This study analyzed microRNA (miRNA) expression in human embryonic stem cells and definitive endoderm cells using next-generation sequencing. Over 700 miRNAs were identified in each cell type, including 77 that were differentially expressed between the two stages. Four of the top five upregulated miRNAs in definitive endoderm cells had not been previously detected. The miR-302a stem-loop produced several abundant miRNA species (isomiRs) with distinct targets, indicating they have unique functions. Tuberous sclerosis 1 was identified as a specific target of the miR-302a-5p(13) isomiR but not the canonical miR-302a-5p, showing isomiRs can have non-overlapping roles

1. sRNA-seq Analysis of Human Embryonic Stem Cells
and Definitive Endoderm Reveals Differentially
Expressed MicroRNAs and Novel IsomiRs with
Distinct Targets
ANDREW HINTON,a
SHAUN E. HUNTER,a
IVKA AFRIKANOVA,a
G. ADAM JONES,a
ANA D. LOPEZ,a
GARY B. FOGEL,b
ALBERTO HAYEK,a
CHARLES C. KING
a
Key Words. Embryonic stem cells • miRNA • Differentiation • Pluripotency
ABSTRACT
MicroRNAs (miRNAs) are noncoding, regulatory RNAs expressed dynamically during differentia-
tion of human embryonic stem cells (hESCs) into defined lineages. Mapping developmental
expression of miRNAs during transition from pluripotency to definitive endoderm (DE) should
help to elucidate the mechanisms underlying lineage specification and ultimately enhance dif-
ferentiation protocols. In this report, next generation sequencing was used to build upon our
previous analysis of miRNA expression in human hESCs and DE. From millions of sequencing
reads, 747 and 734 annotated miRNAs were identified in pluripotent and DE cells, respectively,
including 77 differentially expressed miRNAs. Among these, four of the top five upregulated
miRNAs were previously undetected in DE. Furthermore, the stem-loop for miR-302a, an impor-
tant miRNA for both hESCs self-renewal and endoderm specification, produced several highly
expressed miRNA species (isomiRs). Overall, isomiRs represented >10% of sequencing reads in
>40% of all detected stem-loop arms, suggesting that the impact of these abundant miRNA spe-
cies may have been overlooked in previous studies. Because of their relative abundance, the
role of differential isomiR targeting was studied using the miR-302 cluster as a model system. A
miRNA mimetic for miR-302a-5p, but not miR-302a-5p(13), decreased expression of orthoden-
ticle homeobox 2 (OTX2). Conversely, isomiR 302a-5p(13) selectively decreased expression of
tuberous sclerosis protein 1, but not OTX2, indicating nonoverlapping specificity of miRNA proc-
essing variants. Taken together, our characterization of miRNA expression, which includes novel
miRNAs and isomiRs, helps establish a foundation for understanding the role of miRNAs in DE
formation and selective targeting by isomiRs. STEM CELLS 2014;32:2360–2372
INTRODUCTION
MicroRNAs (miRNAs) are small noncoding
RNAs that regulate expression of protein cod-
ing genes [1]. miRNA biogenesis begins with
transcription of a primary transcript containing
a stem-loop that is subsequently released by
Drosha cleavage [2]. The resulting hairpin RNA
is transported to the cytoplasm by Exportin-5
and processed into a miRNA-5p:miRNA-3p
duplex by Dicer. The mature 22 nucleotide
(nt) miRNA strand of the duplex is then prefer-
entially incorporated into the RNA-induced
silencing complex, and serves as a sequence-
specific guide for the negative regulation of
target mRNAs. There are currently several hun-
dred experimentally validated miRNA genes in
the human genome, predicted to target many
thousands of mRNAs. Investigation of miRNA
function has primarily focused on annotated
miRNAs. However, there remains an active
search for new miRNAs and increasing evi-
dence suggests a biological role for isomiRs,
which result from alternative Drosha and Dicer
processing [3].
Human embryonic stem cells (hESCs) are
defined by their self-renewal and potential to
differentiate into any cell type, and are there-
fore a potential source for therapeutically use-
ful cells, such as b cells for the treatment for
type 1 diabetes. However, a clearer understand-
ing of the molecular mechanisms and signaling
pathways which regulate cell differentiation
and lineage specification is desirable to develop
more efficient protocols for directed differentia-
tion. Temporal expression of various miRNAs
during hESCs differentiation is likely to play a
critical role in cell fate decisions. In fact, miR-
NAs play a role in the maintenance of hESCs
pluripotency and proliferation [4–6], and spe-
cific miRNAs have also been shown to influence
lineage specification, including endocrine
a
Pediatric Diabetes Research
Center, University of
California, San Diego, La
Jolla, California, USA;
b
Natural Selection, Inc., San
Diego, California, USA
Correspondence: Charles C.
King, Ph.D., Pediatric Diabetes
Research Center, University of
California, San Diego, La Jolla,
California 92121, USA.
Telephone: 858-822-4720; Fax:
858-822-1966; e-mail:
chking@ucsd.edu
Received June 13, 2012;
accepted for publication April 9,
2014; first published online in
STEM CELLS EXPRESS May 8,
2014.
VC AlphaMed Press
1066-5099/2014/$30.00/0
http://dx.doi.org/
10.1002/stem.1739
STEM CELLS 2014;32:2360–2372 www.StemCells.com VC AlphaMed Press 2014
EMBRYONIC STEM CELLS/INDUCED
PLURIPOTENT STEM CELLS

2. pancreas [7–11]. To date, the function of the vast majority of
miRNAs detected in stem cells remains poorly characterized.
An important step in elucidating their function is to establish
differences in expression patterns as new cell types are created
throughout differentiation.
In this study, we used a cell culture model-system of in
vitro development for stage-to-stage differentiation of pluripo-
tent hESCs to pure populations of definitive endoderm (DE),
the first stage in pancreatic genesis. Previously, we used micro-
arrays to identify distinctive miRNA expression signatures in
these two cell populations [7]. A limitation of this approach
was the inability to detect novel miRNAs and a large number
of known miRNAs that were not yet available for chip-based
analysis. Here, we extend our findings using next generation
sequencing for both cell populations as the best current
approach to distinguish closely related miRNAs and discover
novel isoforms. We analyzed millions of sequencing reads from
pluripotent and DE cells and identified the differential expres-
sion of 77 significantly expressed miRNAs. Four of the five
most highly upregulated miRNAs (hsa-miR-1263, hsa-miR-1247-
3p, hsa-miR-212-5p, and hsa-miR-132-3p) were previously
undetected in DE. Interestingly, several miRNAs that have been
previously described as hESCs specific were maintained at high
levels specifically in DE following differentiation. We also
observed expression of novel miRNAs and isomiRs. Moreover,
the stem loop for miR-302a, an important miRNA for both
hESCs self-renewal and endoderm specification [10, 12], pro-
duced several highly expressed miRNA species. Specifically,
miR-302a-5p (13) was expressed at much higher levels than
other isomiRs of the miR-302a locus. Tuberous sclerosis 1
(TSC1) and orthodenticle homeobox 2 (OTX2) were identified
as a specific target of miR-302a-5p (13) and 302a-5p, respec-
tively, thus indicating distinct biological roles of these isomiRs.
Taken together, deep sequencing analysis has significantly
expanded the number of annotated miRNAs included in the
molecular profile of hESCs and DE. Given the current lack of
functional characterization of most miRNAs, the abundance and
dynamic expression of unannotated isomiR species suggests an
important biological role also for these novel miRNAs.
MATERIALS AND METHODS
Cell Culture
CyT49 (provided by ViaCyte, San Diego, CA), H1, and H9 cells
were maintained on a sparse layer of mitomycin-C-treated
mouse feeder layers at 37
C, 5% CO2 in Dulbecco’s modified
Eagle’s medium (DMEM)/F-12 supplemented with 20% knock-
out serum replacement, glutamax, nonessential amino acids,
b-mercaptoethanol, and penicillin/streptomycin (Life Technolo-
gies, Carlsbad, CA, http://www.lifetechnologies.com/us/en/
home.html). Medium was replaced daily with 4 ng/ml basic
fibroblast growth factor (Peprotech, Rocky Hill, NJ, http://
www.peprotech.com/en-US) and 10 ng/ml activin A (RD Sys-
tems, Minneapolis, MN, http://www.rndsystems.com/). For
feeder-free cultures, hESCs were plated on BD matrigel and
maintained in medium conditioned by mouse embryonic
fibroblasts as described [13]. Differentiation to DE was carried
out in RPMI (Roswell Park Memorial Institute; Mediatech,
Inc., Manassas, VA, http://www.cellgro.com/) with varying
concentrations of defined FBS (HyClone, Logan, UT, http://
www.thermoscientific.com/content/tfs/en/about-us/general-
landing-page/thermo-scientific-hyclone.html?ca=hyclone): 0%
at days 0–1, 0.2% at days 1–3, and 2% at days 3–4. Addition-
ally, cells were treated with 100 ng/ml Activin A for 4 days,
and 25 ng/ml Wnt3a (RD Systems) from days 0 to 1 only.
