Alternative 3′ UTRs Modify the Localization, Regulatory ...€¦ · niﬁcantly altered by...

NeuroResource

Alternative 30 UTRs Modify
the Localization,Regulatory Potential, Stability, and Plasticity ofmRNAs in Neuronal Compartments
Graphical Abstract

Highlights

d Neuronal mRNAs usually possess multiple 30 UTR isoforms

d 30 UTR isoforms of a transcript family can be enriched in

different compartments

d 30 UTR isoforms enriched in the neuropil have longer half-lives

d Enhanced neural activity alters the 30 UTR isoforms present in

each compartment

Tushev et al., 2018, Neuron 98, 1–17May 2, 2018 ª 2018 Elsevier Inc.https://doi.org/10.1016/j.neuron.2018.03.030

Authors

Georgi Tushev, Caspar Glock,

Maximilian Heum€uller, Anne Biever,

Marko Jovanovic, Erin M. Schuman

[email protected]

In Brief

Tushev, Glock, et al. use 30 endsequencing together with hippocampal

slice microdissection to separately

examine transcripts arising from somata

or the neuropil. They discover a huge

diversity in neuronal mRNA 30 UTRs,which give rise to differences in

localization, stability, translation, and

plasticity.

mailto:[email protected].�de

https://doi.org/10.1016/j.neuron.2018.03.030

Please cite this article in press as: Tushev et al., Alternative 30 UTRs Modify the Localization, Regulatory Potential, Stability, and Plasticity of mRNAs inNeuronal Compartments, Neuron (2018), https://doi.org/10.1016/j.neuron.2018.03.030

Neuron

NeuroResource

Alternative 30 UTRs Modifythe Localization, Regulatory Potential, Stability,and Plasticity of mRNAs in Neuronal CompartmentsGeorgi Tushev,1,3 Caspar Glock,1,3 Maximilian Heum€uller,1 Anne Biever,1 Marko Jovanovic,1,2 and Erin M. Schuman1,4,*1Max Planck Institute for Brain Research, Frankfurt am Main, Germany2Present address: Department of Biology, Columbia University, NY, USA3These authors contributed equally4Lead Contact

*Correspondence: [email protected]


SUMMARY

Neurons localize mRNAs near synapses where theirtranslation can be regulated by synaptic demandand activity. Differences in the 30 UTRs of mRNAscan change their localization, stability, and transla-tional regulation. Using 30 end RNA sequencing ofmi-crodissected rat brain slices, we discovered a hugediversity in mRNA 30 UTRs, with many transcriptsshowing enrichment for a particular 30 UTR isoformin either somata or the neuropil. The 30 UTR isoformsof localized transcripts are significantly longer thanthe 30 UTRs of non-localized transcripts and oftencode for proteins associated with axons, dendrites,and synapses. Surprisingly, long 30 UTRs add notonly new, but also duplicate regulatory elements.The neuropil-enriched 30 UTR isoforms have sig-nificantly longer half-lives than somata-enrichedisoforms. Finally, the 30 UTR isoforms can be sig-nificantly altered by enhanced activity. Most of the30 UTR plasticity is transcription dependent, butintriguing examples of changes that are consistentwith altered stability, trafficking between compart-ments, or local ‘‘remodeling’’ remain.

INTRODUCTION

Neurons distinguish themselves from other cells by their extreme

structural and functional compartmentalization, accomplished

via morphologically complex dendritic and axonal arbors and

synaptic specializations. The number of synapses and the infor-

mation processing accomplished by synaptic protein networks

necessitate an efficient machinery for local information process-

ing (for review, see Hanus and Schuman, 2013). The localization

of mRNAs and regulated local translation allows neurons to

deal rapidly with the compartmentalized function of synapses,

underlying important cellular processes such as memory,

dendrite formation, synapse formation, axon guidance, survival,

and proteostasis (for review, see Holt and Schuman, 2013).

Within neurons, the number of localized mRNA transcripts

numbers in the thousands (Cajigas et al., 2012; Gumy et al.,

2011; Zivraj et al., 2010).

Within mRNA molecules, information in the 30 untranslated re-

gion (30 UTR) can regulate their targeting, translational efficiency,

and stability (Tian and Manley, 2017). As such, an understanding

of the repertoire of 30 UTR isoforms in localized neuronal tran-

scripts is required to understand the logic and function of local

and global protein synthesis in neurons. Studies in worms, flies,

and fish have examined 30 UTR heterogeneity and have

observed that many 30 UTR isoforms are differentially expressed

in a developmentally regulated and/or tissue-specific manner

(Hilgers et al., 2011; Ji et al., 2009; Mangone et al., 2010; Ulitsky

et al., 2012) and that there is an increase in the average 30 UTRlength in neuronal tissue. Similarly, in mouse and human sam-

ples, extended 30 UTRs have been documented in neuronal

tissue (Miura et al., 2013).

In order to investigate the diversity of 30 UTRs in mature

neuronal processes, we microdissected hippocampal slices

to separate the somatic and neuropil layers, giving rise to

RNA samples enriched in cell somata versus neuronal projec-

tions (Cajigas et al., 2012; Mishchenko et al., 2010). Using

30 end sequencing and high-resolution in situ hybridization, we

analyzed the 30 UTR heterogeneity of neuropil and somata

samples and found that localized transcripts, enriched in

either somata or the neuropil, are endowed with longer 30 UTRisoforms that include both novel and repeated regulatory motifs

that can drive localization, mRNA stability, and/or the potential

regulation of translation. Neuropil-localized transcripts also

exhibit longer half-lives, on average, when compared to their

somatic counterparts. Following a period of enhanced activity,

there is an alteration of 30 UTR isoforms present in each

compartment. While most of the 30 UTR plasticity was transcrip-

tion dependent, some 30 UTR isoforms are altered by a transcrip-

tion-independent mechanism, suggesting plasticity-induced

stability changes, trafficking of 30 UTR isoforms between com-

partments, or local remodeling.

RESULTS

30 End Sequencing of Hippocampal TranscriptsTo investigate the diversity of mRNA polyadenylation and

30 UTRs in the rat hippocampus, we isolated RNA after micro-

dissection of the somata and neuropil layers (s. pyramidale

Neuron 98, 1–17, May 2, 2018 ª 2018 Elsevier Inc. 1

mailto:[email protected]


(legend on next page)

2 Neuron 98, 1–17, May 2, 2018



and s. radiatum + s. moleculare) in the CA1 region (Figure 1;

Figure 2A). We purified total RNA and conducted 30 end

sequencing (MACE sequencing, see STAR Methods) of the

somata and neuropil samples, obtaining two replicates (tissue

for each replicate was obtained from nine animals; data from

replicates were highly correlated with one another) (Figure 1;

Figure S1A). In order to quantify the 30 UTR diversity of the

identified transcripts, we developed a pipeline to identify

sequencemotifs within the 30 UTRs that support polyadenylation(poly(A)-supported sites [PASS]; Figures S1B–S1H; see STAR

Methods). For each transcript identified, we counted the number

of distinct 30 UTRs (or PASS) detected. We found that the

majority of hippocampal transcripts possessed more than one

30 UTR isoform, with some transcripts exhibiting three or more

isoforms (Figure 2B).

As the hippocampal somata and neuropil samples comprise

not only neurons, but also glia, we next devised a strategy to

identify 30 UTR isoforms that arise from each of these cell types

by evaluating them in relative isolation via the preparation of cell-

type-enriched cultures (see STAR Methods). We sequenced the

30 ends of polyadenylated RNAs acquired from neuron-enriched

(AraC-treated) mature (Figure S2A) hippocampal cultures or from

glia-enriched cultures (Figure S2B). Combining the neuron- and

glia-enriched datasets, we obtained 21,603 30 UTR isoforms

from 11,301 gene loci (with 82.2% 30 UTR and 92.6% gene loci

overlap with the hippocampal slice data described above)

(Figure S2C; Table S1). Using these data, we developed a clas-

sifier to identify three groups of 30 UTR families: (1) neuron-en-

riched, (2) glia-enriched, or (3) non-enriched (Figure S2B; see

STAR Methods). Most (93.4%) 30 UTR isoform members of a

transcript family showed a faithful enrichment, e.g., all 30 UTRisoforms were enriched in a single (neuronal- or glial-) cell type.

There was, however, a very small subset of transcript families

in which 30 UTR isoforms were split between neurons and glia,

e.g., one 30 UTR isoform enriched in neurons and one enriched

in glia (6.6% of all enriched genes; Figure S2D); these transcripts

were not analyzed further but are identified in Table S1. These

data suggest that most of the transcript diversity between neu-

rons and glia is due to differences in the genes expressed rather

than differences in the 30 UTR isoforms of the same genes. We

validated the relative enrichment and de-enrichment of neural-

andglia-related genes, respectively, usingqRT-PCR (FigureS2E)

and bioinformatically using published datasets (Figures S2F and

S2G). We then evaluated the number of 30 UTR isoforms that

each group possesses and found that neuron- and glia-enriched

transcripts exhibited a significantly higher number of 30 UTR

Figure 1. Scheme of the Study’s Experimental Flow

Description of boxes, starting at the top left and then following the direction indi

(i) Scheme of hippocampal slice showing the regions that were microdissected for

30 UTR isoforms were identified. (iii) The position and relative fraction of the poly(A

of 30 UTR isoforms per transcript family was evaluated. (v) The 30 UTRs enriched i

enriched cultures. (vi) The number of 30 UTR isoforms in cell-type-enriched samp

evaluated. (viii) The localization and enrichment of different 30 UTR isoforms in th

isoforms were evaluated. (x) Changes in 30 UTR isoforms induced by elevated neu

Transcription-dependent and -independent changes in 30 UTR isoforms were dis

changes in 30 UTR isoform: local remodeling (xii) and trafficking/transport betwe

See also Figure S1 and Table S1.

isoforms than the non-enriched transcripts, with neuron-en-

riched transcripts exhibiting the highest number of 30 UTR iso-

forms overall (Figure 2C).

To examine whether 30 UTR diversity is related to protein func-

tion, we used gene ontology (GO) to examine the functional cat-

egories comprising the neuron- or glia-enriched 30 UTR isoforms

and compared these functional categories to the groups of tran-

scripts represented by multiple 30 UTRs or single 30 UTRs as

determined in our analysis of 30 UTRs in hippocampal slices (Fig-

ure 1). Hierarchical clustering of significant functional GO cate-

gories revealed that the family of neuron-enriched transcripts

(closely related to multiple 30 UTR hippocampal slice transcripts;

Figure 2D) was significantly enriched for genes that code for

neuronal and synaptic function (Figure 2E; Table S2). Glia-en-

riched transcripts (with overlap with single 30 UTR transcripts

and, to a lesser extent, multiple 30 UTR transcripts; Figure 2D)

showed enrichment for terms associated with the extracellular

space and adhesion (Figure 2E). Single 30 UTR variants, by

contrast, were enriched for general cellular functions, including

components of the extracellular matrix and basic cellular func-

tions (Figure 2E).

Localization of Neuronal Transcripts in Somataand NeuropilWe used the neuron-enriched transcript dataset identified above

to search for 30 UTR isoforms enriched in the microdissected

somata or neuropil of area CA1 from hippocampal slices. We

found that both the neuropil and the somata contain hundreds

of 30 UTR variants that are significantly enriched (Figure 3A);

these transcripts are in good agreement with previously pub-

lished data (Figure S3A). All 30 UTR isoforms detected and their

relative expression in the somata and neuropil are shown in

Table S1. We used qRT-PCR to validate the enrichment and

de-enrichment of 22 identified transcripts in each compartment

(neuropil or somata) and found very good agreement (R2 = 0.966)

with the sequencing data (Figure 3B). We also used GO to

analyze the somatic- and neuropil-enriched transcripts and

found a significant enrichment of terms associated with axon-,

dendrite-, or synapse-related functions in both compartments

(Figure 3C; Figures S3B and S3C; Table S2). Some somatically

localized transcripts could code for membrane proteins as the

bulk of their synthesis and post-translational modifications might

occur in the abundant somatic endoplasmic reticulum (ER) and

Golgi (Hanus et al., 2016). Indeed, we found that membrane

proteins were often, but not always, represented by somatically

enriched transcripts, whereas cytoplasmic proteins were often

cated by the arrows. Relevant figures are mentioned as indicated.

subsequent 30 end sequencing. (ii) Filtered reads were aligned to genome, and

)-supported sites (PASS) within the gene body were evaluated. (iv) The number

n glia or neurons were evaluated using additional 30 sequencing from cell-type-

les was evaluated. (vii) The regulatory motifs in neuron-enriched 30 UTRs were

e neuropil or somata were evaluated. (ix) The half-lives of the different 30 UTRral activity were evaluated and their transcriptional dependence assessed. (xi)

covered. Two mechanisms were considered for the transcription-independent

en compartments (xiii).

Neuron 98, 1–17, May 2, 2018 3

Figure 2. Neuronal Genes Are Enriched for Multiple 30 UTR Isoforms

(A) Scheme of 30 end mRNA sequencing (30 end seq) in the microdissected CA1 region of the rat hippocampus. 30 end seq was conducted separately for the

somata and the neuropil layers, and 30 UTR isoforms were annotated following alignment of reads to the genome and detection of poly(A)-supported sites (PASS).

