Tissue-specific regulation of mouse microRNA genes in endoderm-derived tissues.

Abstract

MicroRNAs fine-tune the activity of hundreds of protein-coding genes. The identification of tissue-specific microRNAs and their promoters has been constrained by the limited sensitivity of prior microRNA quantification methods. Here, we determine the entire microRNAome of three endoderm-derived tissues, liver, jejunum and pancreas, using ultra-high throughput sequencing. Although many microRNA genes are expressed at comparable levels, 162 microRNAs exhibited striking tissue-specificity. After mapping the putative promoters for these microRNA genes using H3K4me3 histone occupancy, we analyzed the regulatory modules of 63 microRNAs differentially expressed between liver and jejunum or pancreas. We determined that the same transcriptional regulatory mechanisms govern tissue-specific gene expression of both mRNA and microRNA encoding genes in mammals.

Full text

Page 1

Tissue-specific regulation of mouse MicroRNA
genes in endoderm-derived tissues
Yan Gao1, Jonathan Schug2, Lindsay B. McKenna2, John Le Lay2, Klaus H. Kaestner2,*
and Linda E. Greenbaum3,*
1Department of Medicine, University of Pennsylvania School of Medicine, Philadelphia,2Department of Genetics
and Institute for Diabetes, Obesity and Metabolism, University of Pennsylvania School of Medicine,
Philadelphia, PA 19104-6145 and3Departments of Cancer Biology and Medicine, Jefferson Medical College,
Thomas Jefferson University, Philadelphia, PA 19107, USA
Received February 7, 2010; Revised August 14, 2010; Accepted August 19, 2010
ABSTRACT
MicroRNAs fine-tune the activity of hundreds of
protein-coding genes.
tissue-specific microRNAs and their promoters has
been constrained by the limited sensitivity of prior
microRNA quantification methods. Here, we deter-
mine the entire microRNAome of three endoderm-
derived tissues, liver, jejunum and pancreas, using
ultra-high throughput sequencing. Although many
microRNA genes are expressed at comparable
levels, 162 microRNAs exhibited striking tissue-
specificity. After mapping the putative promoters
for these microRNA genes using H3K4me3 histone
occupancy, we analyzed the regulatory modules of
63 microRNAs differentially expressed between liver
and jejunum or pancreas. We determined that
the same transcriptional regulatory mechanisms
govern tissue-specific gene expression of both
mRNA and microRNA encoding genes in mammals.
Theidentification of
INTRODUCTION
MicroRNAs are short non-coding RNAs of 21–23nt that
are present in multiple organisms and that are often evo-
lutionarily conserved (1). MicroRNAs function by sup-
pressing the expression of protein coding genes, with
each microRNA targeting dozens or even hundreds of
mRNAs. In mammals, microRNA function on a global
level has been studied through mutational analysis
of Dicer, an obligate enzyme in the processing of
microRNAprecursors.Thus,
microRNAs are required for ES self-renewal as well as
development and function of tissues including liver (2,3),
intestine (4) and heart (5).
itwasshownthat
There are more than 1000 microRNAs encoded in the
mammalian genome, and these are derived from a
complex series of processing steps. The primary transcript,
or pri-microRNA, synthesized by RNA polymerase II or
III is very labile, and quickly converted to ?70nt precur-
sors, termed pre-microRNA (6). These pre-microRNAs
exist as hairpins and are further processed through a
series of endonuclease digestion steps to the final and func-
tional microRNAs, which are loaded onto the so-called
RNA inducing silencing complex (RISC) to exert their
regulatoryfunctions.Because
sequence, quantification of microRNAs by array-based
technologies has its limitations, as the hybridization con-
ditions used cannot be optimized for all microRNA
probes simultaneously. Previous tissue surveys used
cloning and sequencing to determine the microRNA
abundance in multiple tissues at low sequencing depth.
Whiletheseassays could
microRNAome,they nevertheless
microRNAs are expressed in a tissue-specific manner (7).
Recent studies have demonstrated that transcription
factors can regulate microRNA expression; however,
binding sites have been confirmed experimentally for
only a small number of microRNA promoters, and little
isknown aboutthemechanisms
tissue-specific expression of microRNAs (8–10). In order
toelucidatetheregulatory
tissue-specificexpressionof
determined their complete expression profile by ultra-high
throughput sequencing in three endoderm-derived tissues.
The greatly expanded number of differentially expressed
microRNAs identified through this method provided suf-
ficient sequence depth to determine the cis-regulatory
modulesthatcontrolthe
microRNA genes. Moreover, the results of our analysis
established that microRNA genes are governed by the
oftheirveryshort
notcapture
established
theentire
that
thatinfluence
networks
microRNA
that
genes,
govern
we
differentiallyexpressed
*To whom correspondence should be addressed. Tel: +1 215 503 6345; Fax: +1 215 503 6282; Email: linda.greenbaum@jefferson.edu
Correspondence may also be addressed to Klaus H. Kaestner. Tel: +1 215 898 8759; Fax: +1 215 573 5892; Email: kaestner@mail.med.upenn.edu
454–463
doi:10.1093/nar/gkq782
Nucleic Acids Research, 2011, Vol. 39, No. 2Published online 14 September 2010
? The Author(s) 2010. Published by Oxford University Press.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/
by-nc/2.5), which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Page 2