Differentiation to ectoderm lineage was done with the same
protocol as DE differentiation, except that Wnt3a and Activin
A were replaced with 100 ng/ml Noggin and 5 mM Activin
inhibitor (SB431542; Sigma Aldrich, St. Louis, MO, http://
www.sigmaaldrich.com/united-states.html) for 4 days.
HeLa cells were maintained in DMEM (BioWhittaker; Rad-
nor, PA, https://us.vwr.com/store/catalog/product.jsp?produc-
t_id=4679035) containing 10% heat-inactivated fetal bovine
serum (FBS; Omega Scientific, Inc., Tarzana, CA, http://www.
omegascientific.com/). Cells were grown to 60%–70% conflu-
ence and transfected with HiPerFect transfection reagent
according to the manufacturer’s protocol (Qiagen, Valencia,
CA, http://www.qiagen.com/). MiScript miRNA mimics for
miR-302a-5p (ACUUAAACGUGGAUGUACUUGCU), miR-302a-5p
(13) (UAAACGUGGAUGUACUUGCUUU), and miR-302a-3p
(UAAGUGCUUCCAUGUUUUGGUGA) were purchased from Qia-
gen. Mimics were transfected at 10 nM, and cells were lysed
24 hours post-transfection in buffer containing 150 mM NaCl,
25 mM Tris-HCl, pH 7.5, 1 mM EDTA, 0.1 mM EGTA, 5 mM
MgCl2, 1 mM dithiothreitol, 10% glycerol, 150 IU/ml aprotinin,
2 mg/ml leupeptin, and 1 mM PMSF (phenylmethylsulfonyl
fluoride) with 1% Triton X-100. Cell lysates were centrifuged
at 14,000 rpm for 10 minutes, and the detergent-soluble
supernatants were collected. Proteins were separated by SDS-
polyacrylamide gel electrophoresis (PAGE) and transferred to
PVDF (polyvinylidene difluoride) and Western blotted and
quantified using an AlphaInnotech FlourChemQ. The OTX2
antibody was from Santa Cruz Biotech (Santa Cruz, CA, http://
www.scbt.com/) and the TSC1 antibody was from Cell Signal-
ing (Danvers, MA, http://www.cellsignal.com/).
RNA Preparation
Cells were lysed in Trizol and RNA was extracted by the manu-
facturer’s recommended protocol (Life Technologies). Resultant
RNA was treated with Turbo DNase (Life Technologies) for 30
minutes. DNase-treated RNA was purified by sequential extrac-
tion in acid phenol/chloroform (5:1), followed by chloroform
alone, then precipitated in 4 volumes ethanol. Small RNA libra-
ries were prepared using the Small RNA 1.0 Sample Prepara-
tion Kit (Illumina, Inc., San Diego, CA, http://www.illumina.
com/). A band of RNA ranging from 18 to 30 nt was cut from
a 15% TBE-urea gel and RNA was extracted according to the
manufacturer’s recommended protocol. After ligation of 50
and
30
adaptors, bands of 40–60 and 70–90 nt, respectively, were
cut from the gel and RNA was again extracted as described
above, followed by RT-PCR amplification. Finally, a 92 bp
band of small RNA library was purified from the gel. The library
was validated on an Agilent 2100 Bioanalyzer using the
DNA100 chip and quantified using a Roche LightCycler 480. Ten
picomoles were run per flow cell in an Illumina GAII sequencer
using a v4 Cluster generation kit and a v5 sequencing kit for 36
cycles with Illumina Sequencing Primer Read 1 Mix.
Sequence Data Analysis
Preprocessing of Reads. The microarray data are MIAME com-
pliant and has been deposited at gene expression omnibus
Hinton, Hunter, Afrikanova et al. 2361
www.StemCells.com VC AlphaMed Press 2014

3. (accession# GSE16681). The Illumina output was converted to
FASTA format for analysis. Reads were collapsed into unique
sequences and reads with four or more contiguous As or Ns or six
or more continuous As and Ns in any combination in any location
were discarded. Adapter sequences (P-UCGUAUGCCGUCUUCUG-
CUUGUidT) were then trimmed and resulting reads 17 nucleo-
tides long were grouped by seed region into clusters, and within
each cluster exact duplicates were merged.
Assignment of Reads to miRNA Genes
Nonduplicate reads from each cluster were then assigned to
human miRNA stem-loops (miRBase build 18) through pair-
wise alignment using the following criteria: (a) exact identity
of the seed regions between the two sequences (nucleotides
2–8), (b) 80% sequence identity within nucleotides 9–18
(inclusive), (c) 65% identity for any remaining nucleotides
outside of these regions, (d) when a read aligned to two or
more known stem-loop sequences, the read was assigned to
the stem-loop(s) with the highest sequence similarity, (e) the
following parameters were used for the alignments: match for
non-N nucleotides 5 15, mismatch for non-N nucle-
otides 5 24, match for any nucleotide with N 5 21, gap
open penalty 5 216, gap extension 5 24.
Quantification of miRNA Species
Reads were grouped by the 50
starting position within
each miRNA stem-loop. For each stage, the counts of
uniquely aligned reads were summed for each starting
position. The counts of reads aligning to multiple stem-
loops were partitioned among the matches by the propor-
tion of counts of the unique matches to a given stem-loop
divided by the total number of unique reads to all
matches. The samples were normalized by dividing by the
total number of reads from each stage and expressed as
reads per million (RPM).
Analysis of Annotated Mature miRNAs and isomiRs
The offset of each isomiR was calculated as the difference of
the starting position of the isomiR and the start of the nearest
annotated mature miRNA in the stem-loop. IsomiR sequences
were determined as a fragment of the matching stem-loop
from the isomiR starting position through the average length
of reads with the same starting position weighted by the RPM
of each read. Read clusters beginning with the same 50
starting
position within the stem-loop (offset of 0) as the nearest anno-
tated miRNA were assigned to that miRNA. Alternate starting
positions were considered isomiRs. Expression levels of isomiRs
and annotated miRNAs were determined by summing the RPM
of the isomiRs from all the stem-loops with the same
sequence. For analysis by hairpin arm, human miRNA stem-
loops (miRBase build 18) were folded using the default settings
for RNAfold [14–16] or RNAshapes [17, 18]. Then each struc-
ture was partitioned into 50
arms, 30
arms, and loops. IsomiR
sequences with 75% overlap with the 50
or 30
arms were
assigned to the respective arm.
Identification of Novel miRNAs on Opposite Arms of
Annotated miRNAs
Novel miRNAs were identified from isomiRs on the opposite
arm of stem-loops from the nearest annotated miRNAs by the
following criteria: (a) Expression 1RPM in either time point.
(b) Presence of 0–3nt 30
overhangs with the annotated miRNA
within the stem-loop structure, highest consideration given to
2nt overhangs. (c) Abundance among isomiRs on the same
arm.
Quantitative PCR Analysis
cDNA for mRNA analysis was created using Superscript III reverse
transcriptase (Life Technologies). For analysis of the miR-302a
stem-loop, cDNA was made with the TaqMan MicroRNA Assay kit
from Life Technologies. For analysis of other miRNAs, cDNA was
made with the NCode kit from Life Technologies. Quantitative PCR
was performed on a StepOne Plus thermocycler (Life Technolo-
gies) with SYBR green mastermix or Taqman mastermix (Life Tech-
nologies). mRNA Ct values were normalized to housekeeping
genes Cyclophilin G and TATA-binding protein. miRNA Ct values
were normalized to RNU48 RNA, U6 RNA, or 5S RNA. Oligonucleo-
tide sequences for SYBR green PCR are provided in Supporting
Information Table S1.
Northern Blot Analysis of miRNAs
PAGE northern methods were performed as previously
described [19]. Starfire-labeled DNA oligos from IDT (Coral-
ville, IA, http://www.idtdna.com/site) were used as probes for
miR-302a-5p (AAGTACATCCACGTTTAAGT-Starfire), and miR-
302a-5p (13) (AAAGCAAGTACATCCACGT-Starfire). Ethidium
bromide staining of the gel prior to transfer was used to
detect small rRNAs to assess quality of the total RNA samples.