(B) Relative frequency (percentage) of mRNA 30 UTR isoforms per transcript for all genes (neuropil and somata combined) that passed the data filter criteria (see

STAR Methods).

(C) Bar plot of the relative frequency of 30 UTR isoform numbers per gene for neuron-, glia-, and non-enriched genes (see STAR Methods). For comparison, the

relative frequency of 30 UTR isoforms for all transcripts (overall) is also shown. ***p < 0.001, chi-square test.

(D and E) Hierarchical clustering of significant functional GO categories for transcripts with multiple 30 UTRs and single 30 UTRs as well as for neuron-, glia-, and

non-enriched transcripts. Boxes in (D) indicate GO term clusters with specific enrichment patterns that are shown in (E). Values in (E) indicate –log10 p values of

GO term enrichment.



represented by neuropil-enriched isoforms (Figure S3D).We also

examined the relevance of localized 30 UTR isoforms for protein

families associated with neurodevelopmental, neuropsychiatric,

or neurodegenerative diseases and found a significant enrich-

ment for many diseases (Figure S3E; Table S2).

We next analyzed the length of the 30 UTR isoforms and

found that localized 30 UTR isoforms, enriched in either the

neuropil or the somata, possessed significantly longer

30 UTRs than the non-enriched transcripts, with neuropil

30 UTRs exhibiting the longest 30 UTRs reported to date (Fig-

ure 3D; Figure S3F). Furthermore, neuron-enriched transcripts

tend to have longer 30 UTRs than glia-enriched transcripts,

while both of these families have significantly longer 30 UTRsthan non-enriched isoforms (Figure 3D). Using published

data (Derti et al., 2012; Grillo et al., 2010), we analyzed

30 UTRs across tissues and species and discovered that

30 UTRs are significantly longer for transcripts expressed in

brain relative to other tissues, as observed by others (Hilgers

et al., 2011; Miura et al., 2013; Smibert et al., 2012), and

also noted an increase in 30 UTR length from plants to pri-

4 Neuron 98, 1–17, May 2, 2018

mates (Figure 3D). In contrast to a previous study of devel-

oping neurites or cell lines (Taliaferro et al., 2016), we found

that neuropil-enriched 30 UTRs were not enriched for mRNAs

containing alternative last exons (ALEs) (less than 13% of all

predicted sites; see STAR Methods). Taken together, these

data suggest that longer 30 UTRs are necessary platforms

that might favor both the asymmetric distribution of transcripts

within cells and enhanced regulatory potential.

Significantly enriched among the neuropil-localized long

30 UTRs are transcripts that code for synaptic proteins (Fig-

ure S3C; Table S2). In order to validate the localization of long

30 UTRs in the neuropil, we chose 19 representative transcripts,

which express multiple 30 UTRs as indicated by our sequencing

data. Eight of these transcripts exhibited enriched expression of

the long 30 UTR in the neuropil and 11 exhibited enriched expres-

sion of the long 30 UTR in the somata. We developed in situ

hybridization probes against the unique region of these long

30 UTR isoforms and examined their localization in cultured hip-

pocampal neurons and hippocampal slices (Figures 4A–4F; Fig-

ures S4A–S4D). As shown in Figures 4A–4G and S4, the in situ

Figure 3. Differential Expression of 30 UTRs in Neuronal Compartments, 30 UTR Length, and Protein Functions of Localized Transcripts

(A) MA plot (the average, A, of the log read counts versus the differences in the log read counts, minus, M) showing differentially expressed 30 UTR isoforms

between neuropil and somata. All 30 UTR isoforms are plotted. Red and blue dots indicate 30 UTR isoforms significantly enriched in neuropil or somata,

respectively (edgeR library, with a threshold of 0.05 on the adjusted p value; see STAR Methods). Gray dots represent non-enriched isoforms.

(B) Validation of 30 UTR sequencing data showing the correlated enrichment of somata- (blue) and neuropil- (red) localized 30 UTR isoforms between qRT-PCR

and sequencing measurements. Error bars indicate the standard error of n = 3.

(C) GO terms representing significantly enriched protein function groups for localized (somata and neuropil) and non-localized 30 UTR isoforms.

(D) Average 30 UTR length distributions in different compartments (localized to the neuropil or somata layer compared to non-localized genes), cell types, tissues,

and species.



hybridization signal is strongest for the long 30 UTR isoforms in

the expected compartment (neuropil for CamKIIa and Calm1;

somata forGrin2a; Figure 4) in both dissociated neurons and hip-

pocampal slices. Taken together, the neuropil-enriched 30 UTRisoforms exhibited a significantly higher dendritic in situ hybridi-

zation signal than the somata-enriched 30 UTR isoforms (Fig-

ure 4H; Figure S4C).

We next asked if a 30 UTR isoform of a transcript exhibits local-

ization in either (neuropil or somata) compartment, how do the

other isoforms of the same transcript family behave? To address

this, we plotted the relative expression of short and long isoforms

in the somata and neuropil compartments and evaluated the

prevalence of two different scenarios: the differential expression

of short and long 30 UTR isoform variants in somata versus neu-

ropil (e.g., the short enriched in one compartment while the long

is enriched in the other, or vice versa; Figure 5A, top left and bot-

tom right quadrants) or the coordinate enrichment of both short

and long 30 UTR isoform variants in a compartment (e.g., both the

short and the long enriched in either the soma or the neuropil;

Figure 5A, top right and bottom left quadrants). In the scatterplot,

we observed a positive and significant correlation between the

localization biases of different 30 UTR isoforms of the same

gene (R = 0.3040; p = 3.10E–128), indicating a coordinate enrich-

ment of both long and short isoforms within compartments

(Figure 5A). There exists, however, evidence for the differential

enrichment of isoform variants in different compartments, with

Neuron 98, 1–17, May 2, 2018 5

Figure 4. Validation of 30 UTR IsoformDistri-

butions in Somata and Neuropil Compart-

ments

(A–C) Genome browser views of the sequence

coverage for the different 30 UTR isoforms of

CamKIIa (A), Calm1 (B), and Grin2a (C).

(D–F) Fluorescence in situ hybridization (FISH)

signal in cultured hippocampal neurons (upper

panels) or hippocampal slices (lower panels) using

probes designed to detect the unique sequence of

the long 30 UTR of CamKIIa (D), Calm1 (E), and

Grin2a (F). The mRNA signal in hippocampal cul-

tures was dilated for visualization purposes only.

Dendrites were immunostained with an anti-MAP2

antibody (purple). Scale bars represent 15 and

25 mm for upper and lower panels, respectively.

(G and H) Analysis of in situ hybridization signals in

the dendrites. Violin plots depicting the median

distance to the cell body’s center of mass for the

long 30 UTRs of CamKIIa, Calm1, andGrin2a (G) or

for neuropil- and somata-enriched isoforms (H). n

(number of cells imaged and analyzed and in pa-

rentheses the number of dendritic puncta detected

and analyzed) = 32 (2,013), 29 (845), and 18 (57) for

CamKIIa, Calm1, and Grin2a, respectively. n = 8

(4,967) and 11 (837) for neuropil- and somata-en-

riched genes. ****p < 0.0001, Mann-Whitney U test.

See also Figure S4.


a greater number of transcript families sorting their longest iso-

form to the neuropil rather than to the somata (Figures 5A and

5B). This is illustrated in Figure 5A by a comparison of the number

of paired isoforms in the top left quadrant (274, long enriched in

neuropil and short enriched in somata) and the lower right quad-

rant (122, short enriched in neuropil and long enriched in somata)

as well as the cumulative frequency distribution (Figure 5B). We

evaluated whether the neuropil-localized 30 UTR isoform distri-

bution patterns were enriched for any particular functions using

GO and found a significant enrichment for postsynaptic density

and membrane, dendritic spine, memory, and other groups (Fig-

ure S5A; Table S2). We determined the sufficiency of a long

30 UTR to drive localization of a transcript to the neuropil by clon-

6 Neuron 98, 1–17, May 2, 2018

ing a series of reporter constructs in

which the coding sequence for GFP was

flanked by the three different CamKIIa 30

UTRs, the long 30 UTR for Kcnab2 (en-

riched in the neuropil; Figure 3B; Table

S1), or no 30 UTR as a control. Following

transfection of the reporters into neurons,

we used fluorescence in situ hybridization

(FISH) with probes anti-sense to the GFP

coding sequence to detect the mRNA in

neuronal somata and dendrites (Fig-

ure 5C). We found that the long CamKIIa

30 UTR resulted in the highest number of

transcripts detected throughout the den-

dritic arbor relative to somata; the short

and the middle CamKIIa 30 UTRs resultedin mRNA signal only in the proximal

(approximately <40 mm) segment of the

dendrite (Figure 5D). The long Kcnab2 30 UTR isoform was also

sufficient to drive the localization of the reporter to dendrites

(Figures S5B–S5D).

Properties and Elements within Neuronal 30 UTRsWe analyzed the properties and elements of the diverse 30 UTRsdescribed above, focusing on the localized (neuropil or somata)

versus non-localized transcripts. We first examined sequence

conservation along the length of 30 UTRs of genes with multiple

or single 30 UTRs and compared it to other genomic regions.

We found that brain 30 UTRs are more conserved than intergenic

sequences and introns but are less conserved than exons (Fig-

ure S6A). We next analyzed the guanine-cytosine (GC) content

Figure 5. Differential Expression of 30 UTR Isoform Family Members in Neuronal Compartments

(A) Relative expression of the longer and the shorter isoform of the same transcript family in the neuropil and somata. The schemes within each quadrant illustrate

the distribution of the longer (purple) and shorter (gray) isoform of a given 30 UTR isoform family between neuropil and somata as follows: upper left quadrant, short

isoform enriched in somata and long isoform enriched in neuropil; upper right quadrant, both the long and the short isoforms enriched in neuropil; bottom left

quadrant, both the long and the short isoforms enriched in the somata; and bottom right quadrant, the long isoform enriched in the somata and the short isoform

enriched in the neuropil.

(B) Cumulative distribution of longer (purple) and shorter (gray) 30 UTR isoforms in somata and neuropil. Inset shows boxplot of the observed relative fraction of

longer and shorter isoforms in neuropil and somata. Dashed line indicates expected distribution of longer and shorter isoforms obtained by chance. ***p < 0.001,

Kolmogorov-Smirnov test.

(C) Scheme shows GFP reporter construct using the coding sequence of GFP flanked by the short, middle, or long 30 UTR of CamKIIa. FISH was conducted for

GFP. Images show representative straightened dendrites of transfected neurons. Left, proximal to cell body; right, distal. GFP protein is shown in purple and

mRNA in white. Scale bar, 10 mm.

(D) Analysis of in situ hybridization signals shown in (C). Violin plots depicting the mRNA signal localization along the dendrite normalized by the average soma

intensity. n R 23, ****p < 0.0001, Mann-Whitney U test.



and secondary structure element predictions and observed that,

in comparison to non-localized transcripts, somata-enriched

transcripts possess a lower GC content and less predicted sec-

ondary structure, whereas neuropil-enriched transcripts exhibit

the opposite trend with significantly higher GC content and pre-

dicted secondary structure (Figure 6A). As expected, brain

30 UTRs, in general, also exhibit higher GC content and predicted

secondary structure than intergenic regions (Figures S6B and

S6C). The structure stability evident in the neuropil 30 UTRs likelyprovides binding platforms for RNA-binding proteins to regulate

the localization, translation, and stability (see below) of these

transcripts.

Do the nucleotide sequences added during 30 UTR extension

provide new or redundant regulatory potential for the upstream

transcript? To answer this question, we analyzed all 30 UTR iso-

forms in transcript families withmultiple 30 UTRs for the presenceof 7-mers (see STAR Methods). The presence of these potential

regulatory motifs were then categorized as follows: (1) ‘‘novel,’’ a

motif that is present once and exclusively in the unique sequence

of the long 30 UTR; (2) ‘‘repeat between,’’ a regulatorymotif that is

already present in the shorter 30 UTR isoform and is repeated in

the unique sequence of the longer 30 UTR isoform; or (3) ‘‘repeat

within,’’ a regulatory motif that is introduced and repeated exclu-

sively within the unique sequence of the long 30 UTR isoform

Neuron 98, 1–17, May 2, 2018 7

Figure 6. Regulatory Potential of Localized 30 UTR Isoforms

(A) GC content (left) and secondary structure element prediction (normalized minimum free energy, MFE) of non-, somata-, and neuropil-localized 30 UTRs. TheGC content of somata- and neuropil-localized isoforms was significantly de-enriched or enriched, respectively, relative to non-localized 30 UTR isoforms. The

MFE of neuropil-localized transcripts was significantly lower than either somata- or non-localized 30 UTRs. ****p < 0.0001, Mann-Whitney U test.

(B) Scheme below the plot shows detected regulatory elements that were categorized into novel (square) representing a single introduction of 7-mer in the unique

part of the long 30 UTR, repeat between (circle) representing a duplicate of a motif present in the shared 30 UTR isoform of a family member, and repeat within

(triangle) representing an element introduced and repeated within the unique part of the long 30 UTR. Graph shows the relative fraction of 7-mers (includingmiRNA

binding sites) that fall into the categories: novel, repeat within, or repeat between. To assess expected distribution of these fractions, we repeated the analysis for

intergenic regions (1 kB downstream of the poly(A) site) and dinucleotide shuffled unique 30 UTR isoforms.

