female adult NOD, NOR-gavaged
NOD and NOR mice. Female NOD weanlings were gavaged with caecal bacteria from adult NOR
females and aged to 60 days. qPCR assays were performed as before. Points on the graph represent the
average of three technical replicates; the data presented represent results from 5 mice per genotype and
condition of treatment. Filled circles indicate NOD samples, empty NOR samples, and half-filled indicate
NOD treated by NOR-gavage. The Mann-Whitney U test was used to calculate the P-values presented
between the comparisons indicated.
35
Figure 3-6: T1D incidence in female NOD pup recipients with female NOR-derived intestinal
microbiome. Female NOD weanlings were gavaged with caecal bacteria from adult NOR females. T1D
was assessed in NOR→NOD female recipients (grey line, n = 30) and compared to historical un-
manipulated NOD females (black line, n = 59). T1D survival curves were compared by Log-rank test (p =
0.2541).
36
Taxonomy Sequences Sequence Abundance Fold Difference (NOD/NOR) NOD NOR
Bacteroidetes (Phylum) 21176 18.66% 20.48% 0.911
Clostridia IV (“Family”) 11066 9.99% 8.21% 1.216
Ruminococcus (Genus) 122 0.013% 0.200% 0.068
Clostridia XIVa (“Family”) 33832 37.23% 24.34% 1.595
Helicobacter (Genus) 22901 16.15% 31.25% 0.516
Table 3-1: Taxonomic groups/OTUs with differential abundance selected from NOD vs. NOR 16S
HTS data
37
Taxonomy Fwd. Primer Rev. Primer Amplicon Length
Bacteroidetes (Phylum)
AACGCTTAGCTACAGGCTT
CCACTGTGGGGGACCTTC
271
Clostridia IV (“Family”)
TTACACAATAAGTWATCCACCTGG
ACCTTCCTCCGTTTTGTCAAC
314
Clostridia XIVa (“Family”)
CGGTACCTGACTAAGAAGC
AGTTTYATTCTTGCGAACG
429
Helicobacter (Genus)
ACCAAGCWATGACGGGTATC
CGGAGTTAGCCGGTGCTTATT
220
Ruminococcus (Genus)
GGCGGCYTRCTGGGCTTT
CCAGGTGGATWACTTATTGTGTTAA
157
Table 3-2: Bacterial 16S taxa specific qPCR primers. Bacterial taxa were selected based on their
differential abundance in NOD and NOR mice (Figure 3-1
Figure 3-1: 16S rDNA sequencing of female NOD and NOR caecal flush microbiome. 16S rDNA
PCR libraries were generated from caecal flush extracted from 84 day female NOD and NOR mice.
Libraries were pyrosequenced and the ratio NOD/NOR for each OTU count was calculated. These data
were plotted as bars mapped onto a taxonomical tree to define phylogenetic relationships between OTUs.
The dotted line indicates a ratio of 1. Graphic and data analysis provided by Dan Frank.
). 16S rRNA genes from sequenced members of the target taxa from the RDP database were aligned
conserved regions were found using PRIMROSE[73]. These taxa-specific qPCR primers were then tested
in silico by performing Probe Match searches and in vitro by producing and sequencing amplicons from
NOD and NOR caecal extract DNA.
38
Taxonomy NOD/NOR ratio
Pyrosequencing qPCR (median) qPCR (mean)
Bacteroidetes (Phylum) 0.911 0.304 0.313
Clostridia IV (“Family”) 1.216 16.354 3.404
Ruminococcus (Genus) 0.068 0.745 0.574
Clostridia XIVa (“Family”) 1.595 1.000 1.536
Helicobacter (Genus) 0.516 0.420 0.411
Table 3-3: qPCR quantification of bacterial taxa. The abundance of each taxa in NOD and NOR
samples was extracted and the ratio between the abundance in NOD and NOR was calculated.
39
Chapter 4 The effect of Protecflor probiotic treatment on NOD microbiome
composition and insulitis severity
40
4 The effect of Protecflor probiotic treatment on NOD microbiome composition and insulitis severity
4.1 Background and Rationale
T1D and gut inflammatory disorders have been found to share genetic risk factors, and can occur
as co-morbidities. For example, ZNF365, MST1, ERBB3, and PTPN22 have been identified as a
common susceptibility genes in Crohn’s disease and T1D[17]. T1D and celiac disease share
susceptibility alleles of HLA-DQB1, IL2, RGS1, TAGAP and IL18RAP [18]. Moreover, T1D
patients display a higher risk of developing celiac disease compared to the general
population[96]. Anti-tissue transglutaminase antibodies, a marker for celiac disease, are present
in ~10% of children and ~2% of adults with T1D[19]. These comorbidities suggest shared
genetic and environmental factors in T1D and autoimmune diseases of the gut.
Studies in ulcerative colitis, inflammatory bowel syndrome and T1D have suggested that
probiotic bacteria can have protective effects in these diseases. For example, Lactobacillus
helveticus R0052 and Lactobacillus rhamnosus R0011 have been shown to reduce clinical
symptoms in open label experiments on human adults with inflammatory bowel syndrome[61],
and to reduce epithelial injury due to Escherichia coli O157:H7- and O127:H6- infection in an in
vitro gut epithelial cell co-culture system [70]. Bifidobacterium longum R0175, given in
combination with inulin, a plant polysaccharide, was found to improve self-reported quality of
life, symptoms of blood and mucous in stool, and to reduce relapse rate in a small pediatric study
of ulcerative colitis (UC) [71]. Oral administration of probiotic bacterial strains was reported to
protect NOD mice from T1D. The probiotic mixture VSL#3 ( Bifidobacterium, Lactobacillus
species and Streptococcus salivarius subsp. Thermophiles ) was given by oral gavage three times
a week from 4 to 32 weeks of age, reduced T1D incidence in NOD mice [49]. However, repeated
gavages performed in this study increase the risk of systemic exposure of microbes to the
immune system, by abrasion of the esophageal wall, or aspiration of the gavage material. It is
well documented by multiple studies that bacterial infection or systemic exposure is protective
against T1D in NOD mice. Studies showing deliberate infection with multiple organisms,
including Schistosoma mansoni[42], Mycobacterium avium[43], Trichinella spiralis[44],
Heligmosomoides polygyrus[45] and Salmonella typhimurium[46], can induce protection in
NOD mice. Interperitonial injection of zymosan, a fungal cell wall component, has also been
41
found to be protective in the NOD model[47]. Another study also suggested that treatment with
probiotic strains of Lactobacillus and Bifidobacterium can be protective in T1D, with the caveat
that systemic immune exposure of the mice may have been a confounding factor in their protocol
[49].
Collectively, these studies suggest that long-term treatment with probiotic strains of
Lactobacillus and Bifidobacterium may induce alterations in the gut microbiome and/or produce
anti-inflammatory effects through other mechanisms. We hypothesized that continuous treatment
with Protecflor probiotic, a formulation containing Lactobacillus helveticus R0052,
Lactobacillus rhamnosus R0011, and Bifidobacterium longum R0175 and the yeast
Saccharomyces cerevisiae boulardii, would impact the gut microbial composition of diabetes-
prone NOD mice, and consequently, have an effect on the progression of islet autoimmunity.
Therefore, we established a system to test whether the treatment of NOD animals with Protecflor
altered the recipient’s gut microbiota and whether such changes were associated with an islet
autoimmunity-protective effect on treated NOD mice. To address concerns of systemic immune
exposure to protective organisms, we designed a treatment protocol to minimize gavage
procedures, and maintained continuous Protecflor probiotic dosing by addition to drinking water.
4.2 Results
4.2.1 Detection of Protecflor Probiotic organisms by qPCR
Previous studies of the Protecflor probiotic strains have shown that Lactobacillus helveticus
R0052 was adherent to in vitro gut epithelial cell cultures[97], but there was no evidence that
Protecflor probiotic strains could colonize humans or mice. The related organism, Lactobacillus
casei subsp. rhamnosus Lcr35, ceased to be found in the gut or feces of mice orally fed these
organisms within in 3 days after cessation of treatment [98]. Saccharomyces cerevisiae boulardii
did not colonize or adhere to gut epithelial cell cultures [99]. To test for the retention of
Protecflor organisms during our NOD mouse treatment protocol, we developed a qPCR assay to
quantify the Protecflor organisms. qPCR primers were designed to the Bifidobacterium longum
R0175 rmlB gene and an extrachromosomal plasmid pIR52 in Lactobacillus helveticus R0052.
These primers (Table 4-1) were highly specific for the respective type strains, and were selected
using the full genome sequence of each strain conveyed to us by our collaborators at Lallemand.
42
To evaluate consistency of quantification between qPCR runs, the primers were first used to
derive a set of standards with Protecflor-derived bacterial DNA as the template. Serial dilutions
of Protecflor-derived DNA were quantified by qPCR using each of the primer sets. The copy
number of each template type was determined by a standard curve, and plotted against the
dilution of the template DNA (Figure 4-1). Both primer sets showed linearity between 25ng and
25pg of template DNA (R2 > 0.99). To evaluate reproducibility, the assays were repeated on 3
separate days, and found to display no significant day-effects (p=0.53). These results
demonstrated that the two qPCR assays reliably provided consistent quantification over multiple
experimental replications.
To test the ability of these primers to detect Protecflor strains in the feces of mice, we adopted a
treatment protocol developed by Philip Sherman’s group for the treatment of rats with probiotics
in drinking water[100] to treatment of NOD mice. We recently demonstrated that the gut
microbiome of weanling age mice could be altered by oral gavage with complex commensal
microbes[50]. Therefore weanling (21d old) female NOD mice were gavaged twice with 1x109
CFU Protecflor probiotic separated by a 24h period. The mice were then given free access to
drinking water containing 2 x 108 CFU/mL of Protecflor over 15 weeks. Water consumption was
monitored in treated cages to estimate dosing. To quantify the exposure of treated mice to the
Protecflor organisms, Bifidobacterium longum R0175 and Lactobacillus helveticus R0052 were
quantified using strain-specific qPCR on DNA extracted from fecal samples that were collected
weekly from each cage. Our standard curve quantifying Protecflor-derived template detected
product at 25pg/µL template DNA (Figure 4-1) corresponding to a lower detection limit of
~5000 CFU of Protecflor microbes/g fecal material. Despite this assay sensitivity, we failed
detected either strain in feces, suggesting that if these organisms were present in the feces, their
concentration was below this limit. One possible explanation for these finding was rapid
gastrointestinal transit and clearance of these non-colonizing organisms.
4.2.2 Gastrointestinal Transit times of Protecflor Probiotic organisms in the NOD model
There is no reported evidence that Protecflor probiotic strains can colonize the human or mouse
intestinal tract. Saccharomyces cerevisiae boulardii were found to fall below detectable levels in
2-5 d in humans treated with a single 1g oral dose[99]. Lactobacillus helveticus R0052 to be
adherent to in vitro gut epithelial cell cultures, but there was no in vivo evidence for colonization
43
in this study[97]. One explanation for these findings was rapid gastrointestinal transit and
clearance of these non-colonizing organisms. In mice, a wide range of gastrointestinal transit
times in mice have been reported, depending on the methodology and mouse strain. GI transit
times of 154 ± 24 minutes in BL/6 mice were observed using a contrast MRI-based measurement
method[101]. In contrast, a GI transit time of 6h was reported in 129SvEv mice using a radio-
labeled charcoal pellet[102]. One study in NOD mice observed that a bolus consisting of 10%
charcoal in 5% Arabic gum aqueous suspension had transited 60% of the small intestine 20
minutes after feeding.
To establish the gastrointestinal transit time for our studies, NOD mice were gavaged with
Protecflor probiotic and rested for 1, 2, 3 or 4 hours. At each time point, cohorts of mice were
sacrificed and the GI tract removed and the contents of these organs were recovered. The GI tract
of each animal was divided into stomach, small intestine, and large intestine + caecum, and their
contents recovered by lavage with sterile saline. Bacterial DNA was extracted from each lavage
sample and amplified by qPCR using R0052 and R0175 strain-specific primers (Figure 4-2).
DNA from the Protecflor organisms was evident in each GI tract section obtained from mice
sacrificed 1, 2, and 3 hours after gavage, but was no longer detectable 4 hours post-gavage.
Protecflor organism DNA copy number declined in all GI tract sections over the 4 hour time
course. (Figure 4-2Error! Reference source not found.) These observations suggested that
these probiotic organisms transit rapidly through the GI tract and were below the limit of
detection for the qPCR assay within 4 hours after oral introduction.
4.2.3 Evaluating the Effects of Protecflor Probiotic on Insulitis Severity in 4.2.3 Evaluating the Effects of Protecflor Probiotic on Insulitis Severity in NOD model
Oral administration of probiotic strains can be protective in a number of GI inflammatory
disorders[61][71], and studies have shown genetic and epidemiological data connecting T1D and
GI inflammatory disorders [17][18] [96] [19] (See Background and Rationale). Given this
evidence, we hypothesized that the exposure to Protecflor probiotic would have an effect on the
anti-islet autoimmunity in NOD mice and evaluated islet inflammation severity (insulitis), a T1D
phenotype, as a primary endpoint in a longitudinal treatment study. To test this hypothesis,
weanling (21d) female NOD mice (n=8 each treatment group) were gavaged twice, 24 hours
apart, with 1x109 CFU Protecflor probiotic or a vehicle (the carbohydrate-rich vehicle in which
Protecflor is prepared). The mice were then given free access to drinking water containing with 2
44
x 108 CFU/mL probiotic or an equivalent concentration of vehicle material for an additional 12
weeks. At 15 weeks of age, pancreata were harvested from Protecflor and vehicle treated mice,
fixed, sectioned and stained for histological evaluation. Quantitative insulitis assessment was
done double-blinded using a well-defined scoring method [36][37][50] (Figure 4-3a). The
distribution of insulitis scores was markedly distinct in Protecflor and vehicle-treated mice (Chi-
squared test, Bonferroni-corrected α = 0.0071, p < 0.0001). These data suggested that Protecflor
was protective against insulitis in the NOD model. Insulitis in vehicle-treated animals did not
differ from age and sex-matched untreated historical controls from our mouse colony (Figure
4-3a) (Chi-squared test, Bonferroni-corrected α = 0.0071, p = 0.035).
The impact of Protecflor was also evident when insulitis severity scores were segregated into two
groups: healthy/peri-insulitis (scores 0-1) and infiltrating insulitis (scores 2-4) (Figure 4-3b;
Chi-squared test, p < 0.0001). This observation suggested that exposure to the Protecflor
probiotic organisms beginning in early life is protective against immune infiltration of the islets,
an obligate step in the progression toward T1D in the NOD model.
4.2.4 The Effects of Protecflor Probiotic treatment on the gut microbiome in NOD mice
Animal studies in both agricultural and laboratory settings have shown that treatments with
exogenous microbes can be associated with changes in the microbiome. Administration of a
probiotic containing lactobacilli, bifidobacteria, enterococci, and pediococci improved weight
gain in broiler chickens, which was associated with an increase in Bifidobacterium spp.,
Lactobacilli, and Gram-positive cocci[103]. An increase in the diversity of gut lactobacilli was
observed in fecal as well as intestinal tissue samples in mice treated with L. casei and L.
plantarum. Our lab reported that weanling female NOD mice gavaged with caecal contents from
adult male NOD mice were protected from T1D, and that this protection was associated changes
in testosterone levels provoked by changes in the microbiome[50]. In humans, studies correlating
effects of microbial treatment with changes in the gut microbiome composition are currently few
in number. Kajander et al reported treatment with a multispecies probiotic supplement
(Lactobacillus rhamnosus GG, L. rhamnosus Lc705, Propionibacterium freudenreichii ssp.
shermanii JS and Bifidobacterium animalis ssp. lactis Bb12) was found to improve clinical
scores of treated IBS patients, and was also associated with stabilization of the gut microbiome
composition[104]. We hypothesized that the Protecflor organisms induced a change in gut
45
microbiome of NOD mice, and that these alterations were linked to the insulitis protection we
observed.
To characterize potential changes in the microbiome resulting from Protecflor treatment, we
repeated the oral gavage followed by drinking water exposure protocol in cohorts of NOD
females. DNA was extracted from caecal flush samples from Protecflor- and vehicle-treated
mice at 14 weeks of age, and fecal samples from Protecflor-treated and vehicle-treated cages was
prepared at 4, 6, 10, and 14 weeks of age. 16S rDNA libraries were prepared used bar coded
primers and sequenced on the Illumina MiSeq platform as previously published [50][50][105].
OTUs (operational taxonomical units) were called for the sequences using RDP Classifier[75].
The number of sequence reads per sample is shown in Table 4-2.
4.2.4.1 Differential abundance of bacterial taxa in Protecflor and vehicle treated NOD mice
To determine if the 16S rDNA sequencing provided sufficiently deep coverage, a Good’s
Coverage statistic was calculated for the OTUs from each sample. These data were plotted as a
collector’s curve for each library (Figure 4-4). Good’s coverage approaching 100% suggests that
the sequences were representative samples of the populations they represented, and that
additional sequencing was unlikely to uncover new OTUs. For each 16S library, the sequencing
was found to provide ≥99.97% coverage of the actual sample.