same transcription factor networks that also control
protein-coding genes.
MATERIALS AND METHODS
Processing microRNA reads
Male CD-1 mice aged 8–12-week-old (Charles River
Laboratories) underwent liver harvest between 8 AM
and 12 PM. Small intestinal mucosa was isolated via
mucosal scraping. Pancreata were harvested from the
same mice and snap frozen. Chromatin was later
prepared from a portion of the frozen tissue as described
earlier (11) except that protease inhibitors were also
included in the PBS and cell lysis buffers. H3K4Me3
ChIP was performed as described earlier (12). Total
and small RNAs were extracted using the mirVana
microRNA Isolation kit (Cat. # AM1561, Ambion,
Austin, TX, USA). Small RNAs libraries were generated
using the DGE-small RNA Sample prep kit (Cat. #
FC-102-1009; Illumina, San Diego, CA, USA). Illumina
sequencing libraries were prepared using the ‘long’
Illumina protocol according to the manufacturer’s direc-
tions. Purified PCR product was loaded on an Agilent
Technologies2100Bioanalyzer
quantity and integrity. Reads from six liver, five
small intestinal mucosa, and two pancreas samples were
sequenced on an Illumina GA-II following manufactur-
er’s instructions. The 30-adapter was trimmed from the
end of each read and the frequencies of the resulting
oligos were tabulated for each lane and in total. The
oligos were aligned to precursor hairpins (mirBase 14),
RefSeq sequences, and the mouse genome (NCBI Build
36; mm8) using ELAND and allowing up to two
mismatches. Alignments of reads in the length range
19–25bp were assigned to the mature microRNA that
they overlapped. When mature forms shared an oligo
in this length range, they were merged into an ad hoc
family for reporting read counts and for differential ex-
pression calculations. All high-throughput sequencing
data are accessible from the NCBI Short Read Archive
under accession number SRA023764.
toconfirmsample
Identifying differentially-expressed microRNAs
To identify differentially-expressed microRNAs we used
read counts in reads per million (RPM) from six replicates
from liver, five from small intestine, and two from
pancreas. The RPM values were quantile normalized in
R using the normalizeBetweenArrays function of the
limma package. These values were then analyzed using
SAMR, and microRNAs with an FDR ?10%, a
minimum of 1.5-fold change, and at least 100 RPM
average expression (in the appropriate tissue) were
selected as differentially expressed.
ChIP for histone modifications
Immunoprecipitations were performed as described earlier
(11), except that 4mg of chromatin and 4mg of antibodies
were used for each reaction. Chromatin was immunopre-
cipitated with antibodies for H3K4me3 (Millipore, Cat#
CS200580).
calculating enrichment of control liver, jejunal mucosa
and pancreas expressed genes using control intergenic
regions, by comparing input DNA to ChIP DNA. The
immunoprecipitated DNA was prepared for sequencing
as per Illumina’s instructions (http://www.illumina.com)
andpreviously described
sequencing was performed on an Illumina GA-II follow-
ing manufacturer’s instructions. The 36-bp reads were
aligned to the mouse genome (NCBI Build 36; mm8)
using ELAND and allowing up to two mismatches.
Reads with a unique best alignment were included in
further processing (liver H3K4me3 ChIP: 7720909,
input: 12056786;smallintestine
8492668, input: 11961006, pancreas H3K4me3 ChIP:
20176620).
Immunoprecipitationwasconfirmedby
(13). High-throughput
H3K4me3ChIP:
Identifying regions of significant H3K4me3 modification
We used GLITR (13) to identify regions of significant
enrichment of H3K4me3 as compared to input using a
1% FDR. Adjacent GLITR regions in each tissue were
merged if they were within 1500bp. The merged regions
were considered to be candidate TSSs. We created an
atlas of all H3K4me3 regions by merging overlapping
regions from all three endodermal tissues. To quantify
thestrengthof modification
computed the length-normalized rate of tissue-specific
reads (reads per kilo basepair), then applied quantile nor-
malization to correct for differences in total read count
and ChIP efficiency. The closest H3K4me3 peak with a
normalized intensity of at least 25 reads and within
200000-bp upstream of a pre-miRNA was considered
itsmostlikelytranscriptional
microRNAs with a normalized intensity of at least 32
RPM were considered.
in each tissue,we
startsite. Only
Identifying enriched predicted transcription
factor binding sites
We selected putative TSS for liver-expressed microRNA
genes. We extracted sequence covering ±2kb from the
middleoftheputative
poorly-conserved regions by removing areas with a
phastCons score <0.15 in the UCSC 17-way vertebrate
conservation track. The phastCons score ranges between
0 (not conserved) and 1 (well conserved). Regions with
similar dimensions anchored at the TSS of protein
coding genes were selected as a background set. The back-
ground set was chosen so that it had the same joint dis-
tribution of conserved sequence and base composition as
the microRNA promoters. Receiver-operating character-
istic (ROC) curves were computed for each vertebrate
PWM in TRANSFAC (v2009.2) by varying the scoring
threshold. P-values were computed for each score thresh-
old that yielded a hit in an additional microRNA TSS
using a chi-squared distribution for each threshold that
yielded hits >5 microRNA or background TSSs. The
best chi-squared P-value was tracked for each PWM.
The best P-value was corrected for multiple testing using
a Bonferroni factor equal to the number of tests, which we
took to be the number of positive regions microRNA TSS
TSSand maskedout
Nucleic AcidsResearch, 2011, Vol.39,No. 2 455