(C) Relative expression and significant enrichment of 7-mers in somata and neuropil. 18.4% of all 7-mers show a compartment-specific enrichment in either the

neuropil (red dots) or the somata (blue dots). Also highlighted are the establishedmotifs interacting with the RNA-binding proteins CPE, FMRP, Mbnl1, and Zbp1.

(D) Relative expression and significant enrichment of miRNA binding sites in somata and neuropil. 22% of all binding sites for miRNAs expressed in area CA1

show a compartment-specific enrichment (neuropil, red dots; somata, blue dots) in a 30 UTR isoform.

See also Figure S6 and Tables S3 and S4.


(Figure 6B). As such, novel and repeat-within elements add new

regulatory potential to a transcript. Both repeat-between and

repeat-within elements increase the likelihood of regulation by

competing for a limited pool of regulatory binding elements

(e.g., microRNAs [miRNAs], RNA-binding proteins, etc.) with

repeat within adding this potential exclusively to the long

30 UTR isoform. We found that the dominant type of regulatory

element in a longer isoform was in the novel category, although

the frequency of novel motifs was less than predicted by chance

alone, suggesting an active deselection of these motifs (Fig-

ure 6B). By contrast, the repeat elements, although low in total

8 Neuron 98, 1–17, May 2, 2018

number, were present at significantly higher levels than ex-

pected by chance or by comparison with intergenic sequences

(Figure 6B). This general pattern was also observed when we al-

lowed for a mismatch within the group of significant 7-mers (Fig-

ure S6D). Taken together, these analyses indicate that additional

30 UTR elements add new possibilities for regulation but also

significantly increase regulatory motifs already present, poten-

tially increasing the ‘‘regulatory competitiveness’’ of an mRNA

isoform. We also considered whether differences in secondary

and tertiary structural elements exist between 30 UTR isoforms.

We did not, however, detect any significant differences in GC

(legend on next page)

Neuron 98, 1–17, May 2, 2018 9



content or minimum free energy for the short, middle, or long

30 UTR isoforms (Figures S6B and S6C).

We next addressed whether there are specific 7-mers that are

overrepresented in 30 UTR isoforms localized to the somata or

neuropil. We found 3,017 7-mers, representing 215 families

(allowing an edit distance of 2; see STAR Methods), enriched

in localized transcripts relative to non-localized transcripts,

with 1,637 (138 families) and 1,380 (126 families) 7-mers en-

riched in the neuropil- and somata-localized 30 UTR isoforms,

respectively (Figure 6C; Table S3). We searched for previously

identified consensus binding motifs for several well-known

neuronal RNA-binding proteins and found (Table S4), for

example, that the two motifs identified for the zip-code-binding

protein (Zbp1; Kim et al., 2015) are enriched in neuropil-localized

30 UTRs (Figure 6C). Themotifs for two other neuronal RNA-bind-

ing proteins, fragile X mental retardation protein (FMRP; Ander-

son et al., 2016; Ascano et al., 2012) and muscle-blind (Mbnl1;

Konieczny et al., 2014; Wang et al., 2012), were enriched in

distinct somatic- and neuropil-localized 30 UTRs. The cyto-

plasmic polyadenylation element (CPE) is recognized by the

CPE-binding protein (CPEB) and usually promotes the cyto-

plasmic polyadenylation of mRNAs (de Moor and Richter,

1999) important, for example, for the chemotropic responses in

axons (Lin et al., 2009). Surprisingly, we found that the CPE

motifs were significantly enriched in somata-localized 30 UTRsrather than neuropil 30 UTRs, although we note that several

neuropil 30 UTRs do contain a CPE, for example, CamKIIa

(Figure S5E), as noted by Wu et al. (1998).

We conducted a similar analysis seeking to identify miRNA

seeds present in 30 UTR isoforms (see STAR Methods). In total,

we identified 298 miRNA seed sequences present in the somata

and neuropil (Figure 6D; Table S3). To illustrate the abundance

and diversity of miRNA elements within a single 30 UTR, we

plotted the miRNA seeds and other regulatory motifs within the

CamKIIa complete 30 UTR (Figure S5E). Using Nanostring, we

next addressed whether the specific miRNAs that bind to the

identified seeds are expressed in the somata and neuropil. We

Figure 7. Neuropil-Localized 30 UTR Isoforms Possess Significantly Lo

(A) Approach used to measure half-lives of neuronal 30 UTR isoforms. Samples we

30 end seq. Half-lives were obtained by fitting an exponential degradation curve

(B) Gene set enrichment analysis (GSEA) of neuronal 30 UTR half-lives showing e

(C) Cumulative probability plot showing the distribution of half-lives for neuron

significantly longer lived. *p < 0.05, Kolmogorov-Smirnov test.

(D) Boxplot of the half-life of single, short, middle, and long 30 UTR isoforms (left) a

(right). **p < 0.01, Mann-Whitney U test.

(E) Cumulative distribution function depicting the expected and observed distrib

isoform; the observed half-lives show a rightward shift, indicating a significantly

gorov-Smirnov test.

(F) Motif analysis of stabilizing and destabilizing regulatory elements. Shown are

half-lives, respectively, of 30 UTR isoforms. Highlighted are previously published

(G) Motif analysis of stabilizing and destabilizing miRNA-associated regulatory e

cantly longer or shorter half-lives, respectively, of 30 UTR isoforms.

(H) Approach used to validate the destabilizing effect of somemiRNAs identified in

30 UTR (either containing or lacking a destabilizing miRNA binding site). Transcrip

normalization (see STAR Methods).

(I) Scatterplot showing reporter expression containing the miRNA binding site (re

qRT-PCR. Reporter expressions were normalized to Renilla expression. Data ar

significant (ns), one-way ANOVA followed by Bonferroni’s multiple comparisons

See also Figure S7 and Tables S2 and S3.

10 Neuron 98, 1–17, May 2, 2018

observed that 191 out of the total 423miRNAs that aremeasured

with Nanostring are expressed in the hippocampal CA1 region.

Furthermore, 42 of the detected miRNAs have binding sites

that are significantly enriched in the somata (20) or neuropil

(22) (Figure 6D).

Neuropil Localized Transcripts Are Long LivedDifferences in mRNA half-life might also play an important role in

both the localization and the function of neuropil mRNAs. To

determine the half-lives of neuronal transcript 30 UTR isoforms,

we blocked transcription (e.g., Akbalik et al., 2017) and then

collected cultured hippocampal neuron samples at successive

time points (0 hr, 2 hr, 4 hr, 9 hr, and 16 hr) (Figure 7A). We per-

formed 30 end sequencing on these samples and, after normali-

zation (see STAR Methods and Figures S7A and S7B), fit an

exponential decay function to the time course data of each 30

UTR isoform measured. This allowed us to determine the half-

life of each 30 UTR isoform, yielding a median transcript half-

life of 7.38 hr (Figure 7A; Figure S7C). To determine whether

some groups of biologically related transcript isoforms are

particularly long or short lived, we performed a gene set enrich-

ment analysis (GSEA) on the half-life ranked list of 30 UTR iso-

forms (Figure 7B). Interestingly, we found that the long-lived 30

UTR isoforms are enriched for GSEA terms representing den-

dritic, axonal, or synaptic protein functions, while the GSEA

terms for the less stable transcript groups were enriched for

genes whose products primarily function in the nucleus or nucle-

oplasm, such as transcription factors (Figure 7B; Table S2). We

then examined the relative half-life of synaptic genes (Zhang

et al., 2007) compared to all other neuron-enriched transcripts

and found that synaptic 30 UTR isoforms exhibited significantly

longer half-lives (Figure 7C). We also analyzed the extent to

which the properties of the 30 UTR isoforms described above

(single 30 UTR versus a short, middle, or long family member,

localized or not) affect transcript half-life. Although the different

categories of 30 UTR isoforms have similar median half-lives,

transcripts localized to the neuropil have significantly longer

nger Half-Lives than Somata- or Non-localized 30 UTR Isoforms

re collected at different time points post-transcriptional block and subjected to

(see STAR Methods).

nriched and de-enriched protein function groups as a function of half-life.

al-enriched transcripts and synaptic transcripts; the synaptic transcripts are

s well as the half-life of neuropil-, somata-, and non-localized 30 UTR isoforms

ution of 30 UTR isoform half-lives for transcript families with a short and long

longer half-life for the long 30 UTR isoforms in families. ****p < 0.0001, Kolmo-

significant (blue and red) 7-mers associated with significantly longer or shorter

stabilizing (yellow) and destabilizing (blue) motifs.

lements. Shown are significant (blue and red) miRNAs associated with signifi-

(G). miRNAs were co-transfected with a luciferase reporter with the respective

t levels were analyzed using qRT-PCR. Renilla transcript levels were used for

d) relative to the reporter lacking the miRNA binding site (gray) determined by

e means + SD from six independent experiments. **p < 0.001; *p < 0.05; not

test.

Figure 8. Elevated Neuronal Activity Results in Compartment-Specific Plasticity of 30 UTR Isoforms

(A) Change in 30 UTR isoform expression in neuropil (y axis) versus somata (x axis) following 4 hr of elevated activity (elicited by bicuculline 40 mM). In total, 783 30

UTR isoforms exhibited significant alterations in expression following enhanced activity. In some cases (top left and bottom right quadrants), there was differential

plasticity of 30 UTR isoforms between the somata and the neuropil compartments, whereas in other cases (top right and bottom left quadrants), there was

coordinate regulation in the two compartments. Upregulated 30 UTR isoforms are shown in purple, and downregulated 30 UTR isoforms are shown in gray.

(B) GO terms representing significantly enriched protein function groups for upregulated and downregulated 30 UTR isoforms following elevated activity.

(legend continued on next page)

Neuron 98, 1–17, May 2, 2018 11



half-lives when compared to somata- and non-localized mRNAs

(Figure 7D; Table S1). Following from the above observation,

within a transcript family, the longer 30 UTR isoform possesses

longer half-lives (Figure 7E). We also observed that the half-lives

of short and long isoforms of the same transcript family are not

correlated (Figure S7D). Furthermore, we observed no significant

relationship between the stability of a given 30 UTR isoform and

its structural stability, GC content, level of expression, or length

(Spies et al., 2013).

To determine whether specific sequences influence mRNA

half-life differences in our data, we searched for k-mers signifi-

cantly associated with either shorter- or longer-lived 30 UTR

isoforms. We identified 464 motifs (comprising 121 families,

allowing an edit distance of 2) significantly associated with

shorter 30 UTR half-lives and 344motifs (105 families) associated

with longer half-lives (Figure 7F; Table S3). Previously identified

(Ray et al., 2013; Spies et al., 2013) destabilizing and stabilizing

motifs were also detected and in good correspondence with our

data (Figure 7F).

Similarly, we identified miRNA seed sites within 30 UTRs that

were also significantly associated with either shorter- or longer-

lived transcript isoforms (Figure 7G; Table S3). To validate the

efficacy of some miRNA seed sites in shortening transcript half-

lives, we constructed firefly luciferase 30 UTR reporter constructs

possessing or lacking the seed sites for four different destabilizing

miRNAs (miR-425, miR-146a, miR23a, and miR-154) (Figure 7H).

These constructs, together with their cognatemiRNA, were trans-

fected into cells, and transcript levels were assessed using qRT-

PCR. In three of the four cases examined, the presence of the

miRNA seed togetherwith themiRNAwas sufficient to destabilize

the reporter construct when compared to a 30 UTR lacking the

miRNA seed (Figure 7I). Taken together, our data indicate that

neuropil-localized transcripts are endowed with longer half-lives

and that both somatic-localized and neuropil-localized 30 UTRsare enriched with specific regulatory elements, including miRNA

seed sites, that are important for localization and stability.

30 UTR Plasticity Following Neural ActivityWe next determined whether 30 UTR diversity is altered by

enhanced activity in neural networks. To stimulate neural activ-

(C) Cumulative distribution function depicting the change in 30 UTR isoform length

compartments underwent a significant shortening. ****psomata versus expected < 0.0

(D) Change in 30 UTR isoform expression in neuropil (y axis) versus somata (x a

Transcription inhibition blocked�75%of the 30 UTR isoform plasticity, which is ind

169 30 UTR isoforms exhibited significant coordinate or compartment-specific ch

(E) Implosion index indicating the distance of the data points shown in (A) and (D)

distance to the origin, indicating that the majority (�75%) of the isoform plasticity

U test.

(F) Flow chart indicating the number and percentage of 30 UTR isoforms that chan

and their transcriptional dependence or independence. The example of 30 UTRtransport of 30 UTR isoforms between the somata and the neuropil or the local re

(G) Boxplot showing the transcription-dependent and -independent 30 UTR isofo

control (set to 0) and bicuculline-treated conditions. Red boxes show the averag

conditions in the presence of transcriptional inhibition.

(H) 30 UTR isoform plasticity consistent with transport between the somata and t

neuropil) are accompanied by transcription-independent decreases in compartm

(I) 30 UTR isoform plasticity is consistent with local remodeling in the somata and t

are accompanied by transcription-independent decreases in the longer isoform.