The samples were segregated based upon biogeography (fecal or caecal flush), and the top 20
most abundant taxa for each sample were plotted as stacked bars (Figure 4-5). Fecal samples
were dominated by Bacteroidales S24-7, division TM-7 and Lachnospiraceae, while caecal
samples were dominated by Bacteroidales S24-7, Lachnospiraceae, and Bacteroides. In this
display, we found no obvious relationship between Protecflor and vehicle-treated mice. The
similar taxonomic diversity between Protecflor and vehicle-treated groups suggested that the
probiotics did not cause gross alterations in the microbial biodiversity.
To determine whether specific OTUs were altered in abundance by Protecflor treatment, the
Wilcoxon signed-rank test was used and–Log(p-value) was plotted for each OTU using a
Manhattan plot. OTUs with –Log(pValue) above ~1.30, corresponded to a p-value < 0.05 and
were categorized as differentially abundant between Protecflor and vehicle treated samples. We
46
observed enrichment of the genera Allobaculum, Candidatus-Arthromitus, Roseburia, and the
order Clostridiales in caecal flush and fecal samples from Protecflor-treated animals compared to
controls, and depletion of the genera Helicobacter, Parabacteroides and the order Bacteroidales,
in Protecflor treated vs. controls (Figure 4-6). Despite sustained treatment with Protecflor
Lactobacillus, Bifidobacterium, and Saccharomyces species, we did not observe differential
abundance of any of these taxa supporting our prior evidence that the Protecflor organisms did
not colonize and transited rapidly through the mouse GI tract. However, Protecflor treatment was
associated with observable changes in the microbiome composition of treated mice.
4.2.4.2 β-diversity analysis of 16S rDNA sequencing data
To compare the overall community structure of the microbiomes found in both Protecflor and
vehicle-treated mice, the Morisita-Horn distance was calculated between caecal samples, and
plotted as a heatmap. Visual comparison of the combined heatmap values was then used to
segregate all samples into similar clusters (Figure 4-7). We observed 2 clusters within the set of
samples. Clustering occurred independent of treatment type.
4.2.4.3 Principal components analysis of 16S rDNA sequencing data
To further analyze the relationship between the microbiome composition in Protecflor and
vehicle-treated mice and insulitis severity, we performed PCA (principal components analysis)
on the sequencing dataset. PCA is a multivariate analysis technique which reduces the number of
dimensions/variables in higher dimensional data. This is done by finding a set of orthogonal
linear combinations of the original variables, called principal components, which account for the
greatest variance in a dataset. This can potentially reveal hidden structure in the data. We
hypothesized that differences in the gut microbiomes of Protecflor and vehicle-treated mice
would be represented by these principal components, and these could be correlated to insulitis
severity.
Using Ginkgo, a multivariate statistical analysis software suite [94], we performed principal
components analysis (PCA) on the sequencing data from caecal flush from both Protecflor and
vehicle-treated mice. After inspection of scree plots for the PCA (Figure 4-8), we used the first
two principal components (PC) that together explained 53.47% of the total variance (Figure
4-9). The object loadings for PC1 were ordered from least to greatest (Table 4-3). We found the
47
positive loadings were populated with Order Clostridales, including the families
Ruminococcaceae and Lachnospiraceae. The negative loadings were populated with Phylum
Proteobacteria and Order Bacteroidales. The object loadings for PC2 were ordered from least to
greatest (Table 4-4). We found the positive loadings were populated with family
Erysipelotrichaceae, Order Clostridales, specifically from the Lachnospiraceae family, and
Order Bacteroidales, including the Rikenellaceae and Porphyromonadaceae families. The
negative loadings were populated with phylum Cyanobacteria, and Order Clostridales, including
the Veillonellaceae and Lachnospiraceae families, and Candidatus-Arthromitus (SFB).
We noticed clustering of the caecal samples into 2 distinct groups along PC1 (Figure 4-9). To
examine this possible relationship using independent statistical models, we performed
unsupervised clustering analysis then mapped the clusters on to the PCA components plot
(Figure 4-10). We found 2 clusters: Types 1, 2 that were primarily differentiated by PC1.
Samples from Protecflor-treated animals were not more closely related by PC1 or PC2 than were
vehicle-treated animals, even within the same cluster. These results suggested that the cluster
segregation was not related to Protecflor treatment, so we examined other potential correlations.
4.2.4.4 Correlation of principal components analysis with insulitis
Given the relationship between Protecflor and protection from insulitis in our previous results
(Figure 4-3), we examined the relationship between change in the microbiome composition and
insulitis severity. We performed PCA on the sequencing data from all caecal flush samples,
segregated into Protecflor and vehicle-treated groups, then plotted PC1 or PC2 against overall
insulitis scores for each mouse. In Protecflor-treated mice, we found positive loading in PC1 was
correlated with protection against insulitis (r2 = 0.7408) (Figure 4-11a), while PC2 was not
related to insulitis (r2 = 0.00986). These results suggested that Protecflor treatment may alter
microbiome composition and that this alteration is associated with protection against
autoimmune insulitis. In vehicle-treated mice, positive loading in PC2 was correlated with
protection against insulitis (r2 = 0.7965) (Figure 4-11b), while PC1 was not related to insulitis
(r2 = 0.00181). Taken together, these data suggested that the microbiome changes associated with
lower insulitis progression in vehicle-treated mice were distinct from those associated with
insulitis protection in Protecflor treated animals.
48
4.2.4.5 Correlation of caecal bacterial community richness with principal components analysis and insulitis
Inflammatory phenotypes in GI inflammatory diseases are associated with dysbiosis
characterized by altered microbiome composition and lower gut microbial species
richness[106][107][108]. The Chao1 statistic is a measure of sample richness [67] which
estimates the true total number of OTUs in a sample based on the number of singletons (OTUs
observed once in the sample), and doubletons (OTUs observed twice in the sample) observed
[67]. To compare the richness of samples and correlated these measures with microbiome
composition, we calculated the Chao1 statistic for the samples, then plotted them against PC1
and PC2 (Figure 4-12a). We found that positive PC1 loading was correlated with overall greater
community richness. The linear regression of the Chao1 statistic had an r2 value of 0.6687. An F
Test performed on the linear regression showed that slope is non-zero (p = 0.0012). In contrast,
PC2 was not correlated with species richness (Figure 4-12a). An F Test performed on the linear
regression shows that the slope does not differ from zero (p = 0.9021). Comparing the two
clusters differentiated by PC1, Type 1 samples are overall less rich than Type 2 (Mann Whitney
test, p = 0.0025), with a median Chao1 statistic of 34.15 compared to 51.26 for Type 2 (Figure
4-12b). These results suggest that Type 2 samples represent a more diverse “enterotype” then
type 1 samples.
We then asked whether the difference in community richness was correlated with insulitis
severity. Samples were segregated into Protecflor and vehicle-treated groups, a Chao1 statistic
for community richness[31] was computed for the samples in each group, and plotted against the
average insulitis score for each mouse (Figure 4-13). In Protecflor-treated mice, the Chao1
statistic correlated with protection against insulitis (Figure 4-13; r2 value of 0.8988). An F Test
performed on the linear regression showed that the slope was non-zero (p = 0.0040). In contrast,
the Chao1 statistics were not related to insulitis severity in vehicle-treated mice (Figure 4-13),
where an F Test performed on the linear regression showed the slope was not different from zero
(p = 0.8434). These data suggest that community richness was correlated with decreased insulitis
severity in Protecflor-treated, but not vehicle-treated mice.
49
4.2.4.6 Correlation of caecal bacterial diversity with principal components analysis and insulitis
GI inflammatory diseases are associated with dysbiosis characterized by lower diversity, richness
and altered microbiome composition[106][107][108]. Shannon’s H statistic is measure of the
predictability of a given OTU in a sample[69]. A higher Shannon’s H value indicate higher
entropy or “surprise” in a sample; the next OTU to be called is harder to predict [68]. The
statistic depends on the number of OTU observed in a sample, and the evenness of the
distribution of individual samples across OTU categories and is used to estimate microbial
diversity [68]. To compare the species diversity of samples and correlate these measures with
microbiome composition, samples were segregated into Protecflor and vehicle-treated groups, a
Shannon’s H statistic calculated for each sample, and plotted against the PC1 and PC2 loadings
for each sample (Figure 4-14). In vehicle-treated mice, positive PC2 loading was correlated with
overall greater species diversity (r2 = 0.8491). In contrast, PC1 was not correlated with species
diversity (r2 = 0.00917), and an F Test performed on the linear regression suggested that the
slope did not differ from zero (p = 0.8434). In contrast, in Protecflor-treated mice, neither PC1
nor PC2 were correlated with species diversity (r2 = 0.1330, 0.2138 for PC1 and PC2,
respectively), and an F Test performed on the linear regression suggested that these slopes did
not differ from zero (p =0.4772, 0.8568 for PC1 and PC2, respectively). These results suggest
that although Protecflor and vehicle-treated mice can exhibit microbiome communities with
similar loadings on PC2, Protecflor treatment is associated with loss of correlation between PC2
and community diversity.
To correlate species diversity in samples with insulitis severity, we segregated Protecflor and
vehicle-treated groups, calculated the Shannon’s H statistic for each group, then plotted them
against insulitis scores for each mouse (Figure 4-15). In vehicle-treated mice, Shannon’s H
statistic (species diversity) was correlated with lower insulitis severity (Figure 4-15a). The linear
regression of Shannon’s H statistic with insulitis displayed an r2 value of 0.8491, and an F Test
of the linear regression suggested a non-zero slope (p = 0.0090). In contrast, in Protecflor-treated
mice, Shannon’s H statistic was not related to insulitis (Figure 4-15b) and a F Test on the linear
regression shows that the slope does not differ from zero (p = 0.3559). These results suggest that,
in contrast to community richness, community diversity is correlated with decreased insulitis
severity in vehicle-treated, but not Protecflor-treated mice.
50
4.3 Summary
We present this set of experiments as our effort to probe effects of Protecflor probiotic treatment
on the gut microbiome, and on its effect on T1D development in the NOD mouse model.
By in silico and in vitro methods, we designed and validated the specificity of a set of probiotic
strain-specific qPCR primers targeted to the quantification of Protecflor strains. Using these
primers, we measured the GI transit time of the Protecflor strain in NOD mice gavaged with
Protecflor. We found the probiotic strains rapidly transited the GI tract and could not be detected
after 4 hours post-gavage, suggesting Protecflor consisted of strains which did not colonize the
murine gut.
We modified our previously published gavage treatment protocol and tested the effect of
Protecflor probiotic on autoimmune insulitis in female NOD mice. We found that Protecflor
treatment was protective against insulitis and that Protecflor-treated mice showed less immune
infiltration of the islets compared to vehicle-treated mice. We used Illumina high-throughput
sequencing of the bacterial 16S rDNA gene to quantify changes in the composition of the gut
microbiome associated with Protecflor treatment. Sequencing results showed changes in 7
bacterial taxa in Protecflor-treated animals.
We applied principal components analysis to our sequencing data in order to reduce the
dimensionality of the data, reveal hidden structure and correlate changes in the microbiome with
insulitis. We found that 53.47% of the total variance could be explained by 2 principal
components, PC1 and PC2. The principal components did not differentiate Protecflor and
vehicle-treated mice, but were correlated with insulitis severity in each group. Specifically,
positive loading on PC1 was correlated with decreased insulitis in Protecflor-treated mice, but
not vehicle-treated mice, while, in contrast, positive loading on PC2 was correlated with
decreased insulitis in vehicle-treated mice, but not Protecflor-treated mice. These data suggest
changes in the bacterial community composition can be correlated to protection against islet
autoimmunity in NOD mice.
We applied ecological richness and diversity metrics to the sequencing samples in order to
correlate changes in the community structure with changes in the microbiome OTU composition
and insulitis. We found community richness and diversity were correlated with protection against
51
insulitis, suggesting that changes in overall community structure can be correlated to protection
against islet autoimmunity in NOD mice.
52
A
B
53
Figure 4-1: Quality analysis and validation for Protecflor-specific qPCR assays. (A): Linearity and
technical consistency of Protecflor qPCR assays. DNA was extracted from samples of the Protecflor
formulation. DNA concentrations were determined by Nanodrop (Thermo Fisher Scientific Inc.), and a
serial dilutions (from 1 - 1/10000) of template DNA were prepared. Protecflor-specific qPCR primers (see
Table 4-1) were used to PCR amplify these sequences which were quantified by SYBRGreen. Raw data
were then normalized compared to a PCR reaction quantifying total bacteria, and repeated using the same
template on three separate days. Results from the 3 days are displayed. (B): Efficiency of Protecflor
qPCR assays. Template dilution was plotted against Cthreshold to determine amplification efficiency
reported as the slope. Template DNA serial dilutions and qPCR were performed as in (A). Points on the
graphs represent the average of three technical replicates. Lines represent lines of best fit for each PCR
primer set.
54
Figure 4-2: In vivo transit time for Protecflor probiotic strains. Twelve mice were gavaged with
200uL Protecflor suspension (2 x 108 CFU/mL) then sacrificed 1, 2, 3 or 4 hours post-gavage. The GI
tract was dissected into stomach, small intestine, and caecum + large intestine. Contents of each
compartment were flushed and DNA was extracted from the resulting material. R0052-pIR52 copy
numbers were quantified by qPCR as described above. Data shown is the average of 3 biological
replicates each performed with 3 technical replicates.
55
P < 0.0001
P < 0.0001 A
B
56
Figure 4-3: Insulitis outcomes in female NOD pups with Protecflor probiotic. Female NOD
weanlings were either gavaged with and subsequently, had their drinking water was treated with either
Protecflor probiotic or Protecflor vehicle Treated drinking water was changed every 2 days. Insulitis
severity was assessed by established protocols (see Methods) in female mice either treated with Protecflor
probiotic or Protecflor vehicle at 93-98 days of age (n=8 per treatment group). 1 Protecflor-treated and 2
vehicle-treated mice became diabetic before the treatment cutoff and were excluded. Islet histology and
scoring contributed to by Steve Mortin-Toth (A): Insulitis scores differ significantly between Protecflor
and vehicle-treated mice. Insulitis score distributions were compared by Chi-squared test (p<0.0001). (B):
Protecflor treatment affects pathogenesis transition from peri-insulitis to invasive insulitis. Islets were
segregated into Score 0-1 and Score 2-4 categories. Insulitis score distributions were compared by Chi-
squared test (p<0.0001).
57
Figure 4-4: Quality analysis of Illumina 16S rDNA sequencing of Protecflor and vehicle-treated
mice. The Good’s Coverage statistic was calculated for each OTU for 40 samples of increasing size using
multiple random samplings of the sequencing data. Data was plotted as a collector’s curve showing the
increase in coverage as sample size increases (see Methods) for each 16S rDNA PCR library. Dark line
indicates actual size the sequencing sample with fewest reads.
58
Figure 4-5: Top 20 caecal and fecal microbial taxa in Protecflor and vehicle-treated mice. The top
20 represented taxa from all samples were plotted as stacked bars as a percentage of all counts from each
library. Samples are segregated by: sample type; Protecflor treatment; Cage/mouse number; and age at
sampling. Sequencing and data quality assurance was done by Dan Frank’s Lab.
59
Figure 4-6: Manhattan plot comparing enriched or depleted OTUs in Protecflor-treated vs vehicle-
treated animals. The relative abundance of each OTU was compared by the Wilcoxon signed-rank test,
and –Log(p-value) was plotted for each OTU in the comparison in a Manhattan plot. OTUs with –
Log(pValue) above ~1.30 (indicated by dark line), corresponding to a p-value less than 0.05 were
putatively categorized as those showing differential abundance.
Bacteroidales
Helicobacter
Roseburia SFB Allobaculum
Parabacteroides Clostridiales
60
Figure 4-7: Morisita-Horn distance heatmap of Protecflor-treated vs vehicle-treated caecal samples.
Morisita-Horn distance was calculated between each sample and plotted as a heatmap. Samples were
ordered, grouping samples with similar compositions (high Morisita-Horn similarity). Black/blue
indicates similar compositions, and red indicates dissimilar compositions.
61
Figure 4-8: Scree plots of PCA comparing all bacterial OTUs across caecal flush samples from
Protecflor and vehicle-treated animals. Scree plots were generated from the Eigenvalues of the
principal components.
62
Figure 4-9: PCA comparing all bacterial OTUs across caecal flush samples from Protecflor and
vehicle-treated animals. PCA was performed on the OTU abundance data from sequencing caecal flush
samples from Protecflor and vehicle-treated NOD mice. The samples were separated by principal
components PC1 and PC2, which together explained 53.47% of the total variance between samples.
Protecflor treatment data was mapped onto the PCA components plot. Red symbols represent samples
from Protecflor treated mice, blue symbols vehicle-treated.
63
Figure 4-10: Unsupervised clustering analysis of caecal flush samples from Protecflor and vehicle-
treated animals. A K-medians/fuzzy C-medians clustering algorithm (Dunn’s Coefficient = 0.86180) and
a hierarchical agglomerative method were both used to define clusters across fecal and caecal flush
samples from Protecflor and vehicle-treated animals. Clustering data was mapped onto the PCA
components plot. Blue symbols represent samples placed in group 1, red symbols group 2.
