EpigeneticProfilesRevealThatADCYAP1ServesasKey...

Research ArticleEpigenetic Profiles Reveal That ADCYAP1 Serves as KeyMolecule in Gestational Diabetes Mellitus

Xue Li1 Wenhong Yang2 and Yanning Fang 1

1Department of Obstetrics First Peoplersquos Hospital of Jining Jining Shandong 272000 China2Department of Nursing First Peoplersquos Hospital of Jining Jining Shandong 272000 China

Correspondence should be addressed to Yanning Fang fangyanninggmailcom

Received 12 December 2018 Revised 21 January 2019 Accepted 24 July 2019 Published 14 August 2019

Academic Editor Esmaeil Ebrahimie

Copyright copy 2019 Xue Li et al is is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Gestational diabetes mellitus (GDM) refers to the condition which shows abnormal glucose metabolism that occurs duringpregnancy while normal glucose metabolism before pregnancy In the present study a novel analytical procedure was used toexplore the key molecule of gestational diabetes mellitus First the weighted pathway model was carried out subsequently toeliminate the gene-overlapping effects among pathways Second we assessed the enriched pathways by a combination of Fisherrsquost-test and the MannndashWhitney U test We carried out the functional principal component analysis by estimating F values of genesto identify the hub genes in the enriched pathways Results showed that a total of 4 differential pathways were enriched e keypathway was considered as the insulin secretion pathway F values of each gene in the key pathway were calculated ree hubmolecules were identified as hub differentially methylated genes namely CAMK2B ADCYAP1 and KCNN2 In addition byfurther comparing the gene expression data in a validation cohort one key molecule was obtained ADCYAP1 ereforeADCYAP1 may serve as a potential target for the treatment of GDM

1 Introduction

Gestational diabetes mellitus (GDM) is the most commonmedical complication of pregnancy characterized by glucoseintolerance which did not occur before pregnancy but be-comes clinically apparent in the late stage of fetation [1]Glucose metabolism in most GDM patients returns tonormal after delivery but there is an increased chance ofdeveloping type 2 diabetes metabolic syndrome and ce-rebrovascular disease in the future [2] High blood sugar inthe womb environment is also very unfavorable to the fetusincreasing the risk of macrosomia and premature deliveryand abortion rates fetal growth restriction (FGR) confer-ring a predisposition for obesity neonatal respiratory dis-tress syndrome cardiovascular complications and neonatalhypoglycemia [1]e complication caused by GDM is morecomplicated and harmful to both mothers and their child

In view of the serious consequences of GDM more andmore researchers are committed to the study of the dynamicchanges in pathogenesis andmolecular mechanism of GDM

e occurrence of GDM is related to the formation of insulinresistance in patients It has been reported that the placentasecretes large amounts of anti-insulin hormones duringpregnancy [3] Inflammatory factors such as C-reactiveprotein tumor necrosis factor serum proteins interleukinand other chronic inflammatory reactions were involved inthe occurrence of GDM [4 5] Analysis of gene methylationdata could help us to reveal the relationship betweenmethylation and disease [6] Epigenome analyses showedthat the occurrence of GDM is closely related to changes ingene methylation and expression especially in the humanleukocyte antigens [1] Several recent reports have describedthat gene methylation and its expression profile are alteredin GDM [7ndash9] Research studies showed that distinctivehigh-level expression changes of genes are associated withtheir lower promoter methylation prior to the onset of GDM[8] Although associations between GDM and changes to theepigenomic and genomic profiles for genes have beenstudied the molecular landscape changes for pathway levelduring the GDM are still substantially unknown It has been

HindawiComputational and Mathematical Methods in MedicineVolume 2019 Article ID 6936175 8 pageshttpsdoiorg10115520196936175

elucidated recently that activation of mammalian target ofrapamycin (mTOR) pathway upregulated functions of in-sulin secretion and pancreatic β-cells proliferation Diverseregulation of the mTOR pathway is involved in functions ofpancreatic β-cells as well as the development of the obstetriccomplications studied [10]

Currently researchers use a dynamically modifiedpipeline named FUNNEL-GSEA to analyze the biologicalmechanisms of gene sets is kind of model combined thefunctional principal component analysis and an elastic netregression model [11 12] Different from the classic re-gression analysis which is a predictive modeling techniquethat studies the relationship between dependent and in-dependent variables those techniques were often used forboth predictive analysis and causal inference among vari-ables [13] ere are two advances in the methods currentlyavailable First designed comparisons or regression analyseswere not only applied to the comparison between controland experimental groups but also effectively exploited in-dividual information in the group of transcriptomic mea-surements erefore in this model one could assess thedifferences in gene changes among different individuals ortime points Second it overcomes the problem that over-lapping genes which refers to genes that exist in multiplepathways play multiple roles in hypothesis testing wherethe weight coefficients are overestimated [14]

In our research we first applied the preweight pathwaymodel to find the enriched signaling pathways of GDMusingDNA methylation profiles Second we used functionalprincipal component analysis (FPCA) to identify hub genesin the significantly changed pathways Finally the geneexpression of key molecular pathways in GDM was furthertested in an independent cohort

2 Methods

21 Data Recruitment and Preprocess DNA methylationdata for GDM were deposited in the Gene ExpressionOmnibus database at GSE70453 A total of 82 sample datawere divided into two groups namely cases with gestationaldiabetes and controls without gestational diabetes eGDM cases contained 41 samples (GDM) and the matchedpregnancies (control) contained 41 samples

A total of 10 GDM patientsrsquo tissues and 10 age- and bodymass index-matched normal tissues for further validationwere obtained from the Department of Obstetrics FirstPeoplersquos Hospital of Jining between June 2016 and June 2017e decidua basalis placental samplingmethods were used asBinder et al [1] e umbilical cord was immediately frozenin liquid nitrogen and stored at minus 80degC after delivery isstudy was reviewed and approved by the Ethics Committeeof First Peoplersquos Hospital of Jining (Shandong China)