12 Neuron 98, 1–17, May 2, 2018

ity, we treated hippocampal slices with the GABAA receptor

antagonist bicuculline (40 mM, 4 hr). Following enhanced activity,

hippocampal slices were microdissected to obtain the somata

and neuropil fractions that were then subjected to 30 end

sequencing. We found that increased neural activity led to signif-

icant alterations in 30 UTR isoforms evident in both compart-

ments (Figures S8A and S8B). Enhanced neural activity led to

the upregulation of 367 and 408 30 UTR isoforms and the down-

regulation of 416 and 375 isoforms in the somata and neuropil,

respectively (Figure 8A). Within this dataset, there were exam-

ples of coordinate regulation of the same 30 UTR isoform in

both compartments (Figure 8A, bottom left and top right quad-

rants) as well oppositional regulation in the compartments

(Figure 8A, top left and bottom right quadrants). We used a GO

analysis to evaluate the potential enrichment of protein function

groups represented by altered (up or downregulated) 30 UTRsfollowing bicuculline (Figure 8B; Table S2). Protein groups that

were significantly enriched for downregulated 30 UTR isoforms

included membrane proteins and protein-kinase-associated

proteins, for example. Significantly enriched protein groups

represented by both up- and downregulated 30 UTRs included

long-term memory and ionotropic glutamate receptor binding

(Figure 8B). When analyzed at the population level, we found

that elevated activity resulted in a global and significant short-

ening of the 30 UTR isoforms in both compartments, especially

pronounced for the neuropil-localized isoforms (Figure 8C).

Approximately 14% of the 30 UTRs that were regulated and

18% of the 30 UTRs that were shortened by bicuculline were

also regulated by a chemical long-term potentiation (LTP) proto-

col in a recent study (Fontes et al., 2017); this degree of overlap is

not significantly higher than one would predict by chance

(Figure S8E). We validated the activity-dependent changes in

30 UTRs for several transcripts in each compartment using

qRT-PCR (Figure S8F).

Elevated neural activity is known to elicit changes in gene tran-

scription and polyadenylation site choice (Flavell et al., 2008). As

such, we reasoned that the activity-induced changes in 30 UTRisoforms are most likely the result of regulated alternative poly(A)

site choice following activity-induced changes in transcription. In

order to examine the dependence of the above changes on

following elevated activity. 30 UTR isoforms in both the somata and the neuropil

01, **pneuropil versus somata < 0.01, Kolmogorov-Smirnov test.

xis) following bicuculline treatment in the presence of transcription inhibitors.

icated by the cloud of black dots that cluster around the origin (0,0) of the axes;

anges in 30 UTR isoforms that were resistant to transcription inhibition.

to the origin. Treatment with the transcription inhibitor significantly reduced the

arises from a transcription-dependent process. ****p < 0.0001, Mann-Whitney

ge with enhanced activity, the direction of their change (up- or downregulated),

isoform shortening is considered further, and patterns consistent with either

modeling within a compartment are evaluated.

rm changes. Blue boxes show the average change in a 30 UTR abundance in

e change in a 30 UTR abundance in control (set to 0) and bicuculline-treated

he neuropil. Transcription-independent increases in compartment A (e.g., the

ent B (e.g., somata).

he neuropil. Transcription-independent increases in the shorter 30 UTR isoform


transcription, we blocked transcription during the enhanced ac-

tivity and again conducted 30 end sequencing (Figures S8C and

S8D). As shown in Figures 8D and 8E, the majority (76.6%) of the

activity-induced changes in 30 UTR isoform usage were pre-

vented by transcription inhibition, suggesting a role for transcrip-

tion dependent alternative polyadenylation. Therewas, however,

a substantial population of compartment-specific 30 UTR isoform

plasticity that persisted despite transcriptional inhibition (Figures

8D, 8F, and 8G). These transcription-independent (activity-

dependent) changes might be the result of altered stability, traf-

ficking of 30 UTR isoforms between the two compartments

(somata and neuropil), or local remodeling (e.g., shortening or

lengthening) of 30 UTR isoforms within a compartment. The me-

dian half-life of our transcripts (�7.5 hr) is well beyond the dura-

tion of our activity manipulation, making it unlikely that extending

the half-life can account for increases in most 30 UTR isoforms,

although reductions in 30 UTR isoforms could be accounted for

by enhanced turnover. We further considered the other two pos-

sibilities, regulated trafficking or local remodeling of 30 UTR iso-

forms, with additional criteria to examine their feasibility as ‘‘sim-

ple’’ mechanisms for transcription-independent changes in local

30 UTR composition (Figure 8F; Figure S8G). For the case of

regulated trafficking of 30 UTR isoforms between compartments,

we reasoned that an increase (or decrease) in an isoform in one

compartment (e.g., the neuropil) should be associated with a

decrease (or increase) in that isoform in the other compartment

(e.g., the somata). Applying these criteria, we found 32 (49.2%;

21 in the neuropil and 13 in somata) cases of 30 UTR isoform

regulation consistent with inter-compartmental trafficking (Fig-

ures 8F and 8H). For the case of local remodeling of a 30 UTRisoform, we consider the example case in which there is a tran-

scription-independent increase in a short isoform. We reasoned

that, for this phenotype to be consistent with local remodeling

(e.g., shortening), there should be a commensurate depletion

in any longer isoform family member in the same compartment.

Applying these criteria, we found 22 cases (38.8%; 17 in the neu-

ropil and 5 in somata) of 30 UTR isoform regulation consistent

with local remodeling (Figures 8F and 8I). Taken together, these

data indicate that neural activity can drive 30 UTR isoform

changes in specific compartments; while most (76.6%) of these

changes require transcription, some 30 UTR changes may occur

in the cytoplasm and involve the differential trafficking or the

local remodeling (e.g., shortening) of 30 UTR isoforms.

DISCUSSION

We report, for the first time, the 30 UTR complexity in a mamma-

lian brain tissue, the hippocampal CA1 region, containingmature

neurons with fully developed synapses. We examined the diver-

sity of 30 UTRs in (hippocampal) brain mRNAs to discover differ-

ences that influence the localization, stability, and potential

translational regulation of different mRNA isoforms. We micro-

dissected hippocampal slices and used RNA samples enriched

in cell somata or axons and dendrites (a.k.a. the neuropil; Cajigas

et al., 2012; Mishchenko et al., 2010) for 30 end sequencing. As

previously reported (Derti et al., 2012; Elkon et al., 2013; Shi,

2012; Tian et al., 2005), we found that more than 50% of all

genes and more than 70% of the neuron-enriched genes

express multiple 30 UTR isoforms. Hundreds of 30 UTR mRNA

isoforms show enrichment in either the neuropil or the somatic

compartment. We found that localized transcripts, whether en-

riched in the somata or the neuropil, possessed longer 30 UTRsthan non-localized mRNA isoforms. This is in contrast to a study

of developing neurons or a neural cell line in which no enrichment

for longer isoforms was found in very young, nascent projections

(Taliaferro et al., 2016).

The neuropil-localized 30 UTR isoforms are significantly en-

riched for proteins associated with dendritic spines and

synapses. Prominent among this category is the much-studied

CamKIIa mRNA, one of the most abundant mRNAs in dendrites

(Cajigas et al., 2012). Mayford et al. (1996) originally showed that

the CamKIIa 30 UTR in its entirety is sufficient for dendritic

targeting and then went on to show (Miller et al., 2002) that the

30 UTR is required for the localization of the mRNA and, impor-

tantly, a majority of the synaptically localized CamKIIa protein.

Others have identified distal regions (Blichenberg et al., 2001;

but see Mori et al., 2000) or elements (Huang et al., 2003; Subra-

manian et al., 2011) within the CamKIIa 30 UTR that influence

localization. Notably, the above distal regions and elements

reside in the long, neuropil-localized 30 UTR that we have identi-

fied and characterized here (Figure S5E).

Interestingly, we found that many of the somatically localized

30 UTR isoforms code for dendritic or synaptic membrane pro-

teins, which likely make use of the relatively abundant somatic

ER and Golgi apparatus (Cui-Wang et al., 2012). Perhaps the

most-studied synaptic membrane proteins are the excitatory

neurotransmitter-gated ion channels, the AMPA and NMDA re-

ceptor families. We found that all of the AMPA- and NMDA-

type glutamate receptors express multiple 30 UTRs (except

Grin2d), which under basal conditions are mostly enriched in

the somata or not enriched in either compartment (except

Grin2c, which has an isoform enriched in the neuropil; Table

S1). It is important though to distinguish between enrichment

(a significantly higher quantity present in one compartment

over the other) and presence or ‘‘localization.’’ The mRNAs and

30 UTR isoforms of these receptors are clearly present in the

neuropil, detected in our 30 end sequencing analyses here

(Figure S1) and also by in situ hybridization data shown here

(see Gria1 and -2 and Grin2a and -b puncta in the dendrites

and neuropil; Figure 4F; Figure S4) and in other publications

(e.g., Cajigas et al., 2012; Grooms et al., 2006; Swanger et al.,

2013). In addition, the local translation of some AMPA and

NMDA receptors has also been observed (Swanger et al.,

2013; tom Dieck et al., 2015). Our data show that regulation of

particular 30 UTR isoforms following plasticity can, in principle,

change the local abundance, turnover, and translatability of a

transcript. We speculate that some members of multi-subunit

synaptic membrane complexes may be produced in the soma

and distributed globally but that the final receptor complex

composition can be fine-tuned by local translation of the com-

plex member mRNAs that are enriched in dendrites.

The sequence motifs present in localized 30 UTRs, particularlythe extensions, add new regulatory potential but also exhibit a

significant overrepresentation of redundant motifs, including

miRNA seed sequences. Our previous work indicated that,

in the CA1 region of the hippocampus expressed, miRNAs

Neuron 98, 1–17, May 2, 2018 13


possess, on average, 500 potential mRNA targets (Sambandan

et al., 2017). The addition of redundant motifs may thus enhance

the competitive potential for regulation by a limited pool of indi-

vidual miRNAs. In addition, we searched for 7-mers or miRNA

seeds significantly enriched in either somata- or neuropil-local-

ized 30 UTR isoforms and detected hundreds of motifs that pro-

vide the substrate for regulated localization, stability, and trans-

lational regulation. Kim et al. (2004) previously demonstrated

that, in dissociated cortical neurons, miR-326 localizes with

polyribosomes, the active sites of translation. We found that

the binding site formiR-326was significantly enriched in the neu-

ropil-localized 30 UTRs, suggesting a role for miR-326 in the

regulation of local translation.

We measured the stability of the different 30 UTR isoforms and

found that the 30 UTR isoforms associated with synaptic proteins

are longer lived than the isoforms that code for proteins associ-

ated with transcriptional regulation, for example. In general, the

30 UTRs associated with synaptic proteins and present in the

neuropil were longer lived than other isoforms. Within transcript

families, the longer 30 UTRs were significantly longer lived than

their shorter counterparts, supporting the notion that the longer

half-life might help to establish localization to distal regions of

the dendrite or axon. This is in contrast to a study in fibroblasts,

which found no difference in stability between short and long

30 UTR isoforms (Spies et al., 2013). Interestingly, a potentially

destabilizing 30 UTRGrik4 variant has been associated with neu-

roprotection against bipolar disorder (Pickard et al., 2008). We

also identified hundreds of stabilizing and destabilizing motifs

and miRNA seed sequences, which are significantly associated

with both prolonged and reduced 30 UTR isoform half-life. The

sufficiency of several destabilizing miRNA seed sites was

validated using qRT-PCR. These destabilizing motifs have the

potential advantage of resetting the local mRNA pool via the

degradation of a population of mRNAs with a particular regulato-

ry potential and the subsequent replacement with a different

mRNA population. Alternatively, the use of 30 UTR isoforms

that are inherently unstable can lead to a temporally discrete

epoch of translation, for example, associated with the early

phase of synaptic plasticity.

We found that following an epoch of enhanced neuronal

activity, the 30 UTR isoforms in both compartments, the somata

and the neuropil, were altered, with an overall shortening of

30 UTRs observed. A shortening of 30 UTRswas recently reported

in hippocampal slices following a chemical-LTP protocol (Fontes

et al., 2017). Most of the activity-dependent changes in 30 UTRisoforms were transcription dependent, consistent with alterna-

tive polyadenylation identified as an important mechanism for

diversifying 30 UTRs in both dendrites (Fontes et al., 2017)

and developing axons (Shigeoka et al., 2016; Zhang et al.,

2016). Surprisingly, though, a substantial fraction (�25%) of ac-

tivity-induced changes in 30 UTR isoforms were independent of

transcription. The isoform- and compartment-specific changes

could be due to regulated degradation; the enhanced turnover

of 30 UTR isoforms is certainly a viable mechanism that warrants

further exploration. Alternatively, we considered two additional

possibilities for the transcription-independent changes: regu-

lated trafficking of 30 UTR isoforms between somata and neuropil

and local ‘‘remodeling’’ within a compartment. We found many

14 Neuron 98, 1–17, May 2, 2018

expression profiles consistent with trafficking (decrease of an

isoform in one compartment and the coordinate increase in the

corresponding compartment). We also found many expression

patterns consistent with local remodeling. For example, there

were many cases of 30 UTR isoform plasticity in which enhanced

expression of the shorter 30 UTR isoform was accompanied by a

decrease in the long isoform in the same compartment. The

mechanism for such shortening is unknown but would require

either the excision (e.g., splicing) of 30 UTR elements or the

degradation of nucleotides (e.g., by an exo- or endonuclease)

and the re-adenylation of the newly shortened transcript. Cyto-

plasmic splicing factors (Cajigas et al., 2012; Glanzer et al.,

2005), poly(A) polymerases, and deadenylases have been iden-

tified and, in some cases, visualized in the dendrites of neurons

(Udagawa et al., 2012). The functional consequences of

enhanced expression of short 30 UTRs will also be interesting.