A
64
B
65
Figure 4-11: Correlation of principal components with insulitis protection in Protecflor and vehicle-
treated mice. PCA was performed on the sequencing data from all caecal flush samples. Samples were
segregated into Protecflor and vehicle-treated groups, then PC1 or PC2 was plotted against overall per
mouse insulitis. Red symbols represent samples from Protecflor-treated mice; blue symbols vehicle-
treated mice. (A): positive loading in PC1 was correlated with protection against insulitis in Protecflor-
treated mice, but not vehicle-treated mice. Pearson's r and r2 was calculated for linear regression of each
sample group to determine correlation between insulitis severity and PC1 (r2 = 0.7408, 0.00986 for
Protecflor and vehicle-treated mice, respectively). An F Test was performed on the linear regression of
each sample group to determine of the slope was non-zero. (p = 0.0278, 0.8515 for Protecflor and vehicle-
treated mice, respectively). (B): positive loading in PC2 was correlated with protection against insulitis in
vehicle-treated mice, but not Protecflor-treated mice. Pearson's r and r2 was calculated for linear
regression of each sample group to determine correlation between insulitis severity and PC1 (r2 =
0.00181, 0.7965 for Protecflor and vehicle-treated mice, respectively) An F Test was performed on the
linear regression of each sample group to determine of the slope was non-zero. (p = 0.9362, 0.0167 for
Protecflor and vehicle-treated mice, respectively).
66
A
B
67
Figure 4-12: Correlation of principal components with bacterial community richness. (A): positive
loading in PC1, but not PC2 was correlated with species richness. PCA was performed on the sequencing
data from all caecal flush samples. The Chao1 statistic was calculated for the samples, then plotted
against PC1 and PC2. Red symbols represent samples plotted against PC1, black symbols PC2. Pearson's
r and r2 was calculated for each linear regression to determine correlation between species richness and
PC1 and PC2 (r2 = 0.6687, 0.00158 for PC1 and PC2, respectively) An F Test was performed on the
linear regression of each sample group to determine of the slope was non-zero. (p = 0.0012, 0.9021 for
PC1 and PC2, respectively). (B): Type 1 samples are overall less species rich than Type 2 samples. These
samples were segregated based upon PC1 into Type 1 and Type 2 samples. Chao1 statistics were
calculated and are plotted for each group. The two sample groups were compared by the Mann Whitney
test (p = 0.0025).
68
Figure 4-13: Correlation of bacterial community richness with insulitis protection in Protecflor-
treated mice. Samples segregated into Protecflor and vehicle-treated groups, the Chao1 statistic was
calculated for the samples in each group, and then plotted against overall per mouse insulitis. Red
symbols represent samples from Protecflor-treated mice; blue symbols vehicle-treated mice. Pearson’s r
and r2 was calculated for linear regression of each sample group to determine correlation between Chao1
statistic insulitis severity (r2 = 0.8988, 0.0110 for Protecflor and vehicle-treated mice, respectively). An F
Test was performed on the linear regression of each sample group to determine of the slope was non-zero.
(p = 0.0040, 0.8434 for Protecflor and vehicle-treated mice, respectively
69
A
B
70
Figure 4-14: Correlation of bacterial community diversity with principal components in
Protecflor-treated mice. Samples segregated into Protecflor and vehicle-treated groups, Shannon’s H
statistic was calculated for the samples in each group, then plotted against PC1 and PC2. Red symbols
represent samples from Protecflor-treated mice, blue symbols vehicle-treated mice. (A): PC1 was not
correlated with species diversity in either Protecflor-treated or vehicle-treated mice. Pearson's r and r2 was
calculated for linear regression of each sample group to determine correlation between Shannon’s H
statistic and PC1 (r2 = 0.1330, 0.00917 for Protecflor and vehicle-treated mice, respectively). An F Test
was performed on the linear regression of each sample group to determine of the slope was non-zero. (p =
0.4772, 0.8568 for Protecflor and vehicle-treated mice, respectively). (B): positive loading in PC2 was
correlated with greater species diversity in vehicle-treated mice, but not Protecflor-treated mice. Pearson's
r and r2 was calculated for linear regression of each sample group to determine correlation between
insulitis severity and PC2 (r2 = 0.2138, 0.8491 for Protecflor and vehicle-treated mice, respectively) An F
Test was performed on the linear regression of each sample group to determine of the slope was non-zero.
(p = 0.3559, 0.0090 for Protecflor and vehicle-treated mice, respectively).
71
Figure 4-15: Correlation of bacterial community diversity with insulitis protection in Protecflor-
treated mice. Samples segregated into Protecflor and vehicle-treated groups, Shannon’s H statistic was
calculated for the samples in each group, then plotted against overall per mouse insulitis. Red symbols
represent samples from Protecflor-treated mice; blue symbols vehicle-treated mice. Pearson's r and r2 was
calculated for linear regression of each sample group to determine correlation between Shannon’s H
statistic and insulitis severity (r2 = 0.00001, 0.8011 for Protecflor and vehicle-treated mice, respectively).
An F Test was performed on the linear regression of each sample group to determine of the slope was
non-zero. (p = 0.9949, 0.0159 for Protecflor and vehicle-treated mice, respectively).
72
Target Organism Targeted gene/ sequence
Fwd. Primer Rev. Primer Amplicon Length (bp)
Bifidobacterium longum R0175
rmlB AAGGGACGCTTGGGTGAGA
TCGAATGCGTCAGGATCCT
200
Lactobacillus helveticus R0052
pIR52 AGAATCAAGCAGAGACTGGCTACG
GGACCGGATTTGAGTAGAGGTA
150
Table 4-1: Oligonucleotide primers specific for Protecflor organisms. Primer sequences and other
information were provided by Lallemand (See Methods).
73
Sample Name Number of Sequence Reads Caecal Placebo 1 104,093 Caecal Placebo 2 82,365 Caecal Placebo 3 117,045 Caecal Placebo 4 79,791 Caecal Placebo 5 95,704 Caecal Placebo 6 40,795 Caecal Protecflor 1 83,983 Caecal Protecflor 2 90,416 Caecal Protecflor 3 62,234 Caecal Protecflor 4 95,253 Caecal Protecflor 5 94,558 Caecal Protecflor 6 66,324 Fecal Placebo Cage 1 - 28d 190,856 Fecal Placebo Cage 1 - 42d 119,186 Fecal Placebo Cage 1 - 70d 79,818 Fecal Placebo Cage 1 - 90d 61,661 Fecal Placebo Cage 2 - 28d 73,625 Fecal Placebo Cage 2 - 42d 93,867 Fecal Placebo Cage 2 - 70d 105,389 Fecal Placebo Cage 2 - 90d 38,771 Fecal Protecflor Cage 1 - 28d 137,983 Fecal Protecflor Cage 1 - 42d 75,832 Fecal Protecflor Cage 1 - 70d 77,712 Fecal Protecflor Cage 1 - 90d 43,190 Fecal Protecflor Cage 2 - 28d 82,032 Fecal Protecflor Cage 2 - 42d 69,680 Fecal Protecflor Cage 2 - 70d 96,164 Fecal Protecflor Cage 2 - 90d 104,796
Table 4-2: Number of 16S rDNA sequencing reads per sample. Standard deviation of the number of
sequences read was 30178.27, with a range from 190856 to 38771 (range = 152085).
74
OTUs PC1 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Bacteroidaceae/Bacteroides -0.2 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Prevotellaceae -0.2 Bacteria/Proteobacteria/Betaproteobacteria/Burkholderiales/Alcaligenaceae/ Parasutterella -0.19 Bacteria/Proteobacteria/Alphaproteobacteria/Rhodospirillales/Rhodospirillaceae/ Thalassospira -0.17 Bacteria/Proteobacteria/Epsilonproteobacteria/Campylobacterales/Helicobacteraceae/ Helicobacter -0.16 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Porphyromonadaceae/Parabacteroides -0.14 Bacteria/Proteobacteria/Deltaproteobacteria/Desulfovibrionales/Desulfovibrionaceae/ Desulfovibrio -0.14 Bacteria/Proteobacteria/Gammaproteobacteria/Enterobacteriales/Enterobacteriaceae/ Escherichia-Shigella -0.14 Bacteria/Cyanobacteria/4C0d-2 -0.12 Bacteria/Firmicutes/Clostridia/Clostridiales/Lachnospiraceae/Blautia -0.11 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales -0.09 Bacteria/Verrucomicrobia/Verrucomicrobiae/Verrucomicrobiales/Verrucomicrobiaceae/ Akkermansia -0.09 Bacteria/Firmicutes/Clostridia/Clostridiales/Christensenellaceae/Christensenella -0.08 Bacteria/Firmicutes/Bacilli/Lactobacillales/Streptococcaceae/Streptococcus -0.08 Bacteria/Firmicutes/Bacilli/Lactobacillales/Lactobacillaceae/Lactobacillus -0.06 Bacteria/Firmicutes/Erysipelotrichi/Erysipelotrichales/Erysipelotrichaceae -0.06 Bacteria/Firmicutes/Clostridia/Clostridiales/Lachnospiraceae/Anaerostipes -0.06 Bacteria/Firmicutes/Erysipelotrichi/Erysipelotrichales/Erysipelotrichaceae/Turicibacter -0.03 Bacteria/Firmicutes/Clostridia/Clostridiales/Veillonellaceae/Veillonella -0.01 Bacteria/Cyanobacteria/Chloroplast 0.04 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Prevotellaceae/Prevotella 0.05 Bacteria/Actinobacteria/Coriobacteriia/Coriobacteriales/Coriobacteriaceae 0.05 Bacteria/Firmicutes/Erysipelotrichi/Erysipelotrichales/Erysipelotrichaceae/Allobaculum 0.06 Bacteria/Firmicutes/Clostridia/Clostridiales/Clostridiaceae/Clostridium 0.06 Bacteria/Firmicutes/Bacilli/Bacillales/Staphylococcaceae/Staphylococcus 0.06 Bacteria/Firmicutes/Bacilli/Bacillales/Family-XI-Incertae-Sedis/Gemella 0.06 Bacteria/Firmicutes/Clostridia/Clostridiales/Clostridiaceae/Candidatus-Arthromitus 0.07 Bacteria/Bacteroidetes 0.07 Bacteria/Firmicutes/Clostridia/Clostridiales/Lachnospiraceae 0.09 Bacteria/Firmicutes/Clostridia/Clostridiales/Lachnospiraceae/Roseburia 0.09 Bacteria/Firmicutes/Clostridia/Clostridiales/Peptostreptococcaceae 0.09 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Rikenellaceae/Alistipes 0.1 Bacteria/Tenericutes/Mollicutes/RF9 0.11 Bacteria/Proteobacteria/Gammaproteobacteria/Pseudomonadales/Moraxellaceae/ Acinetobacter 0.11 Bacteria/Firmicutes/Clostridia/Clostridiales/Lachnospiraceae/Coprococcus 0.12
75
Bacteria/Tenericutes/Mollicutes/Anaeroplasmatales/Anaeroplasmataceae/Anaeroplasma 0.13 Bacteria/Deferribacteres/Deferribacteres/Deferribacterales/Deferribacteraceae/ Mucispirillum 0.13 Bacteria/Firmicutes/Clostridia/Clostridiales/Lachnospiraceae/Acetitomaculum 0.13 Bacteria 0.14 Bacteria/Firmicutes 0.14 Bacteria/Firmicutes/Clostridia/Clostridiales/Lachnospiraceae/Marvinbryantia 0.14 Bacteria/Candidate-division-TM7 0.15 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Rikenellaceae/RC9-gut-group 0.15 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Rikenellaceae/Rikenella 0.15 Bacteria/Firmicutes/Clostridia/Clostridiales/Ruminococcaceae/Ruminococcus 0.16 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/S24-7 0.17 Bacteria/Firmicutes/Clostridia/Clostridiales/Ruminococcaceae/Oscillibacter 0.18 Bacteria/Actinobacteria/Coriobacteriia/Coriobacteriales/Coriobacteriaceae/ Enterorhabdus 0.18 Bacteria/Firmicutes/Clostridia/Clostridiales/Christensenellaceae 0.18 Bacteria/Firmicutes/Clostridia/Clostridiales/Peptococcaceae/Peptococcus 0.19 Bacteria/Firmicutes/Clostridia/Clostridiales/Ruminococcaceae 0.2 Bacteria/Firmicutes/Clostridia/Clostridiales 0.2 Bacteria/Firmicutes/Clostridia/Clostridiales/Ruminococcaceae/Anaerotruncus 0.2 Bacteria/Firmicutes/Clostridia/Clostridiales/Peptococcaceae 0.2 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Porphyromonadaceae 0.2 Bacteria/Firmicutes/Clostridia/Clostridiales/Family-XIII-Incertae-Sedis 0.2
Table 4-3: PC1 OTU object loadings on Protecflor treated mouse caecal samples. PCA was
performed on Protecflor and vehicle-treated caecal samples using Ginkgo. OTU object loadings on PC1
were extracted from Ginkgo and ordered from least to greatest.
76
OTUs PC2 Bacteria/Cyanobacteria/Chloroplast -0.23 Bacteria/Firmicutes/Clostridia/Clostridiales/Veillonellaceae/Veillonella -0.17 Bacteria/Firmicutes/Bacilli/Bacillales/Family-XI-Incertae-Sedis/Gemella -0.12 Bacteria/Firmicutes/Clostridia/Clostridiales/Lachnospiraceae/Anaerostipes -0.11 Bacteria/Firmicutes/Clostridia/Clostridiales/Clostridiaceae/Candidatus-Arthromitus -0.1 Bacteria/Proteobacteria/Alphaproteobacteria/Rhodospirillales/Rhodospirillaceae/ Thalassospira -0.09 Bacteria/Firmicutes/Clostridia/Clostridiales/Lachnospiraceae/Roseburia -0.08 Bacteria/Proteobacteria/Gammaproteobacteria/Pseudomonadales/Moraxellaceae/ Acinetobacter -0.08 Bacteria/Firmicutes/Clostridia/Clostridiales/Ruminococcaceae/Oscillibacter -0.07 Bacteria -0.07 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Prevotellaceae/Prevotella -0.07 Bacteria/Firmicutes/Clostridia/Clostridiales/Christensenellaceae -0.02 Bacteria/Firmicutes/Clostridia/Clostridiales/Peptococcaceae -0.02 Bacteria/Proteobacteria/Gammaproteobacteria/Enterobacteriales/Enterobacteriaceae/ Escherichia-Shigella -0.01 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Bacteroidaceae/Bacteroides -0.01 Bacteria/Bacteroidetes 0 Bacteria/Firmicutes/Clostridia/Clostridiales/Lachnospiraceae/Marvinbryantia 0 Bacteria/Cyanobacteria/4C0d-2 0.01 Bacteria/Proteobacteria/Betaproteobacteria/Burkholderiales/Alcaligenaceae/ Parasutterella 0.01 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Prevotellaceae 0.01 Bacteria/Firmicutes/Clostridia/Clostridiales/Ruminococcaceae/Ruminococcus 0.01 Bacteria/Firmicutes/Clostridia/Clostridiales/Lachnospiraceae/Coprococcus 0.01 Bacteria/Firmicutes/Bacilli/Lactobacillales/Streptococcaceae/Streptococcus 0.02 Bacteria/Firmicutes/Clostridia/Clostridiales/Family-XIII-Incertae-Sedis 0.02 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Rikenellaceae/RC9-gut-group 0.03 Bacteria/Firmicutes/Clostridia/Clostridiales 0.03 Bacteria/Tenericutes/Mollicutes/Anaeroplasmatales/Anaeroplasmataceae/Anaeroplasma 0.04 Bacteria/Firmicutes/Clostridia/Clostridiales/Ruminococcaceae 0.04 Bacteria/Actinobacteria/Coriobacteriia/Coriobacteriales/Coriobacteriaceae 0.04 Bacteria/Firmicutes/Clostridia/Clostridiales/Ruminococcaceae/Anaerotruncus 0.06 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Porphyromonadaceae 0.07 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/S24-7 0.07 Bacteria/Firmicutes/Clostridia/Clostridiales/Peptococcaceae/Peptococcus 0.08 Bacteria/Actinobacteria/Coriobacteriia/Coriobacteriales/Coriobacteriaceae/ Enterorhabdus 0.1 Bacteria/Firmicutes/Clostridia/Clostridiales/Peptostreptococcaceae 0.11 Bacteria/Tenericutes/Mollicutes/RF9 0.12 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Rikenellaceae/Alistipes 0.13
77
Bacteria/Firmicutes 0.13 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Rikenellaceae/Rikenella 0.14 Bacteria/Proteobacteria/Epsilonproteobacteria/Campylobacterales/Helicobacteraceae/ Helicobacter 0.14 Bacteria/Deferribacteres/Deferribacteres/Deferribacterales/Deferribacteraceae/ Mucispirillum 0.14 Bacteria/Firmicutes/Erysipelotrichi/Erysipelotrichales/Erysipelotrichaceae/Allobaculum 0.15 Bacteria/Firmicutes/Clostridia/Clostridiales/Clostridiaceae/Clostridium 0.16 Bacteria/Firmicutes/Bacilli/Bacillales/Staphylococcaceae/Staphylococcus 0.16 Bacteria/Candidate-division-TM7 0.18 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales/Porphyromonadaceae/Parabacteroides 0.18 Bacteria/Proteobacteria/Deltaproteobacteria/Desulfovibrionales/Desulfovibrionaceae/ Desulfovibrio 0.19 Bacteria/Firmicutes/Clostridia/Clostridiales/Lachnospiraceae/Acetitomaculum 0.2 Bacteria/Firmicutes/Erysipelotrichi/Erysipelotrichales/Erysipelotrichaceae 0.2 Bacteria/Firmicutes/Clostridia/Clostridiales/Lachnospiraceae/Blautia 0.23 Bacteria/Bacteroidetes/Bacteroidia/Bacteroidales 0.23 Bacteria/Verrucomicrobia/Verrucomicrobiae/Verrucomicrobiales/Verrucomicrobiaceae/ Akkermansia 0.24 Bacteria/Firmicutes/Bacilli/Lactobacillales/Lactobacillaceae/Lactobacillus 0.24 Bacteria/Firmicutes/Clostridia/Clostridiales/Christensenellaceae/Christensenella 0.25 Bacteria/Firmicutes/Erysipelotrichi/Erysipelotrichales/Erysipelotrichaceae/Turicibacter 0.29 Table 4-4: PC2 OTU object loadings on Protecflor treated mouse caecal samples. PCA was
performed on Protecflor and vehicle-treated caecal samples using Ginkgo. OTU object loadings on PC2
were extracted from Ginkgo and ordered from least to greatest.