Page 3

regions (between 25 and 32 depending on the compari-
son), multiplied by the number of PWMs (644 vertebrate
PWMs). This is a conservative correction as many of the
TRANSFAC PWMs are similar and therefore do not con-
stitute independent trials which the Bonferroni correction
assumes. We also tested HNF4a sites using the SVM
(support vector machine) prediction method developed
by Sladek and colleagues (14). We submitted all sequences
to the website and tabulated the best score for each
masked sequence. The chi-squared P-value was 0.00013
which we corrected to 0.00013 * 39=0.0052.
Associating Foxa2-binding sites with TSS regions
We assumed that a Foxa2-binding event is likely to
regulate the gene or microRNA associated with the
closest H3K4me3 region to the Foxa2 site. However, in
cases where a binding site was located in the common
promoterofdivergently
array with overlapping genes, or in an intergenic region
adjacent to multiple genes, identifying the target gene was
less straightforward. We assigned experimentally-defined
Foxa2-binding events to the gene with a H3K4me3 region
that was closest to the Foxa2 site, as well as any additional
genes with a H3K4me3 region that was no >50% further
than the closest region.
transcribedgenes, in an
Visualizing genomic data
The positions of expressed miRNAs, genes, H3K4me3
regions and profiles, miRNA-TSS associations from this
and previous work were visualized using the TessLA
system which consists of a genome browser and a data
analysis tool kit (unpublished data).
RESULTS AND DISCUSSION
The microRNAome of liver, jejunum and pancreas
We isolated small RNA fractions from six livers, six
samples of small intestinal mucosa (from the jejunum)
and two pancreata, converted them into libraries, and
obtained 26574536, 55 885 851 and 38613301 sequence
reads for liver, jejunal mucosa and pancreas, respectively.
The resulting sequence reads were aligned to known
microRNA precursor genes, obtained from miRBase v14
(15), in order to assess the abundance of each mature
microRNA. Next, we verified that our sequence reads rep-
resented microRNAs and not degraded mRNAs by
aligning them to the RefSeq database as well. As shown
in Supplementary Figure S1, <10% of the reads in the
microRNA size range (21–23nt) aligned to mRNAs,
while >80% matched to precursor microRNAs, indicating
that our small RNA preparation was highly enriched for
true microRNAs. In total, we generated 19 754 019, 45 949
823 and 13 487 288 trimmed reads in the range of 19–25nt
that aligned to precursor microRNAs for liver, jejunum
and pancreas, respectively. Using these reads, we found
evidence for expression of 769 of the 1094 (69.9%)
known or predicted mature microRNAs, corresponding
to 459 of 547 (83.9%) pre-microRNAs. (Supplementary
TableS1).Weconfirmed high-levelexpression of
previously known abundant microRNAs in the respective
tissue, such as mir-122 and mir-192 in the liver, miR-215
and miR-192 in the intestine, and miR-375 and miR-152
in the pancreas (Supplementary Table S1) (16–20). The
let-7 family was highly expressed in all tissues. The extra-
ordinary dynamic range (about six orders of magnitude)
of the technology used allowed us to detect and quantify
microRNAs present in a few copies per million as well as
those that contribute up to ?44% of the total microRNA
pool, i.e. miR-122 in the liver. Because of technical limi-
tations of prior efforts, many of the microRNAs identified
here had been missed in previous studies (7,21).
Tissue-specific microRNA expression
Next, we determined the differential microRNA gene ex-
pression between the three tissues. To this end, we
employed computational tools established previously for
the analysis of microarray expression profiling (for details,
see ‘Materials and Methods’ section). After quantile nor-
malization, expression levels of the independent samples
for each tissue were compared. As shown in Figure 1,
most microRNA genes are expressed at similar levels
between any two tissues, suggesting that organs of
related developmental origin such as liver and intestinal
mucosa co-express many microRNA genes, just as they
co-express many protein-coding genes. However, 162
microRNA genes exhibited statistically significant enrich-
ment in either liver (63), small intestine (65) or pancreas
(96), with differential expression of up to 120000-fold.
The top 30 microRNAs enriched in each organ versus
the other two are listed in Tables 1–3. Full lists are avail-
able inSupplementary
Supplementary Figure S2.
TablesS2–S4;seealso
Mapping transcriptional start sites of microRNAs
The analysis of cis-regulatory elements requires know-
ledge of the promoter used for the microRNA gene in
question in the liver, small intestine or pancreas. To this
end, we took advantage of the recent discovery that tran-
scriptional start sites in the mammalian genome are
marked by trimethylated histone H3 (H3K4me3), often
in a characteristic double peak pattern (22). We performed
ChIP-Seq experiments for H3K4me3 in liver, small intes-
tine and pancreas, identified areas of H3K4me3 enrich-
ment, and then mapped these putative transcriptional
start sites to expressed microRNAs by associating each
microRNAwiththenearest
H3K4me3 occupancy. To increase the chances that a
microRNA would have an H3K4me3-marked TSS, we
only processed microRNAs that had an expression level
of at least 32 RPM, and required that the H3K4me3 en-
richment levels reached at least 25 (see ‘Materials and
Methods’ section for details.) Of a total of 17505
H3K4me3 regions present in at least one organ, we
identified 106 as putative TSS for a total of 128
pre-miRNAs (Supplementary Table S5). About 80% of
theTSSwerewithin
(Supplementary Figure S3). As found previously for the
analysis of microRNA promoters in embryonic stem cells,
between 73.3% (jejunum) and 77.6% (pancreas) (Table 4)
upstreamregionof
50kbofthemiRNA
456Nucleic Acids Research, 2011,Vol. 39,No. 2