22 Screening for Differentially Methylated Genes Microarraydata contain 473864 CpG of methylation sites CpG siteswere eliminated when they met the following three types ofprobes (1) distance from CpG to single-nucleotide poly-morphism (SNP) is less than or equal to 2 (2)

minimum equipotential frequency (MAF) less than 005and (3) cross-hybridized probes and probes on sex chro-mosomes A total of 426693 CpGs were kept for furtherstudy In this paper β values represented percentagemethylation changes ranging between 0 and 1 Mean βvalues of GDM and normal population were calculatedrespectively We took the mean value of all the relatedCpGs as the methylation level of a gene since differentiallymethylated probes may functionally implicate more distalgenes Differentially methylated CpGs were identified at thethreshold of plt 001 and then kept for further study

23 Estimating the Weights of Genes in Each PathwayGenes that present in multiple pathways case the ldquoover-lapping problemsrdquo which were often overestimated in theenrichment analysis [13] We basically used the FUNNEL-GSEA model with some modifications to deal with ourmethylation data [15 16] We used the Kyoto Encyclopedia ofGenes and Genomes (KEGG) pathway as our database [17]e weights of the overlapping genes can be obtained by

wik 1

1113936kisinkiβk

i

(1)

where ki is all the pathways which contain gene k β is thevector of gene coefficients which is set as 1 here ereforethe weights of the overlapped genes would be estimated emodel above would decompose an overlapping gene be-tween gene sets and eliminate the effects of overlappinggenes [14]

24 Assessment of Significant Enriched Pathways Pathwayanalysis was used to find out significant pathways of theGDM In this study Fisherrsquos exact test and the MannndashWhitney U (MWU) test were carried out to select enrichedpathways e MannndashWhitney U (MWU) test is a rank-based nonparametric test that usually is used in a com-petitive gene set enrichment analysiseMWU test utilizedthe gene weight value to test whether the weight of this geneis significantly greater than other genes of the differentialpathway (background genes) [18] Combined with Fisherrsquosexact test the final p value was calculated as

Pi PMWU times PFisher

21113968

(2)

where Pi is the final p value of a pathway i while PMWU andPFisher represent the p values calculated from the MWU testand Fisherrsquos exact test respectively e list of differentialmethylated genes was assessed as gene list while the wholegenome was set as the background Pathways with plt 005and gene count gt1 were extracted and were considered asenriched pathways

25 Estimating the F-Statistics of Genes in the EnrichedPathways Using the FPCA Model In this model each genegets an F value [14] e mean methylation of each gene issubtracted and FPCA is adopted across all the centeredmethylation values Each gene methylation value is calcu-lated according to the following functions

2 Computational and Mathematical Methods in Medicine

1113955Xi (t) 1113954μi + 1113944L

l1

1113955ξil1113955Φl(t) (3)

In the above formula 1113955ξil is the FPC score which couldquantify how much 1113955Xi (t) can be explained by 1113955Φl(t) 1113954μi

represents the temporal sample average expression and1113955Φl(t) represents the lth eigenfunction

We net use functional F-statistics to summarize the genepattern information for each gene

Fi RSS0i minus RSS1iRSS1i + δ

(4)

where RSS0i is the residual sum of squares of null hypothesesRSS1i represents the residual sum of squares of alternativehypotheses δ could be considered as a ldquosignal-to-noiserdquoratio and Fi revealed the importance of genes [19] Geneswith higher F value indicate higher importance

26 Identification of Gene Expression of Hub Genes UsingFresh-Frozen Umbilical Cord Tissue by Reverse Transcription-Quantitative Polymerase Chain Reaction (RT-qPCR)Total RNA was extracted from cells or tissues using TRIzol(Invitrogen ermo Fisher Scientific Inc Waltham MAUSA) cDNA synthesis was performed at 37degC for 15minand then 85degC for 5 sec using reverse transcriptase (AppliedBiosystems ermo Fisher Scientific Inc) following themanufacturer protocol qPCR was conducted with the ABI7500 system (Applied Biosystemsermo Fisher ScientificInc) using SYBR-Green (Takara Biotechnology Co LtdJinan China) PCR was performed for 25 cycles of 10 sec at98degC 10 sec at 55degC and 20 sec at 72degC e primer se-quences used were as follows CAMK2B forward 5prime-TACGAGGATATTGGCAAGGG-3prime and reverse 5prime-GCTTCT GGT GAT AGT GTG C-3prime ADCYAP1 forward 5prime-ATC CTT AAC GAG GCC TAC C-3prime and reverse 5prime-CATTTG TTT CCG GTAGCGG-3prime KCNN2 forward 5prime-CCAGGA ACT GTA CTC TTG GT-3prime and reverse 5prime-ATCATGGTACCTTTCACAAGC-3prime GAPDH forward5prime-ACA CCC ACT CCT CCA CCT TT-3prime and reverse 5prime-TTA CTC CTT GGA GGC CAT GT-3prime mRNA expressionlevels were normalized using GAPDH Fold changes werecounted using the 2-ΔΔCt method

3 Results

31 Identification of Differentially Methylated GenesWith the threshold of plt 001 a total of 2310 differentiallymethylated CpGs (covering 1520 genes) were obtainedAmong the 2310 methylated CpGs 851 of the CpGs weredown-methylated and 1459 of the CpGs were up-meth-ylated in the GDM group Figure 1(a) shows the volcanicmap of differentially methylated CpGs According to thethreshold 2310 differentially methylated CpGs initiallyextracted were subjected to further filtering to obtain thehigh differentially methylated CpGs CpGs meeting Sge 01were retained resulting in 87 differentially methylatedCpGs covered 87 genes e top 10 differentially methyl-ated CpGs are shown in Figure 1(b)

32 Screening for Significantly Enriched Pathways Using aPreweighted Pathway Database Given a gene associatedwith multiple gene sets we assume that the overlappinggenes are activated by all gene sets to which they belongEstimated weights were assigned as 1n where n is thenumber of gene sets that this gene is associated withPathway enrichment analysis of GDM was conducted onthe basis of the KEGG pathway database A total of 286pathways covered 6893 genes were obtained Figure 2(a)shows the distribution of weights of all pathway geneswhile Figure 2(b) shows the distribution of sum weights ofall pathways and Figure 2(c) shows the weights of genes inthe 4 enriched pathways Based on the preweightedpathway database 4 differential pathways were yieldedTable 1 shows the differential signaling pathways in as-cending order based on the final p value