Others have reported relatively enhanced translation of short

30 UTRs (Mayr and Bartel, 2009; Sandberg et al., 2008), presum-

ably owing to the loss of miRNA-mediated repression. In our

data, the local conversion (shortening) of a long 30 UTR isoform

would allow for a given isoform to be first localized (using infor-

mation present in the long isoform) but then exhibit the relatively

faster turnover and translational regulation inherent to the short

isoform.

Lastly, we note that the diversity of 30 UTRs discovered here

represents an almost entirely unexplored landscape for dysregu-

lation during neurodevelopmental, psychiatric, and degenerative

disorders. We found that disease-related proteins are signifi-

cantly more likely to make use of mRNAs with multiple

30 UTRs. Indeed, a recent study reported that the Huntingtin tran-

script, Htt, contains multiple 30 UTRs, and expression of the

shorter 30 UTR is associated with more protein aggregates and

cell death in cultured cells (Xu et al., 2017). As most disease-

related sequencing analyses have focused on protein-coding

exons and often failed to find mutations, more attention should

be focused on the 30 UTRs of these genes.

STAR+METHODS

Detailed methods are provided in the online version of this paper

and include the following:

d KEY RESOURCES TABLE

d CONTACT FOR REAGENT AND RESOURCE SHARING

d EXPERIMENTAL MODEL AND SUBJECT DETAILS

B Acute Hippocampal Slices

B Primary Hippocampal Cultures

B Primary Glia-Enriched Cultures

B Cell Lines

d METHOD DETAILS

B Tissue Microdissection, RNA Isolation, and 30 End

Sequencing

B Pharmacological Treatments

B High-Resolution In SituHybridization in Primary Hippo-

campal Neurons and Slices

B GFP Reporter Experiments

B cDNA Synthesis and Quantitative Real-Time PCR

B Transfection of MicroRNA Mimics


d QUANTIFICATION AND STATISTICAL ANALYSIS

B Image Analysis

B Detection and Annotation of poly(A) Supported Sites

B Consolidated Set of Experimental Data

B Classification of Cell-Enriched Genes

B Structural and Sequence-Specific Properties of

Neuronal 30 UTRsB miRNA Site Prediction

B K-mer Site Prediction

B Localization of Neuronal 30 UTRs in Subcellular Com-

partments

B K-mer Content

B Classifying Protein Products

B Stability of Neuronal 30 UTRsB Neuronal 30 UTR Plasticity upon Altered Neuronal Ac-

tivity

d DATA AND SOFTWARE AVAILABILITY

SUPPLEMENTAL INFORMATION

Supplemental Information includes eight figures and five tables and can be

found with this article online at https://doi.org/10.1016/j.neuron.2018.03.030.

ACKNOWLEDGMENTS

We thank Gueney Akbalik, Irina Epstein, Mantian Wang, Irena Vlatkovic, Xin-

tian You, and Wei Chen for discussions. We thank Mantian Wang for assis-

tance with microdissections. We thank Tristan Will and Ivan Cajigas for intel-

lectual contributions and pilot experiments at the beginning of the study.

Work in the laboratory of E.M.S. is supported by the Max Planck Society,

the European Research Council, DFGCRC 902 and 1080, and the DFGCluster

of Excellence for Macromolecular Complexes.

AUTHOR CONTRIBUTIONS

G.T., C.G., and E.M.S. designed experiments; C.G., M.H., and A.B. conducted

experiments; G.T., C.G., and M.J. analyzed experiments; E.M.S. wrote the

manuscript; and all authors edited and revised the manuscript.

DECLARATION OF INTERESTS

The authors declare no competing interests.

Received: June 13, 2017

Revised: January 20, 2018

Accepted: March 16, 2018

Published: April 12, 2018

SUPPORTING CITATIONS

The following reference appears in the Supplemental Information: Zeisel

et al. (2015).

REFERENCES

Aakalu, G., Smith, W.B., Nguyen, N., Jiang, C., and Schuman, E.M. (2001).

Dynamic visualization of local protein synthesis in hippocampal neurons.

Neuron 30, 489–502.

Agarwal, V., Bell, G.W., Nam, J.W., and Bartel, D.P. (2015). Predicting effective

microRNA target sites in mammalian mRNAs. eLife 4, 4.

Akbalik, G., Langebeck-Jensen, K., Tushev, G., Sambandan, S., Rinne, J.,

Epstein, I., Cajigas, I., Vlatkovic, I., and Schuman, E.M. (2017). Visualization

of newly synthesized neuronal RNA in vitro and in vivo using click-chemistry.

RNA Biol. 14, 20–28.

Anderson, B.R., Chopra, P., Suhl, J.A., Warren, S.T., and Bassell, G.J. (2016).

Identification of consensus binding sites clarifies FMRP binding determinants.

Nucleic Acids Res. 44, 6649–6659.

Ascano, M., Jr., Mukherjee, N., Bandaru, P., Miller, J.B., Nusbaum, J.D.,

Corcoran, D.L., Langlois, C., Munschauer, M., Dewell, S., Hafner, M., et al.

(2012). FMRP targets distinct mRNA sequence elements to regulate protein

expression. Nature 492, 382–386.

Bailey, T.L., Boden, M., Buske, F.A., Frith, M., Grant, C.E., Clementi, L., Ren,

J., Li, W.W., and Noble, W.S. (2009). MEME SUITE: tools for motif discovery

and searching. Nucleic Acids Res. 37, W202-8.

Blichenberg, A., Rehbein, M., M€uller, R., Garner, C.C., Richter, D., and Kindler,

S. (2001). Identification of a cis-acting dendritic targeting element in the mRNA

encoding the alpha subunit of Ca2+/calmodulin-dependent protein kinase II.

Eur. J. Neurosci. 13, 1881–1888.

Cajigas, I.J., Tushev, G., Will, T.J., tom Dieck, S., Fuerst, N., and Schuman,

E.M. (2012). The local transcriptome in the synaptic neuropil revealed by

deep sequencing and high-resolution imaging. Neuron 74, 453–466.

Chi, S.W., Zang, J.B., Mele, A., and Darnell, R.B. (2009). Argonaute HITS-CLIP

decodes microRNA-mRNA interaction maps. Nature 460, 479–486.

Cui-Wang, T., Hanus, C., Cui, T., Helton, T., Bourne, J., Watson, D., Harris,

K.M., and Ehlers, M.D. (2012). Local zones of endoplasmic reticulum

complexity confine cargo in neuronal dendrites. Cell 148, 309–321.

de Moor, C.H., and Richter, J.D. (1999). Cytoplasmic polyadenylation ele-

ments mediate masking and unmasking of cyclin B1 mRNA. EMBO J. 18,

2294–2303.

Derti, A., Garrett-Engele, P., Macisaac, K.D., Stevens, R.C., Sriram, S., Chen,

R., Rohl, C.A., Johnson, J.M., and Babak, T. (2012). A quantitative atlas of pol-

yadenylation in five mammals. Genome Res. 22, 1173–1183.

Dobin, A., Davis, C.A., Schlesinger, F., Drenkow, J., Zaleski, C., Jha, S., Batut,

P., Chaisson,M., andGingeras, T.R. (2013). STAR: ultrafast universal RNA-seq

aligner. Bioinformatics 29, 15–21.

Elkon, R., Ugalde, A.P., and Agami, R. (2013). Alternative cleavage and poly-

adenylation: extent, regulation and function. Nat. Rev. Genet. 14, 496–506.

Enright, A.J., John, B., Gaul, U., Tuschl, T., Sander, C., andMarks, D.S. (2003).

MicroRNA targets in Drosophila. Genome Biol. 5, R1.

Flavell, S.W., Kim, T.K., Gray, J.M., Harmin, D.A., Hemberg, M., Hong, E.J.,

Markenscoff-Papadimitriou, E., Bear, D.M., and Greenberg, M.E. (2008).

Genome-wide analysis of MEF2 transcriptional program reveals synaptic

target genes and neuronal activity-dependent polyadenylation site selection.

Neuron 60, 1022–1038.

Fontes, M.M., Guvenek, A., Kawaguchi, R., Zheng, D., Huang, A., Ho, V.M.,

Chen, P.B., Liu, X., O’Dell, T.J., Coppola, G., et al. (2017). Activity-dependent

regulation of alternative cleavage and polyadenylation during hippocampal

long-term potentiation. Sci. Rep. 7, 17377.

Glanzer, J., Miyashiro, K.Y., Sul, J.Y., Barrett, L., Belt, B., Haydon, P., and

Eberwine, J. (2005). RNA splicing capability of live neuronal dendrites. Proc.

Natl. Acad. Sci. USA 102, 16859–16864.

Grillo, G., Turi, A., Licciulli, F., Mignone, F., Liuni, S., Banfi, S., Gennarino, V.A.,

Horner, D.S., Pavesi, G., Picardi, E., and Pesole, G. (2010). UTRdb and

UTRsite (RELEASE 2010): a collection of sequences and regulatory motifs of

the untranslated regions of eukaryotic mRNAs. Nucleic Acids Res. 38,

D75–D80.

Grooms, S.Y., Noh, K.M., Regis, R., Bassell, G.J., Bryan, M.K., Carroll, R.C.,

and Zukin, R.S. (2006). Activity bidirectionally regulates AMPA receptor

mRNA abundance in dendrites of hippocampal neurons. J. Neurosci. 26,

8339–8351.

Gumy, L.F., Yeo, G.S., Tung, Y.C., Zivraj, K.H., Willis, D., Coppola, G., Lam,

B.Y., Twiss, J.L., Holt, C.E., and Fawcett, J.W. (2011). Transcriptome analysis

of embryonic and adult sensory axons reveals changes in mRNA repertoire

localization. RNA 17, 85–98.

Hanus, C., and Schuman, E.M. (2013). Proteostasis in complex dendrites. Nat.

Rev. Neurosci. 14, 638–648.

Neuron 98, 1–17, May 2, 2018 15


http://refhub.elsevier.com/S0896-6273(18)30236-8/sref1









































































Hanus, C., Geptin, H., Tushev, G., Garg, S., Alvarez-Castelao, B., Sambandan,

S., Kochen, L., Hafner, A.S., Langer, J.D., and Schuman, E.M. (2016).

Unconventional secretory processing diversifies neuronal ion channel proper-

ties. eLife 5, 5.

Hilgers, V., Perry, M.W., Hendrix, D., Stark, A., Levine, M., and Haley, B. (2011).

Neural-specific elongation of 30 UTRs during Drosophila development. Proc.

Natl. Acad. Sci. USA 108, 15864–15869.

Hobohm, U., Scharf, M., Schneider, R., and Sander, C. (1992). Selection of

representative protein data sets. Protein Sci. 1, 409–417.

Holt, C.E., and Schuman, E.M. (2013). The central dogma decentralized: new

perspectives on RNA function and local translation in neurons. Neuron 80,

648–657.

Huang, Y.S., Carson, J.H., Barbarese, E., and Richter, J.D. (2003). Facilitation

of dendritic mRNA transport by CPEB. Genes Dev. 17, 638–653.

Ji, Z., Lee, J.Y., Pan, Z., Jiang, B., and Tian, B. (2009). Progressive lengthening

of 30 untranslated regions of mRNAs by alternative polyadenylation during

mouse embryonic development. Proc. Natl. Acad. Sci. USA 106, 7028–7033.

Jiang, M., Anderson, J., Gillespie, J., and Mayne, M. (2008). uShuffle: a useful

tool for shuffling biological sequences while preserving the k-let counts. BMC

Bioinformatics 9, 192.

Kang, H., and Schuman, E.M. (1996). A requirement for local protein synthesis

in neurotrophin-induced hippocampal synaptic plasticity. Science 273,

1402–1406.

Karolchik, D., Hinrichs, A.S., Furey, T.S., Roskin, K.M., Sugnet, C.W.,

Haussler, D., and Kent, W.J. (2004). The UCSC Table Browser data retrieval

tool. Nucleic Acids Res. 32, D493–D496.

Kim, J., Krichevsky, A., Grad, Y., Hayes, G.D., Kosik, K.S., Church, G.M., and

Ruvkun, G. (2004). Identification of many microRNAs that copurify with polyri-

bosomes in mammalian neurons. Proc. Natl. Acad. Sci. USA 101, 360–365.

Kim, H.H., Lee, S.J., Gardiner, A.S., Perrone-Bizzozero, N.I., and Yoo, S.

(2015). Different motif requirements for the localization zipcode element of

b-actin mRNA binding by HuD and ZBP1. Nucleic Acids Res. 43, 7432–7446.

Konieczny, P., Stepniak-Konieczna, E., and Sobczak, K. (2014). MBNL pro-

teins and their target RNAs, interaction and splicing regulation. Nucleic

Acids Res. 42, 10873–10887.

Kozomara, A., and Griffiths-Jones, S. (2014). miRBase: annotating high confi-

dence microRNAs using deep sequencing data. Nucleic Acids Res. 42,

D68–D73.