Page 4

of the H3K4me3 loci covering putative transcriptional
start sites overlapped CpG islands, a DNA sequence
feature frequently associated
compared microRNA expression levels to the degree of
H3K4me3 occupancy at their promoters, but did not
observe any obvious correlation (data not shown).
Comparing our results with previous efforts that had
mapped transcriptional start sites in ES cells (23), we
found agreement at 59 of 106 (55.7%) of our H3K4me3
loci. Reasons for this discrepancy might include the
reliance on historic H3K4me3 mapping data in the
previous study, and the use of different transcriptional
start sites in liver and intestinal mucosa compared to em-
bryonic stem cells. We observe that the average concord-
ance rate within a given organ (64.5%) is higher than the
rate over all identified miRNA/TSS pairs, which is con-
sistent with the idea that the TSS concordant between ES
cells and a given organ or tissue correspond to ‘housekeep-
ing’ miRNAs which are used in most tissues. As we deter-
mine TSS indifferent
we identify additional tissue-specific microRNA/TSS,
whilestill re-identifying
with promoters.We
endoderm-derivedtissues,
thesamehousekeeping
microRNA/TSS, thus lowering the total concordance
rate. A few examples of this latter category are discussed
below. Figures 2 and 3 illustrate the advantage of identify-
ing tissue-specific microRNA transcriptional start sites
using H3K4me3 mapping in tissues that express the
microRNA in question, and also illustrate the regulatory
complexity of miRNAs. In Figure 4 we show strong local
promoters active in liver and small intestine where
miRNAs miR-192 and miR-194 are highly-expressed. In
pancreas, the expression of these miRNAs is about 10?
lower, though still strong, but the level of H3K4me3 has
been reduced to background levels. Marson and col-
leagues identified a promoter region further upstream in
the Atg2a gene, which is also at background levels in
pancreas (23). However, the CpG-containing promoter
for Atg2a is active, suggesting that it may be the source
of miR-192 and miR-194 in the pancreas.
Cis-regulatory elements of differentially expressed
microRNA genes
Next we employed positional weight matrices (PWMs)
to identify potential binding sites of tissue-enriched
Figure 1. Differential expression of microRNAs in three endoderm-derived tissues. microRNAs were identified as differentially expressed in (A) liver
versus small intestinal epithelium, (B) liver versus pancreas and (C) small intestine versus pancreas using an FDR of 10% and a minimum fold
change of 1.5?. Several abundant microRNAs such as mmu-miR-192 are expressed at similar levels in both tissues (dots near the orange line). Each
tissue has one highly-expressed, highly-differential microRNA as well as ?30 other microRNAs that exhibit significant differential expression, which
are highlighted in green (down) and red (up).
Nucleic AcidsResearch, 2011, Vol.39,No. 2457