Luciferase Assays
The putative target site from the human OTX2 30
UTR was
PCR amplified using the following primers: 50
-CTAGTAAGGAG
TCAATATGTAGTTTAAGAGAA-30
and 50
-AGCTTTCTCTTAAACTACAT
ATTGACTCCTTA-30
(synthesized by Valuegene, San Diego, CA,
http://www.valuegene.com/) and cloned downstream of the
stop codon in pMIRR-Luc (Ambion, Austin, TX) to generate
EFLuc-OTX. This LucOTX construct was used to generate the
mutant LucOTX plasmid (EFLuc-OTXmut) through site-directed
mutagenesis of three nucleotides in the miRNA seed
sequence. The putative target site from the human TSC1 30
UTR was PCR amplified using the following primers: 50
-
CTAGTACAGGAGGTGTGAATGCACGTTTCAAA-30
and 50
-AGCTT
TTGAAACGTGCATTCACACCTCCTGTA-30
and cloned downstream
of the stop codon in pMIRR-Luc (Ambion, http://www.lifetech-
nologies.com/us/en/home/brands/ambion.html) to generate
EFLuc-TSC. This LucTSC construct was used to generate the
mutant LucTSC plasmid (EFLuc-TSCmut) through site-directed
mutagenesis of three nucleotides in the miRNA seed sequence.
Cyt49 cells were cultured in 12-well plates and each well trans-
fected using the Neon electroporator (Life Technologies, Inc.)
with 450 ng of luciferase plasmid and 25 ng of CMV-b-
galactosidase vector for normalization. Cells were harvested and
assayed 45–48 hours after transfection. Results represent three
independent experiments over all time points. Luciferase assays
were performed as previously described [20].
RESULTS
Differentiation of Pluripotent hESCs into DE
The CyT49 hESCs line has been differentiated previously on
mouse embryonic feeders with high efficiency to DE and
2362 sRNA-seq Analysis of Human Embryonic Stem Cells
VC AlphaMed Press 2014 STEM CELLS

4. subsequent pancreatic stages [7, 21]. To analyze miRNA expres-
sion during DE differentiation, CyT49 cells were treated with
Activin A in low serum for 4 days while samples were collected
on day 0 (hESCs) and day 4 (DE) for RNA analysis. Differentia-
tion was monitored by qRT-PCR for linage-specific markers.
Pluripotency markers decreased while DE markers increased
(Fig. 1A). Size-selected fractions of the RNA were then used
for analysis of small RNAs. Analysis of small RNA libraries
yielded 3,183,398 small RNA reads from hESCs and 4,111,732
small RNA reads from DE, representing 812 annotated miRNAs.
Six hundred and sixty-nine miRNAs were detected in both
hESCs and DE, while 78 miRNA were detected only in hESCs
and 65 miRNAs were detected only in DE (Supporting Informa-
tion Table S2). Approximately 20% of annotated miRNAs (miR-
Base release 18) were highly expressed at 100 RPM, 50%
were abundantly expressed at 10 RPM, and 75% were
expressed at 1 RPM.
Differential miRNA Expression Occurs During DE
Formation
Similar expression levels in both hESCs and DE for most anno-
tated miRNAs are indicated by proximity to the line of corre-
lation in Figure 1B. Outliers significantly above or below the
line of correlation represent differentially regulated miRNAs.
We chose to focus on highly expressed miRNAs (100 RPM)
for further differential expression analysis. Thirty-seven miR-
NAs were upregulated in DE more than twofold relative to
hESCs (Table 1), 40 were downregulated more than twofold
(Table 2), and 74 changed by less than twofold (Supporting
Information Table S3). Putative targets of the 37 upregulated
and 40 downregulated were analyzed and compiled using
miRWalk (http://www.umm.uni-heidelberg.de/apps/zmf/mir-
walk/index.html) (Supporting Information Files S1 and S2)
[22]. Figure 1C shows qRT-PCR validation in two separate
hESCs lines (Cyt49 and H1) of the top four upregulated and
top four downregulated miRNAs detected by deep sequenc-
ing at greater than 150 RPM. Several miRNAs were highly
expressed in both pluripotent hESCs and in DE. Furthermore,
analysis of the representation of seed sequences, which
include nucleotides two to eight of the mature miRNAs and
mediate most target recognition [1], revealed that the two
stages shared 6 of the 10 most abundant seed sequences
(Supporting Information Tables S4, S5), including AAGUGC,
previously described as a common hESCs seed sequence
[23].
Comparison of Annotated miRNAs from Microarray
and Deep Sequencing
We previously profiled miRNA expression in hESCs and DE
using a microarray platform [7]. While the previous analysis
detected only 229 annotated miRNAs in hESCs and 200 miR-
NAs in DE, the current analysis identified 747 miRNAs in
hESCs and 744 miRNAs in DE (Figure 1D). Of the 602 miRNAs
detected by sequencing but not by microarray, 550 were
annotated recently in miRBase and therefore were not origi-
nally present on the microarray platform. Furthermore,
approximately 20% of the microarray queried miRNAs that
were detected by sequencing alone were present at 50
RPM. Thus, the discovery of novel miRNAs and limited sensi-
tivity account for the majority of discrepancies between these
two methods for miRNA analysis.
Characterization of isomiRs
We noted that approximately 68,000 reads did not share the
same starting position as the nearest annotated miRNA, yet
accounted for 14% of all reads aligning to annotated miRNA
stem-loops (Supporting Information Table S6). We chose to
focus on the 50
end heterogeneity because the 50
end deter-
mines the seed sequence and as such isomiRs with different
50
ends are likely to bind a distinct range of mRNA targets.
For simplicity, we refer to isomiRs as sequences with a 50
end
offset from a reference miRNA. While the majority of the
reads from a stem-loop began at the annotated miRNA, a sig-
nificant fraction of reads began at alternate positions on both
the 50
and 30
arms (Fig. 2A, 2B). Because the annotated
miRNA was not always the most abundant sequence in each
arm, we then compared the starting position of reads relative
to the most abundant starting position. Analysis of the 30
arm
revealed strong consistency between the annotated start site
and the most abundant start site (Fig. 2A). The accuracy of
annotation of the most abundant species for this arm sug-
gests that the other isomiRs are minor processing products;
however, they constitute more than 10% of the reads aligning
to the 30
arm. The reads from the 50
arm of the miRNA stem-
loop began from the most abundant starting position 92%
of the time, compared to 83% for the annotated start site
(Fig. 2B). This shift suggests mis-annotation of the starting
positions of the most abundant isomiR for the differential
miRNAs.
However, miRNAs with lower expression levels are better
represented by analyzing isomiR abundance within each stem-
loop arm. IsomiRs constituted 10% of the reads in 40% of
all detected miRNA stem-loop arms (Fig. 2C). IsomiR abun-
dance was lower within stem-loop arms of highly expressed
miRNAs (100 RPM) yet isomiRs constituted 10% of the
reads in 20% of stem-loop arms (Fig. 2C). In total, 3,333 dis-
tinct isomiRs were detected (Supporting Information Table
S6). Among these, 52 isomiRs were detected at 100 RPM
(Supporting Information Table S7), and 66 isomiRs expressed
10 RPM were also expressed at higher levels than the near-
est annotated miRNA on the same stem-loop arm (Supporting
Information Table S6). While a few of the highly expressed
isomiRs were unannotated star forms for annotated miRNAs
(Supporting Information Table S8), the majority of isomiRs
likely arose from alternative Drosha/Dicer processing.
Noting that several potential isomiRs were located within
a stem-loop with only one annotated arm, we surmised that
these were likely miRNAs that would previously have been
annotated as miRNA*s. Upon further examination of these
species, we were able to identify 41 novel miRNAs within
unannotated stem-loop arms expressed over 1 RPM (Support-
ing Information Table S8). Interestingly, several of these miR-
NAs were expressed higher than the canonical miRNA,
highlighting the importance of miRNAs from both strands.
Furthermore, miRNA-5p and miRNA-3p (formerly miRNA and
miRNA*) expression profiles are poorly correlated during
differentiation, suggesting frequent, independent post-
transcriptional regulation (Fig. 2D). The miR-302–367 cluster
stands out, as four of five members with detectable expres-
sion from both strands showed more than twofold difference
in fold change (DE/hESCs) between the 5p and 3p strand of
the stem-loop.