Langmead, B., Trapnell, C., Pop, M., and Salzberg, S.L. (2009). Ultrafast and

memory-efficient alignment of short DNA sequences to the human genome.

Genome Biol. 10, R25.

Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., Marth, G.,

Abecasis, G., and Durbin, R.; 1000 Genome Project Data Processing

Subgroup (2009). The Sequence Alignment/Map format and SAMtools.

Bioinformatics 25, 2078–2079.

Lin, A.C., Tan, C.L., Lin, C.L., Strochlic, L., Huang, Y.S., Richter, J.D., and Holt,

C.E. (2009). Cytoplasmic polyadenylation and cytoplasmic polyadenylation

element-dependent mRNA regulation are involved in Xenopus retinal axon

development. Neural Dev. 4, 8.

Lorenz, R., Bernhart, S.H., Honer Zu Siederdissen, C., Tafer, H., Flamm, C.,

Stadler, P.F., and Hofacker, I.L. (2011). ViennaRNA package 2.0. Algorithms

Mol. Biol. 6, 26.

Mangone, M., Manoharan, A.P., Thierry-Mieg, D., Thierry-Mieg, J., Han, T.,

Mackowiak, S.D., Mis, E., Zegar, C., Gutwein, M.R., Khivansara, V., et al.

(2010). The landscape of C. elegans 3’UTRs. Science 329, 432–435.

Mayford, M., Baranes, D., Podsypanina, K., and Kandel, E.R. (1996). The

30-untranslated region of CaMKII alpha is a cis-acting signal for the localization

and translation of mRNA in dendrites. Proc. Natl. Acad. Sci. USA 93,

13250–13255.

Mayr, C., and Bartel, D.P. (2009). Widespread shortening of 3’UTRs by alterna-

tive cleavage and polyadenylation activates oncogenes in cancer cells. Cell

138, 673–684.

16 Neuron 98, 1–17, May 2, 2018

McCarthy, D.J., Chen, Y., and Smyth, G.K. (2012). Differential expression anal-

ysis of multifactor RNA-seq experiments with respect to biological variation.

Nucleic Acids Res. 40, 4288–4297.

Miller, S., Yasuda, M., Coats, J.K., Jones, Y., Martone, M.E., and Mayford, M.

(2002). Disruption of dendritic translation of CaMKIIalpha impairs stabilization

of synaptic plasticity and memory consolidation. Neuron 36, 507–519.

Mishchenko, Y., Hu, T., Spacek, J., Mendenhall, J., Harris, K.M., and

Chklovskii, D.B. (2010). Ultrastructural analysis of hippocampal neuropil

from the connectomics perspective. Neuron 67, 1009–1020.

Miura, P., Shenker, S., Andreu-Agullo, C., Westholm, J.O., and Lai, E.C. (2013).

Widespread and extensive lengthening of 30 UTRs in the mammalian brain.

Genome Res. 23, 812–825.

Mori, Y., Imaizumi, K., Katayama, T., Yoneda, T., and Tohyama,M. (2000). Two

cis-acting elements in the 30 untranslated region of alpha-CaMKII regulate its

dendritic targeting. Nat. Neurosci. 3, 1079–1084.

M€uller, S., Rycak, L., Afonso-Grunz, F., Winter, P., Zawada, A.M., Damrath, E.,

Scheider, J., Schm€ah, J., Koch, I., Kahl, G., and Rotter, B. (2014). APADB: a

database for alternative polyadenylation and microRNA regulation events.

Database (Oxford) 2014, 2014.

Pickard, B.S., Knight, H.M., Hamilton, R.S., Soares, D.C., Walker, R., Boyd,

J.K., Machell, J., Maclean, A., McGhee, K.A., Condie, A., et al. (2008). A com-

mon variant in the 3’UTR of the GRIK4 glutamate receptor gene affects tran-

script abundance and protects against bipolar disorder. Proc. Natl. Acad.

Sci. USA 105, 14940–14945.

Ray, D., Kazan, H., Cook, K.B., Weirauch, M.T., Najafabadi, H.S., Li, X.,

Gueroussov, S., Albu, M., Zheng, H., Yang, A., et al. (2013). A compendium

of RNA-binding motifs for decoding gene regulation. Nature 499, 172–177.

Sambandan, S., Akbalik, G., Kochen, L., Rinne, J., Kahlstatt, J., Glock, C.,

Tushev, G., Alvarez-Castelao, B., Heckel, A., and Schuman, E.M. (2017).

Activity-dependent spatially localized miRNA maturation in neuronal den-

drites. Science 355, 634–637.

Sandberg, R., Neilson, J.R., Sarma, A., Sharp, P.A., and Burge, C.B. (2008).

Proliferating cells express mRNAs with shortened 30 untranslated regions

and fewer microRNA target sites. Science 320, 1643–1647.

Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch,

T., Preibisch, S., Rueden, C., Saalfeld, S., Schmid, B., et al. (2012). Fiji: an

open-source platform for biological-image analysis. Nat. Methods 9, 676–682.

Shi, Y. (2012). Alternative polyadenylation: new insights from global analyses.

RNA 18, 2105–2117.

Shigeoka, T., Jung, H., Jung, J., Turner-Bridger, B., Ohk, J., Lin, J.Q., Amieux,

P.S., and Holt, C.E. (2016). Dynamic axonal translation in developing and

mature visual circuits. Cell 166, 181–192.

Smibert, P., Miura, P., Westholm, J.O., Shenker, S., May, G., Duff, M.O.,

Zhang, D., Eads, B.D., Carlson, J., Brown, J.B., et al. (2012). Global patterns

of tissue-specific alternative polyadenylation in Drosophila. Cell Rep. 1,

277–289.

Spies, N., Burge, C.B., and Bartel, D.P. (2013). 30 UTR-isoform choice has

limited influence on the stability and translational efficiency of most mRNAs

in mouse fibroblasts. Genome Res. 23, 2078–2090.

Subramanian, A., Tamayo, P., Mootha, V.K., Mukherjee, S., Ebert, B.L.,

Gillette, M.A., Paulovich, A., Pomeroy, S.L., Golub, T.R., Lander, E.S., and

Mesirov, J.P. (2005). Gene set enrichment analysis: a knowledge-based

approach for interpreting genome-wide expression profiles. Proc. Natl.

Acad. Sci. USA 102, 15545–15550.

Subramanian, M., Rage, F., Tabet, R., Flatter, E., Mandel, J.L., and Moine, H.

(2011). G-quadruplex RNA structure as a signal for neurite mRNA targeting.

EMBO Rep. 12, 697–704.

Swanger, S.A., He, Y.A., Richter, J.D., and Bassell, G.J. (2013). Dendritic

GluN2A synthesis mediates activity-induced NMDA receptor insertion.

J. Neurosci. 33, 8898–8908.

Taliaferro, J.M., Vidaki, M., Oliveira, R., Olson, S., Zhan, L., Saxena, T., Wang,

E.T., Graveley, B.R., Gertler, F.B., Swanson, M.S., and Burge, C.B. (2016).







































































































































Distal alternative last exons localize mRNAs to neural projections. Mol. Cell 61,

821–833.

Tian, B., and Manley, J.L. (2017). Alternative polyadenylation of mRNA precur-

sors. Nat. Rev. Mol. Cell Biol. 18, 18–30.

Tian, B., Hu, J., Zhang, H., and Lutz, C.S. (2005). A large-scale analysis of

mRNA polyadenylation of human and mouse genes. Nucleic Acids Res. 33,

201–212.

tom Dieck, S., Kochen, L., Hanus, C., Heum€uller, M., Bartnik, I., Nassim-Assir,

B., Merk, K., Mosler, T., Garg, S., Bunse, S., et al. (2015). Direct visualization of

newly synthesized target proteins in situ. Nat. Methods 12, 411–414.

Trotta, E. (2014). On the normalization of the minimum free energy of RNAs by

sequence length. PLoS ONE 9, e113380.

Tyner, C., Barber, G.P., Casper, J., Clawson, H., Diekhans, M., Eisenhart, C.,

Fischer, C.M., Gibson, D., Gonzalez, J.N., Guruvadoo, L., et al. (2017). The

UCSC Genome Browser database: 2017 update. Nucleic Acids Res. 45 (D1),

D626–D634.

Udagawa, T., Swanger, S.A., Takeuchi, K., Kim, J.H., Nalavadi, V., Shin, J.,

Lorenz, L.J., Zukin, R.S., Bassell, G.J., and Richter, J.D. (2012). Bidirectional

control of mRNA translation and synaptic plasticity by the cytoplasmic polya-

denylation complex. Mol. Cell 47, 253–266.

Ulitsky, I., Shkumatava, A., Jan, C.H., Subtelny, A.O., Koppstein, D., Bell,

G.W., Sive, H., and Bartel, D.P. (2012). Extensive alternative polyadenylation

during zebrafish development. Genome Res. 22, 2054–2066.

Wang, E.T., Cody, N.A., Jog, S., Biancolella, M., Wang, T.T., Treacy, D.J., Luo,

S., Schroth, G.P., Housman, D.E., Reddy, S., et al. (2012). Transcriptome-wide

regulation of pre-mRNA splicing and mRNA localization by muscleblind pro-

teins. Cell 150, 710–724.

Wu, L., Wells, D., Tay, J., Mendis, D., Abbott, M.A., Barnitt, A., Quinlan, E.,

Heynen, A., Fallon, J.R., and Richter, J.D. (1998). CPEB-mediated cytoplasmic

polyadenylation and the regulation of experience-dependent translation of

alpha-CaMKII mRNA at synapses. Neuron 21, 1129–1139.

Xu, H., An, J.J., and Xu, B. (2017). Distinct cellular toxicity of two mutant hun-

tingtin mRNA variants due to translation regulation. PLoS ONE 12, e0177610.

Zeisel, A., Munoz-Manchado, A.B., Codeluppi, S., Lonnerberg, P., La Manno,

G., Jureus, A., Marques, S., Munguba, H., He, L., Betsholtz, C., et al. (2015).

Brain structure. Cell types in the mouse cortex and hippocampus revealed

by single-cell RNA-seq. Science 347, 1138–1142.

Zhang, W., Zhang, Y., Zheng, H., Zhang, C., Xiong, W., Olyarchuk, J.G.,

Walker, M., Xu, W., Zhao, M., Zhao, S., et al. (2007). SynDB: a synapse protein

DataBase based on synapse ontology. Nucleic Acids Res. 35, D737–D741.

Zhang, K.X., Tan, L., Pellegrini, M., Zipursky, S.L., and McEwen, J.M. (2016).

Rapid changes in the translatome during the conversion of growth cones to

synaptic terminals. Cell Rep. 14, 1258–1271.

Zivraj, K.H., Tung, Y.C., Piper, M., Gumy, L., Fawcett, J.W., Yeo, G.S., and

Holt, C.E. (2010). Subcellular profiling reveals distinct and developmentally

regulated repertoire of growth cone mRNAs. J. Neurosci. 30, 15464–15478.

Neuron 98, 1–17, May 2, 2018 17

















































STAR+METHODS

KEY RESOURCES TABLE

REAGENT or RESOURCE SOURCE IDENTIFIER

Antibodies

Guinea pig polyclonal anti-MAP 2 Synaptic Systems Cat#188004; RRID: AB_2138181

Mouse monoclonal anti-NeuN (clone A60) Millipore Cat#MAB377; RRID: AB_2298772

Rabbit polyclonal anti-PSD95 Abcam Cat#ab18258; RRID: AB_444362

Goat anti-Guinea pig DyLight 405 Jackson ImmunoResearch Cat#106-475-003; RRID: AB_2337432

Goat anti-Guinea pig Alexa Fluor 488 Invitrogen Cat#A11073; RRID: AB_142018

Goat anti-Rabbit Alexa Fluor 488 Invitrogen Cat#A11008; RRID: AB_143165

Chemicals, Peptides, and Recombinant Proteins

RNAlater Thermo Fisher Scientific Cat#AM7020

TRIzol Reagent Thermo Fisher Scientific Cat#15596026

Bicuculline Tocris Cat#0131

Actinomycin D Sigma Cat#A1410

5,6-Dichlorobenzimidazole riboside (DRB) Sigma Cat#1916

Triptolide Sigma Cat#T3652

Cytosine b-D-arabinofuranoside (AraC) Sigma Cat#C1768

Pepsin Sigma Cat#P7012

Magnetofectamine OZ Biosciences Cat#MTX0750

Lipofectamine 2000 Thermo Fisher Scientific Cat#11668027

SYBR Green PCR Master Mix Thermo Fisher Scientific Cat#4309155

Critical Commercial Assays

RNeasy Mini Kit QIAGEN Cat#74104

QuantiTect Reverse Transcription Kit QIAGEN Cat#205310

Direct-zol Zymo Research Cat#R2051

TURBO DNA-free Kit Thermo Fisher Scientific Cat#AM1907

MACE sequencing kit GenXPro Cat#160424

Quantigene ViewRNA ISH Cell Assay Thermo Fisher Scientific QVC0001

Deposited Data

Sequencing NCBI BioProject: PRJNA390472

Rattus norvegicus reference genome, build rn5 UCSC http://genome.ucsc.edu/cgi-bin/hgGateway?db=rn5