Page 5

transcription factors. We compared the occurrence of
the best match to each PWM in the conserved regions
surrounding the TSSs (for simplicity referred to as ‘pro-
moters’ below) with their occurrence in randomly se-
lected promoters from protein-coding genes. We used
the sets of promoters that were associated with liver-,
small intestine-,orpancreas-enriched
Because we were interested in identifying motifs that
may cover only a subset of microRNA promoters, we
developed a novel enrichment statistic that empha-
sizes the enrichment of high-scoring motif matches in
agroup ofpromoters,
strong motifs in all or even most of the tissue-specific
promoters (see ‘Materials and Methods’ section for
details).
In the set of liver-enriched microRNA genes, predicted
sites for the factors MYB (M00183, pv=3.1e–5), and
SREBP (M01168, pv=4.8e–3) were enriched within
1000bp of H3K4me3 regions. Considering just the
miRNA-only H3K4me3regions,
(M00916pv=1.2e–4),
pv=1.8e–4 or M00237 pv=2.6e–3) and E2F (M00938
pv=6.8e–3) to be enriched within 3000–8000bp. All of
these factors are known to be relevant to the function of
the liver. Because of the major role played by the nuclear
microRNAs.
butdoes notrequire
wefoundCREB
(M00778AHR/ARNT
receptor HNF4a in gene regulation in hepatocytes, we
assessed the potential for regulation of liver microRNAs
by HNF4a using a recently developed support vector
machine-based prediction method (14), and found the
HNF4a
motifsignificantly
P-value=0.0052) among the liver-expressed microRNA
promoters. In fact, eight microRNA genes expressed pref-
erentially in the liver contain predicted HNF4a-binding
sites above the recommended threshold. We note that
HNF4a is a hepatocyte-specific transcription factor
which is not expressed in all cell types in the liver; there-
fore, there may be a set of liver-specific miRNAs that are
not expressed hepatocytes, for which HNF4a is irrelevant.
Next, we sought experimental evidence of binding of
tissue-specific transcriptional regulators near microRNA
TSSs. Foxa2 is an important transcriptional regulator of
liver development and function (11,24–29) so we checked
previously published experimental data for Foxa2 binding
(13) and found that indeed seven of the putative
liver-enriched microRNAs TSSs (associated with 10
microRNAs) were within 2KB of experimentally defined
Foxa2 sites (Table 5). A further eight miRNA had a
Foxa2 site within 15kb of the TSS. Four of the miRNA
TSS were not associated with a protein-coding gene and
thus represent evidence of miRNA-specific regulation by
enriched (corrected
Table 1. Top 20 microRNAs enriched in liver versus jejunum and/or
Pancreas
miRNALiver
[RPM]
Ratio
[log2]
Jejunum
[RPM]
Ratio
[log2]
Pancreas
[RPM]
mmu-miR-122
mmu-miR-122-3p
mmu-miR-1948
mmu-miR-485*
mmu-miR-455
mmu-miR-21
mmu-miR-340-5p
mmu-miR-193
mmu-miR-101b
mmu-miR-1937a
mmu-miR-30a*
mmu-miR-30c-2*
mmu-miR-107
mmu-miR-22*
mmu-miR-31
mmu-miR-99a
mmu-miR-221
mmu-miR-29b-2
mmu-miR-192
mmu-miR-98
mmu-miR-127
mmu-miR-486
mmu-miR-194-2
mmu-miR-451
mmu-miR-125b-1-5p
mmu-miR-541
mmu-miR-142-5p
mmu-miR-125a-5p
mmu-miR-130a
mmu-miR-126-5p
447350.96
188.71
138.99
102.56
199.55
27344.49
332.16
116.31
6708.91
86.26
881.83
244.84
7795.86
269.87
578.92
394.09
1992.34
294.92
136451.85
770.18
352.69
185.32
719.8
730.88
320.5
101.72
228.22
627.32
429.6
120.17
13.69
9.53
6.96
8.57
6.16
–
2.26
5.4
2.79
5.26
5.03
4.95
2.92
4.62
–
4.48
4.48
–
–
2.51
3.96
3.95
–
3.89
3.8
3.8
–
3.63
3.45
3.37
33.75
0.25
1.12
0.27
2.78
–
69.11
2.76
968.95
2.25
26.96
7.94
1032.13
10.96
–
17.63
89.2
–
–
135.44
22.7
11.98
–
49.29
23.05
7.32
–
50.61
39.31
11.66
15.99
9.96
8.86
6.22
6.49
5.53
5.53
–
5.29
5
1.64
–
4.82
4.45
4.58
–
–
4.46
4.31
4
–
1.66
3.93
–
–
1.63
3.63
–
3.61
3.54
6.86
0.19
0.3
1.37
2.23
591.84
7.21
–
171.02
2.69
282.37
–
276.37
12.38
24.21
–
–
13.43
6889.72
48.17
–
58.51
47.28
–
–
32.79
18.38
–
35.08
10.33
All expression levels are expressed as reads per million. Selection was
based on a fold change of at least 1.5 and a false discovery rate of
10%.
Table 2. microRNAs enriched in Small Intestine versus Liver and/or
Pancreas
miRNA Jejunum
[RPM]
Ratio
[log2]
Liver
[RPM]
Ratio
[log2]
Pancreas
[RPM]
mmu-miR-215
mmu-miR-215-3p
mmu-miR-194-1
mmu-miR-200c
mmu-miR-375
mmu-miR-145
mmu-miR-194-1-3p
mmu-miR-141
mmu-miR-363-5p
mmu-miR-429
mmu-miR-1-1
mmu-miR-200a
mmu-miR-200b
mmu-miR-194-2
mmu-miR-33
mmu-miR-200b*
mmu-miR-31
mmu-miR-21
mmu-miR-142-5p
mmu-miR-130b
mmu-miR-1-2
mmu-miR-142-3p
mmu-miR-20a
mmu-miR-872
mmu-miR-1274a
mmu-miR-182
mmu-miR-192
mmu-miR-143
mmu-miR-29b-2
mmu-miR-203
447350.96
2221.8
3448.07
3495.91
408.84
4903.86
63.24
68
68.34
693.71
1429.3
4802.45
3905.74
3769.81
136.43
401.44
1343.17
31885.58
986.36
82.83
880.44
108.57
94.27
58
61.77
63.73
102132.57
1357.07
172.11
1660.92
14.71
13.49
9.99
9.33
8.78
8.49
8.36
7.86
5.88
7.3
6.34
6.53
6.43
2.39
1.53
5.93
1.21
–
2.11
5.68
5.64
4.96
3.56
–
4.31
4.07
–
3.79
–
2.66
16.73
0.19
3.38
5.41
0.93
13.6
0.19
0.29
1.16
4.4
17.67
51.99
45.17
719.8
47.12
6.6
578.92
–
228.22
1.61
17.67
3.49
8.02
–
3.12
3.8
–
98.12
–
262.14
14.21
13.4
11.36
–
–
–
8.39
2.61
7.31
2.55
7.13
2.03
–
6.32
6.23
1.48
5.79
5.75
5.75
–
5.61
4
4.6
4.58
–
–
3.89
–
3.68
3.63
23.66
0.21
1.31
–
–
–
0.19
11.12
0.43
118.23
10.18
1179.74
–
47.28
1.81
143.63
24.21
591.84
18.38
–
18.04
6.78
3.89
2.42
–
–
6889.72
–
13.43
134.5
All expression levels are in RPM. Selection was based on a fold change
of at least 1.5 and a false discovery rate of 10%.
458Nucleic Acids Research, 2011,Vol. 39,No. 2