78
Chapter 5 Discussion
79
5 Discussion
5.1 The effect of oral gavage transfer of female NOR caecal microbiota into female NOD recipients on microbiome composition and T1D incidence
5.1.1 Summary
We present this set of experiments as our effort to validate qPCR as viable method for the
detection and quantification of members of specific bacterial taxa, and the application of such an
assay panel to probe the effect of the microbiome on T1D development in the NOD mouse
model. Specifically, we applied the assay to quantify changes in the gut microbiome of female
NOD mice from weaning to adulthood. We quantified changes associated with gavage treatment
of weanling female NOD mice with caecal material from donor adult NOR mice. We followed
up on the changes we detected by looking at the effect of gavage treatment with NOR-derived
material in NOD mice on the incidence and progression to T1D.
5.1.2 Use of “indicator” taxa
The microbiome is a complex community of organisms, and the ecology and interaction of these
organisms is only beginning to be understood. High-throughput 16S sequencing provides deep
taxonomical overview of the target microbiome. However, despite continuing advances in
sequencing technology, the cost of HTS can be prohibitive for large sample sets, longitudinal
studies, and routine verification of experimental conditions. In experimental comparisons where
prior sequencing data has distinguished microbes which are consistently indicators of differences
in the microbiome, we suggest that taxa-specific qPCR assays can be used to supplement
sequencing technology in a more cost-effective manner.
In our study, prior sequencing data from our lab had shown the microbiomes of NOD and NOR
animals consistently differ in five bacterial taxa. We designed and validated the specificity of a
set of taxa-specific qPCR primers by in silico and in vitro methods. We determined ranges of
linear amplification and amplification efficiency of these assays, and validated the ability of
designed taxa-specific qPCR assays to produce consistent and sensitive quantification of
bacterial OTUs present in the microbiome of NOD mice. We found that these 5 taxa could act as
low resolution indicator taxa of the state of the microbiome; differentially abundant taxa we
80
previously saw by sequencing the microbiome of NOD and NOR mice were reflected by changes
in relative abundance in the indicator taxa as quantified by qPCR. We also detected changes in
the microbiome of mice as they aged from weaning to adulthood, suggesting that these 5 taxa
could also serve as rough indicators of the community state of the microbiome.
5.1.3 Comparison of abundance of caecal bacterial taxa in NOD mice from weaning to adulthood
When we applied our panel of assays to tracking changes in the composition of the microbiome
from weaning to adulthood, we found that changes in the gut microbiome from weaning to
adulthood could be detected and quantified by taxa-specific qPCR, and a major shift in the
microbiome composition was observed between 25d and 41d female NOD mice.
Our data shows major shifts in the qPCR profiles of the gut microbiome at 25, 33 and 41d mice,
with the profile stabilizing between 41d and 55d animals. Previously, our lab showed the ability
to perturb the developing mouse microbiome by the introduction of male donor caecal contents
to weanling (21 days of age) female mice, suggesting that the gut microbiome at weaning is
alterable by the introduction of exogenous bacteria[50]. Ivanov et al. also showed successful
transfer of SFB and subsequent effects on the TH-17 cell population by gavage at 28d[60]. Our
results corroborate the unstable state of the gut microbiome between weaning at 21d and puberty
at 35-42d. The stabilization of the gut microbiome after puberty at ~40d is in agreement with
previous work in our lab showing that puberty is associated with a divergence of the
microbiomes of male and female mice[50]. Together, these data suggests a consistent shift in the
microbiome of NOD mice around the 30-40 day time point.
We speculate, in agreement with previous work in our lab, that this may be related to the onset of
puberty in mice[50], although further evidence is needed to determine the correlation between
puberty and this microbiome shift. Previous work in our lab found that the microbiome of male
and female NOD mice diverged in composition at the onset of puberty[50]. Other studies have
provided evidence that the puberty and sex hormones may have an effect on microbiome
composition. In humans, studies have shown changes in the microbiome of other body sites
(gingiva, nares, skin, vaginal canal) associated with the onset of puberty[109][110][111]. A study
on salivary bacterial communities also found the abundance of selected bacterial taxa
(Veillonellaceae and Actinomycetaceae) were correlated with age from adolescence to early
81
adulthood[112]. The gut microbiome of human adolescents has also been found to differ from
that of adults, with significantly higher abundance of genera Bifidobacterium and Clostridium
among adolescent samples[113]. Other studies have also shown that the gut microbiome
develops in conjunction with human host development from infancy to adulthood, modified by
geography[56]; most notably in infancy at weaning and the introduction of solid food[114].
5.1.4 Comparison of abundance of caecal bacterial taxa in NOD, NOR and NOR-gavaged NOD Mice
We applied the qPCR panel to quantifying bacterial taxa in unmanipulated NOD and NOR mice.
We observed differences in the microbiome composition of NOD and NOR mice. Bacteroidetes
and Helicobacter were enriched in NOR animals and Clostridia IV, and Ruminococcus were
enriched in NOD animals. These differences were in agreement in directionality, but not
magnitude with previous sequencing data in our lab comparing the microbiomes of NOD and
NOR animals. We also found that variance of abundance in Clostridia IV, Helicobacter, and
Ruminococcus was higher between NOR animals compared to NOD. Both the differential
abundance and differential variance suggest NOR animals may have differing microbiome
communities than NOD.
When we compared bacterial taxa in unmanipulated NOD, NOD recipients of NOR caecal
material by gavage (NOR-gavaged NOD), and NOR mice, we found significant differences
between untreated NOD and NOR-gavaged NOD mice in the Clostridia IV and Clostridia XIVa.
These differences suggest that the gavage treatment has some effect on the composition of
microbiome of treated NOD mice. Data also showed statistically significant differences between
NOR and NOR-gavaged NOD mice in the Clostridia IV, Clostridia XIVa, and Helicobacter
groups. These data suggested that the resulting microbiome of recipient mice is not identical to
that of the donor, nor is it clearly partially composed of NOD and NOR contributions. This is
consistent with data from previous experiment in our lab comparing female, male-gavaged
female and male NOD mice by 16S sequencing. Male-gavaged female NOD mice were found to
have microbiomes which were significantly different from either unmanipulated female or male
NOD mice[50].
Except in the Clostridia IV and Ruminococcus, samples from the same strain (NOD/NOR) were
clustered. These results are consistent with previous findings in our lab; previous sequencing
82
results comparing the gut microbiome of NOD and NOR mice found inter-strain differences
between NOD and NOR, and similarity between individuals from the same strain. These data
support the idea of a stable microbiome “archetype” which is mostly dictated by the genotype of
the host mouse. Other studies suggest that microbiomes do develop similarly within mouse
strains; Hildebrand et al. found genetic distance was associated with microbiota distance between
four strains of inbred mice (B6, BALB/c, FVB and NOD)[115]. Other studies also support the
idea that genotypic differences are a significant determinant in microbiome composition. Benson
et al. described quantitative trait loci in mice associated with relative abundances of specific
microbial taxa in an intercross line between C57BL/6J and an ICR-derived outbred line [116].
Alexander et al. showed intra-strain differences in Altered Schaedler Flora (ASF) colonization
between multiple barrier-raised inbred strains were maintained despite co-housing[117].
We suggest that the effects of genetics on the composition or functional capacity of the gut
microbiome may be an important phenotypic outcome of genotypic differences. Other studies
have also suggested genetic differences amongst inbred mouse strains are an important factor in
maintaining distinct microbiome profiles under similar environmental exposures[117]. The NOD
mouse, when contrasted with NOR and NOD.NOR congenic animals, provides a good
experimental system in which to study the effects of genetics on microbiome composition.
Previous work in our lab and others has generated an extensive number of congenic strains in the
course of dissecting the genetic contribution of IDD loci to autoimmune defects in the NOD
mouse. Analysis of the microbiome, both by composition and functional differences, of these
closely related congenic strains may shed more light on the microbiome phenotypic outcomes of
T1D risk loci.
5.1.5 Interstrain microbiome variation
Despite most members of a strain showing similar microbiome composition, significant variation
was also observed, most notably in the Clostridia IV group for NOR mice. Microbiome
divergence between NOR mice was greater than that in NOD. This data is consistent with data
presented by Alexander et al., showing that the degree of microbiome divergence between non-
littermates depended on the genotype of the mouse[117]. However, other studies have suggested
that maternal transmission and housing conditions are more important in determining the
composition of the microbiome. Specifically, Benson et al. found in an advanced intercross line
83
derived from C57BL/6J and an ICR-derived HR line that variation in microbiome composition
could be segregated by a mixed-model analysis into parental, sex and cohousing effects[116].
Ubeda et al. also found that genetic defects in innate immune sensing (MyD88-/-, TLR2-/-, TLR4-
/-, TLR5-/-, and TLR9-/-) in C57BL/6 mice did not have a significant impact on the composition,
diversity, or antibiotic recovery of the microbiome when compared to wildtype littermates. In
contrast, clear maternal and litter effects were observed[118].
Further experiments are needed to narrow down the cause of the intra-strain microbiome
variation; but we hypothesize the variation could be due to differences in the composition of the
initial colonizing organisms. This hypothesis would include differences in this composition
caused by variations in environment, such as those caused by differences in cage environment. In
this case, we would expect variations in the microbiome to segregate along these environmental
variations. Assorting the data by cage/littermates, we found no clustering of cage/littermates into
groups of similar relative abundance. This suggests that housing cages were not related to the
intra-strain variation we observed. Of course, our observations were limited by the sample size,
in this case. Further experiments or meta-analysis of collected T1D natural history data may shed
more on the issue of cage environment.
Other studies have shown that housing environment and cohousing can have a significant effect
on the composition of the microbiome, and moreover, these alterations of the microbiome can be
linked to inflammatory phenotypes. Elinav et al. found that the colitogenic microbiome of
NLRP6-/- mice was transferable by cohousing with wildtype mice, and furthermore, in cohoused
wildtype mice, the colitogenicity of the dysbiosis was recapitulated[119].
5.2 The effect of Protecflor probiotic treatment on NOD microbiome composition and insulitis severity
5.2.1 Summary
We present this set of experiments as our effort to probe effect of Protecflor probiotic treatment
on the gut microbiome, and on its effect on T1D development in the NOD mouse model.
Specifically, we validated qPCR as viable method for the detection and quantification of
members of specific bacterial taxa, and the applied this to determining the gastrointestinal transit
time of the Protecflor organisms in female NOD mice to validate the gavage treatment of mice
84
with Protecflor. We then looked at the effect of gavage and water treatment with Protecflor in
NOD mice on insulitis progression and severity. We followed up with 16S high-throughput
sequencing of caecal and fecal samples from these mice to quantify changes in the gut
microbiome associated with Protecflor treatment.
By in silico and in vitro methods, we validated the specificity of a set of probiotic strain-specific
qPCR primers conveyed to us by our collaborators at Lallemand. We determined ranges of linear
amplification and amplification efficiency of these assays, and validated the ability of designed
taxa-specific qPCR assays to produce consistent and sensitive quantification of probiotic strain
present in the guts of Protecflor-treated mice.
5.2.2 Gastrointestinal Transit times of Protecflor Probiotic organisms in the NOD model
We applied our assays to tracking the gastrointestinal transit of the Protecflor probiotic strains in
NOD mice gavaged with Protecflor. We found that Protecflor strains could be detected in the
stomach, small intestine and, caecum and large intestine of treated animals after gavage.
Protecflor organisms were rapidly cleared from the GI tract, and could not be detected by qPCR
4 hours after gavage treatment. Our data shows that Protecflor strains could be detected in the
stomach, small intestine and, caecum and large intestine of treated animals by qPCR up to 4
hours after gavage.
In other studies, a wide range of gastrointestinal transit times in mice have been observed,
depending on the methodology and mouse strain used. Schwarz et al. found GI transit times of
154 ± 24 minutes in BL/6 mice using a contrast MRI-based measurement method[101].
Padmanabhan et al. found a GI transit time of 6h in 129SvEv mice using a radiolabelled charcoal
pellet[102]. One study found, in NOD mice, a bolus consisting of 10% charcoal in 5% Arabic
gum aqueous suspension had transited 60% of the small intestine 20 minutes after feeding[120].
Our results corroborate this range of transit times, and add to the evidence that a liquid bolus has
faster transit times compared to a solid bolus.
85
5.2.3 Evaluating the Effects of Protecflor Probiotic on Insulitis Severity in NOD model
To test the effect of Protecflor probiotic on autoimmune insulitis, we assessed insulitis severity
in NOD mice treated with Protecflor and vehicle. We found that Protecflor treatment was
protective against insulitis and that Protecflor-treated mice showed less immune infiltration of
the islets compared to vehicle-treated mice.
Other studies have shown that probiotic formulations containing the organisms in Protecflor are
beneficial in irritable bowel syndrome and ulcerative colitis. Lactobacillus helveticus R0052 and
Lactobacillus rhamnosus R0011 have been shown to reduce clinical symptoms of irritable bowel
syndrome in open label experiments on human adults with irritable bowel syndrome [121].
Bifidobacterium longum R0175, given in combination with inulin in a small pediatric study of
ulcerative colitis (UC) was found to improve self-reported quality of life, UC symptoms of blood
and mucous in stool, and to reduce relapse rate[71]. Oral administration of probiotic bacterial
strains related to the Protecflor organisms has also been found to protect NOD mice from T1D.
Specifically, the probiotic mixture VSL#3 (Bifidobacterium, Lactobacillus species and
Streptococcus salivarius subsp. Thermophiles) was given by oral gavage three times a week from
4 to 32 weeks of age, reduced T1D incidence in NOD mice [49].
Our results demonstrate the protective effect of the Protecflor organisms against insulitis, a pre-
T1D phenotype, in a single-gavage experimental setting. This is novel compared to three times a
week gavage treatment tested using VSL#3. Of particular importance is that multiple gavages
increase the risk of systemic exposure of microbes to the immune system, by abrasion of the
esophageal wall, or aspiration of the gavage material. It is well documented by multiple studies
that bacterial infection or systemic exposure is protective against T1D in NOD mice. Studies
showing deliberate infection with multiple organisms, including Schistosoma mansoni[42],
Mycobacterium avium[43], Trichinella spiralis[44], Heligmosomoides polygyrus[45] and
Salmonella typhimurium[46], can induce protection in NOD mice. Interperitonial injection of
zymosan, a fungal cell wall component, has also been found to be protective in the NOD
model[47]. In contrast, our protocol limits the frequency of the gavage procedure, and probiotic
dosing is maintained by addition to drinking water.
86
5.2.4 The Effects of Protecflor Probiotic treatment on the gut microbiome in NOD mice
We used Illumina high-throughput sequencing of the bacterial 16S rDNA gene to quantify
changes in the composition of the gut microbiome associated with Protecflor treatment.
Sequencing results showed enrichment of the genera Allobaculum, Candidatus Arthromitus,
Roseburia, and the order Clostridiales in caecal flush and fecal samples from Protecflor-treated
animals compared to controls, and depletion of the genera Helicobacter, Parabacteroides and the
order Bacteroidales, in Protecflor treated vs. controls.
5.2.4.1 Taxa enriched in Protecflor treated mice
Overall, the OTUs enriched in samples from Protecflor-treated mice were depleted in TH1-
associated inflammatory disorders or linked to protection from T1D in mice or humans. Our
results corroborate the association of some bacterial taxa, in particular SFB and Roseburia, with
protection in the NOD T1D setting.