After Fisherrsquos exact test and the MannndashWhitneyU (MWU) test 4 members of the pathway are shown inTable 1 olfactory transduction prostate cancer insulinsecretion and amphetamine addiction ese 4 signalingpathways may play important roles in the occurrence ofGDM e insulin secretion pathway was considered as themost important pathway and kept for further identificationof key molecules in this pathway since it has been widelyapproved to be associated with GDM [3 20 21] e insulinsecretion pathway here contained 16 differentially methyl-ated genes Figure 3 shows the heatmap of DNAmethylationlevel of genes in the insulin secretion pathway

33 FPCA Analysis of Expression Profile for Hub Genes in theEnriched Pathways e FPCA model was used to identifyhub genes in the enriched pathways FPCA could effectivelyutilize the time series information and overcome the tradi-tional control design deficiencies [14] Each gene got an Fvalue Higher F value indicated a higher activation in theirpathways Figure 4(a) shows the distribution of the F value ofall pathway genes Figure 4(b) shows the F value of all genes inthe insulin secretion pathway Genes and their F values werelisted in descending order e top 3 genes (CAMK2BADCYAP1 and KCNN2) with high F values in the insulinsecretion pathway were selected for further validation

34 Validation of the Gene Expression of Hub Genes Toinvestigate the relationship between the methylation andthe gene expression of hub genes in GDM the additionalcohort was used to identify the expressions of hub genes inGDM e expression profiles of GDM and normal controlgroups of our cohort were used By assessing the RNAexpression data of hub genes significant changes betweenGDM and normal control groups were found in the geneexpression of ADCYAP1 e expression levels ofCAMK2B ADCYAP1 and KCNN2 between GDM andnormal control groups are shown in Figure 5

4 Discussion

GDM refers to a varying degree of impaired glucose tol-erance occurring for the first time during pregnancy

Computational and Mathematical Methods in Medicine 3

excluding patients who were with diabetes previous togestation but were first diagnosed during pregnancy [1] InChina the incidence of GDM is about 5ndash7 and there isan increasing trend [22] According to the results of tra-ditional research methods the etiology of GDM is closelyrelated to insulin resistance In recent years with the de-velopment of molecular genetics molecular immunologyand bioinformatics more and more studies have shown thatmany factors such as life style β-cell dysfunction in-flammatory factors and adipokines are involved in thedevelopment of GDM [23ndash25]

Epigenetics refers to heritable changes in gene functionthat occurs under the condition of not changing the DNAsequence including DNA methylation genomic imprintingmaternal effects gene silencing and RNA editing [26] As oneof the important epigenetic phenomena DNA methylationplays an important regulatory role in the gene expressionDNA methylation is closely related to the occurrence anddevelopment of many diseases such as type 2 diabetes [27]autoimmune diseases and various cancers [28 29]

In this study GDM pathogenesis was analyzed usingbioinformatics including KEGG enrichment methodfunctional principal component analysis (FPCA) elasticnet regression and the MannndashWhitney U test Accordingto this new analytical procedure four signaling pathwaysfor olfactory transduction prostate cancer insulin secre-tion and amphetamine addiction were found out erewere some genes involved in the enriched pathways whichwere related to GDM Herein one differentially methylated

key molecule was identified adenylate cyclase activatingpolypeptide 1 (ADCYAP1) ADCYAP1 gene encodes apituitary adenylate cyclase activating polypeptide(PACAP) PACAP is a secreted proprotein with the abilityto activate adenylyl cyclase which is a membrane-boundenzyme that converts ATP to cAMP ADCY3 is one of theadenylate cyclases that participates in the insulin secretionpathway and also identified as a key moleculegene by ouranalysis ose results provide a novel insight into GDMdiagnosis and therapy Numerous studies demonstrate thatPACAP and adenylyl cyclase have a potential role in isletphysiology and as a basis for development of islet-pro-moting therapy in diabetes [30ndash32] Adenylyl cyclase is aneffector in the G protein-coupled system [33] and itsenzymatic activity is under the control of several hormonesincluding insulin [34 35] PACAP and adenylyl cyclasewere capable of influencing pancreatic islet function bystimulating pancreatic beta cells to secrete insulin andglucagon [36] For the clinical treatment of type 2 diabetesPACAP and adenylyl cyclase were also thought to be ef-fective due to stimulation of insulin secretion [37] andincreased proliferation and differentiation of β cells [38] Inaddition PACAP is also a neurotransmitter and a memberof the vasoactive intestinal peptidesecretinglucagonpeptide superfamily PACAP shows highly potent neuro-protective and general cytoprotective effects [39] PACAPis also protective in diabetes-induced pathologies likeretinopathy and nephropathy [40ndash42] Consistent withthose studies our results showed that several signal

ndash015 ndash010 ndash005 000 005 010 015

5

4

3

2

0

1

Volcano plot for differentially methylated genes

log2 fold change

ndashlog

10 (p

val

ue)

(a)cg

0157

0480

cg07

3045

26

cg08

1470

94

cg18

3514

06

cg20

3009

11

cg22

8684

33

cg22

9844

39

cg24

4987

60

cg27

0356

78

cg27

0888

22

Symbol

01

00

ndash01

log2

fold

chan

geUpDown

(b)

Figure 1 (a) e volcanic map of differentially methylated CpGs All points in the figure represent all the 426693 methylated CpGs ex-axis represents methylation differences between GDM and normal (log2-transformed fold change) Y-axis was log10-transformed p

values (b) e top 10 differentially methylated CpGs


0 50 100 150 200 250Pathways

14

12

10

8

6

4

2

0

Sum

wei

ght

(a)

1000

800

600

400

200

0

00 02 04 06 08 10Weight

Freq

uenc

y

(b)