Page 6

Foxa2. In Figure 2A and B we illustrate two examples
where wehaveidentified
microRNA TSS that is occupied by Foxa2 within a few
hundred base pairs of the TSS.
anovel liver-specific
CONCLUSIONS
Here we have taken advantage of the substantial dynamic
range of ultra-high throughput sequencing to detect and
quantify microRNAs in three endoderm-derived tissues.
Many of the microRNAs identified here had been
missed in previous studies (21). However, despite the
fact that we obtained >90 million sequence reads for
small RNAs, no new microRNAs were discovered,
strongly suggesting that mirBase covers all or nearly all
existing microRNA genes.
As one might expect for developmentally related tissues
such as liver, jejunum and pancreas, most of the
microRNA genes are expressed at roughly similar levels
in the three tissues. However, we also discovered
tissue-specific expression between pairs of tissues, in
some cases over several orders of magnitude. We utilized
experimental mapping of transcriptional start sites in
order tolocalize thepromoters
tissue-specific gene activation. Importantly, we found
multiple cases where different transcriptional start sites
are used by microRNA genes in endoderm-derived
tissues as opposed to embryonic stem cells. These
findings suggest that understanding microRNA gene pro-
moters requires their experimental validation in each
tissue of interest. We did not measure repressive chroma-
tin marks, e.g. H3K27me3, in this work, so potentially
active H3K4me3-marked TSS need to be evaluated for
absence of repressive marks to further validate their
activity.
Finally, we analyzed the cis-regulatory elements that
contribute to the regulation of microRNA gene expres-
sion in liver, jejunum, and pancreas. We find that the
same major transcription factors that regulate the
tissue-specific expression of protein-coding genes also
contribute to the regulation
suggesting that both classes of genes utilize the same
fundamentalregulatory mechanisms.
extend earlierfindings by
that directthis
ofmicroRNA genes,
These
For
results
example, others.
Table 3. microRNAs enriched in pancreas versus liver and/or small intestine
miRNAPancreas
[RPM]
Ratio
[log2]
Liver
[RPM]
Ratio
[log2]
Jejunum
[RPM]
mmu-miR-375
mmu-miR-217
mmu-miR-216b
mmu-miR-200c
mmu-miR-672
mmu-miR-138-1
mmu-miR-676
mmu-miR-184
mmu-miR-148a
mmu-miR-148a*
mmu-miR-99a
mmu-miR-145
mmu-miR-200b
mmu-miR-183
mmu-miR-10b
mmu-miR-7b
mmu-miR-155
mmu-miR-2143-3
mmu-miR-130b
mmu-miR-224
mmu-miR-141
mmu-miR-802
mmu-miR-676*
mmu-miR-152
mmu-miR-2133-2-3p
mmu-miR-802-3p
mmu-miR-1195-3p
mmu-miR-429
mmu-miR-34a
mmu-miR-1274a
115969.49
3177.41
257.11
2528.08
226.77
89.86
779.03
173.65
24381.18
63.48
1814.3
1353.92
3572.31
72.59
50.92
16.55
20.46
494.87
79.92
13.32
11.12
78.2
39.2
18175.19
86.25
112.04
254.2
118.23
25.21
78.33
16.93
10.54
8.88
8.87
8.23
8.22
3.98
7.44
3.82
4.77
2.2
6.64
6.31
6.21
6.2
6.02
5.88
5.78
5.63
5.3
5.25
5.25
2.06
3.15
4.96
4.88
4.77
4.75
4.74
4.65
0.93
2.13
0.55
5.41
0.75
0.3
49.21
1
1721.49
2.33
394.09
13.6
45.17
0.98
0.69
0.25
0.35
9
1.61
0.34
0.29
2.06
9.4
2053.99
2.78
3.81
9.35
4.4
0.95
3.12
8.15
12.77
9.26
–
4.29
4.12
8
5.4
6.86
6.76
6.69
–
–
–
2.46
408.84
0.46
0.42
–
11.62
5.16
3.05
4.1
209.89
0.58
17.63
–
–
–
9.27
2.95
4.17
–
–
–
1.3
5.17
5.04
4.97
1.55
3.86
–
4.22
–
2.65
27.48
–
–
–
31.67
1.09
552.97
2.75
38.37
17.51
–
1.35
–
All expression levels are in RPM. Selection was based on a fold change of at least 1.5 and a false discovery rate of 10%.
Table 4. Overlap between putative miRNA TSS and other regulatory
features
Organ TSS
Regions
CpG Island,
%
ESC Promoters,
%
Foxa2 Sites,
%
Liver
Jejunum
Pancreas
All
74
75
76
77.0
73.3
77.6
74.5
62.2
66.7
64.5
55.7
10.8
8.0
9.2
7.5106
Nucleic AcidsResearch, 2011, Vol.39,No. 2 459

Page 7

Figure 2. Transcriptional start sites of microRNAs enriched in the liver. (A) Shows mmu-miR-122 that is highly expressed in liver and expressed
about 1000–2000? lower in jejunal epithelium and pancreas. Each section (liver, jejunum, and pancreas) shows the profile of H3K4me3, regions of
significant levels of H3K4me3, the link from the miRNA to the nearest H3K4me3 peak, and the log2normalized expression level of miRNAs. Note
the presence of H3K4me3-modified histones at the transcriptional start site in the liver (olive). No H3K4me3 is present in jejunum (purple) or
pancreas (yellow). Additionally, this TSS was not identified in mouse embryonic stem cells [TSS from (23) in magenta]. This locus also includes an
example of Foxa2 binding. In (B), mmu-miR-101b is located within the protein-coding Rcl1 gene, which is not highly expressed in liver, yet the
microRNA TSS shows clear evidence of liver-specific H3K4me3. mmu-miR-101b is expressed at lower levels in jejunum and pancreas and may the
use Rcl1 promoter.
460Nucleic Acids Research, 2011,Vol. 39,No. 2

Page 8

Figure 3. Transcriptional start sites of microRNAs enriched in the small intestinal epithelium or pancreas. (A) mmu-miR-215 and mmu-miR-194-1
are specific to jejunum and have a strong proximal jejunum-specific H3K4me3 region which is internal to the Iars2 gene. (B) mmu-miR-375, situated
just downstream of Ccdc108 gene, has an adjacent region of H3K4me3 [also identified in ref. (23)] that is absent in liver and jejunum. mmu-miR-375
is expressed in jejunum, but the promoter identified is at the Ihh gene about 50-kb upstream.
Nucleic AcidsResearch, 2011, Vol.39,No. 2461