Hinton, Hunter, Afrikanova et al. 2363
www.StemCells.com VC AlphaMed Press 2014

5. Figure 1. Differences in miRNA expression of human ESC (hESCs) and DE. (A; top) Outline of experimental design for DE formation in
CyT49, H1, and H9 cells. (A; bottom) qRT-PCR analysis of lineage-specific markers on day 0 and day 4 of differentiation. (B): Scatter analysis
of expression levels of all annotated miRNAs. Gray points indicate miRNAs expressed 100 RPM and colored dots indicate miRNAs
detected at 100 RPM. Green dots indicate more than twofold increase in DE. Blue dots indicate less than twofold change. Red dots indi-
cate more than twofold decrease in DE. (C): Quantitative PCR analysis of differentially expressed miRNAs between undifferentiated hESCs
and definitive endoderm (day 0 vs. day 4 of differentiation) in two human embryonic stem cell lines (Cyt49 and H1). Data are shown for
the four miRNAs with largest upregulation (top) and the four miRNAs with the largest downregulation (bottom) detected by next genera-
tion sequencing (150 RPM). (D): Comparison of annotated miRNA detection between microarray analysis and NextGen sequencing analy-
sis of hESCs differentiated into DE. Abbreviations: DE, definitive endoderm; ESC, embryonic stem cell; RPM, reads per million.
2364 sRNA-seq Analysis of Human Embryonic Stem Cells
VC AlphaMed Press 2014 STEM CELLS

6. The miR-302a Stem-Loop Produces Several Independ-
ently Regulated Species of miRNA
The complexity of mature miRNAs coming from a single stem-
loop is well illustrated by that of miR-302a. Three major spe-
cies (miRNA-3p, miRNA-5p, and isomer) are highly expressed
with distinct 50
ends (Fig. 3). Surprisingly, our sequencing data
indicated that the canonical miR-302a and miR-302a* were
not the majority species from this locus. The majority of the
reads aligning to the miR-302a stem-loop appeared to be an
isomiR, miR-302a-5p (13), with the 50
end shifted three
nucleotides to the 30
of miR-302a* within the stem-loop, and
therefore containing a new seed sequence (Fig. 3A, 3B). To
confirm that isomiR detection was not an artifact from mouse
embryonic fibroblasts, DE differentiation was repeated under
feeder-free conditions and miR-302a-5p (13) expression was
then confirmed by qPCR in two independent hESCs lines
(Fig. 3C). This isomiR has been previously reported as a minor-
ity species [24], yet in our experiments for both undifferenti-
ated hESCs and differentiated DE cells this isomiR is
predominant.
During miRNA biogenesis, the RNase III enzyme Drosha
releases the miRNA precursor hairpin leaving a two base 30
overhang [2]. Given the structure of the stem-loop for miR-
302a (Fig. 3A), it is unlikely that miR-302a-5p (13) would
arise from the same Drosha processing event that gave rise
to miR-302a-3p, suggesting an independent cleavage. Multi-
ple Drosha cleavage products were confirmed by northern
blot analysis. Probing for miR-302a-5p (13) revealed two
precursor bands, and as miR-302a-5p (13) is 30
of miR-
302a-5p within the stem-loop, the probe hybridizes to the
Drosha cleavage products that give rise to both isomiRs, and
only the larger species hybridizes to a probe for miR-302a-
5p (Fig. 3D).
mir-302a-5p (13) Contains a Unique Seed Sequence
that Targets TSC1
Seed sequences analysis suggested that miR-302a-5p and
miR-302a-5p (13) have different, nonoverlapping targets.
Although, many groups have identified targets within the
miR-302 locus [25–28], none of the studies have identified a
unique target for miR-302a-5p. Therefore, TargetScan (version
5.2) was used to identify potential targets for miR-302a-5p
and miR-302a-5p (13). Of the potential targets generated,
OTX2 was selected as a potential miR-302a-5p target based
on both seed sequence pairing and because the protein is a
marker for stem cells entering neuroectoderm lineage [29].
TSC1 was tested as a selective miR-302a-5p (13) target based
Table 1. miRNAs upregulated during differentiation from human ESC to DE
miRNA Mature sequence ESC RPM DE RPM Fold change
hsa-miR-1263 AUGGUACCCUGGCAUACUGAGU 0.6 1870.7 2977.66
hsa-miR-1247-3p CGGGAACGUCGAGACUGGAGC 0.3 693.9 2208.86
hsa-miR-375 UUUGUUCGUUCGGCUCGCGUGA 101.5 108112.8 1065.53
hsa-miR-212-5p ACCUUGGCUCUAGACUGCUUACUG 0.6 121.1 192.78
hsa-miR-132-3p UAACAGUCUACAGCCAUGGUCG 0.9 154.9 164.39
hsa-miR-146b-5p UGAGAACUGAAUUCCAUAGGCU 30.8 1377.5 44.75
hsa-miR-708-5p AAGGAGCUUACAAUCUAGCUGGG 60.6 2573.9 42.45
hsa-miR-489 GUGACAUCACAUAUACGGCAGC 5.7 121.1 21.42
hsa-miR-452-5p AACUGUUUGCAGAGGAAACUGA 54.0 932.5 17.26
hsa-miR-9-5p AUAAAGCUAGAUAACCGAAAGU 12.9 143.5 11.14
hsa-miR-224-5p CAAGUCACUAGUGGUUCCGUU 61.9 489.8 7.92
hsa-miR-1246 AAUGGAUUUUUGGAGCAGG 33.3 202.3 6.08
hsa-miR-26b-5p UUCAAGUAAUUCAGGAUAGGU 490.4 2910.9 5.94
hsa-miR-1290 UGGAUUUUUGGAUCAGGGA 102.4 586.7 5.73
hsa-miR-210 CUGUGCGUGUGACAGCGGCUGA 144.5 803.3 5.56
hsa-miR-374b-5p AUAUAAUACAACCUGCUAAGUG 163.7 867.5 5.30
hsa-miR-9-5p UCUUUGGUUAUCUAGCUGUAUGA 143.1 726.9 5.08
hsa-miR-371-3p AAGUGCCGCCAUCUUUUGAGUGU 91.1 406.9 4.47
hsa-miR-340-5p UUAUAAAGCAAUGAGACUGAUU 1344.8 5577.0 4.15
hsa-miR-320c AAAAGCUGGGUUGAGAGGGU 115.6 473.2 4.09
hsa-miR-373-3p GAAGUGCUUCGAUUUUGGGGUGU 1285.7 4582.5 3.56
hsa-miR-31-5p AGGCAAGAUGCUGGCAUAGCU 416.5 1440.5 3.46
hsa-miR-361-5p UUAUCAGAAUCUCCAGGGGUAC 50.9 172.4 3.39
hsa-let-7e-5p UGAGGUAGGAGGUUGUAUAGUU 143.7 447.9 3.12
hsa-miR-23b-3p AUCACAUUGCCAGGGAUUACC 36.3 110.3 3.04
hsa-miR-331-3p GCCCCUGGGCCUAUCCUAGAA 88.0 243.2 2.77
hsa-miR-26a-5p UUCAAGUAAUCCAGGAUAGGCU 1336.3 3549.4 2.66
hsa-miR-423-5p UGAGGGGCAGAGAGCGAGACUUU 3247.5 8496.7 2.62
hsa-miR-181d AACAUUCAUUGUUGUCGGUGGGU 162.0 418.0 2.58
hsa-miR-125a-5p UCCCUGAGACCCUUUAACCUGUGA 102.7 255.1 2.48
hsa-miR-130a-3p CAGUGCAAUGUUAAAAGGGCAU 1955.1 4590.0 2.35
hsa-miR-532-5p CAUGCCUUGAGUGUAGGACCGU 53.1 124.0 2.34
hsa-miR-191-5p CAACGGAAUCCCAAAAGCAGCUG 4062.0 9150.4 2.25
hsa-miR-302b-3p UAAGUGCUUCCAUGUUUUAGUAG 3359.1 6898.8 2.05
hsa-miR-27b-3p UUCACAGUGGCUAAGUUCUGC 170.9 346.8 2.03
hsa-miR-193b-5p CGGGGUUUUGAGGGCGAGAUGA 85.8 173.9 2.03
hsa-miR-302a-3p UAAGUGCUUCCAUGUUUUGGUGA 3205.1 6460.2 2.02
Annotated miRNAs detected at 100 reads per million (RPM), 2 fold upregulation, and p-values 0.01 are listed above.
Abbreviations: DE, definitive endoderm; ESC, embryonic stem cell; RPM, reads per million.
Hinton, Hunter, Afrikanova et al. 2365
www.StemCells.com VC AlphaMed Press 2014

7. on a strong seed pairing sequence (Fig. 4A, 4B). To explore
whether the isomiRs generated at the miR-302a stem loop tar-
geted these different mRNAs, specific miRNA mimetics were
transfected into HeLa cells and protein expression of OTX2 and
TSC1 was measured (Fig. 4C, 4D). OTX2 protein levels dropped
by 40% in HeLa cells transfected with the miR-302a-5p mimetic
(lane 2), but were unchanged in cells transfected with miR-
302a-5p (13) (lane 3). The miR-302a-5p (13) mimetic
decreased expression of TSC1 by 25% (lane 3), but had no
effect on OTX2 levels (lane 2). Transfection with a control
mimetic did not alter expression of either protein (lane 1).