PDGFDCCNE2TCF7E2F3RB1CDKN1BPDGFAPDGFBPDGFRAPDPK1CREB5CREB3L2CREB3L3EGFCREBBPGSK3BEGFRNFKBIAPIK3CD

010009008008007007006006006005005005005005004003003003001

100100100100100050017014004003002

050033033009

009011020

007 007005005005 005

005006

008

005004004004004 004

004004

003003003

003003

002002

OR4C6OR2K2OR2H1OR2F2OR5H2CNGA3GNALARRB2CAMK2BADCY3PRKACA

ADCYAP1KCNN2PDX1RYR2ATF6BCREB5CREB3L2CREB3L3CACNA1CCACNA1DCAMK2BITPR3ADCY3GNASADCY5PRKACA

SLC6A3CAMK4GRIA1GRIA2

GRIN2AATF6B

CREB5CREB3L2CREB3L3CACNA1CCACNA1DCAMK2BGNASADCY5PRKACA

Mem

bers

hip

Membership

Mem

bers

hip

Mem

bers

hipM

embe

rshi

p

hsa0

5215

pro

stat

eca

ncer

hsa0

4911

ins

ulin

secr

etio

n

hsa0

4740

olfa

ctor

ytr

ansd

uctio

n

hsa0

5031

am

phet

amin

ead

ditio

n

1

08

06

04

02

0

ExculsiveOverlapping

(c)

Figure 2 Plot of sumweight (a) and weight distribution of genes (b) for KEGG pathways in GDM and the weights of genes in the 4 enrichedpathways (c)


transduction pathways such as olfactory transductionpathway were enriched as well

ere are limitations present in our study Un-fortunately the methylation data along with the expression

data for samples were unprovided Further study will bedirectly tested for the most significant CpGs that impactupon each hub gene since multiple CpGs could impact thesame gene

Table 1 Significant enriched pathways of GDM

Pathway name Fisherrsquos exact test MWU test Final p value Count Totalhsa04740 olfactory transduction 00014 0126083253 00013 11 408hsa05215 prostate cancer 00043 0164116273 00026 19 89hsa04911 insulin secretion 00008 0094356684 00091 16 86hsa05031 amphetamine addiction 00001 049776561 00092 15 68Count number of genes in a pathway Total total number of genes in a pathway

CACNA1D

KCNN2

PRKACA

CACNA1C

PDX1

CAMK2B

ITPR3

CREB3L2

CREB3L3

ADCYAP1

ADCY3

CREB5

GNAS

RYR2

ADCY5

ATF6B

ID08

07

06

05

04

03

02

Figure 3 Heatmap of methylation levels of the insulin secretion pathway

120

100

80

60

40

20

0

Freq

uenc

y

000 005 010 015 020 025

(a)

ITPR

3

ATF6

B

RYR2

GN

AS

PDX1

PRKA

CACA

MK2

BA

DCY

5

CACN

A1C

CACN

A1D

CREB

3L2

CREB

5CR

EB3L

3

AD

CYA

P1

KCN

N2

AD

CY3

Genes in the insulin secretion pathway

015

010

005

000

F va

lue

(b)

Figure 4 e distribution of F value of pathway genes from all pathway genes (a) and insulin secretion pathway (b) Gene methylation datawere analyzed by FPCA and each gene got an F value (x-coordinate F value) Y-axis represents gene density


5 Conclusions

Based on the analytical results of the present study there wassignificant ADCYAP1 methylation and gene expressiondifferences between GDM and normal control groups GDMwas associated with insulin resistance and insulin-signalingsystem may require ADCYAP1 participation [43] Wespeculate that ADCYAP1 may be related to the GDM andmore experimental data were needed to support ourprediction

Data Availability

e datasets used and analyzed during the current study areavailable from the corresponding author on reasonablerequest

Conflicts of Interest

e authors declare that they have no conflicts of interest

References

[1] A M Binder J Larocca C Lesseur C J Marsit andK B Michels ldquoEpigenome-wide and transcriptome-wideanalyses reveal gestational diabetes is associated with alter-ations in the human leukocyte antigen complexrdquo ClinicalEpigenetics vol 7 no 1 pp 1ndash12 2015

[2] A Vaag C Broslashns L Gillberg et al ldquoGenetic nongenetic andepigenetic risk determinants in developmental programmingof type 2 diabetesrdquo Acta Obstetricia et Gynecologica Scandi-navica vol 93 no 11 pp 1099ndash1108 2014

[3] E P Gunderson M M Hedderson V Chiang et al ldquoLac-tation intensity and postpartum maternal glucose tolerance

and insulin resistance in women with recent GDM theSWIFTcohortrdquo Diabetes Care vol 35 no 1 pp 50ndash56 2012

[4] A Syngelaki G H A Visser K Krithinakis A Wright andK H Nicolaides ldquoFirst trimester screening for gestationaldiabetes mellitus by maternal factors and markers of in-flammationrdquo Metabolism vol 65 no 3 pp 131ndash137 2016

[5] Y Yang J Yang G Zhong and W Xu ldquoPotential risk factorof pre-eclampsia among healthy Chinese women a retro-spective case control studyrdquo Biomedical Research vol 28no 3 pp 1183ndash1188 2017

[6] H Lee andM Shin ldquoMining pathway associations for disease-related pathway activity analysis based on gene expression andmethylation datardquo Biodata Mining vol 10 no 1 p 3 2017

[7] L Bouchard S ibault S-P Guay et al ldquoLeptin geneepigenetic adaptation to impaired glucose metabolism duringpregnancyrdquo Diabetes Care vol 33 no 11 pp 2436ndash24412010

[8] J Alexander A M Teague J Chen et al ldquoOffspring seximpacts DNA methylation and gene expression in placentaefrom women with diabetes during pregnancyrdquo PLoS Onevol 13 no 2 Article ID e0190698 2018

[9] P Wu W E Farrell K E Haworth et al ldquoMaternal genome-wide DNAmethylation profiling in gestational diabetes showsdistinctive disease-associated changes relative to matchedhealthy pregnanciesrdquo Epigenetics vol 13 no 2 pp 122ndash1282018

[10] K Xu D Bian L Hao et al ldquoMicroRNA-503 contribute topancreatic beta cell dysfunction by targeting the mTORpathway in gestational diabetes mellitusrdquo EXCLI Journalvol 16 pp 1177ndash1187 2017

[11] F Al-Shahrour P Minguez J Tarraga et al ldquoBABELOMICSa systems biology perspective in the functional annotation ofgenome-scale experimentsrdquo Nucleic Acids Research vol 34pp W472ndashW476 2006

[12] D Tabas-Madrid R Nogales-Cadenas and A Pascual-Montano ldquoGeneCodis3 a non-redundant and modular en-richment analysis tool for functional genomicsrdquoNucleic AcidsResearch vol 40 no W1 pp W478ndashW483 2012