Page 9

SREBP-1c
miRNAs in skeletal muscle in response to insulin signal-
ing (30), so it is exciting to find evidence for a direct link
in liver between SREBP-1c and microRNA regulation, as
this is a tissue which also responds to insulin. Similarly,
c-Myb has been shown to be both a target and a
regulator of miRNA-15a in K562 myeloid leukemia
cells (31).
hasbeenshown toindirectlyregulate
SUPPLEMENTARY DATA
Supplementary Data are available at NAR Online.
ACKNOWLEDGEMENTS
The authors thank Alan Fox and Olga Smirnova for
expert technical assistance
Figure 4. Transcriptional start sites for miR-192 and miR-194 which are expressed in all three tissues, but at decreased levels in pancreas where the
H3K4me3 region at chr19:6263000 is nearly absent. The ES-cell promoter from (23) does not appear to be active in any of these tissues, but a CpG+
promoter for the Atg2 gene is active and may be the source of these miRNAs in the pancreas.
Table 5. Location of Foxa2 sites within 10kb of miRNA/TSS
Foxa2 LocationTSS/miRNA LocationDistance [bp]miRNA
chr11:86403625–86403809
chr6:31152438–31152697
chr1:196718801–196719044
chr1:196718801–196719044
chr5:138400937–138401132
chr5:138400937–138401132
chr4:100854070–100854278
chr19:29194307–29194636
chr5:138400704–138400930
chr5:138400704–138400930
chr4:100853701–100853887
chr16:18255150–18255448
chr5:119931308–119931521
chr13:48564064–48564222
chr13:48564064–48564222
chr13:48564064–48564222
chr11:79470698–79470956
chr11:86400262–86403879
chr6:30992823–31151892
chr1:196717762–196737844
chr1:196717762–196738358
chr5:138395109–138402675
chr5:138395311–138402675
chr4:100844877–100855891
chr19:29192749–29201372
chr5:138395109–138402675
chr5:138395311–138402675
chr4:100844877–100855891
chr16:18240964–18257550
chr5:119786217–119936556
chr13:48547949–48558966
chr13:48549766–48558966
chr13:48550116–48558966
chr11:79476766–79542706
162
675
1160
1160
1641
1641
1717
1722
1858
1858
2097
2251
5142
?5177
?5177
?5177
?5939
mmu-mir-21
mmu-mir-29a
mmu-mir-29b-2
mmu-mir-29c
mmu-mir-25
mmu-mir-93
mmu-mir-101a
mmu-mir-101b
mmu-mir-25
mmu-mir-93
mmu-mir-101a
mmu-mir-185
mmu-mir-1959
mmu-let-7d
mmu-let-7f-1
mmu-let-7a-1
mmu-mir-365-2
462Nucleic Acids Research, 2011,Vol. 39,No. 2