5.2.4.1.1 Allobaculum
Allobaculum is a newly recognized genus with a single species first isolated from canine
feces[122]. The type strain of Allobaculum stercoricanis was found to produce butyrate in
glucose and other hexose metabolism. This data suggests that Allobaculum may be able to
modulate host mucosal immunity from SCFA metabolites (see below). There is also evidence
Allobaculum is associated with Th2 immune skewing. Allobaculum was found to be the most
abundant OTU in arthritis-susceptible HLA transgenic DRB1*0401 mice, compared to arthritis-
resistant HLA transgenic DRB1*0402 mice[123]. Allobaculum was also enriched in control diet
fed mice and depleted in diet-induced obese mice[124].
5.2.4.1.2 Segmented Filamentous Bacteria (SFB)
The association of SFB with Protecflor treatment and protection is consistent with evidence from
previous studies. Kriegel et al. found the presence of SFB in feces was correlated with diabetes
protection in female NOD mice, but insulitis did not depend on SFB colonization[62].
Yurkovetskiy et al. found SFB monocolonization was protective in male by not female NOD
mice, compared to germ-free mice [125]. SFB-positive fecal transfer suppressed the insulitis and
T1D incidence in NOD mice that were on acidified (pH 3.2) water but not in those on neutral
87
water (pH 7.0)[126]. Interestingly, bacterial taxa which were enriched in NOD mice acidified
(pH 3.2) water in another study[127] were also loaded positively on PC1 and were associated
with protection from insulitis in Protecflor-treated mice. However, it should be noted that these
two studies suggest opposite effects of acidified water in NOD T1D incidence[127][126].
5.2.4.1.3 Roseburia
Roseburia is a butyrate-producing genus in the Lachnospiraceae family. Roseburia has been
associated in previous studies with T1D-protected populations in both the NOD mouse model
and in human T1D case-control studies. Previous work in our lab by Markle et al. found
Roseburia was enriched in male vs. female NOD mice, and also in female recipients of gavage
material derived from male caecal contents vs. female caecal contents[50]. Roseburia was found
to be enriched in control vs. T1D case subjects, along with other butyrate-producing genera[128].
Goffau et al. found that Roseburia was enriched in healthy controls vs. children which developed
anti-islet autoantibodies[129].
5.2.4.2 Taxa depleted in Protecflor treated mice
Overall, the OTUs enriched in samples from vehicle-treated mice were linked in small studies to
T1D in mice or humans. Our results corroborate the association of some bacterial taxa, in
particular SFB and Roseburia, with protection in the NOD T1D setting.
5.2.4.2.1 Helicobacter
Helicobacter species have been associated with both protection and disease progression in
previous studies. One Japanese case report reported the eradication of a Helicobacter pylori
infection in an adult (42 year old) patient with a high risk HLA haplotype was followed by the
diagnosis with T1D[130].
Other studies have suggested that Helicobacter is a bystander to T1D, and changes in host
physiology affect Helicobacter survival. Gasbarrini et al. found Helicobacter pylori infection in
T1D patients was more resistant to eradication by triple antibiotic therapy. Other studies have
suggested Helicobacter seropositive status has a time-dependent relationship with T1D. de Luis
et al. found that seroprevalence of Helicobacter pylori was higher among younger (<24 years
old) T1D patients vs. healthy controls, but lower in older (<24 years old) T1D patients[131].
88
One study found seroprevalence of Helicobacter pylori was significantly higher in patients with
T1D than in healthy controls[132]. It has been suggested that Helicobacter is a possible trigger
for gastric autoimmunity by molecular mimicry, as it expresses lipopolysaccharides which mimic
Lewis y, Lewis x, and H type I blood group structures similar to those commonly occurring in
gastric mucosa. One study found FUT2 non-secretor status, resulting in a lack of Lewis antigen
expression, was linked to T1D susceptibility, as well as resistance to pathogens which mimic the
Lewis antigens, such as H. pylori[133].
5.2.4.2.2 Parabacteroides
Previous work in our lab by Markle et al. found Parabacteroides was enriched in 14 week old
female NOD mice vs. male NOD mice and 14 week old vs. 34 week old NOD female recipients
of male caecal contents[50].
5.2.4.2.3 Bacteroidales
Multiple studies have found Bacteroides, a genus within Bacteroidales, is enriched in patients
with T1D vs. healthy controls. Brown et al. found Bacteroides were enriched in T1D patients vs.
healthy controls in a small cohort of age/HLA matched children[128]. Similarly, Mejía-León et
al. found Bacteroides were enriched in newly-diagnosed Mexican T1D patients vs. healthy
controls[134], and Murri et al. found Bacteroides were enriched in T1D cases vs. healthy
controls by PCR-DGGE[135]. Giongo et al., in a Finnish T1D case-control study, found
Bacteroides were enriched in T1D patients vs. healthy controls. Bacteroides enrichment was also
correlated with a pre-T1D phenotype, anti-islet autoantibodies, in a study by Goffau et al[129].
5.2.5 Principal components analysis of Protecflor and vehicle-treated caecal samples
We applied principal components analysis to our sequencing data in order to reduce the
dimensionality of the data, reveal hidden structure and correlate changes in the microbiome with
insulitis. We found that 53.47% of the total variance could be explained by 2 principal
components, PC1 and PC2. Positive loading on PC1 were populated with Order Clostridales,
including the Ruminococcaceae and Lachnospiraceae families. The negative loadings were
populated with Order Bacteroidales, and Phylum Proteobacteria. Samples were distributed
along PC1 in two clusters, regardless of treatment. Positive loadings for PC2 were populated
89
with family Erysipelotrichaceae, Order Clostridales, specifically from the Lachnospiraceae
family, and Order Bacteroidales, including the Rikenellaceae and Porphyromonadaceae
families. The negative loadings were populated with phylum Cyanobacteria, and Order
Clostridales, including the Veillonellaceae and Lachnospiraceae families, and Candidatus-
Arthromitus (SFB). Samples were more evenly distributed along PC2, also independent of
treatment. Positive loading on PC1 was correlated with decreased insulitis in Protecflor-treated
mice, but not vehicle-treated mice. In contrast, positive loading on PC2 was correlated with
decreased insulitis in vehicle-treated mice, but not Protecflor-treated mice.
5.2.5.1 Bacterial taxa associated with T1D in NOD mice and humans
Bacterial taxa found to significantly load in our principal components analysis have been linked
in other studies to T1D in other murine experimental settings. Roseburia, Coprococcus,
Peptococcus, all members of Order Clostridales, found to load positively on PC1 and be
associated with protection from insulitis in Protecflor-treated mice in our study, were found to be
enriched in NOD female recipients of male-derived microbiome transplant compared to
untreated females at 14 weeks of age[50]. Members of Family Ruminococcaceae, Rikenellaceae
and Erysipelotrichaceae were found to load positively on PC1 and be associated with protection
from insulitis in Protecflor-treated mice in our study, were all also found to be enriched in T1D-
protected NOD mice treated with acidified (pH 3.2) water vs. controls treated with neutral (pH 7)
water[127]. Lactobacillaceae, Erysipelotrichaceae, and Rikenellaceae, found to load positively
on PC2 and be associated with protection from insulitis in untreated mice in our study, were all
also found to be enriched in T1D-protected NOD.MyD88-/- vs. SPF NOD controls[48].
Bacterial taxa found to significantly load in our principal components analysis have been linked
in other studies to T1D in humans. Familes Ruminococcaceae, Family XIII Incertae Sedis, and
Lachnospiraceae, all members of Order Clostridales, found to load positively on PC1 and be
associated with protection from insulitis in Protecflor-treated mice in our study were found to be
enriched in healthy controls vs. T1D patients in a small cohort of age/HLA matched
children[128].
90
5.2.5.2 Proteolytic Bacteria
Dietary triggers have been implicated in the etiology of T1D in humans, particularly in
association with cow’s milk proteins[136] and gluten, although conflicting evidence has been
presented. The epidemiological link between T1D and celiac disease is also suggestive of gluten
as a dietary trigger. Previous studies have suggested that microbial peptidases present in the gut
may degrade immunogenic peptides and reduce autoimmunity. In our study,
Peptostreptococcaceae, Clostridium, Peptococcus, taxa with known proteolytic/amino acid
fermenting ability[137][138], were found to load positively on both PC1 and PC2 and be
associated with protection from insulitis in all experimental groups in our study.
5.2.5.3 Short Chain Fatty Acid Producers
Short chain fatty acids (SCFAs) such as butyrate and propionate are products of the anaerobic
bacterial fermentation of dietary fiber in the host gut[139]. Previous studies have highlighted the
phylogenetic relationships between the major bacterial taxa involved in SFCA production in the
gut[140][141][142]. SCFAs bind and signal through previously orphaned G Protein-coupled
Receptors (GPRs) GPR41, GPR43[143][144] and GPR109A[145]. SFCA signaling through
GPR43 has been implicated in regulating the gut mucosa-associated immune system. SFCA
added to drinking water was found to increase colonic Treg frequency and number in germ free
mice, and stimulate in vitro production of IL-10, and this effect is thought to be mediated by
GPR43 signaling[146]. Bacterially-derived SCFAs were also found to stimulate TGF-β
production by the gut epithelium[147].
Propionate and butyrate have also been found to have Histone De-ACetylase (HDAC) inhibitory
activity[148]. This HDAC inhibitory activity may also have a role in regulating mucosal
immunity in the gut. Treatment of macrophages with butyrate led to the down-regulation of pro-
inflammatory mediators, including nitric oxide, IL-6, and IL-12 in a GPR and TLR independent
mechanism[149]. Lin− bone marrow cells treated with SCFAs were blocked from DC, but not
granulocyte development[150].
Although butyrate and propionate have similar signaling activity in many studies, some have
suggested differential activity in these SCFAs. Butyrate and propionate act in an opposed fashion
in IL-2 and IFN-γ production in cultured mesenteric lymph node lymphocytes, with butyrate
91
promoting IFN-γ production , and propionate partially rescuing IL-2 inhibition by butyrate[151].
There is evidence suggesting that GPR109A interacts with butyrate by not propionate, providing
another source of differential signaling[145].
In this study, order Clostridiales members Clostridium, Roseburia, Coprococcus, Peptococcus,
and Anaerotruncus, known to specifically produce butyrate (butanoic acid)[140][141], were
found to load positively on PC1 and be associated with protection from insulitis in Protecflor-
treated mice in our study. In contrast, Bacteroides, Escherichia, Veillonella, Prevotella, bacterial
genera known to specifically produce propionate (propanoic acid)[152], were found to load
negatively on PC1 and PC2 and be associated with insulitis in both vehicle and Protecflor-treated
mice in our study. Together, this evidence suggests there may be a role of the differential
regulation of mucosal immunity by propionate and butyrate in the pathogenesis of T1D.
5.2.5.4 Enterotypes in NOD and T1D
Hildebrand et al. observed in their study the segregation of other inbred mouse strains into
archetypes or “enterotypes,” which were mostly independent of genotype, but not in NOD[115].
They also found that fecal calprotectin levels, an indication of intestinal inflammation in
IBD[153] and IBS[154], were elevated in mice harbouring the low-richness, Bacteroides-rich
enterotype. In their study, calprotectin levels were significantly negatively correlated to
Lachnospiraceae, Rikenellaceae, Ruminococcaceae as well as Prevotellaceae, while the positive
correlation to Bacteroidaceae, Verrucomicrobiaceae, Enterobacteriaceae and
Burkholderiales[115].
The clustering of our NOD-derived samples into two clusters, primarily differentiated by PC1,
provides initial evidence that these enterotypes, with microbial compositions corresponding
those found in Hildebrand et al., also exist in NOD mice. Cluster 1 and Cluster 2 correspond by
composition to the low-richness, Bacteroides-rich enterotype, and high-richness, Ruminococcus-
rich enterotypes in Hildebrand et al.
Based on the presence of similar “enterotypes” in untreated adult mice, both NOD and other
genotypes, we suggest that these archetypes represent two basal microbiota communities which
can stabilize in the un-manipulated mouse gut. Furthermore, having found protection from
insulitis is well-correlated with cluster membership in Protecflor-treated animals, we speculate
92
that the Cluster 2-type “enterotype” may be a “receptive” microbiome to Protecflor treatment, or
may be indicative of a similarly “receptive” host.
The correlation between PC loading organisms and insulitis in untreated NOD mice also raises
the intriguing possibility that the abundance of some group of bacterial taxa are directly
proportional to insulitis severity, either causally or in response to host factors.
5.2.6 Species richness and diversity in Protecflor and vehicle-treated animals
We applied ecological richness and diversity metrics to the sequencing samples in order to
correlate changes in the community structure with changes in the microbiome OTU composition
and insulitis. We found that community richness, as measured by the Chao1 statistic, was
correlated with changes in microbiome composition associated with positive loadings on PC1 in
all mice, and with decreased insulitis severity in Protecflor, but not vehicle-treated mice. We
found that community diversity, as measured by Shannon’s H statistic, was correlated with
changes in microbiome composition associated with positive loadings on PC2 and with
decreased insulitis in vehicle, but not Protecflor-treated mice.
Other studies have related both changes in community richness and diversity with dysbiosis
associated with T1D in both NOD mice and humans. A study on NOD mice treated with gluten-
free chow found T1D incidence was lower in mice given gluten-free chow, and these mice
exhibited increased gut microbial community richness, as measured by the Chao1 statistic.
However, community diversity, as measured by Shannon’s H statistic was not altered by gluten-
free chow[155]. Wolf et al. also found the consumption of acidic (pH 3.2) water by NOD mice
altered both community richness and diversity, although not statistically significantly[127]. In
humans, a number of small studies have found correlations between community richness and
diversity and T1D. Giongo et al. found microbial community diversity increased over time in
both healthy children and T1D patients, but much more rapidly in controls, while diversity in
T1D patients plateaued as they approached clinical autoimmunity and T1D diagnosis[156].
Together, these results suggest that diet-based treatment in the host can alter the overall gut
microbiome community structure, and these alterations in community richness and diversity may
be correlated with T1D pathogenesis.
93
References
[1] M. A. Atkinson and E. H. Leiter, “The NOD mouse model of type 1 diabetes: As good as it gets?,” Nat. Med., vol. 5, no. 6, pp. 601–604, Jun. 1999.
[2] M. G. Walsh, J. Zgibor, K. Borch-Johnsen, and T. J. Orchard, “A multinational assessment of complications in type 1 diabetes: the DiaMond substudy of complications (DiaComp) Level 1,” Diab. Vasc. Dis. Res., vol. 3, no. 2, pp. 84–92, Sep. 2006.
[3] E. A. M. Gale, “The Rise of Childhood Type 1 Diabetes in the 20th Century,” Diabetes, vol. 51, no. 12, pp. 3353–3361, Dec. 2002.
[4] M. Karvonen, M. Viik-Kajander, E. Moltchanova, I. Libman, R. LaPorte, and J. Tuomilehto, “Incidence of childhood type 1 diabetes worldwide. Diabetes Mondiale (DiaMond) Project Group.,” Diabetes Care, vol. 23, no. 10, pp. 1516–1526, Oct. 2000.
[5] P. Onkamo, S. Väänänen, M. Karvonen, and J. Tuomilehto, “Worldwide increase in incidence of Type I diabetes – the analysis of the data on published incidence trends,” Diabetologia, vol. 42, no. 12, pp. 1395–1403, Dec. 1999.
[6] S. T. Bennett and and J. A. Todd, “HUMAN TYPE 1 DIABETES AND THE INSULIN GENE: Principles of Mapping Polygenes,” Annu. Rev. Genet., vol. 30, no. 1, pp. 343–370, 1996.
[7] N. Risch, “Assessing the role of HLA-linked and unlinked determinants of disease.,” Am. J. Hum. Genet., vol. 40, no. 1, pp. 1–14, Jan. 1987.
[8] J. Kaprio, J. Tuomilehto, M. Koskenvuo, K. Romanov, A. Reunanen, J. Eriksson, J. Stengård, and Y. A. Kesäniemi, “Concordance for type 1 (insulin-dependent) and type 2 (non-insulin-dependent) diabetes mellitus in a population-based cohort of twins in Finland,” Diabetologia, vol. 35, no. 11, pp. 1060–1067, Nov. 1992.
[9] J. C. Barrett, D. G. Clayton, P. Concannon, B. Akolkar, J. D. Cooper, H. A. Erlich, C. Julier, G. Morahan, J. Nerup, C. Nierras, V. Plagnol, F. Pociot, H. Schuilenburg, D. J. Smyth, H. Stevens, J. A. Todd, N. M. Walker, and S. S. Rich, “Genome-wide association study and meta-analysis find that over 40 loci affect risk of type 1 diabetes,” Nat. Genet., vol. 41, no. 6, pp. 703–707, Jun. 2009.
[10] J. A. Todd, J. I. Bell, and H. O. McDevitt, “HLA-DQβ gene contributes to susceptibility and resistance to insulin-dependent diabetes mellitus,” Nature, vol. 329, no. 6140, pp. 599–604, Oct. 1987.
[11] S. Onengut-Gumuscu, K. G. Ewens, R. S. Spielman, and P. Concannon, “A functional polymorphism (1858C/T) in the PTPN22 gene is linked and associated with type I diabetes in multiplex families,” Genes Immun., vol. 5, no. 8, pp. 678–680, Dec. 2004.