[13] C J Lu J Y Wu and T S Lee ldquoApplication of independentcomponent analysis preprocessing and support vector re-gression in time series predictionrdquo in Proceedings of the 2009International Joint Conference on Computational Sciences andOptimization pp 468ndash471 IEEE Sanya China April 2009

[14] Y Zhang D J Topham J akar and X Qiu ldquoFUNNEL-GSEA functional elastic-net regression in time-course geneset enrichment analysisrdquo Bioinformatics vol 33 no 13pp 1944ndash1952 2017

[15] J O Ogutu T Schulz-Streeck and H-P Piepho ldquoGenomicselection using regularized linear regression models ridgeregression lasso elastic net and their extensionsrdquo BMCProceedings vol 6 no S2 p S10 2012

[16] X Li and Y Fang ldquoBioinformatics identification of potentialgenes and pathways in preeclampsia based on functional geneset enrichment analysesrdquo Experimental and HerapeuticMedicine vol 18 no 3 pp 1837ndash1844 2019

[17] J Zhu and X Yao ldquoUse of DNA methylation for cancerdetection promises and challengesrdquoHe International Journalof Biochemistry amp Cell Biology vol 41 no 1 pp 147ndash1542009

[18] C-A Brandsma M van den Berge D S Postma et al ldquoA largelung gene expression study identifying fibulin-5 as a novel playerin tissue repair in COPDrdquoHorax vol 70 no 1 pp 21ndash32 2015

[19] S Wu and H Wu ldquoMore powerful significant testing for timecourse gene expression data using functional principal

10

8

6

4

2

0

KCNN2

KCNN2

CAMK2

B

CAMK2

B

ADCY

AP1

ADCY

AP1

Genes

lowastlowast

Rela

tive e

xpre

ssio

n

Control groupGDM group

Figure 5 Expression levels of CAMK2B ADCYAP1 and KCNN2Grey color represents the control group and black color representsthe GDM group


component analysis approachesrdquo BMC Bioinformaticsvol 14 no 1 p 6 2013

[20] B Liu Y Xu C Voss et al ldquoAltered protein expression ingestational diabetes mellitus placentas provides insight intoinsulin resistance and coagulationfibrinolysis pathwaysrdquoPLoS One vol 7 no 9 Article ID e44701 2012

[21] V Pisprasert K H Ingram M F Lopez-Davila A J Munozand W T Garvey ldquoLimitations in the use of indices usingglucose and insulin levels to predict insulin sensitivity impactof race and gender and superiority of the indices derived fromoral glucose tolerance test in African Americansrdquo DiabetesCare vol 36 no 4 pp 845ndash853 2013

[22] M Anand M Anand and D S Mahajan ldquoTo study theincidence of gestational diabetes mellitus and risk factorsassociated with GDMrdquo International Journal of Advances inMedicine vol 4 no 1 p 112 2017

[23] P Moleda A Fronczyk K Safranow and L MajkowskaldquoAdipokines and β-cell dysfunction in normoglycemicwomen with previous gestational diabetes mellitusrdquo PolishArchives of Internal Medicine vol 125 no 9 pp 641ndash6482015

[24] S K Abell B D Courten J A Boyle and H Teede ldquoIn-flammatory and other biomarkers role in pathophysiologyand prediction of gestational diabetes mellitusrdquo InternationalJournal of Molecular Sciences vol 16 no 12 pp 13442ndash134732015

[25] M R Al-Badri M S Zantout and S T Azar ldquoe role ofadipokines in gestational diabetes mellitusrdquo HerapeuticAdvances in Endocrinology amp Metabolism vol 6 no 3pp 103ndash108 2015

[26] A P Feinberg andM D Fallin ldquoEpigenetics at the crossroadsof genes and the environmentrdquo JAMA vol 314 no 11pp 1129-1130 2015

[27] H R Elliott H A Shihab G A Lockett et al ldquoRole of DNAmethylation in type 2 diabetes etiology using genotype as acausal anchorrdquo Diabetes vol 66 no 6 pp 1713ndash1722 2017

[28] M J Ziller H Gu F Muller et al ldquoCharting a dynamic DNAmethylation landscape of the human genomerdquo Naturevol 500 no 7463 pp 477ndash481 2013

[29] B Sun L Hu Z-Y Luo X-P Chen H-H Zhou andW Zhang ldquoDNA methylation perspectives in the patho-genesis of autoimmune diseasesrdquo Clinical Immunologyvol 164 pp 21ndash27 2016

[30] M Basille B J Gonzalez L Desrues M Demas A Fournierand H Vaudry ldquoPituitary adenylate cyclase-activating poly-peptide (PACAP) stimulates adenylyl cyclase and phospho-lipase C activity in rat cerebellar neuroblastsrdquo Journal ofNeurochemistry vol 65 no 3 pp 1318ndash1324 2002

[31] M Solymar I Ivic M Balasko et al ldquoPituitary adenylatecyclase-activating polypeptide ameliorates vascular dysfunc-tion induced by hyperglycaemiardquo Diabetes and VascularDisease Research vol 15 no 4 pp 277ndash285 2018

[32] B Ahren ldquoRole of pituitary adenylate cyclase-activatingpolypeptide in the pancreatic endocrine systemrdquo Annals of theNew York Academy of Sciences vol 1144 no 1 pp 28ndash35 2008

[33] J Shengunther C M Wang G M Poage et al ldquoMolecularpap smear HPV genotype and DNA methylation of ADCY8CDH8 and ZNF582 as an integrated biomarker for high-grade cervical cytologyrdquo Clinical Epigenetics vol 8 no 1p 96 2016

[34] D Vaudry B J Gonzalez M Basille L Yon A Fournier andH Vaudry ldquoPituitary adenylate cyclase-activating polypeptideand its receptors from structure to functionsrdquo PharmacologicalReviews vol 52 no 2 pp 269ndash324 2000

[35] L A Kuznetsova S A Plesneva T S SharovaM N Pertsevaand A O Shpakov ldquoRegulation of adenylyl cyclase signalingsystem by insulin biogenic amines and glucagon at theirseparate and combined action in muscle membranes ofmollusc Anodonta cygneardquo Journal of Evolutionary Bio-chemistry and Physiology vol 49 no 2 pp 145ndash152 2013