Experimental Models: Cell Lines

Human: 293 Cell Line Sigma Cat#85120602

Rat: P0-1 hippocampal primary neuron culture Charles River RRID: RGD_734476

Experimental Models: Organisms/Strains

Rattus norvegicus: Crl:CD(SD) Charles River RRID: RGD_734476

Oligonucleotides

mirVana miRNA Mimic: rno-miR-425-5p Thermo Fisher Scientific MC11575

mirVana miRNA Mimic: rno-miR-146a Thermo Fisher Scientific MC10722

mirVana miRNA Mimic: rno-miR-23a-5 Thermo Fisher Scientific MC10644

mirVana miRNA Mimic: rno-miR-154-5p Thermo Fisher Scientific MC10086

Primers for qRT-PCR, see Table S5 Eurofins/QIAGEN N/A

Recombinant DNA

Plasmid: AcGFP noUTR control This paper N/A

Plasmid: AcGFP Camk2a long 30 UTR This paper N/A

Plasmid: AcGFP Camk2a medium 30 UTR This paper N/A

(Continued on next page)

e1 Neuron 98, 1–17.e1–e6, May 2, 2018

http://genome.ucsc.edu/cgi-bin/hgGateway?db=rn5

Continued

REAGENT or RESOURCE SOURCE IDENTIFIER

Plasmid: AcGFP Camk2a short 30 UTR This paper N/A

Plasmid: AcGFP Kcnab2 long 30 UTR This paper N/A

Plasmid: pmirGLO_ATP6v0c-30 UTR-miR425 This paper N/A

Plasmid: pmirGLO_ATP6c0c-30 UTR-miR425del This paper N/A

Plasmid: pmirGLO_Cdk16-30 UTR-miR146 This paper N/A

Plasmid: pmirGLO_Cdk16-30 UTR-miR146del This paper N/A

Plasmid: pmirGLO_Ctxn1-30 UTR-miR23 This paper N/A

Plasmid: pmirGLO_Ctxn1-30 UTR-miR23del This paper N/A

Plasmid: pmirGLO_Arl2bp-30 UTR-miR154 This paper N/A

Plasmid: pmirGLO_Arl2bp-30 UTR-miR154del This paper N/A

Software and Algorithms

Fiji Schindelin et al., 2012 https://imagej.net/Welcome

MATLAB MathWorks https://de.mathworks.com/

NeuroBits This paper https://github.molgen.mpg.de/MPIBR/NeuroBits

STAR Dobin et al., 2013 https://github.com/alexdobin/STAR

Bowtie Langmead et al., 2009 http://bowtie-bio.sourceforge.net/index.shtml

Samtools Li et al., 2009 http://samtools.sourceforge.net/

Tabix Li et al., 2009 https://github.com/samtools/tabix

FastSeqStats This paper https://software.scic.brain.mpg.de/projects/MPIBR-

Bioinformatics/fastqSeqStats

PASSFinder This paper https://software.scic.brain.mpg.de/projects/MPIBR-

Bioinformatics/PASSFinder

ClusterPASS This paper https://github.molgen.mpg.de/MPIBR-Bioinformatics/

NeuroUTRs/tree/master/step01_PASSAnotation/

step04_ClusterPASS

UCSC Table Browser Karolchik et al., 2004 http://genome.ucsc.edu/cgi-bin/hgTables

PASSAnnotator This paper https://github.molgen.mpg.de/MPIBR-Bioinformatics/

PASSAnnotator

PASSCountExpression This paper https://github.molgen.mpg.de/MPIBR-Bioinformatics/

NeuroUTRs/blob/master/step01_PASSAnotation/

step06_ExpressionPASS/PASSCountExpression.pl

edgeR library (v.3.18) McCarthy et al., 2012 https://bioconductor.org/packages/release/bioc/html/

edgeR.html

RNAfold, ViennaRNA v.2.0 Lorenz et al., 2011 https://www.tbi.univie.ac.at/RNA/#download

MEME v.4.11 Bailey et al., 2009 http://meme-suite.org/

mkseed https://github.com/takayasaito/mkseed

TargetScan Agarwal et al., 2015 http://www.targetscan.org

miRanda Enright et al., 2003 http://www.microrna.org/

Hobohm clustering algorithm Hobohm et al., 1992 http://search.cpan.org/�brunov/String-Cluster-Hobohm-

0.121330/lib/String/Cluster/Hobohm.pm

GOAnalysis This paper https://software.scic.brain.mpg.de/projects/MPIBR-

Bioinformatics/GOAnalysis

uShuffle Jiang et al., 2008 http://digital.cs.usu.edu/�mjiang/ushuffle/

Other

LSM880 Zeiss https://www.zeiss.de/mikroskopie/produkte/confocal-

microscopes/lsm-880.html

StepOnePlus Real-Time PCR System Thermo Fisher Scientific Cat#4376600

MACE sequencing GenXPro https://genxpro.net/

Neuron 98, 1–17.e1–e6, May 2, 2018 e2


https://imagej.net/Welcome

https://de.mathworks.com/

https://github.molgen.mpg.de/MPIBR/NeuroBits

https://github.com/alexdobin/STAR

http://bowtie-bio.sourceforge.net/index.shtml

http://samtools.sourceforge.net/

https://github.com/samtools/tabix

https://software.scic.brain.mpg.de/projects/MPIBR-Bioinformatics/fastqSeqStats

https://software.scic.brain.mpg.de/projects/MPIBR-Bioinformatics/fastqSeqStats

https://software.scic.brain.mpg.de/projects/MPIBR-Bioinformatics/PASSFinder

https://software.scic.brain.mpg.de/projects/MPIBR-Bioinformatics/PASSFinder

https://github.molgen.mpg.de/MPIBR-Bioinformatics/NeuroUTRs/tree/master/step01_PASSAnotation/step04_ClusterPASS



http://genome.ucsc.edu/cgi-bin/hgTables

https://github.molgen.mpg.de/MPIBR-Bioinformatics/PASSAnnotator

https://github.molgen.mpg.de/MPIBR-Bioinformatics/PASSAnnotator

https://github.molgen.mpg.de/MPIBR-Bioinformatics/NeuroUTRs/blob/master/step01_PASSAnotation/step06_ExpressionPASS/PASSCountExpression.pl



https://bioconductor.org/packages/release/bioc/html/edgeR.html

https://bioconductor.org/packages/release/bioc/html/edgeR.html

https://www.tbi.univie.ac.at/RNA/#download

http://meme-suite.org/

https://github.com/takayasaito/mkseed

http://www.targetscan.org

http://www.microrna.org/

http://search.cpan.org/%7Ebrunov/String-Cluster-Hobohm-0.121330/lib/String/Cluster/Hobohm.pm



https://software.scic.brain.mpg.de/projects/MPIBR-Bioinformatics/GOAnalysis

https://software.scic.brain.mpg.de/projects/MPIBR-Bioinformatics/GOAnalysis

http://digital.cs.usu.edu/%7Emjiang/ushuffle/

http://digital.cs.usu.edu/%7Emjiang/ushuffle/

https://www.zeiss.de/mikroskopie/produkte/confocal-microscopes/lsm-880.html

https://www.zeiss.de/mikroskopie/produkte/confocal-microscopes/lsm-880.html

https://genxpro.net/


CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Erin M.

Schuman ([email protected]).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

The procedures involving animal treatment and care were conducted in conformity with the institutional guidelines that are in compli-

ance with national and international laws and policies (DIRECTIVE 2010/63/EU; German animal welfare law; FELASA guidelines). The

animals were euthanized according to annex 2 of x 2 Abs. 2 Tierschutz-Versuchstier-Verordnung.

Acute Hippocampal SlicesSprague Dawley rats were housed in standard cages and fed standard lab chow and water ad libitum. Rats Hippocampal slices

(500 mm) were prepared from four-week-old male animals as previously described (Kang and Schuman, 1996).

Primary Hippocampal CulturesDissociated rat hippocampal neurons were prepared from P0-1 day-old rat pups as previously described (Aakalu et al., 2001). For

neuron-enriched cultures and cultures used for half-live measurements, neurons were plated at a density of 36,000 cells/cm2 onto

poly-d-lysine-coated 60 mm cell culture dishes. At one day in vitro (DIV), AraC was added to a final concentration of 3 mM. After

2 days, medium was exchanged to pre-conditioned growth medium (Neurobasal-A supplemented with B27 and GlutaMAX, 30%

glia-culture supernatant, 15% cortex-culture supernatant) and neurons were cultured until DIV21. For fluorescence in situ hybridiza-

tion, 30,000 cells were plated onto poly-d-lysine coated glass-bottom Petri dishes (MatTek) and cultured for 21 days in growth me-

dium (Neurobasal-A supplemented with B27 and GlutaMAX). The cultures weremaintained in a humidified incubator at 37�C and 5%

CO2. The sex of cells was not determined.

Primary Glia-Enriched CulturesThe hippocampi of P0-1 day old rat pups were isolated and triturated after digestion with papain. Cells were plated on uncoated

60mmcell culture dishes and grown at 37�C and 5%CO2 inminimal essential medium supplemented with GlutaMAX and 10%horse

serum. At DIV7, medium was changed to growth medium (Neurobasal-A supplemented with B27 and GlutaMAX) and cells were

cultured until DIV21. Glial enrichment was confirmed by the over-representation of glial markers (Figures S2F and S2G). The sex

of cells was not determined.

Cell LinesHEK293 cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% FCS at 37�C in a 5% CO2

atmosphere.

METHOD DETAILS

Tissue Microdissection, RNA Isolation, and 30 End SequencingThe somatic and the neuropil layer of the CA1 region weremicrodissected by hand from each slice. Tissue pieces were first collected

in RNAlater to stabilize and protect RNA from degradation. Total RNA was extracted using TRIzol reagent and purified using the

RNeasy Mini Kit following the manufacturer’s instructions including the DNaseI digestion. Successful isolation of the neuropil was

confirmed by de-enrichment of NeuN, a nuclear marker, using an anti-NeuN antibody (1:5000) as previously described in Cajigas

et al. (2012). Total RNA was subjected to 30 end sequencing using the MACE sequencing kit (M€uller et al., 2014).

Pharmacological TreatmentsRat hippocampal slices (500 mm) were treated with bicuculline (40 mM) or water control for 4 hr on filter paper in a recovery chamber.

For transcription inhibition, Actinomycin D (10 ug/ml), DRB (D-ribofuranosylbenzimidazole) (100 mM) and Triptolide (1 mM)were added

to the ACSF 30 min before the addition of bicuculline. For RNA half-life measurements, Actinomycin D (10 ug/ml), DRB (100 mM) and

Triptolide (1 mM) were added to the culturing medium.

High-Resolution In Situ Hybridization in Primary Hippocampal Neurons and SlicesIn situ hybridization was performed using the Quantigene ViewRNA ISH Cell Assay for Fluorescence with probes targeting the unique

sequence of the long 30 UTR. The manufacturer’s protocol was applied with the following modifications: Primary hippocampal neu-

rons (DIV 21) were fixed for 20 min at room temperature using a 4% paraformaldehyde solution (4% paraformaldehyde, 5.4%

glucose, 0.01M sodiummetaperiodate, in lysine-phosphate buffer). The Proteinase K treatment was substitutedwith a Pepsin diges-

tion. For this, neurons were kept for no more than 45 s in 0.01 mg/ml Pepsin, 10 mMHCl in water, pH 2.5. Enzyme activity was halted

with PBS pH 7.2 and washed thoroughly. After completion of in situ hybridization, cells were washed with PBS and incubated in

e3 Neuron 98, 1–17.e1–e6, May 2, 2018

mailto:[email protected]


blocking buffer (4% goat serum in PBS) for 1 hr. Dendrites were stained for 1 hr at room temperature using an anti-MAP2 antibody

(SySy, 188004; 1:1000), washed three times with 1x PBS and incubated with the secondary antibody (Jackson IR, 106-475-003,

DyLight 405; 1:1,000 dilution) for 1 hr at room temperature.

For in situ hybridization in hippocampal slices, 4-week-old male rats were perfused with 1x PBS and 4% (v/v) paraformaldehyde

solution in PBS. The hippocampi were dissected, sliced to 2mm and fixed for 3 hr at room temperature. Slices were cryoprotected in

20% (w/v) sucrose in PBS (DEPC-treated) overnight at 4�Cand cryosectioned at 12 mm thickness. In situ hybridization was performed

as described above with the following modifications. Probes were diluted to 1:50. After completion of the in situ hybridization, the

slices were blocked for 1 hr in blocking buffer. Dendrites were stained overnight at 4�C using an anti-MAP2 antibody (SySy,

188004; 1:1000), washed three times with 1x PBS and incubated with the secondary antibody (Invitrogen, A11073, Alexa 488;

1:1,000) for 2 hr at room temperature. Slices were mounted in AquaPolymount. Fluorescence imaging was performed with a

LSM880 laser scanning confocal microscope with x40 (NA 1.4) objectives with appropriate excitation laser lines and spectral

detection windows. In Figures 4D–4F and Figures S4A and S4B, themRNA signal in hippocampal cultures was dilated for better visu-

alization. The raw, undilated images were used for analysis.

GFP Reporter ExperimentsCultured hippocampal neurons were transfected (using Magnetofectamine) at DIV11-16 with 0.2-1 mg per dish of the reporter

construct (CaMK2a promoter-GFP-CaMK2a short 30 UTR, middle 30 UTR, long 30 UTR, Kcnab2 long 30 UTR or no UTR, respectively).