[12] J. Yamanouchi, D. Rainbow, P. Serra, S. Howlett, K. Hunter, V. E. S. Garner, A. Gonzalez-Munoz, J. Clark, R. Veijola, R. Cubbon, S.-L. Chen, R. Rosa, A. M. Cumiskey,
94
D. V. Serreze, S. Gregory, J. Rogers, P. A. Lyons, B. Healy, L. J. Smink, J. A. Todd, L. B. Peterson, L. S. Wicker, and P. Santamaria, “Interleukin-2 gene variation impairs regulatory T cell function and causes autoimmunity,” Nat. Genet., vol. 39, no. 3, pp. 329–337, Mar. 2007.
[13] L. Nisticò, R. Buzzetti, L. E. Pritchard, B. V. der Auwera, C. Giovannini, E. Bosi, M. T. M. Larrad, M. S. Rios, C. C. Chow, C. S. Cockram, K. Jacobs, C. Mijovic, S. C. Bain, A. H. Barnett, C. L. Vandewalle, F. Schuit, F. K. Gorus, R. Tosi, P. Pozzilli, and J. A. Todd, “The CTLA-4 Gene Region of Chromosome 2q33 Is Linked to, and Associated with, Type 1 Diabetes,” Hum. Mol. Genet., vol. 5, no. 7, pp. 1075–1080, Jul. 1996.
[14] D. J. Smyth, J. D. Cooper, R. Bailey, S. Field, O. Burren, L. J. Smink, C. Guja, C. Ionescu-Tirgoviste, B. Widmer, D. B. Dunger, D. A. Savage, N. M. Walker, D. G. Clayton, and J. A. Todd, “A genome-wide association study of nonsynonymous SNPs identifies a type 1 diabetes locus in the interferon-induced helicase (IFIH1) region,” Nat. Genet., vol. 38, no. 6, pp. 617–619, Jun. 2006.
[15] H.-C. So, A. H. S. Gui, S. S. Cherny, and P. C. Sham, “Evaluating the heritability explained by known susceptibility variants: a survey of ten complex diseases,” Genet. Epidemiol., vol. 35, no. 5, pp. 310–317, Jul. 2011.
[16] K. A. Hunt, V. Mistry, N. A. Bockett, T. Ahmad, M. Ban, J. N. Barker, J. C. Barrett, H. Blackburn, O. Brand, O. Burren, F. Capon, A. Compston, S. C. L. Gough, L. Jostins, Y. Kong, J. C. Lee, M. Lek, D. G. MacArthur, J. C. Mansfield, C. G. Mathew, C. A. Mein, M. Mirza, S. Nutland, S. Onengut-Gumuscu, E. Papouli, M. Parkes, S. S. Rich, S. Sawcer, J. Satsangi, M. J. Simmonds, R. C. Trembath, N. M. Walker, E. Wozniak, J. A. Todd, M. A. Simpson, V. Plagnol, and D. A. van Heel, “Negligible impact of rare autoimmune-locus coding-region variants on missing heritability,” Nature, vol. 498, no. 7453, pp. 232–235, Jun. 2013.
[17] G. Lettre and J. D. Rioux, “Autoimmune diseases: insights from genome-wide association studies,” Hum. Mol. Genet., vol. 17, no. R2, pp. R116–121, Oct. 2008.
[18] D. J. Smyth, V. Plagnol, N. M. Walker, J. D. Cooper, K. Downes, J. H. M. Yang, J. M. M. Howson, H. Stevens, R. McManus, C. Wijmenga, G. A. Heap, P. C. Dubois, D. G. Clayton, K. A. Hunt, D. A. van Heel, and J. A. Todd, “Shared and distinct genetic variants in type 1 diabetes and celiac disease,” N. Engl. J. Med., vol. 359, no. 26, pp. 2767–2777, Dec. 2008.
[19] M. Mäki, K. Mustalahti, J. Kokkonen, P. Kulmala, M. Haapalahti, T. Karttunen, J. Ilonen, K. Laurila, I. Dahlbom, T. Hansson, P. Höpfl, and M. Knip, “Prevalence of Celiac Disease among Children in Finland,” N. Engl. J. Med., vol. 348, no. 25, pp. 2517–2524, 2003.
[20] V. Hyttinen, J. Kaprio, L. Kinnunen, M. Koskenvuo, and J. Tuomilehto, “Genetic liability of type 1 diabetes and the onset age among 22,650 young Finnish twin pairs: a nationwide follow-up study,” Diabetes, vol. 52, no. 4, pp. 1052–1055, Apr. 2003.
[21] D. Kumar, N. S. Gemayel, D. Deapen, D. Kapadia, P. H. Yamashita, M. Lee, J. H. Dwyer, P. Roy-Burman, G. A. Bray, and T. M. Mack, “North-American twins with IDDM.
95
Genetic, etiological, and clinical significance of disease concordance according to age, zygosity, and the interval after diagnosis in first twin,” Diabetes, vol. 42, no. 9, pp. 1351–1363, Sep. 1993.
[22] S. Srikanta, O. P. Ganda, R. A. Jackson, R. E. Gleason, A. Kaldany, M. R. Garovoy, E. L. Milford, C. B. Carpenter, J. S. Soeldner, and G. S. Eisenbarth, “Type I diabetes mellitus in monozygotic twins: chronic progressive beta cell dysfunction,” Ann. Intern. Med., vol. 99, no. 3, pp. 320–326, Sep. 1983.
[23] A. Guttmann, M. Nakhla, M. Henderson, T. To, D. Daneman, K. Cauch-Dudek, X. Wang, K. Lam, and J. Hux, “Validation of a health administrative data algorithm for assessing the epidemiology of diabetes in Canadian children,” Pediatr. Diabetes, vol. 11, no. 2, pp. 122–128, 2010.
[24] J. A. Staples, A.-L. Ponsonby, L. L.-Y. Lim, and A. J. McMichael, “Ecologic analysis of some immune-related disorders, including type 1 diabetes, in Australia: latitude, regional ultraviolet radiation, and disease prevalence,” Environ. Health Perspect., vol. 111, no. 4, pp. 518–523, Apr. 2003.
[25] R. G. Feltbower, H. J. Bodansky, P. A. McKinney, J. Houghton, C. R. Stephenson, and D. Haigh, “Trends in the incidence of childhood diabetes in south Asians and other children in Bradford, UK,” Diabet. Med. J. Br. Diabet. Assoc., vol. 19, no. 2, pp. 162–166, Feb. 2002.
[26] N. T. Raymond, J. R. Jones, P. G. Swift, M. J. Davies, G. Lawrence, P. G. McNally, M. L. Burden, R. Gregory, A. C. Burden, and J. L. Botha, “Comparative incidence of Type I diabetes in children aged under 15 years from South Asian and White or Other ethnic backgrounds in Leicestershire, UK, 1989 to 1998,” Diabetologia, vol. 44 Suppl 3, pp. B32–36, Oct. 2001.
[27] A. Kondrashova, A. Reunanen, A. Romanov, A. Karvonen, H. Viskari, T. Vesikari, J. Ilonen, M. Knip, and H. Hyöty, “A six-fold gradient in the incidence of type 1 diabetes at the eastern border of Finland,” Ann. Med., vol. 37, no. 1, pp. 67–72, 2005.
[28] A. F. Nakhooda, A. A. Like, C. I. Chappel, C. N. Wei, and E. B. Marliss, “The spontaneously diabetic Wistar rat (the ‘BB’ rat). Studies prior to and during development of the overt syndrome,” Diabetologia, vol. 14, no. 3, pp. 199–207, Mar. 1978.
[29] S. Makino, K. Kunimoto, Y. Muraoka, Y. Mizushima, K. Katagiri, and Y. Tochino, “Breeding of a non-obese, diabetic strain of mice,” Jikken Dobutsu, vol. 29, no. 1, pp. 1–13, Jan. 1980.
[30] P. Pozzilli, A. Signore, A. J. Williams, and P. E. Beales, “NOD mouse colonies around the world--recent facts and figures,” Immunol. Today, vol. 14, no. 5, pp. 193–196, May 1993.
[31] J. A. Todd and L. S. Wicker, “Genetic protection from the inflammatory disease type 1 diabetes in humans and animal models,” Immunity, vol. 15, no. 3, pp. 387–395, Sep. 2001.
96
[32] L. S. Wicker, J. A. Todd, and L. B. Peterson, “Genetic control of autoimmune diabetes in the NOD mouse,” Annu. Rev. Immunol., vol. 13, pp. 179–200, 1995.
[33] D. Miao, L. Yu, and G. S. Eisenbarth, “Role of autoantibodies in type 1 diabetes,” Front. Biosci. J. Virtual Libr., vol. 12, pp. 1889–1898, 2007.
[34] S. Ghosh, S. M. Palmer, N. R. Rodrigues, H. J. Cordell, C. M. Hearne, R. J. Cornall, J.-B. Prins, P. McShane, G. M. Lathrop, L. B. Peterson, L. S. Wicker, and J. A. Todd, “Polygenic control of autoimmune diabetes in nonobese diabetic mice,” Nat. Genet., vol. 4, no. 4, pp. 404–409, Aug. 1993.
[35] M. Prochazka, D. V. Serreze, W. N. Frankel, and E. H. Leiter, “NOR/Lt mice: MHC-matched diabetes-resistant control strain for NOD mice,” Diabetes, vol. 41, no. 1, pp. 98–106, Jan. 1992.
[36] E. A. Ivakine, C. J. Fox, A. D. Paterson, S. M. Mortin-Toth, A. Canty, D. S. Walton, K. Aleksa, S. Ito, and J. S. Danska, “Sex-specific effect of insulin-dependent diabetes 4 on regulation of diabetes pathogenesis in the nonobese diabetic mouse,” J. Immunol. Baltim. Md 1950, vol. 174, no. 11, pp. 7129–7140, Jun. 2005.
[37] C. J. Fox, A. D. Paterson, S. M. Mortin-Toth, and J. S. Danska, “Two genetic loci regulate T cell-dependent islet inflammation and drive autoimmune diabetes pathogenesis,” Am. J. Hum. Genet., vol. 67, no. 1, pp. 67–81, Jul. 2000.
[38] D. V. Serreze, M. Bridgett, H. D. Chapman, E. Chen, S. D. Richard, and E. H. Leiter, “Subcongenic analysis of the Idd13 locus in NOD/Lt mice: evidence for several susceptibility genes including a possible diabetogenic role for beta 2-microglobulin,” J. Immunol. Baltim. Md 1950, vol. 160, no. 3, pp. 1472–1478, Feb. 1998.
[39] D. V. Serreze, M. Prochazka, P. C. Reifsnyder, M. M. Bridgett, and E. H. Leiter, “Use of recombinant congenic and congenic strains of NOD mice to identify a new insulin-dependent diabetes resistance gene,” J. Exp. Med., vol. 180, no. 4, pp. 1553–1558, Oct. 1994.
[40] C. J. Fox and J. S. Danska, “Independent genetic regulation of T-cell and antigen-presenting cell participation in autoimmune islet inflammation,” Diabetes, vol. 47, no. 3, pp. 331–338, Mar. 1998.
[41] J.-F. Bach, “The effect of infections on susceptibility to autoimmune and allergic diseases,” N. Engl. J. Med., vol. 347, no. 12, pp. 911–920, Sep. 2002.
[42] A. Cooke, P. Tonks, F. M. Jones, H. O’Shea, P. Hutchings, A. J. Fulford, and D. W. Dunne, “Infection with Schistosoma mansoni prevents insulin dependent diabetes mellitus in non-obese diabetic mice,” Parasite Immunol., vol. 21, no. 4, pp. 169–176, Apr. 1999.
[43] A. Brás and A. P. Aguas, “Diabetes-prone NOD mice are resistant to Mycobacterium avium and the infection prevents autoimmune disease,” Immunology, vol. 89, no. 1, pp. 20–25, Sep. 1996.
97
[44] K. A. Saunders, T. Raine, A. Cooke, and C. E. Lawrence, “Inhibition of Autoimmune Type 1 Diabetes by Gastrointestinal Helminth Infection,” Infect. Immun., vol. 75, no. 1, pp. 397–407, Jan. 2007.
[45] Q. Liu, K. Sundar, P. K. Mishra, G. Mousavi, Z. Liu, A. Gaydo, F. Alem, D. Lagunoff, D. Bleich, and W. C. Gause, “Helminth Infection Can Reduce Insulitis and Type 1 Diabetes through CD25- and IL-10-Independent Mechanisms,” Infect. Immun., vol. 77, no. 12, pp. 5347–5358, Dec. 2009.
[46] T. Raine, P. Zaccone, P. Mastroeni, and A. Cooke, “Salmonella typhimurium infection in nonobese diabetic mice generates immunomodulatory dendritic cells able to prevent type 1 diabetes,” J. Immunol. Baltim. Md 1950, vol. 177, no. 4, pp. 2224–2233, Aug. 2006.
[47] S. Karumuthil-Melethil, N. Perez, R. Li, and C. Vasu, “Induction of innate immune response through TLR2 and dectin 1 prevents type 1 diabetes,” J. Immunol. Baltim. Md 1950, vol. 181, no. 12, pp. 8323–8334, Dec. 2008.
[48] L. Wen, R. E. Ley, P. V. Volchkov, P. B. Stranges, L. Avanesyan, A. C. Stonebraker, C. Hu, F. S. Wong, G. L. Szot, J. A. Bluestone, J. I. Gordon, and A. V. Chervonsky, “Innate immunity and intestinal microbiota in the development of Type 1 diabetes,” Nature, vol. 455, no. 7216, pp. 1109–1113, Oct. 2008.
[49] F. Calcinaro, S. Dionisi, M. Marinaro, P. Candeloro, V. Bonato, S. Marzotti, R. B. Corneli, E. Ferretti, A. Gulino, F. Grasso, C. De Simone, U. Di Mario, A. Falorni, M. Boirivant, and F. Dotta, “Oral probiotic administration induces interleukin-10 production and prevents spontaneous autoimmune diabetes in the non-obese diabetic mouse,” Diabetologia, vol. 48, no. 8, pp. 1565–1575, Aug. 2005.
[50] J. G. M. Markle, D. N. Frank, S. Mortin-Toth, C. E. Robertson, L. M. Feazel, U. Rolle-Kampczyk, M. von Bergen, K. D. McCoy, A. J. Macpherson, and J. S. Danska, “Sex Differences in the Gut Microbiome Drive Hormone-Dependent Regulation of Autoimmunity,” Science, vol. 339, no. 6123, pp. 1084–1088, Mar. 2013.
[51] H. J. Bodansky, A. Staines, C. Stephenson, D. Haigh, and R. Cartwright, “Evidence for an environmental effect in the aetiology of insulin dependent diabetes in a transmigratory population.,” BMJ, vol. 304, no. 6833, pp. 1020–1022, Apr. 1992.
[52] C. K. W. Lai, R. Beasley, J. Crane, S. Foliaki, J. Shah, and S. Weiland, “Global variation in the prevalence and severity of asthma symptoms: phase three of the International Study of Asthma and Allergies in Childhood (ISAAC),” Thorax, vol. 64, no. 6, pp. 476–483, Jun. 2009.
[53] J. Penders, E. E. Stobberingh, P. A. van den Brandt, and C. Thijs, “The role of the intestinal microbiota in the development of atopic disorders,” Allergy, vol. 62, no. 11, pp. 1223–1236, Nov. 2007.
[54] F. Bäckhed, R. E. Ley, J. L. Sonnenburg, D. A. Peterson, and J. I. Gordon, “Host-Bacterial Mutualism in the Human Intestine,” Science, vol. 307, no. 5717, pp. 1915–1920, Mar. 2005.
98
[55] M. G. Dominguez-Bello, E. K. Costello, M. Contreras, M. Magris, G. Hidalgo, N. Fierer, and R. Knight, “Delivery mode shapes the acquisition and structure of the initial microbiota across multiple body habitats in newborns,” Proc. Natl. Acad. Sci., Jun. 2010.
[56] T. Yatsunenko, F. E. Rey, M. J. Manary, I. Trehan, M. G. Dominguez-Bello, M. Contreras, M. Magris, G. Hidalgo, R. N. Baldassano, A. P. Anokhin, A. C. Heath, B. Warner, J. Reeder, J. Kuczynski, J. G. Caporaso, C. A. Lozupone, C. Lauber, J. C. Clemente, D. Knights, R. Knight, and J. I. Gordon, “Human gut microbiome viewed across age and geography,” Nature, vol. 486, no. 7402, pp. 222–227, Jun. 2012.
[57] R. E. Ley, C. A. Lozupone, M. Hamady, R. Knight, and J. I. Gordon, “Worlds within worlds: evolution of the vertebrate gut microbiota,” Nat. Rev. Microbiol., vol. 6, no. 10, pp. 776–788, Oct. 2008.
[58] J. L. Round and S. K. Mazmanian, “The gut microbiota shapes intestinal immune responses during health and disease,” Nat. Rev. Immunol., vol. 9, no. 5, pp. 313–323, May 2009.
[59] E. B. Troy and D. L. Kasper, “Beneficial effects of Bacteroides fragilis polysaccharides on the immune system,” Front. Biosci. J. Virtual Libr., vol. 15, pp. 25–34, Jan. 2010.