[36] K Filipsson F Sundler J Hannibal and B Ahren ldquoPACAPand PACAP receptors in insulin producing tissues locali-zation and effectsrdquo Regulatory Peptides vol 74 no 2-3pp 167ndash175 1998

[37] M S Winzell and B Ahren ldquoRole of VIP and PACAP in isletfunctionrdquo Peptides vol 28 no 9 pp 1805ndash1813 2007

[38] K Yamamoto H Hashimoto S Tomimoto et al ldquoOver-expression of PACAP in transgenic mouse pancreatic beta-cells enhances insulin secretion and ameliorates streptozo-tocin-induced diabetesrdquo Diabetes vol 52 no 5 pp 1155ndash1162 2003

[39] D Reglodi T Atlasz A Jungling et al ldquoAlternative routes ofadministration of the neuroprotective pituitary adenylatecyclase activating polypeptiderdquo Current Pharmaceutical De-sign vol 24 no 33 pp 3892ndash3904 2018

[40] K Szabadfi A Szabo P Kiss et al ldquoPACAP promotes neuronsurvival in early experimental diabetic retinopathyrdquo Neuro-chemistry International vol 64 pp 84ndash91 2014

[41] E Banki P Degrell P Kiss et al ldquoEffect of PACAP treatmenton kidney morphology and cytokine expression in rat diabeticnephropathyrdquo Peptides vol 42 pp 125ndash130 2013

[42] A G DrsquoAmico G Maugeri D M Rasa et al ldquoModulation ofIL-1β and VEGF expression in rat diabetic retinopathy afterPACAP administrationrdquo Peptides vol 97 pp 64ndash69 2017

[43] Y Li C Hadden P Singh et al ldquoGDM-associated insulindeficiency hinders the dissociation of SERT from ERp44 anddown-regulates placental 5-HT uptakerdquo Proceedings of theNational Academy of Sciences vol 111 no 52 pp E5697ndashE5705 2014


Stem Cells International

Hindawiwwwhindawicom Volume 2018


MEDIATORSINFLAMMATION

of

EndocrinologyInternational Journal of



Disease Markers


BioMed Research International

OncologyJournal of



Oxidative Medicine and Cellular Longevity


PPAR Research

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Immunology ResearchHindawiwwwhindawicom Volume 2018

Journal of

ObesityJournal of



Computational and Mathematical Methods in Medicine


Behavioural Neurology

OphthalmologyJournal of


Diabetes ResearchJournal of



Research and TreatmentAIDS


Gastroenterology Research and Practice


Parkinsonrsquos Disease

Evidence-Based Complementary andAlternative Medicine

Volume 2018Hindawiwwwhindawicom

Submit your manuscripts atwwwhindawicom

elucidated recently that activation of mammalian target ofrapamycin (mTOR) pathway upregulated functions of in-sulin secretion and pancreatic β-cells proliferation Diverseregulation of the mTOR pathway is involved in functions ofpancreatic β-cells as well as the development of the obstetriccomplications studied [10]

Currently researchers use a dynamically modifiedpipeline named FUNNEL-GSEA to analyze the biologicalmechanisms of gene sets is kind of model combined thefunctional principal component analysis and an elastic netregression model [11 12] Different from the classic re-gression analysis which is a predictive modeling techniquethat studies the relationship between dependent and in-dependent variables those techniques were often used forboth predictive analysis and causal inference among vari-ables [13] ere are two advances in the methods currentlyavailable First designed comparisons or regression analyseswere not only applied to the comparison between controland experimental groups but also effectively exploited in-dividual information in the group of transcriptomic mea-surements erefore in this model one could assess thedifferences in gene changes among different individuals ortime points Second it overcomes the problem that over-lapping genes which refers to genes that exist in multiplepathways play multiple roles in hypothesis testing wherethe weight coefficients are overestimated [14]

In our research we first applied the preweight pathwaymodel to find the enriched signaling pathways of GDMusingDNA methylation profiles Second we used functionalprincipal component analysis (FPCA) to identify hub genesin the significantly changed pathways Finally the geneexpression of key molecular pathways in GDM was furthertested in an independent cohort

2 Methods

21 Data Recruitment and Preprocess DNA methylationdata for GDM were deposited in the Gene ExpressionOmnibus database at GSE70453 A total of 82 sample datawere divided into two groups namely cases with gestationaldiabetes and controls without gestational diabetes eGDM cases contained 41 samples (GDM) and the matchedpregnancies (control) contained 41 samples

A total of 10 GDM patientsrsquo tissues and 10 age- and bodymass index-matched normal tissues for further validationwere obtained from the Department of Obstetrics FirstPeoplersquos Hospital of Jining between June 2016 and June 2017e decidua basalis placental samplingmethods were used asBinder et al [1] e umbilical cord was immediately frozenin liquid nitrogen and stored at minus 80degC after delivery isstudy was reviewed and approved by the Ethics Committeeof First Peoplersquos Hospital of Jining (Shandong China)

22 Screening for Differentially Methylated Genes Microarraydata contain 473864 CpG of methylation sites CpG siteswere eliminated when they met the following three types ofprobes (1) distance from CpG to single-nucleotide poly-morphism (SNP) is less than or equal to 2 (2)

minimum equipotential frequency (MAF) less than 005and (3) cross-hybridized probes and probes on sex chro-mosomes A total of 426693 CpGs were kept for furtherstudy In this paper β values represented percentagemethylation changes ranging between 0 and 1 Mean βvalues of GDM and normal population were calculatedrespectively We took the mean value of all the relatedCpGs as the methylation level of a gene since differentiallymethylated probes may functionally implicate more distalgenes Differentially methylated CpGs were identified at thethreshold of plt 001 and then kept for further study

23 Estimating the Weights of Genes in Each PathwayGenes that present in multiple pathways case the ldquoover-lapping problemsrdquo which were often overestimated in theenrichment analysis [13] We basically used the FUNNEL-GSEA model with some modifications to deal with ourmethylation data [15 16] We used the Kyoto Encyclopedia ofGenes and Genomes (KEGG) pathway as our database [17]e weights of the overlapping genes can be obtained by

wik 1

1113936kisinkiβk

i

(1)

where ki is all the pathways which contain gene k β is thevector of gene coefficients which is set as 1 here ereforethe weights of the overlapped genes would be estimated emodel above would decompose an overlapping gene be-tween gene sets and eliminate the effects of overlappinggenes [14]