In situ hybridization against the coding sequence of GFP was performed 24 hr post-transfection, as described above.

cDNA Synthesis and Quantitative Real-Time PCRTotal RNA was reverse-transcribed with the QuantiTect Reverse Transcription Kit. qRT-PCR was performed using the 2X SYBR

Green Master Mix. Reaction setup and cycling parameters were used according the QuantiTect primer assay recommendations.

qRT-PCR was run on a StepOnePlus Real-Time PCR System. The primer sequences used in this study are indicated in Table S5.

Neuron- and glia enriched genes (Figure S2E) were normalized to the geometric mean of GAPDH and 18 s rRNA. Somata- and neuro-

pil-enriched genes (Figure 3B) were normalized to the geometric mean of 18 s rRNA and Cdc73. Long 30 UTR isoforms that were

altered with neural activity (Figure S8E) were normalized to GAPDH.

Transfection of MicroRNA MimicsmirVanamiRNAMimics and the respective 30 UTR-reporter (pmirGLODual-Luciferase vector (Promega) expressing firefly flanked by

the 30 UTR containing or lacking the miRNA binding site and renilla) were co-transfected with Lipofectamine 2000 according to the

manufacturer’s instructions. For each co-transfection 1 mg of the 30 UTR-reporter and a final concentration of 30 nM of miRNAmimic

was used. 24 hr post-transfection, RNA was isolated using Direct-zol and treated with DNase using the TURBO DNA-free kit to

remove potential DNA contaminations. qRT-PCR was performed as described above. Firefly Ct-values were normalized to renilla

Ct-values. Then the deltaCt method was used to give the fold change between a reporter with the miRNA binding site and a reporter

lacking the miRNA binding site.

QUANTIFICATION AND STATISTICAL ANALYSIS

Image AnalysisNeuronal cell soma and processes were segmented using NeuroBits. FISH puncta were detected by locating the center of mass of

local maxima along the segmented regions. For each puncta the path traveled, and hence the distance to the soma, was calculated

with the help of the segmented neuronal tree.

Puncta traveled distances were grouped by in situ probe and a non-parametric test was applied. Analysis of FISH probes against

GFP followed a different approach due to the less discrete character of the signal in the GFP-mRNA channel. The signal was

smoothed by local averaging along the segmented dendrites and normalized to the average GFP signal in the cell soma. A median

of the distance along the dendrite was weighted by the GFP-mRNA signal to produce a final measure of signal localization. Again,

each distance measure was grouped by in situ probe and a non-parametric test was applied to assess group differences.

Detection and Annotation of poly(A) Supported SitesThe pipeline for the detection and annotation of poly(A) supported sites (PASS) consisted of the following steps:

i) Quality assessment.

The tool fastqSeqStats was used to assess read quality. The sequenced reads were fragmented in 300-500 nt regions and

sequenced from the 50 side. The expected elevation of poly(A/T) content toward the 30 end was observed (Figure S1).

ii) Genome Alignment

Reads alignment was conducted with STAR aligner (Dobin et al., 2013), version (2.5.2) with the following parameters:

STAR–runMode alignReads–runThreadN 6–genomeDir $2–readFilesIn $file–readFilesCommand zcat–outSAMattributes All–

outStd Log–outSAMtype BAM SortedByCoordinate–outSAMstrandField intronMotif–outFilterIntronMotifs RemoveNoncanonical–

alignSoftClipAtReferenceEnds No–outFilterScoreMinOverLread 0.25–outFilterMatchNminOverLread 0.25

Neuron 98, 1–17.e1–e6, May 2, 2018 e4


iii) Poly(A) supported site prediction.

To detect the peaks of poly(A) supported sites we used the in-house developed tool PASSFinder. It relies on theHTSlib API (Li et al.,

2009) version (1.4.1).

A genomemap of potential internal priming regions was created by aligning a string of 10 consecutives As using Bowtie. The result

was converted to a BED record and the mask file was indexed and randomly accessed with Tabix. The 30 base coverage indicated

potential 30 UTR sites. In order to identify the positions of polyadenylation we implemented a base clustering technique that merges

positions in a user-defined window. During the grouping process, information about the best expressed base and best expressed

seed (30 base of reads containing poly(A) tail) is maintained by ClusterPASS. Alternative last exons (ALE) were defined as poly(A) sites

detected in intronic regions. Only �13% of the identified PASS fell into this category. ALEs showed relatively low expression

compared to PASS detected in annotated 30 UTRs, accumulating less than 3% of all reads (Figure S1G).

iv) Annotation assignment.

Annotationwas based on aBEDmap that was downloadedwith theUCSCTable Browser tool for Rattus Norvegicus, build rn5. The

PASSAnnotator tool used the map and assigned gene features based on upstream proximity.

v) Expression counting.

Expression was determined by read coverage in proximity to each of the predicted poly(A) supported sites. The provided

PASSCountExpression.pl script automated that routine.

Consolidated Set of Experimental DataExperimental data were sequenced from tissue and cell-culture samples. To facilitate analysis and maintain data integrity, a consol-

idated set of predicted poly(A) supported site was created. The required conditions were minimum site expression of 2 reads per

million in at least one sample and a site position hitting within the 30 UTR or extended 30 UTR region (< 20 kb) of detected genes.

We intentionally excluded sites falling in introns or coding sequence exons as those propose a model of truncated protein products

and are out of the scope of this study.

Classification of Cell-Enriched GenesDifferential 30 UTR isoforms expression between neuron-enriched and glia-enriched samples was determined using the edgeR library

(version 3.18) in R (McCarthy et al., 2012). The trimmed mean of M-values (TMM) normalization was combined with frame design to

avoid batch effects. A gene-wise negative binomial generalized linear model with quasi-likelihood test was used to conduct differ-

ential expression (DE) analysis. The steps followed closely examples given in the edgeR manual.

The results of neuron- or glia-enriched 30 UTR isoforms were projected back on the gene level, where a gene was labeled as

‘‘neuron-enriched’’ if at least one isoform showed DE in a neuron-enriched sample. The same logic was applied for glia-enriched iso-

forms. Genes with predictedmultiple 30 UTR isoforms, for which one isoformwas DE in the neuron-enriched sample and another was

DE in the glia-enriched sample were labeled as ‘‘shared’’ and removed from downstream analysis.

Structural and Sequence-Specific Properties of Neuronal 30 UTRsDNA sequence was extracted from the Rattus Norvegicus rn5 version genome, starting from the closest stop codon to a predicted

poly(A) supported site. Only sequences with a size ranging from 100 to 5000 nucleotides were included in the analysis. The GC con-

tent was assessed by counting the number of G or C bases in the sequence and then divided by the number of bases in the predicted

30 UTR. The ViennaRNA package version 2.0 with RNAfold was used to calculate theminimum free energy per 30 UTR sequence (Lor-

enz et al., 2011). A method described by Trotta (2014) was adapted to normalize minimum free energy units for varying 30 UTR length.

Sequences were shuffled preserving the di-nucleotide content by a routine implemented in the Motif based sequence analysis tool

MEME, version 4.11 (Bailey et al., 2009).

miRNA Site PredictionmiRNA sites were predicted using three target prediction programs:

d mkseed

d targetscan (Agarwal et al., 2015)

d miRanda (Enright et al., 2003) (pairing score > 155 and energy score < �20)

ReferencemiRNAswere downloaded frommiRBase release 21 (Kozomara andGriffiths-Jones, 2014). The union of the predictions

was further compared against Ago-Clip binding sites (Chi et al., 2009) and a score based on the number of Ago-Clip reads spanning

each prediction site was calculated. A conservation score was calculated by averaging the base-wise phyloPhast score from 13 an-

imals available for Rattus norvegicus genome version 5 (Tyner et al., 2017). We measured the expression of miRNAs with Nanostring

in rat hippocampus (Sambandan et al., 2017) and cross-referenced that datawith the predicted seed. Our final predictedmiRNA sites

were combination of the following criteria: more than one prediction program recognized the site, there are more than 10 Ago-Clip

reads spanning the site, average phyloPhast score for the site seed is more than 0.2, and the miRNA is in hippocampus above back-

ground levels as reported from Nanostring counter (Sambandan et al., 2017).

e5 Neuron 98, 1–17.e1–e6, May 2, 2018


K-mer Site PredictionTo define a set of predicted k-mers per 30 UTR region, a sliding window of 7-mers was applied. With each step of the window an

average conservation score was calculated based on the phyloPhast Rattus norvegicus version 5. A threshold of 0.2 was applied

and each k-mer was recorded in hash data structure. Every time a k-mer was found the localization fold change and stability of

the corresponding 30 UTR was recorded to the hashed data. The k-mer localization and stability measure was determined by aver-

aging those results and the difference between one k-mer and the rest of the candidates defined the enrichment score and statistics.

The Levenshtein distance of one was used to allow for fuzzy k-mer matches. The k-mers were clustered with the Hobohm clustering

algorithm (Hobohm et al., 1992).

Localization of Neuronal 30 UTRs in Subcellular CompartmentsTo address localization, only genes classified as neuron-enriched (see Classification of Cell-Enriched Genes) were used. Differential

expression was determined using the edgeR library as described above (see Classification of Cell-Enriched Genes). Gene ontology

analysis was conducted on two unranked lists. The target list comprised all localized genes or CA1 layer-specific genes and the back-

ground list comprised all neuron-enriched classified genes. We used the GOAnalysis tool developed in our Bioinformatics source

library.

K-mer ContentThe k-mer content was investigated in neuron-enriched geneswithmultiple 30 UTRs. The top two expressed isoformswere labeled as

‘‘shorter’’ and ‘‘longer’’ representing their proximity to the stop codon. We extracted each 7-mer from the ‘‘longer’’ isoform using a

sliding window. Three classes were defined describing the content options. Class ‘‘repeat between’’ was counted when the 7-mer is

also present in the ‘‘shorter’’ 30 UTR isoform. Class ‘‘novel’’ and ‘‘repeat within’’ were counted in the case of lacking a match in the

‘‘shorter’’ isoform but being present either one or multiple times in the ‘‘longer’’ isoform, respectively. To address the expected k-mer

content we conducted two controls. First we used di-nucleotide shuffling of the ‘‘longer’’ isoform and repeated the procedure 3 times

per gene. Di-nucleotide shuffling was done using the uShuffle procedure (Jiang et al., 2008). Second, we extracted the intergenic

sequence 1kb downstream of the last predicted PASS and repeated the same analysis as for the observed 30 UTRs. The Levenshteindistance of one was used to allow for fuzzy k-mer matches. The k-mers were clustered with the Hobohm clustering algorithm (Ho-

bohm et al., 1992).

Classifying Protein ProductsProtein products were classified as described in Hanus et al. (2016).

Stability of Neuronal 30 UTRsRaw counts were first represented as reads per million with minimum threshold of 2 reads per million. In a sequencing experiment

with fixed depth in the same RNA pool, stable messages will exhibit an apparent increase in expression over time when transcription

is blocked, owing to their enhanced representation in the pool. We used that property to define a normalization scheme. Each time

point was individually ranked and ranks per gene were summed to determine an overall rank score. 100 genes with the highest overall

rank score were chosen and used for calculating a normalization ratio. To predict 30 UTR stability a non-linear regression with an

exponential decay function was applied to each biological replicate. The average of the two replicas was used for half-life value.

Only 30 UTR isoforms with stability between 0 and 24 hr were included in our final analysis as higher predictions were unreliable

with the chosen time points. GSEA was implemented and conducted as previously described in Subramanian et al. (2005). Gene

ontology terms were used for grouping 30 UTR isoforms. To investigate k-mers or miRNAs seeds associated with 30 UTR stability,

we conducted an analysis adapted from the above GSEA analysis. Each k-mer or miRNA represented a group in a ranked list of

30 UTR targets. The median half-life in the groups was compared against the median half-life outside the group using the non-para-

metric Mann-Whitney U-test. The skewedness of the distribution was taken to indicate either a stabilizing or de-stabilizing influence.

The analysis was corrected for multiple comparisons (Benjamini-Hochberg) and the results were visualized using volcano plots.

Neuronal 30 UTR Plasticity upon Altered Neuronal ActivityRaw expression counts were normalized using the TMM procedure and replica batch effects were removed by edgeR library. To

determine differential expression a simpler procedure was applied. 30 UTR isoforms were sorted according to their expression

and data points were binned in blocks of size 150. For each bin, data points were Z-transformed and outliers were determined

with p value % 0.05. Multiple hypothesis correction was applied at the end to account for false discoveries.

DATA AND SOFTWARE AVAILABILITY

The accession number for the raw sequencing data reported in this paper is NCBI BioProject: PRJNA390472. Processed data used

for analyses in this manuscript are included in Table S1. All scripts used in this study are deposited in https://github.molgen.mpg.de/

MPIBR-Bioinformatics/NeuroUTRs.

Neuron 98, 1–17.e1–e6, May 2, 2018 e6

https://github.molgen.mpg.de/MPIBR-Bioinformatics/NeuroUTRs

https://github.molgen.mpg.de/MPIBR-Bioinformatics/NeuroUTRs

Alternative 3′ UTRs Modify the Localization, Regulatory ...€¦ · niﬁcantly altered by...

Documents

Transcript of Alternative 3′ UTRs Modify the Localization, Regulatory ...€¦ · niﬁcantly altered by...