[60] I. I. Ivanov, K. Atarashi, N. Manel, E. L. Brodie, T. Shima, U. Karaoz, D. Wei, K. C. Goldfarb, C. A. Santee, S. V. Lynch, T. Tanoue, A. Imaoka, K. Itoh, K. Takeda, Y. Umesaki, K. Honda, and D. R. Littman, “Induction of intestinal Th17 cells by segmented filamentous bacteria,” Cell, vol. 139, no. 3, pp. 485–498, Oct. 2009.
[61] K. Atarashi, T. Tanoue, T. Shima, A. Imaoka, T. Kuwahara, Y. Momose, G. Cheng, S. Yamasaki, T. Saito, Y. Ohba, T. Taniguchi, K. Takeda, S. Hori, I. I. Ivanov, Y. Umesaki, K. Itoh, and K. Honda, “Induction of Colonic Regulatory T Cells by Indigenous Clostridium Species,” Science, vol. 331, no. 6015, pp. 337–341, Jan. 2011.
[62] M. A. Kriegel, E. Sefik, J. A. Hill, H.-J. Wu, C. Benoist, and D. Mathis, “Naturally transmitted segmented filamentous bacteria segregate with diabetes protection in nonobese diabetic mice,” Proc. Natl. Acad. Sci., vol. 108, no. 28, pp. 11548–11553, Jul. 2011.
[63] H.-J. Wu, I. I. Ivanov, J. Darce, K. Hattori, T. Shima, Y. Umesaki, D. R. Littman, C. Benoist, and D. Mathis, “Gut-Residing Segmented Filamentous Bacteria Drive Autoimmune Arthritis via T Helper 17 Cells,” Immunity, vol. 32, no. 6, pp. 815–827, Jun. 2010.
[64] H. Sokol, B. Pigneur, L. Watterlot, O. Lakhdari, L. G. Bermúdez-Humarán, J.-J. Gratadoux, S. Blugeon, C. Bridonneau, J.-P. Furet, G. Corthier, C. Grangette, N. Vasquez, P. Pochart, G. Trugnan, G. Thomas, H. M. Blottière, J. Doré, P. Marteau, P. Seksik, and P. Langella, “Faecalibacterium prausnitzii is an anti-inflammatory commensal bacterium identified by gut microbiota analysis of Crohn disease patients,” Proc. Natl. Acad. Sci., vol. 105, no. 43, pp. 16731–16736, Oct. 2008.
99
[65] C. R. Woese, O. Kandler, and M. L. Wheelis, “Towards a natural system of organisms: proposal for the domains Archaea, Bacteria, and Eucarya,” Proc. Natl. Acad. Sci. U. S. A., vol. 87, no. 12, pp. 4576–4579, Jun. 1990.
[66] J.-M. Neefs, Y. Van De Peer, L. Hendriks, and R. De Wachter, “Compilation of small ribosomal subunit RNA sequences,” Nucleic Acids Res., vol. 18, no. Suppl, pp. 2237–2317, Apr. 1990.
[67] A. Chao, “Nonparametric Estimation of the Number of Classes in a Population,” Scand. J. Stat., vol. 11, no. 4, pp. 265–270, Jan. 1984.
[68] I. F. Spellerberg and P. J. Fedor, “A tribute to Claude Shannon (1916–2001) and a plea for more rigorous use of species richness, species diversity and the ‘Shannon–Wiener’ Index,” Glob. Ecol. Biogeogr., vol. 12, no. 3, pp. 177–179, 2003.
[69] C. E. Shannon, “A Mathematical Theory of Communication,” SIGMOBILE Mob Comput Commun Rev, vol. 5, no. 1, pp. 3–55, Jan. 2001.
[70] P. M. Sherman, K. C. Johnson-Henry, H. P. Yeung, P. S. C. Ngo, J. Goulet, and T. A. Tompkins, “Probiotics Reduce Enterohemorrhagic Escherichia coli O157:H7- and Enteropathogenic E. coli O127:H6-Induced Changes in Polarized T84 Epithelial Cell Monolayers by Reducing Bacterial Adhesion and Cytoskeletal Rearrangements,” Infect. Immun., vol. 73, no. 8, pp. 5183–5188, Aug. 2005.
[71] N. Haskey and W. J. Dahl, “Synbiotic Therapy Improves Quality of Life and Reduces Symptoms in Pediatric Ulcerative Colitis,” ICAN Infant Child Adolesc. Nutr., vol. 1, no. 2, pp. 88–93, Apr. 2009.
[72] J. R. Cole, B. Chai, R. J. Farris, Q. Wang, S. A. Kulam, D. M. McGarrell, G. M. Garrity, and J. M. Tiedje, “The Ribosomal Database Project (RDP-II): sequences and tools for high-throughput rRNA analysis,” Nucleic Acids Res., vol. 33, no. Database issue, pp. D294–D296, Jan. 2005.
[73] K. E. Ashelford, A. J. Weightman, and J. C. Fry, “PRIMROSE: a computer program for generating and estimating the phylogenetic range of 16S rRNA oligonucleotide probes and primers in conjunction with the RDP-II database,” Nucleic Acids Res., vol. 30, no. 15, pp. 3481–3489, Aug. 2002.
[74] K. Tamura, J. Dudley, M. Nei, and S. Kumar, “MEGA4: Molecular Evolutionary Genetics Analysis (MEGA) Software Version 4.0,” Mol. Biol. Evol., vol. 24, no. 8, pp. 1596–1599, Aug. 2007.
[75] Q. Wang, G. M. Garrity, J. M. Tiedje, and J. R. Cole, “Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy,” Appl. Environ. Microbiol., vol. 73, no. 16, pp. 5261–5267, Aug. 2007.
[76] M. J. Claesson, O. O’Sullivan, Q. Wang, J. Nikkilä, J. R. Marchesi, H. Smidt, W. M. de Vos, R. P. Ross, and P. W. O’Toole, “Comparative Analysis of Pyrosequencing and a
100
Phylogenetic Microarray for Exploring Microbial Community Structures in the Human Distal Intestine,” PLoS ONE, vol. 4, no. 8, p. e6669, Aug. 2009.
[77] D. N. Frank, “BARCRAWL and BARTAB: software tools for the design and implementation of barcoded primers for highly multiplexed DNA sequencing,” BMC Bioinformatics, vol. 10, p. 362, Oct. 2009.
[78] J. A. Frank, C. I. Reich, S. Sharma, J. S. Weisbaum, B. A. Wilson, and G. J. Olsen, “Critical Evaluation of Two Primers Commonly Used for Amplification of Bacterial 16S rRNA Genes,” Appl. Environ. Microbiol., vol. 74, no. 8, pp. 2461–2470, Apr. 2008.
[79] J. G. Caporaso, C. L. Lauber, W. A. Walters, D. Berg-Lyons, C. A. Lozupone, P. J. Turnbaugh, N. Fierer, and R. Knight, “Global patterns of 16S rRNA diversity at a depth of millions of sequences per sample,” Proc. Natl. Acad. Sci., p. 201000080, Jun. 2010.
[80] D. N. Frank, L. M. Feazel, M. T. Bessesen, C. S. Price, E. N. Janoff, and N. R. Pace, “The Human Nasal Microbiota and Staphylococcus aureus Carriage,” PLoS ONE, vol. 5, no. 5, p. e10598, May 2010.
[81] E. P. Nawrocki, D. L. Kolbe, and S. R. Eddy, “Infernal 1.0: inference of RNA alignments,” Bioinformatics, vol. 25, no. 10, pp. 1335–1337, May 2009.
[82] B. J. Haas, D. Gevers, A. M. Earl, M. Feldgarden, D. V. Ward, G. Giannoukos, D. Ciulla, D. Tabbaa, S. K. Highlander, E. Sodergren, B. Methé, T. Z. DeSantis, J. F. Petrosino, R. Knight, and B. W. Birren, “Chimeric 16S rRNA sequence formation and detection in Sanger and 454-pyrosequenced PCR amplicons,” Genome Res., vol. 21, no. 3, pp. 494–504, Mar. 2011.
[83] E. Li, C. M. Hamm, A. S. Gulati, R. B. Sartor, H. Chen, X. Wu, T. Zhang, F. J. Rohlf, W. Zhu, C. Gu, C. E. Robertson, N. R. Pace, E. C. Boedeker, N. Harpaz, J. Yuan, G. M. Weinstock, E. Sodergren, and D. N. Frank, “Inflammatory Bowel Diseases Phenotype, C. difficile and NOD2 Genotype Are Associated with Shifts in Human Ileum Associated Microbial Composition,” PLoS ONE, vol. 7, no. 6, p. e26284, Jun. 2012.
[84] W. Ludwig, O. Strunk, R. Westram, L. Richter, H. Meier, Yadhukumar, A. Buchner, T. Lai, S. Steppi, G. Jobb, W. Förster, I. Brettske, S. Gerber, A. W. Ginhart, O. Gross, S. Grumann, S. Hermann, R. Jost, A. König, T. Liss, R. Lüßmann, M. May, B. Nonhoff, B. Reichel, R. Strehlow, A. Stamatakis, N. Stuckmann, A. Vilbig, M. Lenke, T. Ludwig, A. Bode, and K.-H. Schleifer, “ARB: a software environment for sequence data,” Nucleic Acids Res., vol. 32, no. 4, pp. 1363–1371, Mar. 2004.
[85] E. Pruesse, J. Peplies, and F. O. Glockner, “SINA: Accurate high-throughput multiple sequence alignment of ribosomal RNA genes,” Bioinformatics, vol. 28, no. 14, pp. 1823–1829, Jul. 2012.
[86] E. Pruesse, C. Quast, K. Knittel, B. M. Fuchs, W. Ludwig, J. Peplies, and F. O. Glöckner, “SILVA: a comprehensive online resource for quality checked and aligned ribosomal RNA sequence data compatible with ARB,” Nucleic Acids Res., vol. 35, no. 21, pp. 7188–7196, Dec. 2007.
101
[87] B. Ewing, L. Hillier, M. C. Wendl, and P. Green, “Base-Calling of Automated Sequencer Traces UsingPhred. I. Accuracy Assessment,” Genome Res., vol. 8, no. 3, pp. 175–185, Mar. 1998.
[88] B. Ewing and P. Green, “Base-Calling of Automated Sequencer Traces Using Phred. II. Error Probabilities,” Genome Res., vol. 8, no. 3, pp. 186–194, Mar. 1998.
[89] D. Gordon, C. Abajian, and P. Green, “Consed: A Graphical Tool for Sequence Finishing,” Genome Res., vol. 8, no. 3, pp. 195–202, Mar. 1998.
[90] R. C. Edgar, B. J. Haas, J. C. Clemente, C. Quince, and R. Knight, “UCHIME improves sensitivity and speed of chimera detection,” Bioinformatics, vol. 27, no. 16, pp. 2194–2200, Aug. 2011.
[91] P. D. Schloss, D. Gevers, and S. L. Westcott, “Reducing the Effects of PCR Amplification and Sequencing Artifacts on 16S rRNA-Based Studies,” PLoS ONE, vol. 6, no. 12, p. e27310, Dec. 2011.
[92] H. S. Horn, “Measurement of ‘Overlap’ in Comparative Ecological Studies,” Am. Nat., vol. 100, no. 914, pp. 419–424, Sep. 1966.
[93] C. E. Robertson, J. K. Harris, B. D. Wagner, D. Granger, K. Browne, B. Tatem, L. M. Feazel, K. Park, N. R. Pace, and D. N. Frank, “Explicet: graphical user interface software for metadata-driven management, analysis and visualization of microbiome data,” Bioinformatics, vol. 29, no. 23, pp. 3100–3101, Dec. 2013.
[94] G. Bouxin, “Ginkgo, a multivariate analysis package,” J. Veg. Sci., vol. 16, no. 3, pp. 355–359, 2005.
[95] J. R. Bray and J. T. Curtis, “An Ordination of the Upland Forest Communities of Southern Wisconsin,” Ecol. Monogr., vol. 27, no. 4, pp. 325–349, Oct. 1957.
[96] G. Barera, R. Bonfanti, M. Viscardi, E. Bazzigaluppi, G. Calori, F. Meschi, C. Bianchi, and G. Chiumello, “Occurrence of Celiac Disease After Onset of Type 1 Diabetes: A 6-Year Prospective Longitudinal Study,” Pediatrics, vol. 109, no. 5, pp. 833–838, May 2002.
[97] E. Wine, M. G. Gareau, K. Johnson-Henry, and P. M. Sherman, “Strain-specific probiotic (Lactobacillus helveticus) inhibition of Campylobacter jejuni invasion of human intestinal epithelial cells,” FEMS Microbiol. Lett., vol. 300, no. 1, pp. 146–152, 2009.
[98] C. De Champs, N. Maroncle, D. Balestrino, C. Rich, and C. Forestier, “Persistence of Colonization of Intestinal Mucosa by a Probiotic Strain, Lactobacillus casei subsp. rhamnosus Lcr35, after Oral Consumption,” J. Clin. Microbiol., vol. 41, no. 3, pp. 1270–1273, Mar. 2003.
[99] L. V. McFarland and P. Bernasconi, “Saccharomyces boulardii’. A Review of an Innovative Biotherapeutic Agent,” 11-Jul-2009. [Online]. Available:
102
http://informahealthcare.com/doi/abs/10.3109/08910609309141323. [Accessed: 12-Nov-2013].
[100] D. M. Rodrigues, A. J. Sousa, K. C. Johnson-Henry, P. M. Sherman, and M. G. Gareau, “Probiotics Are Effective for the Prevention and Treatment of Citrobacter rodentium–Induced Colitis in Mice,” J. Infect. Dis., vol. 206, no. 1, pp. 99–109, Jul. 2012.
[101] R. Schwarz, A. Kaspar, J. Seelig, and B. Künnecke, “Gastrointestinal transit times in mice and humans measured with 27Al and 19F nuclear magnetic resonance,” Magn. Reson. Med., vol. 48, no. 2, pp. 255–261, 2002.
[102] P. Padmanabhan, J. Grosse, A. B. M. A. Asad, G. K. Radda, and X. Golay, “Gastrointestinal transit measurements in mice with 99mTc-DTPA-labeled activated charcoal using NanoSPECT-CT,” EJNMMI Res., vol. 3, no. 1, p. 60, Aug. 2013.
[103] K. C. Mountzouris, P. Tsirtsikos, E. Kalamara, S. Nitsch, G. Schatzmayr, and K. Fegeros, “Evaluation of the Efficacy of a Probiotic Containing Lactobacillus, Bifidobacterium, Enterococcus, and Pediococcus Strains in Promoting Broiler Performance and Modulating Cecal Microflora Composition and Metabolic Activities,” Poult. Sci., vol. 86, no. 2, pp. 309–317, Feb. 2007.
[104] K. Kajander, E. Myllyluoma, M. Rajilić-Stojanović, S. Kyrönpalo, M. Rasmussen, S. Järvenpää, E. G. Zoetendal, W. M. De Vos, H. Vapaatalo, and R. Korpela, “Clinical trial: multispecies probiotic supplementation alleviates the symptoms of irritable bowel syndrome and stabilizes intestinal microbiota,” Aliment. Pharmacol. Ther., vol. 27, no. 1, pp. 48–57, 2008.
[105] M. Tong, X. Li, L. Wegener Parfrey, B. Roth, A. Ippoliti, B. Wei, J. Borneman, D. P. B. McGovern, D. N. Frank, E. Li, S. Horvath, R. Knight, and J. Braun, “A Modular Organization of the Human Intestinal Mucosal Microbiota and Its Association with Inflammatory Bowel Disease,” PLoS ONE, vol. 8, no. 11, p. e80702, Nov. 2013.
[106] J. Dicksved, J. Halfvarson, M. Rosenquist, G. Järnerot, C. Tysk, J. Apajalahti, L. Engstrand, and J. K. Jansson, “Molecular analysis of the gut microbiota of identical twins with Crohn’s disease,” ISME J., vol. 2, no. 7, pp. 716–727, Apr. 2008.
[107] P. Lepage, R. Häsler, M. E. Spehlmann, A. Rehman, A. Zvirbliene, A. Begun, S. Ott, L. Kupcinskas, J. Doré, A. Raedler, and S. Schreiber, “Twin Study Indicates Loss of Interaction Between Microbiota and Mucosa of Patients With Ulcerative Colitis,” Gastroenterology, vol. 141, no. 1, pp. 227–236, Jul. 2011.
[108] A. W. Walker, J. D. Sanderson, C. Churcher, G. C. Parkes, B. N. Hudspith, N. Rayment, J. Brostoff, J. Parkhill, G. Dougan, and L. Petrovska, “High-throughput clone library analysis of the mucosa-associated microbiota reveals dysbiosis and differences between inflamed and non-inflamed regions of the intestine in inflammatory bowel disease,” BMC Microbiol., vol. 11, no. 1, p. 7, Jan. 2011.
[109] P. S. Kumar, “Sex and the subgingival microbiome: Do female sex steroids affect periodontal bacteria?,” Periodontol. 2000, vol. 61, no. 1, pp. 103–124, 2013.