Expression of selected miRNAs was examined in hESCs
undergoing differentiation into DE and ectoderm was exam-
ined to explore the role of selected miRNAs in hESCs cell fate
decisions (Fig. 5A). Expression of highly expressed miRNAs,
miR-1263, miR-1247, miR-375, miR-132-3p, and miR-302a was
found to be selectively enriched in DE compared with ecto-
derm, suggesting a possible role in targeting mRNAs that
must be downregulated during DE specification. Targeting of
OTX2 and TSC1 by miR-302a-5p and miR-302a-5p (13) during
differentiation was further explored during DE formation. Con-
sistent with a role for miRNAs in inhibition of protein transla-
tion, mRNA levels of TSC1 and OTX2 increased during DE
formation (data not shown); however, protein levels of TSC1
dropped by 30% and OTX2 levels decreased by more than
40% during the 4-day differentiation protocol (Fig. 5B, 5C).
Incubation of CyT49 cells with MG-132, an inhibitor of the
proteasome, had no effect on the decrease in TSC1 or OTX2
degradation, suggesting that increased protein degradation
was not responsible for the observed decrease in protein
(data not shown). To further support the direct targeting of
OTX2 and TSC1 in hESCs, the putative binding sites for miR-
302a isomiRs from each gene was inserted into the 30
UTR of
a luciferase reporter plasmid vector. When compared with
control luciferase activity, inclusion of the putative miR-302a
binding site from the 30
UTR of OTX and TSC both resulted in
decreased reporter activity in transfected CyT49 cells.
Mutation of three nucleotides in the seed sequence of each
putative binding site resulted in derepression of reporter
activity (Fig. 5D). Taken together, these results suggest that
isomiRs with unique seed sequences differentially target
mRNAs to help regulate differentiation to DE.
Table 2. miRNAs downregulated during differentiation from human ESC (hESCs) to DE
miRNA Mature sequence RPM ESC DE RPM Fold change
hsa-miR-486-3p CGGGGCAGCUCAGUACAGGAU 251.3 14.6 17.22
hsa-miR-520a-3p AAAGUGCUUCCCUUUGGACUGU 169.9 13.6 12.47
hsa-miR-498 UUUCAAGCCAGGGGGCGUUUUUC 251.3 20.2 12.45
hsa-miR-512-3p AAGUGCUGUCAUAGCUGAGGUC 1130.9 98.5 11.48
hsa-miR-1323 UCAAAACUGAGGGGCAUUUUCU 8095.8 830.3 9.75
hsa-miR-518b CAAAGCGCUCCCCUUUAGAGGU 165.2 17.0 9.71
hsa-miR-517a-3p AUCGUGCAUCCCUUUAGAGUGU 127.7 18.4 6.95
hsa-miR-92b-3b UAUUGCACUCGUCCCGGCCUCC 9425.8 1356.4 6.95
hsa-miR-124-3p UAAGGCACGCGGUGAAUGCC 952.6 149.2 6.38
hsa-miR-486-5p UCCUGUACUGAGCUGCCCCGAG 143.9 23.8 6.04
hsa-miR-1 UGGAAUGUAAAGAAGUAUGUAU 1074.3 190.2 5.65
hsa-let-7a-5pa
UGAGGUAGUAGGUUGUAUAGUU 4838.8 985.0 4.91
hsa-miR-222-3p AGCUACAUCUGGCUACUGGGU 3581.1 757.8 4.73
hsa-miR-221-3p AGCUACAUUGUCUGCUGGGUUUC 15764.0 3370.6 4.68
hsa-miR-199a-3p ACAGUAGUCUGCACAUUGGUUA 357.6 81.1 4.41
hsa-miR-589-5p UGAGAACCACGUCUGCUCUGAG 105.5 25.3 4.17
hsa-miR-516b-5p AUCUGGAGGUAAGAAGCACUUU 370.0 89.8 4.12
hsa-miR-363-3p CGGGUGGAUCACGAUGCAAUUU 154.6 43.3 3.57
hsa-miR-92a-l-5p AGGUUGGGAUCGGUUGCAAUGCU 1033.5 320.5 3.22
hsa-miR-1257 AGUGAAUGAUGGGUUCUGACC 144.5 45.0 3.21
hsa-miR-3168 GAGUUCUACAGUCAGAC 182.5 58.9 3.10
hsa-miR-15 l-5p UCGAGGAGCUCACAGUCUAGU 628.9 203.3 3.09
hsa-miR-28-5p AAGGAGCUCACAGUCUAUUGAG 156.1 54.0 2.89
hsa-miR-30c-5p UGUAAACAUCCUACACUCUCAGC 267.0 93.6 2.85
hsa-miR-21-5p CAACACCAGUCGAUGGGCUGU 197.6 70.5 2.80
hsa-miR-148a-3p UCAGUGCACUACAGAACUUUGU 1836.4 664.7 2.76
hsa-miR-335-5p UCAAGAGCAAUAACGAAAAAUGU 307.8 112.1 2.75
hsa-miR-129-5p CUUUUUGCGGUCUGGGCUUGC 125.0 47.2 2.65
hsa-miR-432-5p UCUUGGAGUAGGUCAUUGGGUGG 185.0 70.3 2.63
hsa-miR-1298 UUCAUUCGGCUGUCCAGAUGUA 295.3 112.8 2.62
hsa-miR-92b-3b AGGGACGGGACGCGGUGCAGUG 534.0 208.2 2.57
hsa-miR-7-5p UGGAAGACUAGUGAUUUUGUUGU 1023.7 401.5 2.55
hsa-miR-128 UCACAGUGAACCGGUCUCUUU 2033.0 812.8 2.50
hsa-miR-302a-5p ACUUAAACGUGGAUGUACUUGCU 8992.0 3986.6 2.26
hsa-miR-21-5p UAGCUUAUCAGACUGAUGUUGA 34900.4 15510.5 2.25
hsa-miR-18lb-5p AACAUUCAUUGCUGUCGGUGGGU 467.4 209.1 2.23
hsa-miR-19b-3p UGUGCAAAUCCAUGCAAAACUGA 625.9 297.2 2.11
hsa-miR-148b-3p UCAGUGCAUCACAGAACUUUGU 333.1 163.4 2.04
hsa-miR-30a-5p UGUAAACAUCCUCGACUGGAAG 1359.3 669.9 2.03
hsa-miR-941 CACCCGGCUGUGUGCACAUGUGC 151.1 75.0 2.01
Annotated miRNAs detected at 100 RPM, more than or equal to twofold upregulation, and p-values .01 are listed above.
a
hsa-let-7a detection in hESCs is likely due to contamination from mouse feeder layers, as described in Hinton et al. [7].
Abbreviations: DE, definitive endoderm; ESC, embryonic stem cell; RPM, reads per million.
2366 sRNA-seq Analysis of Human Embryonic Stem Cells
VC AlphaMed Press 2014 STEM CELLS

8. DISCUSSION
Through self-renewal and broad differentiation capacity, hESCs
provide the potential for an unlimited supply of cells for cell-
based therapies for many diseases, including type I diabetes.
For the generation of pure hESCs-derived, functional b cells
the major challenges of low yield of glucose-responsive insu-
lin-positive cells and high risk of tumor formation remain [30].
As key regulators of cell-fate and function, characterization of
miRNA expression profiles at all stages of hESCs differentia-
tion is important for improvements in the production of fully
differentiated and functional therapeutic cells.
This study focuses on the formation of DE, the first step
in b-cell differentiation. Consistent with our previously iden-
tified DE miRNA signature [7], miR-375, miR-708-5p, miR-
371-3p, and miR-373-3p were found to be upregulated more
than twofold in this study. miR-375, described previously as
a regulator of endocrine pancreas differentiation [9], was
the second most differentially expressed miRNA and the
most highly expressed miRNA in DE. Several additional miR-
NAs were also highly upregulated during hESCs differentia-
tion to DE (Table 1). Among them, miR-1263 exhibited the
highest fold change, but was not assayed in the previous
study.
Several miRNAs (miR-200c-3p, miR-302a-3p, miR-371-5p,
miR-372, and miR-373-3p) have been shown previously to be
downregulated during undirected hESCs differentiation and
thus described as being “ESC-specific” (ESCC) [4, 24]. How-
ever, comparing our dataset (hESCs vs. DE) to other cells
types resulting from nondirected differentiation of hESCs [31],
it is clear that several ESC-specific miRNAs persist or are
upregulated specifically in DE, suggesting an important role in
endoderm formation. Furthermore, Laurent et al. [23]
described a common seed sequence for hESCs that is also a
predominant seed in DE (Supporting Information Table S5). It
is possible that the significant overlap of miRNA and seed sig-
natures in pluripotency and DE is a result of activin A signal-
ing, a known regulatory factor required for maintenance of
pluripotency, DE formation, and miRNA regulation [32–34].