Page 10

FUNDING
National Institutes of Health (DK056669 to L.E.G.,
DK049210 and DK053839 to K.H.K.); American Liver
Foundation Innovative Seed grant (to L.E.G.); Core
services provided by the DERC at the University of
Pennsylvania from a grant sponsored by National
Institutes of Health (P30-DK19525). Funding for open
access charge: DK049210 grant.
Conflict of interest statement. None declared.
REFERENCES
1. Shabalina,S.A. and Koonin,E.V. (2008) Origins and evolution of
eukaryotic RNA interference. Trends Ecol. Evol., 23, 578–587.
2. Sekine,S., Ogawa,R., Ito,R., Hiraoka,N., McManus,M.T.,
Kanai,Y. and Hebrok,M. (2009) Disruption of Dicer1 induces
dysregulated fetal gene expression and promotes
hepatocarcinogenesis. Gastroenterology, 136, 2304–2315
e2301–e2304.
3. Sekine,S., Ogawa,R., McManus,M.T., Kanai,Y. and Hebrok,M.
(2009) Dicer is required for proper liver zonation. J. Pathol., 219,
365–372.
4. McKenna,L.B., Schug,J., Vourekas,A., McKenna,J.B.,
Bramswig,N., Friedman,J.R. and Kaestner,K.H. (2010)
MicroRNAs control intestinal epithelial differentiation,
architecture, and barrier function. Gastroenterology, doi:10.1053/
j.gastro.2010.07.040.
5. Chen,J.F., Murchison,E.P., Tang,R., Callis,T.E., Tatsuguchi,M.,
Deng,Z., Rojas,M., Hammond,S.M., Schneider,M.D.,
Selzman,C.H. et al. (2008) Targeted deletion of Dicer in the heart
leads to dilated cardiomyopathy and heart failure. Proc. Natl
Acad. Sci. USA, 105, 2111–2116.
6. Lee,Y., Kim,M., Han,J., Yeom,K.H., Lee,S., Baek,S.H. and
Kim,V.N. (2004) MicroRNA genes are transcribed by RNA
polymerase II. EMBO J., 23, 4051–4060.
7. Landgraf,P., Rusu,M., Sheridan,R., Sewer,A., Iovino,N.,
Aravin,A., Pfeffer,S., Rice,A., Kamphorst,A.O., Landthaler,M.
et al. (2007) A mammalian microRNA expression atlas based on
small RNA library sequencing. Cell, 129, 1401–1414.
8. Grove,C.A. and Walhout,A.J. (2008) Transcription factor
functionality and transcription regulatory networks. Mol. Biosyst.,
4, 309–314.
9. O’Donnell,K.A., Wentzel,E.A., Zeller,K.I., Dang,C.V. and
Mendell,J.T. (2005) c-Myc-regulated microRNAs modulate E2F1
expression. Nature, 435, 839–843.
10. Martinez,N.J., Ow,M.C., Barrasa,M.I., Hammell,M., Sequerra,R.,
Doucette-Stamm,L., Roth,F.P., Ambros,V.R. and Walhout,A.J.
(2008) A C. elegans genome-scale microRNA network contains
composite feedback motifs with high flux capacity. Genes Dev.,
22, 2535–2549.
11. Tuteja,G., Jensen,S.T., White,P. and Kaestner,K.H. (2008)
Cis-regulatory modules in the mammalian liver: composition
depends on strength of Foxa2 consensus site. Nucleic Acids Res.,
36, 4149–4157.
12. Bhandare,R., Schug,J., Le Lay,J., Fox,A., Smirnova,O., Liu,C.,
Naji,A. and Kaestner,K.H. Genome-wide analysis of histone
modifications in human pancreatic islets. Genome Res., 20,
428–433.
13. Tuteja,G., White,P., Schug,J. and Kaestner,K.H. (2009)
Extracting transcription factor targets from ChIP-Seq data.
Nucleic Acids Res., 37, e113.
14. Bolotin,E., Liao,H., Ta,T.C., Yang,C., Hwang-Verslues,W.,
Evans,J.R., Jiang,T. and Sladek,F.M. (2010) Integrated approach
for the identification of human hepatocyte nuclear factor 4alpha
target genes using protein binding microarrays. Hepatology, 51,
642–653.
15. Griffiths-Jones,S., Bateman,A., Marshall,M., Khanna,A. and
Eddy,S.R. (2003) Rfam: an RNA family database. Nucleic Acids
Res., 31, 439–441.
16. Akao,Y., Nakagawa,Y. and Naoe,T. (2006) let-7 microRNA
functions as a potential growth suppressor in human colon cancer
cells. Biol. Pharm. Bull., 29, 903–906.
17. Gregory,P.A., Bracken,C.P., Bert,A.G. and Goodall,G.J. (2008)
MicroRNAs as regulators of epithelial-mesenchymal transition.
Cell Cycle, 7, 3112–3118.
18. Hino,K., Fukao,T. and Watanabe,M. (2007) Regulatory
interaction of HNF1-alpha to microRNA-194 gene during
intestinal epithelial cell differentiation. Nucleic Acids Symp. Ser.,
415–416.
19. Tzur,G., Levy,A., Meiri,E., Barad,O., Spector,Y., Bentwich,Z.,
Mizrahi,L., Katzenellenbogen,M., Ben-Shushan,E., Reubinoff,B.E.
et al. (2008) MicroRNA expression patterns and function in
endodermal differentiation of human embryonic stem cells.
PLoS One, 3, e3726.
20. Wang,K., Zhang,S., Marzolf,B., Troisch,P., Brightman,A., Hu,Z.,
Hood,L.E. and Galas,D.J. (2009) Circulating microRNAs,
potential biomarkers for drug-induced liver injury. Proc. Natl
Acad. Sci. USA, 106, 4402–4407.
21. Cummins,J.M., He,Y., Leary,R.J., Pagliarini,R., Diaz,L.A. Jr,
Sjoblom,T., Barad,O., Bentwich,Z., Szafranska,A.E., Labourier,E.
et al. (2006) The colorectal microRNAome. Proc. Natl Acad. Sci.
USA, 103, 3687–3692.
22. Barski,A., Cuddapah,S., Cui,K., Roh,T.Y., Schones,D.E.,
Wang,Z., Wei,G., Chepelev,I. and Zhao,K. (2007) High-resolution
profiling of histone methylations in the human genome. Cell, 129,
823–837.
23. Marson,A., Levine,S.S., Cole,M.F., Frampton,G.M.,
Brambrink,T., Johnstone,S., Guenther,M.G., Johnston,W.K.,
Wernig,M., Newman,J. et al. (2008) Connecting microRNA genes
to the core transcriptional regulatory circuitry of embryonic stem
cells. Cell, 134, 521–533.
24. Bochkis,I.M., Rubins,N.E., White,P., Furth,E.E., Friedman,J.R.
and Kaestner,K.H. (2008) Hepatocyte-specific ablation of Foxa2
alters bile acid homeostasis and results in endoplasmic reticulum
stress. Nat. Med., 14, 828–836.
25. Bochkis,I.M., Schug,J., Rubins,N.E., Chopra,A.R., O’Malley,B.W.
and Kaestner,K.H. (2009) Foxa2-dependent hepatic gene
regulatory networks depend on physiological state. Physiol.
Genomics, 38, 186–195.
26. Lee,C.S., Friedman,J.R., Fulmer,J.T. and Kaestner,K.H. (2005)
The initiation of liver development is dependent on Foxa
transcription factors. Nature, 435, 944–947.
27. Li,Z., White,P., Tuteja,G., Rubins,N., Sackett,S. and
Kaestner,K.H. (2009) Foxa1 and Foxa2 regulate bile duct
development in mice. J. Clin. Invest., 119, 1537–1545.
28. Sund,N.J., Ang,S.L., Sackett,S.D., Shen,W., Daigle,N.,
Magnuson,M.A. and Kaestner,K.H. (2000) Hepatocyte nuclear
factor 3beta (Foxa2) is dispensable for maintaining the
differentiated state of the adult hepatocyte. Mol. Cell Biol., 20,
5175–5183.
29. Zhang,L., Rubins,N.E., Ahima,R.S., Greenbaum,L.E. and
Kaestner,K.H. (2005) Foxa2 integrates the transcriptional
response of the hepatocyte to fasting. Cell Metab., 2, 141–148.
30. Granjon,A., Gustin,M.P., Rieusset,J., Lefai,E., Meugnier,E.,
Guller,I., Cerutti,C., Paultre,C., Disse,E., Rabasa-Lhoret,R. et al.
(2009) The microRNA signature in response to insulin reveals its
implication in the transcriptional action of insulin in human
skeletal muscle and the role of a sterol regulatory element-binding
protein-1c/myocyte enhancer factor 2C pathway. Diabetes, 58,
2555–2564.
31. Zhao,H., Kalota,A., Jin,S. and Gewirtz,A.M. (2009) The c-myb
proto-oncogene and microRNA-15a comprise an active
autoregulatory feedback loop in human hematopoietic cells.
Blood, 113, 505–516.
Nucleic AcidsResearch, 2011, Vol.39,No. 2463