103
[110] J. Oh, S. Conlan, E. C. Polley, J. A. Segre, and H. H. Kong, “Shifts in human skin and nares microbiota of healthy children and adults,” Genome Med., vol. 4, no. 10, p. 77, Oct. 2012.
[111] T. Yamamoto, X. Zhou, C. J. Williams, A. Hochwalt, and L. J. Forney, “Bacterial Populations in the Vaginas of Healthy Adolescent Women,” J. Pediatr. Adolesc. Gynecol., vol. 22, no. 1, pp. 11–18, Feb. 2009.
[112] S. S. Stahringer, J. C. Clemente, R. P. Corley, J. Hewitt, D. Knights, W. A. Walters, R. Knight, and K. S. Krauter, “Nurture trumps nature in a longitudinal survey of salivary bacterial communities in twins from early adolescence to early adulthood,” Genome Res., vol. 22, no. 11, pp. 2146–2152, Nov. 2012.
[113] R. Agans, L. Rigsbee, H. Kenche, S. Michail, H. J. Khamis, and O. Paliy, “Distal gut microbiota of adolescent children is different from that of adults,” FEMS Microbiol. Ecol., vol. 77, no. 2, pp. 404–412, 2011.
[114] J. E. Koenig, A. Spor, N. Scalfone, A. D. Fricker, J. Stombaugh, R. Knight, L. T. Angenent, and R. E. Ley, “Succession of microbial consortia in the developing infant gut microbiome,” Proc. Natl. Acad. Sci., p. 201000081, Jul. 2010.
[115] F. Hildebrand, T. L. A. Nguyen, B. Brinkman, R. G. Yunta, B. Cauwe, P. Vandenabeele, A. Liston, and J. Raes, “Inflammation-associated enterotypes, host genotype, cage and inter-individual effects drive gut microbiota variation in common laboratory mice,” Genome Biol., vol. 14, no. 1, p. R4, Jan. 2013.
[116] A. K. Benson, S. A. Kelly, R. Legge, F. Ma, S. J. Low, J. Kim, M. Zhang, P. L. Oh, D. Nehrenberg, K. Hua, S. D. Kachman, E. N. Moriyama, J. Walter, D. A. Peterson, and D. Pomp, “Individuality in gut microbiota composition is a complex polygenic trait shaped by multiple environmental and host genetic factors,” Proc. Natl. Acad. Sci., vol. 107, no. 44, pp. 18933–18938, Nov. 2010.
[117] A. D. Alexander, R. P. Orcutt, J. C. Henry, J. B. Jr, A. C. Bissahoyo, and D. W. Threadgill, “Quantitative PCR assays for mouse enteric flora reveal strain-dependent differences in composition that are influenced by the microenvironment,” Mamm. Genome, vol. 17, no. 11, pp. 1093–1104, Nov. 2006.
[118] C. Ubeda, L. Lipuma, A. Gobourne, A. Viale, I. Leiner, M. Equinda, R. Khanin, and E. G. Pamer, “Familial transmission rather than defective innate immunity shapes the distinct intestinal microbiota of TLR-deficient mice,” J. Exp. Med., vol. 209, no. 8, pp. 1445–1456, Jul. 2012.
[119] E. Elinav, T. Strowig, A. L. Kau, J. Henao-Mejia, C. A. Thaiss, C. J. Booth, D. R. Peaper, J. Bertin, S. C. Eisenbarth, J. I. Gordon, and R. A. Flavell, “NLRP6 Inflammasome Regulates Colonic Microbial Ecology and Risk for Colitis,” Cell, vol. 145, no. 5, pp. 745–757, May 2011.
104
[120] M. El-Salhy, “Gastrointestinal transit in nonobese diabetic mouse: An animal model of human diabetes Type 1,” J. Diabetes Complications, vol. 15, no. 5, pp. 277–284, Sep. 2001.
[121] L. M. Foster, T. A. Tompkins, and W. J. Dahl, “A comprehensive post-market review of studies on a probiotic product containing Lactobacillus helveticus R0052 and Lactobacillus rhamnosus R0011,” Benef. Microbes, vol. 2, no. 4, pp. 319–334, Dec. 2011.
[122] H. L. Greetham, G. R. Gibson, C. Giffard, H. Hippe, B. Merkhoffer, U. Steiner, E. Falsen, and M. D. Collins, “Allobaculum stercoricanis gen. nov., sp. nov., isolated from canine feces,” Anaerobe, vol. 10, no. 5, pp. 301–307, Oct. 2004.
[123] A. Gomez, D. Luckey, C. J. Yeoman, E. V. Marietta, M. E. Berg Miller, J. A. Murray, B. A. White, and V. Taneja, “Loss of Sex and Age Driven Differences in the Gut Microbiome Characterize Arthritis-Susceptible *0401 Mice but Not Arthritis-Resistant *0402 Mice,” PLoS ONE, vol. 7, no. 4, p. e36095, Apr. 2012.
[124] Y. Ravussin, O. Koren, A. Spor, C. LeDuc, R. Gutman, J. Stombaugh, R. Knight, R. E. Ley, and R. L. Leibel, “Responses of Gut Microbiota to Diet Composition and Weight Loss in Lean and Obese Mice,” Obesity, vol. 20, no. 4, pp. 738–747, 2012.
[125] L. Yurkovetskiy, M. Burrows, A. A. Khan, L. Graham, P. Volchkov, L. Becker, D. Antonopoulos, Y. Umesaki, and A. V. Chervonsky, “Gender Bias in Autoimmunity Is Influenced by Microbiota,” Immunity, vol. 39, no. 2, pp. 400–412, Aug. 2013.
[126] M. H. Sofi, R. Gudi, S. Karumuthil-Melethil, N. Perez, B. M. Johnson, and C. Vasu, “pH of Drinking Water Influences the Composition of Gut Microbiome and Type 1 Diabetes Incidence,” Diabetes, vol. 63, no. 2, pp. 632–644, Feb. 2014.
[127] K. J. Wolf, J. G. Daft, S. M. Tanner, R. Hartmann, E. Khafipour, and R. G. Lorenz, “Consumption of Acidic Water Alters the Gut Microbiome and Decreases the Risk of Diabetes in NOD Mice,” J. Histochem. Cytochem. Off. J. Histochem. Soc., Jan. 2014.
[128] C. T. Brown, A. G. Davis-Richardson, A. Giongo, K. A. Gano, D. B. Crabb, N. Mukherjee, G. Casella, J. C. Drew, J. Ilonen, M. Knip, H. Hyöty, R. Veijola, T. Simell, O. Simell, J. Neu, C. H. Wasserfall, D. Schatz, M. A. Atkinson, and E. W. Triplett, “Gut Microbiome Metagenomics Analysis Suggests a Functional Model for the Development of Autoimmunity for Type 1 Diabetes,” PLoS ONE, vol. 6, no. 10, p. e25792, Oct. 2011.
[129] M. C. de Goffau, K. Luopajärvi, M. Knip, J. Ilonen, T. Ruohtula, T. Härkönen, L. Orivuori, S. Hakala, G. W. Welling, H. J. Harmsen, and O. Vaarala, “Fecal Microbiota Composition Differs Between Children With β-Cell Autoimmunity and Those Without,” Diabetes, vol. 62, no. 4, pp. 1238–1244, Apr. 2013.
[130] A. Tada, N. Komiya, A. Shimada, T. Katsuki, Y. Oikawa, and H. Itoh, “Eradication of Helicobactor pylori May Trigger Onset of Type 1 Diabetes,” Ann. N. Y. Acad. Sci., vol. 1150, no. 1, pp. 208–209, Dec. 2008.
105
[131] D. A. de Luis, H. de la Calle, G. Roy, C. Martin de Argila, S. Valdezate, R. Canton, and D. Boixeda, “Helicobacter pylori infection and insulin-dependent diabetes mellitus,” Diabetes Res. Clin. Pract., vol. 39, no. 2, pp. 143–146, Feb. 1998.
[132] M. M. El-Eshmawy, A. K. El-Hawary, S. S. Abdel Gawad, and A. A. El-Baiomy, “Helicobacter pylori infection might be responsible for the interconnection between type 1 diabetes and autoimmune thyroiditis,” Diabetol. Metab. Syndr., vol. 3, p. 28, Oct. 2011.
[133] D. J. Smyth, J. D. Cooper, J. M. M. Howson, P. Clarke, K. Downes, T. Mistry, H. Stevens, N. M. Walker, and J. A. Todd, “FUT2 Nonsecretor Status Links Type 1 Diabetes Susceptibility and Resistance to Infection,” Diabetes, vol. 60, no. 11, pp. 3081–3084, Nov. 2011.
[134] M. E. Mejía-León, J. F. Petrosino, N. J. Ajami, M. G. Domínguez-Bello, and A. M. C. de la Barca, “Fecal microbiota imbalance in Mexican children with type 1 diabetes,” Sci. Rep., vol. 4, Jan. 2014.
[135] M. Murri, I. Leiva, J. M. Gomez-Zumaquero, F. J. Tinahones, F. Cardona, F. Soriguer, and M. I. Queipo-Ortuño, “Gut microbiota in children with type 1 diabetes differs from that in healthy children: a case-control study,” BMC Med., vol. 11, no. 1, pp. 1–12, Dec. 2013.
[136] M. M. Lamb, M. Miller, J. A. Seifert, B. Frederiksen, M. Kroehl, M. Rewers, and J. M. Norris, “The effect of childhood cow’s milk intake and HLA-DR genotype on risk of islet autoimmunity and type 1 diabetes: The Diabetes Autoimmunity Study in the Young,” Pediatr. Diabetes, p. n/a–n/a, Jan. 2014.
[137] L. Dethlefsen, P. B. Eckburg, E. M. Bik, and D. A. Relman, “Assembly of the human intestinal microbiota,” Trends Ecol. Evol., vol. 21, no. 9, pp. 517–523, Sep. 2006.
[138] J. h. Cummings and G. t. Macfarlane, “The control and consequences of bacterial fermentation in the human colon,” J. Appl. Bacteriol., vol. 70, no. 6, pp. 443–459, 1991.
[139] T. L. Miller and M. J. Wolin, “Pathways of acetate, propionate, and butyrate formation by the human fecal microbial flora.,” Appl. Environ. Microbiol., vol. 62, no. 5, pp. 1589–1592, May 1996.
[140] A. Barcenilla, S. E. Pryde, J. C. Martin, S. H. Duncan, C. S. Stewart, C. Henderson, and H. J. Flint, “Phylogenetic Relationships of Butyrate-Producing Bacteria from the Human Gut,” Appl. Environ. Microbiol., vol. 66, no. 4, pp. 1654–1661, Apr. 2000.
[141] P. Louis and H. J. Flint, “Diversity, metabolism and microbial ecology of butyrate-producing bacteria from the human large intestine,” FEMS Microbiol. Lett., vol. 294, no. 1, pp. 1–8, 2009.
[142] N. Reichardt, S. H. Duncan, P. Young, A. Belenguer, C. McWilliam Leitch, K. P. Scott, H. J. Flint, and P. Louis, “Phylogenetic distribution of three pathways for propionate production within the human gut microbiota,” ISME J., Feb. 2014.
106
[143] A. J. Brown, S. M. Goldsworthy, A. A. Barnes, M. M. Eilert, L. Tcheang, D. Daniels, A. I. Muir, M. J. Wigglesworth, I. Kinghorn, N. J. Fraser, N. B. Pike, J. C. Strum, K. M. Steplewski, P. R. Murdock, J. C. Holder, F. H. Marshall, P. G. Szekeres, S. Wilson, D. M. Ignar, S. M. Foord, A. Wise, and S. J. Dowell, “The Orphan G Protein-coupled Receptors GPR41 and GPR43 Are Activated by Propionate and Other Short Chain Carboxylic Acids,” J. Biol. Chem., vol. 278, no. 13, pp. 11312–11319, Mar. 2003.
[144] E. L. Poul, C. Loison, S. Struyf, J.-Y. Springael, V. Lannoy, M.-E. Decobecq, S. Brezillon, V. Dupriez, G. Vassart, J. V. Damme, M. Parmentier, and M. Detheux, “Functional Characterization of Human Receptors for Short Chain Fatty Acids and Their Role in Polymorphonuclear Cell Activation,” J. Biol. Chem., vol. 278, no. 28, pp. 25481–25489, Jul. 2003.
[145] M. Thangaraju, G. A. Cresci, K. Liu, S. Ananth, J. P. Gnanaprakasam, D. D. Browning, J. D. Mellinger, S. B. Smith, G. J. Digby, N. A. Lambert, P. D. Prasad, and V. Ganapathy, “GPR109A Is a G-protein–Coupled Receptor for the Bacterial Fermentation Product Butyrate and Functions as a Tumor Suppressor in Colon,” Cancer Res., vol. 69, no. 7, pp. 2826–2832, Apr. 2009.
[146] P. M. Smith, M. R. Howitt, N. Panikov, M. Michaud, C. A. Gallini, M. Bohlooly-Y, J. N. Glickman, and W. S. Garrett, “The Microbial Metabolites, Short-Chain Fatty Acids, Regulate Colonic Treg Cell Homeostasis,” Science, vol. 341, no. 6145, pp. 569–573, Aug. 2013.
[147] K. Atarashi, T. Tanoue, K. Oshima, W. Suda, Y. Nagano, H. Nishikawa, S. Fukuda, T. Saito, S. Narushima, K. Hase, S. Kim, J. V. Fritz, P. Wilmes, S. Ueha, K. Matsushima, H. Ohno, B. Olle, S. Sakaguchi, T. Taniguchi, H. Morita, M. Hattori, and K. Honda, “Treg induction by a rationally selected mixture of Clostridia strains from the human microbiota,” Nature, vol. 500, no. 7461, pp. 232–236, Aug. 2013.
[148] M. Waldecker, T. Kautenburger, H. Daumann, C. Busch, and D. Schrenk, “Inhibition of histone-deacetylase activity by short-chain fatty acids and some polyphenol metabolites formed in the colon,” J. Nutr. Biochem., vol. 19, no. 9, pp. 587–593, Sep. 2008.
[149] P. V. Chang, L. Hao, S. Offermanns, and R. Medzhitov, “The microbial metabolite butyrate regulates intestinal macrophage function via histone deacetylase inhibition,” Proc. Natl. Acad. Sci., p. 201322269, Jan. 2014.
[150] N. Singh, M. Thangaraju, P. D. Prasad, P. M. Martin, N. A. Lambert, T. Boettger, S. Offermanns, and V. Ganapathy, “Blockade of Dendritic Cell Development by Bacterial Fermentation Products Butyrate and Propionate through a Transporter (Slc5a8)-dependent Inhibition of Histone Deacetylases,” J. Biol. Chem., vol. 285, no. 36, pp. 27601–27608, Sep. 2010.
[151] C. R. Cavaglieri, A. Nishiyama, L. C. Fernandes, R. Curi, E. A. Miles, and P. C. Calder, “Differential effects of short-chain fatty acids on proliferation and production of pro- and anti-inflammatory cytokines by cultured lymphocytes,” Life Sci., vol. 73, no. 13, pp. 1683–1690, Aug. 2003.
107
[152] E. Hosseini, C. Grootaert, W. Verstraete, and T. Van de Wiele, “Propionate as a health-promoting microbial metabolite in the human gut,” Nutr. Rev., vol. 69, no. 5, pp. 245–258, May 2011.
[153] G. D’Haens, M. Ferrante, S. Vermeire, F. Baert, M. Noman, L. Moortgat, P. Geens, D. Iwens, I. Aerden, G. Van Assche, G. Van Olmen, and P. Rutgeerts, “Fecal calprotectin is a surrogate marker for endoscopic lesions in inflammatory bowel disease,” Inflamm. Bowel Dis., vol. 18, no. 12, pp. 2218–2224, Dec. 2012.
[154] R. J. Shulman, M. N. Eakin, D. I. Czyzewski, M. Jarrett, and C.-N. Ou, “Increased Gastrointestinal Permeability and Gut Inflammation in Children with Functional Abdominal Pain and Irritable Bowel Syndrome,” J. Pediatr., vol. 153, no. 5, pp. 646–650, Nov. 2008.
[155] E. V. Marietta, A. M. Gomez, C. Yeoman, A. Y. Tilahun, C. R. Clark, D. H. Luckey, J. A. Murray, B. A. White, Y. C. Kudva, and G. Rajagopalan, “Low Incidence of Spontaneous Type 1 Diabetes in Non-Obese Diabetic Mice Raised on Gluten-Free Diets Is Associated with Changes in the Intestinal Microbiome,” PLoS ONE, vol. 8, no. 11, p. e78687, Nov. 2013.
[156] A. Giongo, K. A. Gano, D. B. Crabb, N. Mukherjee, L. L. Novelo, G. Casella, J. C. Drew, J. Ilonen, M. Knip, H. Hyöty, R. Veijola, T. Simell, O. Simell, J. Neu, C. H. Wasserfall, D. Schatz, M. A. Atkinson, and E. W. Triplett, “Toward defining the autoimmune microbiome for type 1 diabetes,” ISME J., vol. 5, no. 1, pp. 82–91, Jan. 2011.