While miRNA expression had been previously profiled in
hESCs and DE using a microarray platform [7], next generation
sequencing has several advantages compared to probe-based
expression analysis platforms. First, sRNA-seq offers increased
sensitivity. Thus miRNAs expressed at very low levels can be
Figure 2. Analysis of 50
end variability. Frequency of reads in the (A) 30
arms or (B) 50
arms of annotated stem-loops with 50
end rela-
tive to that of the annotated miRNA (red) or the most abundant isomiR (green). (C): miRNAs were ranked in decreasing order by the
abundance (%) of isomiR reads from the same stem-loop arm. The cumulative frequency of all miRNAs (purple) or highly expressed miR-
NAs (green) was then plotted against the isomiR abundance. (D): Correlation of expression changes from miRNAs and cognate forms on
opposite arms. The fold change of miRNA-5p versus miRNA-3p was plotted above. miRNA-5p:miRNA-3p pairs were excluded if not
expressed in both time points and expressed 10 RPM at any time point. The line of correlation is shown with the solid line. The
dashed and dashed-dot lines indicate a 2-fold and 10-fold difference in fold change, respectively. Members of the miR-302/367 cluster
are highlighted in red. Abbreviation: RPM, reads per million.
Hinton, Hunter, Afrikanova et al. 2367
www.StemCells.com VC AlphaMed Press 2014

9. more readily detected. Second, obtaining the exact sequences
with deep sequencing provides unparalleled specificity with
the ability to distinguish closely related miRNAs. For example,
there are 148 pairs from 168 human miRNAs (miRBase Build
16) with only single nucleotide differences. In addition to
closely related miRNAs, many isomiRs are likely to hybridize
to probes for canonical miRNAs and distort proper quantita-
tion (e.g., miR-653 discussed below). Finally, miRNAs detected
by microarray were limited to annotated miRNAs at the time
of array construction. Therefore, neither novel miRNA genes,
such as miR-1263, nor novel isomiRs could be detected using
that approach.
Most miRNA functional analyses focus on the canonical
mature sequence curated in miRBase, which in general corre-
sponds to the most abundant species from a hairpin in high-
throughput cloning/sequencing studies [35]. However, recent
reports demonstrate more complex populations of miRNA spe-
cies derived from single hairpins [3, 36, 37]. Mature miRNAs
from opposite strands of a stem-loop regulate a distinct set of
targets [38, 39], and the relative abundance of miRNA-5p ver-
sus miRNA-3p species can be tissue specific, as demonstrated
by the miR-302a stem-loop. In addition to miRNA-5p and
miRNA-3p, alternative processing by Drosha or Dicer within a
stem loop can generate readily detectable isomiR species with
variability at the 50
ends, such as miR-302-5p (13). Changes at
the 50
end, especially those that modify the seed sequence,
are likely to have large effects on miRNA function.
The growing complexity of small RNA species from miRNA
genes calls for updates to miRNA nomenclature with regard
to annotation of arm position within a stem-loop and the
presence of multiple isomiRs. Until recently miRNAs detected
on both arms of a stem-loop were often annotated as miRNA
versus miRNA* with the latter being considered a minor
byproduct of miRNA biogenesis. miRBase has recently retired
this nomenclature with release 18, in favor of the miRNA-5p/
3p notation that annotates one miRNA from each arm accord-
ing to the position of the arm within the stem-loop. This
removes the abundance assumption and provides additional
positional information to distinguish the two duplex strands.
Unfortunately, miRNA stem-loops with only one arm repre-
sented by an annotated miRNA were not updated with the
25p, 23p annotation (e.g., miR-372). When later efforts dis-
cover the miRNAs arising from the alternative strands, and
the names are then changed (e.g., miR-372-3p), it becomes a
cumbersome effort to follow the literature using the original
names. For example, in this report, we detected 41 miRNAs
(Supporting Information Table S8) in arms of stem-loops for
which the opposite arm was previously annotated but not yet
given the 25p, 23p designation. Simply updating all the
current miRNAs now, as well as adding all new miRNAs with
the updated notation, largely mitigates this problem.
Another limitation of the current nomenclature is that
only a single miRNA species is currently annotated for each
arm of the miRNA stem-loop. We and others have shown that
alternative Dicer or Drosha processing events can generate
multiple miRNAs with distinct 50
ends from single miRNA
genes [3] and these miRNAs should be recognized as distinct
regulatory RNAs. The use of a single mature sequence to rep-
resent an entire arm of a stem-loop may be especially inap-
propriate given the fact that the assignment of the annotated
mature sequence can change each time new tissues are ana-
lyzed. For example, the currently annotated mature miR-653
sequence was not detected in this study, although previous
microarray analysis indicated that it was highly upregulated
during DE formation [7]. Closer analysis of deep sequencing
data from the miR-653 stem loop revealed an isomiR with a
50
end shifted three bases 30
of the annotated miRNA at miR-
Base. This miR-653 isomiR was upregulated 4.6-fold during DE
differentiation (Supporting Information Table S6). Interestingly,
this isomiR sequence was identical to the originally cloned
Figure 3. Expression and quantification of multiple independently regulated isomiRs produced from the miR-302a stem-loop. (A): Pre-
dicted hairpin structure for miR-302a. Green lines indicate the canonical miR-302a-5p and miR-302a-3p. Red lines indicate the isomiRs
miR-302a-5p (13) and miR-302a-3p (13). Colored arrows show the predicted sites for Drosha cleavage for the production of these iso-
miRs. (B): IsomiRs coming from miR-302a are highlighted with boxes to indicate the cluster of reads beginning at the same 50
starting
position relative to the miR-302a hairpin sequence. Vertical lines indicate the 30
ends of the reads clustered into the boxes. Listed to
the right of each box is the expression level in RPM for hESCs and DE samples. (C): Quantitative PCR analysis of miR-302a isomiRs
between undifferentiated hESCs and DE (day 0 vs. day 4 of differentiation) in two human embryonic stem cell lines (Cyt49 and H9). (D):
Northern blots showing mature and precursor signals for miR-302a-5p (13) and miR-302a-5p. Green arrows indicate the mature and
precursor form giving rise to mature miR-302a-5p. Red arrows indicate the mature and precursor form giving rise to mature miR-302a-
5p (13). The doublet detected with the probe for miR-302a-5p (13) shows the independent Drosha products giving rise to miR-302a-
5p and miR-302a-5p (13) species. Abbreviations: DE, definitive endoderm; ESC, embryonic stem cell; RPM, reads per million.
2368 sRNA-seq Analysis of Human Embryonic Stem Cells
VC AlphaMed Press 2014 STEM CELLS

10. sequence [40]. The mature miR-653 sequence was later rean-
notated due to a subsequent publication reporting the cur-
rent isomiR as the mature form [41]. Although Landgraf et al.
[41] only detected a single read for the modified miR-653, the
methodology used gave this isomiR precedence over the origi-
nal report. Thus, proper quantitation and differential expres-
sion analysis are limited when dependent upon annotation of
only a single isomiR per stem-loop arm, and comparison of
data between studies is compromised when the annotation is
subject to change.
One method for distinguishing isomiRs enumerates them
by their 50
position (e.g., miR-210.1 for the 50
most start site
and miR-210.2 for the next most 50
start site, etc. [42]). While
this system encodes useful information in the name, the iso-
miRs may need to be renumbered as new datasets become
available. A more flexible system numbers isomiRs in the
order in which they were discovered, losing the positional
information yet retaining a consistent nomenclature. Thus, for
miR-653 the originally cloned miRNA sequence would be miR-
653-5p.1 and the more 50
isomiR currently annotated in miR-
Base would be miR-653-5p.2.
Any system recognizing isomiRs needs to take care, how-
ever, to minimize the incorporation of artifacts. Expression-
based filters are one step. Expression can be filtered by ana-
lyzing the absolute expression of a potential isomiR or the
fraction of reads coming from the same arm of the stem-loop
corresponding to the potential isomiR. We chose to focus on
absolute expression because some isomiRs of highly expressed
genes were a small fraction of reads from the stem-loop arm,
yet were expressed higher than many annotated miRNAs.