24 Assessment of Significant Enriched Pathways Pathwayanalysis was used to find out significant pathways of theGDM In this study Fisherrsquos exact test and the MannndashWhitney U (MWU) test were carried out to select enrichedpathways e MannndashWhitney U (MWU) test is a rank-based nonparametric test that usually is used in a com-petitive gene set enrichment analysiseMWU test utilizedthe gene weight value to test whether the weight of this geneis significantly greater than other genes of the differentialpathway (background genes) [18] Combined with Fisherrsquosexact test the final p value was calculated as

Pi PMWU times PFisher

21113968

(2)

where Pi is the final p value of a pathway i while PMWU andPFisher represent the p values calculated from the MWU testand Fisherrsquos exact test respectively e list of differentialmethylated genes was assessed as gene list while the wholegenome was set as the background Pathways with plt 005and gene count gt1 were extracted and were considered asenriched pathways

25 Estimating the F-Statistics of Genes in the EnrichedPathways Using the FPCA Model In this model each genegets an F value [14] e mean methylation of each gene issubtracted and FPCA is adopted across all the centeredmethylation values Each gene methylation value is calcu-lated according to the following functions


1113955Xi (t) 1113954μi + 1113944L

l1

1113955ξil1113955Φl(t) (3)





(4)



3 Results






4 Discussion








5

4

3

2

0

1


log2 fold change

ndashlog

10 (p

val

ue)

(a)cg

0157

0480

cg07

3045

26

cg08

1470

94

cg18

3514

06

cg20

3009

11

cg22

8684

33

cg22

9844

39

cg24

4987

60

cg27

0356

78

cg27

0888

22

Symbol

01

00

ndash01

log2

fold

chan

geUpDown

(b)




0 50 100 150 200 250Pathways

14

12

10

8

6

4

2

0

Sum

wei

ght

(a)

1000

800

600

400

200

0

00 02 04 06 08 10Weight

Freq

uenc

y

(b)


010009008008007007006006006005005005005005004003003003001

100100100100100050017014004003002

050033033009

009011020

007 007005005005 005

005006

008

005004004004004 004

004004

003003003

003003

002002




GRIN2AATF6B


Mem

bers

hip

Membership

Mem

bers

hip

Mem

bers

hipM

embe

rshi

p

hsa0

5215

pro

stat

eca

ncer

hsa0

4911

ins

ulin

secr

etio

n

hsa0

4740

olfa

ctor

ytr

ansd

uctio

n

hsa0

5031

am

phet

amin

ead

ditio

n

1

08

06

04

02

0


(c)








CACNA1D

KCNN2

PRKACA

CACNA1C

PDX1

CAMK2B

ITPR3

CREB3L2

CREB3L3

ADCYAP1

ADCY3

CREB5

GNAS

RYR2

ADCY5

ATF6B

ID08

07

06

05

04

03

02


120

100

80

60

40

20

0

Freq

uenc

y

000 005 010 015 020 025

(a)

ITPR

3

ATF6

B

RYR2

GN

AS

PDX1

PRKA

CACA

MK2

BA

DCY

5

CACN

A1C

CACN

A1D

CREB

3L2

CREB

5CR

EB3L

3

AD

CYA

P1

KCN

N2

AD

CY3


015

010

005

000

F va

lue

(b)



5 Conclusions


Data Availability




References





















10

8

6

4

2

0

KCNN2

KCNN2

CAMK2

B

CAMK2

B

ADCY

AP1

ADCY

AP1

Genes

lowastlowast

Rela

tive e

xpre

ssio

n


































of




Disease Markers



OncologyJournal of





PPAR Research



Volume 2018


Journal of

ObesityJournal of



















1113955Xi (t) 1113954μi + 1113944L

l1

1113955ξil1113955Φl(t) (3)





(4)



3 Results






4 Discussion








5

4

3

2

0

1


log2 fold change

ndashlog

10 (p

val

ue)

(a)cg

0157

0480

cg07

3045

26

cg08

1470

94

cg18

3514

06

cg20

3009

11

cg22

8684

33

cg22

9844

39

cg24

4987

60

cg27

0356

78

cg27

0888

22

Symbol

01

00

ndash01

log2

fold

chan

geUpDown

(b)




0 50 100 150 200 250Pathways

14

12

10

8

6

4

2

0

Sum

wei

ght

(a)

1000

800

600

400

200

0

00 02 04 06 08 10Weight

Freq

uenc

y

(b)


010009008008007007006006006005005005005005004003003003001

100100100100100050017014004003002

050033033009

009011020

007 007005005005 005

005006

008

005004004004004 004

004004

003003003

003003

002002




GRIN2AATF6B


Mem

bers

hip

Membership

Mem

bers

hip

Mem

bers

hipM

embe

rshi

p

hsa0

5215

pro

stat

eca

ncer

hsa0

4911

ins

ulin

secr

etio

n

hsa0

4740

olfa

ctor

ytr

ansd

uctio

n

hsa0

5031

am

phet

amin

ead

ditio

n

1

08

06

04

02

0


(c)








CACNA1D

KCNN2

PRKACA

CACNA1C

PDX1

CAMK2B

ITPR3

CREB3L2

CREB3L3

ADCYAP1

ADCY3

CREB5

GNAS

RYR2

ADCY5

ATF6B

ID08

07

06

05

04

03

02


120

100

80

60

40

20

0

Freq

uenc

y

000 005 010 015 020 025

(a)

ITPR

3

ATF6

B

RYR2

GN

AS

PDX1

PRKA

CACA

MK2

BA

DCY

5

CACN

A1C

CACN

A1D

CREB

3L2

CREB

5CR

EB3L

3

AD

CYA

P1

KCN

N2

AD

CY3


015

010

005

000

F va

lue

(b)