Annotation of isomiRs captures an important and currently
underappreciated aspect of miRNA biology, although it creates
a higher quality control burden on submitters and curators.
miRNAs from the miR-302/367 cluster containing five stem-
loops are highly expressed in hESCs, and are downregulated dur-
ing nondirected differentiation [24, 31, 37, 43]. Expression of the
miR-302/367 cluster depends on the activity of transcription fac-
tors Oct4, Sox2, and nanog, all of which bind to the predicted
promoter region of the miR-302/367 cluster [27]. A subset of the
mature miRNAs from the miR-302/367 cluster are part of the
ESCC regulating miRNAs described by Wang et al. [6], and the
abundance of miRNAs in this group with a common seed
sequence suggests that they regulate similar targets that help
maintain pluripotency. Differentiation into a neurectodermal lin-
eage is considered by many to be the default differentiation
pathway in the absence of cues that direct cell fate [44, 45].
Here, we demonstrate that several of the highly upregulated
miRNAs in DE formation are expressed at relatively low levels in
ectoderm formation, including the miR-302a stem-loop. The miR-
302/367 cluster has previously been shown to negatively regu-
late expression of genes involved in early ectoderm specification,
including NR2F2 [28]. Here, we expand the known targets of the
miR-302/367 cluster to include a novel target, OTX2, a gene
expressed early in hESCs differentiation [46]. Unlike other identi-
fied targets of the miR-302/367 cluster, we used mimetics to
selectively identify miR-302a-5p as the miRNA that targets this
gene. This approach was also used to determine whether
another predicted target of the miR-302/367 cluster, an isomiR
Figure 4. IsomiRs of miR-302a have different, nonoverlapping seed sequences that differentially regulate expression of OTX2 and TSC1.
(A): Seed pairing alignment of miR-302a-5p isomiRs to putative target site in 30
UTR of OTX2. (B): Seed pairing alignment of miR-302a-
5p isomiRs to putative target site in 30
UTR of TSC1. (C): Western analysis of HeLa cell extracts 24 hours following transfection with
miRNA mimetic of miR-302a-5p, miR-302a-5p (13), or siRNA control. GSK3a/b and Hsp90 proteins were used as loading controls. (D):
Densitometry of Western blots was used to quantify protein levels normalized to loading controls. Abbreviations: OTX2, orthodenticle
homeobox 2; TSC1, tuberous sclerosis 1.
Hinton, Hunter, Afrikanova et al. 2369
www.StemCells.com VC AlphaMed Press 2014

11. called miR-302a-5p (13), had overlapping specificity. Based on
TargetScan 5.2, the GTPase activating protein TSC1 was predicted
to be a target of miR-302a-5p (13). This protein regulates forma-
tion of the TORC1 and TORC2 complex through catalytic removal
of the terminal phosphate from the Rheb GTPase [47]. Transfec-
tion of the isomiR mimetic resulted in a decrease in expression
of the TSC1 protein, but not OTX2, indicating specificity in isomiR
selection. The observation that miR-302a-5p and miR-302a-5p
(13) are the predominant miRNAs expressed in pluripotent
hESCs suggests that differential targeting of miRNAs from this
complex acts to coordinate decisions about cell fate.
A previous report has already implicated miR-302a in the
promotion of mesendoderm formation [10]. In this report,
they identified Lefty, a regulator of the Nodal signaling path-
way, as a target of miR-302a. Additionally, they stably
expressed the miR-302a stem-loop in differentiating hESCs,
which resulted in upregulation of mesendoderm markers in
favor of ectoderm marker expression. In this study, we did
not attempt to replicate results from this work. Although sta-
ble expression of the miR-302a stem-loop would induce endo-
derm formation, the roles of individual isomiRs cannot be
isolated using a stable expression vector. Here, we provide
additional information about the role of the miR-302a stem-
loop by implicating additional targets OTX and TSC1. In a
recent report by Easley et al. [48], the authors found that
hESCs have high levels of TSC1/TSC2 which act to block
mTORC1/p70S6 kinase signaling. siRNA-mediated knockdown
of TSC1/TSC2 increased p70S6 kinase activation and induced
differentiation. We find that OTX2 and TSC1 increase in steady
state RNA levels during the transition from pluripotent ESCs
to DE, but the protein levels decrease in DE. This observation
is consistent with the hypothesis that miRNAs are regulating
the gene expression specifically by inhibiting translation. The
transfection of the miR-302a isomiRs into a heterologous cell
line resulted in decreased protein levels of each target, but
no significant decrease in RNA levels, which further promotes
the model in which some miRNAs specifically regulate target
genes via translational inhibition.
Figure 5. isomiR expression and regulation of definitive endoderm (DE) formation. (A): Lineage-specific induction during directed differen-
tiation of hESCs. Quantitative PCR analysis of selected miRNAs in undifferentiated hESCs, definitive endoderm, and neurectodermal lineages.
hESCs (Cyt49) were differentiated for 4 days while treated with Activin (Endoderm) or Noggin 1 Activin inhibitor (Ectoderm). (B): Western
analysis of TSC1 and OTX2 expression in CyT49 cell extracts at 24 hour intervals during DE formation. Hsp70 expression was used as a loading
control. (C): Densitometry of Western blots was used to quantify protein levels normalized to loading controls. * Significantly different from
control, p .05, n 5 4. (D): Luciferase reporter activity for plasmids transfected into Cyt49 cells. Control luciferase construct (EFLuc) activity
was compared to reporter plasmids with various oligos inserted into 30
UTR of luciferase cDNA (EFLuc-302a-5p: full complement of miR-302a-
5p sequence; EFLuc-OTX: putative 302a binding site in OTX 30
UTR, EFLuc-OTXmut: putative OTX binding site with 3 bases mutated within the
seed sequence; EFLuc-TSC: putative 302a binding site in TSC 30
UTR; EFLuc-TSCmut: putative TSC binding site with 3 bases mutated within the
seed sequence. Abbreviations: hESCs, human embryonic stem cell; OTX2, orthodenticle homeobox 2; TSC1, tuberous sclerosis 1.
2370 sRNA-seq Analysis of Human Embryonic Stem Cells
VC AlphaMed Press 2014 STEM CELLS

12. CONCLUSION
In conclusion, we have refined the description of a molecu-
lar signature associated with DE differentiation, a critical
step in the pathway to b-cell development. A significant por-
tion of this signature comes from novel miRNAs and alterna-
tive isomiRs. Given that the seed sequence, and thus target
specificity, is defined by the 50
end, these isomiRs would be
expected to have altered functions. Furthermore, several
isomiRs were expressed at higher levels than the nearest
annotated miRNA and therefore likely to be significant regu-
lators of cellular differentiation. In addition to the data pre-
sented here, a previous report described functional
significance to isomiR expression in mouse [3]. Moreover, as
most miRNAs annotated at miRBase have not been
described functionally, we propose that abundant isomiRs
detected by biochemical means should be given the same
consideration as annotated miRNAs with potential biological
functions.
ACKNOWLEDGMENTS
This work was supported by a grant from the California Insti-
tute for Regenerative Medicine (CIRM; www.cirm.ca.gov/) to
C.K., the Larry L. Hillblom Foundation (http://www.llhf.org/) to
C.K. and A. Hayek, and the Garb Foundation to A. Hayek. A.
Hinton is supported by a grant from the Larry L. Hillblom
Foundation. We are grateful to Dr. William Strauss for his
advice on miRNA extraction and analysis, Dr. Amy Pasquinelli
for her helpful comments and revision of the manuscript, Dr.
Sevan Ficici for assistance with programming, and Chris
Cowing-Zitron for advice and assistance with bioinformatic
tools. The funders had no role in study design, data collection
and analysis, decision to publish, or preparation of the
manuscript.
AUTHOR CONTRIBUTIONS
A. Hinton and S.E.H.: conception and design, collection and/or
assembly of data, data analysis and interpretation, and
manuscript writing; I.A. and G.A.J.: collection of data; G.B.F.:
conception and design, data analysis and interpretation, and
manuscript writing; A. Hayek: conception and design, manu-
script writing, and financial support; C.C.K.: conception and
design, data analysis and interpretation, manuscript writing,
and financial support. A. Hinton and S.E.H. contributed
equally to this work.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.
REFERENCES
1 Bartel DP. MicroRNAs: Target recognition
and regulatory functions. Cell 2009;136:215–
233.
2 Bartel DP. MicroRNAs: Genomics, bio-
genesis, mechanism, and function. Cell 2004;
116:281–297.
3 Chiang HR, Schoenfeld LW, Ruby JG,
et al. Mammalian microRNAs: Experimental
evaluation of novel and previously annotated
genes. Genes Dev 2010;24:992–1009.
4 Mallanna SK, Rizzino A. Emerging roles
of microRNAs in the control of embryonic
stem cells and the generation of induced
pluripotent stem cells. Dev Biol 2010;344:
16–25.
5 Melton C, Blelloch R. MicroRNA regula-
tion of embryonic stem cell self-renewal and
differentiation. Adv Exp Med Biol 2010;695:
105–117.
6 Wang Y, Baskerville S, Shenoy A, et al.
Embryonic stem cell-specific microRNAs
regulate the G1-S transition and promote
rapid proliferation. Nat Genet 2008;40:1478–
1483.
7 Hinton A, Afrikanova I, Wilson M, et al.
A distinct microRNA signature for definitive
endoderm derived from human embryonic
stem cells. Stem Cells Dev 2010;19:797–807.
8 Ivey KN, Muth A, Arnold J, et al. Micro-
RNA regulation of cell lineages in mouse and
human embryonic stem cells. Cell Stem Cell
2008;2:219–229.
9 Poy MN, Hausser J, Trajkovski M, et al.
miR-375 maintains normal pancreatic alpha-
and beta-cell mass. Proc Natl Acad Sci USA
2009;106:5813–5818.
10 Rosa A, Spagnoli FM, Brivanlou AH. The
miR-430/427/302 family controls mesendo-
dermal fate specification via species-specific
target selection. Dev Cell 2009;16:517–527.
11 Hinton A, Hunter S, Reyes G, et al. From
pluripotency to islets: miRNAs as critical reg-
ulators of human cellular differentiation. Adv
Genet 2012;79:1–34.
12 Melton C, Judson RL, Blelloch R. Oppos-
ing microRNA families regulate self-renewal
in mouse embryonic stem cells. Nature 2010;
463:621–626.
13 Xu C, Inokuma MS, Denham J, et al.
Feeder-free growth of undifferentiated
human embryonic stem cells. Nat Biotechnol
2001;19:971–974.
14 Hofacker IL. Vienna RNA secondary
structure server. Nucleic Acids Res 2003;31:
3429–3431.
15 McCaskill JS. The equilibrium partition
function and base pair binding probabilities
for RNA secondary structure. Biopolymers
1990;29:1105–1119.
16 Zuker M, Stiegler P. Optimal computer
folding of large RNA sequences using ther-
modynamics and auxiliary information.
Nucleic Acids Res 1981;9:133–148.
17 Janssen S, Giegerich R. Faster computa-
tion of exact RNA shape probabilities. Bioin-
formatics 2010;26:632–639.
18 Steffen P, Voss B, Rehmsmeier M, et al.
RNAshapes: An integrated RNA analysis pack-
age based on abstract shapes. Bioinformatics
2006;22:500–503.
19 Reinhart BJ, Slack FJ, Basson M, et al.
The 21-nucleotide let-7 RNA regulates devel-
opmental timing in Caenorhabditis elegans.
Nature 2000;403:901–906.
20 Brasier AR, Tate JE, Habener JF. Opti-
mized use of the firefly luciferase assay as a
reporter gene in mammalian cell lines. Bio-
techniques 1989;7:1116–1122.
21 D’Amour KA, Agulnick AD, Eliazer S,
et al. Efficient differentiation of human
embryonic stem cells to definitive endoderm.
Nat Biotechnol 2005;23:1534–1541.
22 Dweep H, Sticht C, Pandey P, et al. miR-
Walk—Database: Prediction of possible
miRNA binding sites by “walking” the genes
of three genomes. J Biomed Inform 2011;44:
839–847.
23 Laurent LC, Chen J, Ulitsky I, et al. Com-
prehensive microRNA profiling reveals a
unique human embryonic stem cell signature
dominated by a single seed sequence. Stem
Cells 2008;26:1506–1516.
24 Suh MR, Lee Y, Kim JY, et al. Human
embryonic stem cells express a unique set
of microRNAs. Dev Biol 2004;270:488–
498.
25 Barroso-del Jesus A, Lucena-Aguilar G,
Menendez P. The miR-302–367 cluster as a
potential stemness regulator in ESCs. Cell
Cycle 2009;8:394–398.
26 Barroso-delJesus A, Lucena-Aguilar G,
Sanchez L, et al. The Nodal inhibitor Lefty is
negatively modulated by the microRNA miR-
302 in human embryonic stem cells. FASEB J
2011;25:1497–1508.
27 Card DA, Hebbar PB, Li L, et al. Oct4/
Sox2-regulated miR-302 targets cyclin D1 in
human embryonic stem cells. Mol Cell Biol
2008;28:6426–6438.
28 Rosa A, Brivanlou AH. A regulatory cir-
cuitry comprised of miR-302 and the tran-
scription factors OCT4 and NR2F2 regulates
human embryonic stem cell differentiation.
EMBO J 2011;30:237–248.
29 Acampora D, Mazan S, Lallemand Y,
et al. Forebrain and midbrain regions are
deleted in Otx2-/- mutants due to a defec-
tive anterior neuroectoderm specification
Hinton, Hunter, Afrikanova et al. 2371
www.StemCells.com VC AlphaMed Press 2014

13. during gastrulation. Development 1995;121:
3279–3290.
30 Blum B, Benvenisty N. The tumorigenic-
ity of diploid and aneuploid human pluripo-
tent stem cells. Cell Cycle 2009;8:3822–
3830.
31 Bar M, Wyman SK, Fritz BR, et al. Micro-
RNA discovery and profiling in human embry-
onic stem cells by deep sequencing of small
RNA libraries. Stem Cells 2008;26:2496–2505.
32 Beattie GM, Lopez AD, Bucay N, et al.
Activin A maintains pluripotency of human
embryonic stem cells in the absence of
feeder layers. Stem Cells 2005;23:489–495.
33 James D, Levine AJ, Besser D, et al.
TGFbeta/activin/nodal signaling is necessary
for the maintenance of pluripotency in
human embryonic stem cells. Development
2005;132:1273–1282.
34 Tsai ZY, Singh S, Yu SL, et al. Identifica-
tion of microRNAs regulated by activin A in
human embryonic stem cells. J Cell Biochem
2010;109:93–102.
35 Kozomara A, Griffiths-Jones S. miRBase:
Integrating microRNA annotation and deep-
sequencing data. Nucleic Acids Res 2011;
39(Database issue):D152–157.
36 Lee LW, Zhang S, Etheridge A, et al.
Complexity of the microRNA repertoire
revealed by next-generation sequencing. RNA
2010;16:2170–2180.
37 Morin RD, O’Connor MD, Griffith M,
et al. Application of massively parallel
sequencing to microRNA profiling and discov-
ery in human embryonic stem cells. Genome
Res 2008;18:610–621.
38 Jeon HM, Sohn YW, Oh SY, et al. ID4
imparts chemoresistance and cancer stem-
ness to glioma cells by derepressing miR-9*-
mediated suppression of SOX2. Cancer Res
2011;71:3410–3421.
39 Saunders LR, Sharma AD, Tawney J,
et al. miRNAs regulate SIRT1 expression dur-
ing mouse embryonic stem cell differentia-
tion and in adult mouse tissues. Aging
(Albany NY) 2010;2:415–431.
40 Cummins JM, He Y, Leary RJ, et al. The
colorectal microRNAome. Proc Natl Acad Sci
USA 2006;103:3687–3692.
41 Landgraf P, Rusu M, Sheridan R, et al. A
mammalian microRNA expression atlas based
on small RNA library sequencing. Cell 2007;
129:1401–1414.
42 Ruby JG, Stark A, Johnston WK, et al.
Evolution, biogenesis, expression, and target
predictions of a substantially expanded set of
Drosophila microRNAs. Genome Res 2007;17:
1850–1864.
43 Houbaviy HB, Murray MF, Sharp PA.
Embryonic stem cell-specific MicroRNAs. Dev
Cell 2003;5:351–358.
44 Munoz-Sanjuan I, Brivanlou AH. Neural
induction, the default model and embryonic
stem cells. Nat Rev Neurosci 2002;3:271–280.
45 Vallier L, Reynolds D, Pedersen RA.
Nodal inhibits differentiation of human
embryonic stem cells along the neuroecto-
dermal default pathway. Dev Biol 2004;275:
403–421.
46 Erceg S, Lukovic D, Moreno-Manzano V,
et al. Derivation of cerebellar neurons from
human pluripotent stem cells. Curr Protoc
Stem Cell Biol 2012;Chapter 1:Unit 1H 5.
47 Laplante M, Sabatini DM. mTOR signal-
ing in growth control and disease. Cell 2012;
149:274–293.
48 Easley CAT, Ben-Yehudah A, Redinger CJ,
et al. mTOR-mediated activation of p70 S6K
induces differentiation of pluripotent human
embryonic stem cells. Cell Reprogram 2010;
12:263–273.
See www.StemCells.com for supporting information available online.
2372 sRNA-seq Analysis of Human Embryonic Stem Cells
VC AlphaMed Press 2014 STEM CELLS