5 Conclusions


Data Availability




References





















10

8

6

4

2

0

KCNN2

KCNN2

CAMK2

B

CAMK2

B

ADCY

AP1

ADCY

AP1

Genes

lowastlowast

Rela

tive e

xpre

ssio

n


































of




Disease Markers



OncologyJournal of





PPAR Research



Volume 2018


Journal of

ObesityJournal of
























5

4

3

2

0

1


log2 fold change

ndashlog

10 (p

val

ue)

(a)cg

0157

0480

cg07

3045

26

cg08

1470

94

cg18

3514

06

cg20

3009

11

cg22

8684

33

cg22

9844

39

cg24

4987

60

cg27

0356

78

cg27

0888

22

Symbol

01

00

ndash01

log2

fold

chan

geUpDown

(b)




0 50 100 150 200 250Pathways

14

12

10

8

6

4

2

0

Sum

wei

ght

(a)

1000

800

600

400

200

0

00 02 04 06 08 10Weight

Freq

uenc

y

(b)


010009008008007007006006006005005005005005004003003003001

100100100100100050017014004003002

050033033009

009011020

007 007005005005 005

005006

008

005004004004004 004

004004

003003003

003003

002002




GRIN2AATF6B


Mem

bers

hip

Membership

Mem

bers

hip

Mem

bers

hipM

embe

rshi

p

hsa0

5215

pro

stat

eca

ncer

hsa0

4911

ins

ulin

secr

etio

n

hsa0

4740

olfa

ctor

ytr

ansd

uctio

n

hsa0

5031

am

phet

amin

ead

ditio

n

1

08

06

04

02

0


(c)








CACNA1D

KCNN2

PRKACA

CACNA1C

PDX1

CAMK2B

ITPR3

CREB3L2

CREB3L3

ADCYAP1

ADCY3

CREB5

GNAS

RYR2

ADCY5

ATF6B

ID08

07

06

05

04

03

02


120

100

80

60

40

20

0

Freq

uenc

y

000 005 010 015 020 025

(a)

ITPR

3

ATF6

B

RYR2

GN

AS

PDX1

PRKA

CACA

MK2

BA

DCY

5

CACN

A1C

CACN

A1D

CREB

3L2

CREB

5CR

EB3L

3

AD

CYA

P1

KCN

N2

AD

CY3


015

010

005

000

F va

lue

(b)



5 Conclusions


Data Availability




References





















10

8

6

4

2

0

KCNN2

KCNN2

CAMK2

B

CAMK2

B

ADCY

AP1

ADCY

AP1

Genes

lowastlowast

Rela

tive e

xpre

ssio

n


































of




Disease Markers



OncologyJournal of





PPAR Research



Volume 2018


Journal of

ObesityJournal of



















0 50 100 150 200 250Pathways

14

12

10

8

6

4

2

0

Sum

wei

ght

(a)

1000

800

600

400

200

0

00 02 04 06 08 10Weight

Freq

uenc

y

(b)


010009008008007007006006006005005005005005004003003003001

100100100100100050017014004003002

050033033009

009011020

007 007005005005 005

005006

008

005004004004004 004

004004

003003003

003003

002002




GRIN2AATF6B


Mem

bers

hip

Membership

Mem

bers

hip

Mem

bers

hipM

embe

rshi

p

hsa0

5215

pro

stat

eca

ncer

hsa0

4911

ins

ulin

secr

etio

n

hsa0

4740

olfa

ctor

ytr

ansd

uctio

n

hsa0

5031

am

phet

amin

ead

ditio

n

1

08

06

04

02

0


(c)








CACNA1D

KCNN2

PRKACA

CACNA1C

PDX1

CAMK2B

ITPR3

CREB3L2

CREB3L3

ADCYAP1

ADCY3

CREB5

GNAS

RYR2

ADCY5

ATF6B

ID08

07

06

05

04

03

02


120

100

80

60

40

20

0

Freq

uenc

y

000 005 010 015 020 025

(a)

ITPR

3

ATF6

B

RYR2

GN

AS

PDX1

PRKA

CACA

MK2

BA

DCY

5

CACN

A1C

CACN

A1D

CREB

3L2

CREB

5CR

EB3L

3

AD

CYA

P1

KCN

N2

AD

CY3


015

010

005

000

F va

lue

(b)



5 Conclusions


Data Availability




References





















10

8

6

4

2

0

KCNN2

KCNN2

CAMK2

B

CAMK2

B

ADCY

AP1

ADCY

AP1

Genes

lowastlowast

Rela

tive e

xpre

ssio

n


































of




Disease Markers



OncologyJournal of





PPAR Research



Volume 2018


Journal of

ObesityJournal of
























CACNA1D

KCNN2

PRKACA

CACNA1C

PDX1

CAMK2B

ITPR3

CREB3L2

CREB3L3

ADCYAP1

ADCY3

CREB5

GNAS

RYR2

ADCY5

ATF6B

ID08

07

06

05

04

03

02


120

100

80

60

40

20

0

Freq

uenc

y

000 005 010 015 020 025

(a)

ITPR

3

ATF6

B

RYR2

GN

AS

PDX1

PRKA

CACA

MK2

BA

DCY

5

CACN

A1C

CACN

A1D

CREB

3L2

CREB

5CR

EB3L

3

AD

CYA

P1

KCN

N2

AD

CY3


015

010

005

000

F va

lue

(b)



5 Conclusions


Data Availability




References





















10

8

6

4

2

0

KCNN2

KCNN2

CAMK2

B

CAMK2

B

ADCY

AP1

ADCY

AP1

Genes

lowastlowast

Rela

tive e

xpre

ssio

n


































of




Disease Markers



OncologyJournal of





PPAR Research



Volume 2018


Journal of

ObesityJournal of



















5 Conclusions


Data Availability




References





















10

8

6

4

2

0

KCNN2

KCNN2

CAMK2

B

CAMK2

B

ADCY

AP1

ADCY

AP1

Genes

lowastlowast

Rela

tive e

xpre

ssio

n


































of




Disease Markers



OncologyJournal of





PPAR Research



Volume 2018


Journal of

ObesityJournal of

















































of




Disease Markers



OncologyJournal of





PPAR Research



Volume 2018


Journal of

ObesityJournal of



















EpigeneticProfilesRevealThatADCYAP1ServesasKey...

Documents

Transcript of EpigeneticProfilesRevealThatADCYAP1ServesasKey...