S em Cell Repo s
Resou ce
RNA Polyme ase III Subuni POLR3G Regula es Speci ic Subse s o PolyA
+
and SmallRNA T ansc ip omes and Splicing in Human Plu ipo en S em Cells
Riikka J. Lund,
1,5,
*Nelly Rahkonen,
1,5
Maia Malonzo,
2,5
Leni Kauko,
1
Maheswa a Reddy Emani,
1
Vi pi Ki inen,
3
Elisa Na
¨ a
¨,
1
Esko Kemppainen,
1
As a Laiho,
1
Heli Sko man,
3
Ou i Ho a a,
4
Omid Rasool,
1
Ma i Nyk e ,
1,3
Ha i La
¨hdesma
¨ki,
1,2
and Rii a Lahesmaa
1
1
Tu ku Cen e o Bio echnology, Uni e si y o Tu ku and A
˚bo Akademi Uni e si y, Tu ku 20520, Finland
2
Depa men o Compu e Science, Aal o Uni e si y, Espoo 02150, Finland
3
Facul y o Medicine and Li e Sciences, BioMediTech, Uni e si y o Tampe e, Tampe e 33014, Finland
4
Depa men CLINTEC, Ka olinska Ins i u e , Ka olinska Uni e si y Hospi al Huddinge, S ockholm 171 77, Sweden
5
Co- i s au ho
*Co espondence: iikka.lund@u u. i
h p://dx.doi.o g/10.1016/j.s emc .2017.04.016
SUMMARY
POLR3G is exp essed a high le els in human plu ipo en s em cells (hPSCs) and is equi ed o main enance o s em cell s a e
h ough mechanisms no known in de ail. To explo e how POLR3G egula es s em cell s a e, we ca ied ou deep-sequencing anal-
ysis o polyA
+
and smallRNA ansc ip omes p esen in hPSCs and egula ed in POLR3G-dependen manne . Ou da a e eal
ha POLR3G egula es a speci ic subse o he hPSC ansc ip ome, including mul iple ansc ip ypes, such as p o ein-coding
genes, long in e ening non-coding RNAs, mic oRNAs and small nucleola RNAs, and a ec s RNA splicing. The p ima y unc ion
o POLR3G is in he main enance a he han ep ession o ansc ip ion. The majo i y o POLR3G polyA
+
ansc ip ome is egu-
la ed du ing di e en ia ion, and he key plu ipo ency ac o s bind o he p omo e s o a leas 30% o he POLR3G- egula ed an-
sc ip s. Among he di ec a ge s o POLR3G, POLG is po en ially impo an in sus aining s em cell s a us in a POLR3G-dependen
manne .
INTRODUCTION
Human emb yonic s em cells (hESCs) a e plu ipo en cells
ha ing a unique capaci y o sel - enew and di e en ia e
in o all specialized cell ypes ound in soma ic issues
(Thomson e al., 1998). Simila p ope ies a e gained by
human induced plu ipo en s em cells (hiPSCs) ep og-
ammed om cells ound in adul issues (Yu e al., 2007;
Takahashi e al., 2007). The co e ac o s known o be c ucial
o main enance and con ol o plu ipo ency include
POU5F1, SOX2, and NANOG (Boye e al., 2005; Chambe s
and Tomlinson, 2009). These ansc ip ion ac o s ope a e
in ac i a ion o ep ession o genes impo an o plu ipo-
ency and di e en ia ion, and o m an au o egula o y
loop o posi i ely egula e hei own exp ession. In addi-
ion, POU5F1 and SOX2, in combina ion wi h ei he
KLF4 and CMYC o NANOG and LIN28, we e used in
he i s s udies ep og amming adul cells back o he
plu ipo en s a e, highligh ing he impo ance o hese
ac o s in he egula ion o plu ipo ency (Yu e al., 2007;
Takahashi e al., 2007). Mo eo e , epigene ic egula ion
and pos - ansc ip ional mechanisms, such as mic oRNAs
(miRNAs), a e impo an in he egula ion o plu ipo ency
(Young, 2011). Howe e , he mechanisms unde lying he
unique p ope y o plu ipo ency a e s ill no comple ely
unde s ood.
In euka yo ic cells, h ee DNA-di ec ed RNA polyme -
ases, polyme ases I, II, and III (Pol I–III), egula e an-
sc ip ion o di e en se s o a ge genes. Pol III egula es
ansc ip ion o s uc u al RNAs ( RNAs and 5S RNAs)
and se e al small non-p o ein-coding RNAs (ncRNAs)
including miRNAs (Ozsolak e al., 2008; Bo che e al.,
2006; Ole e al., 2010; Dieci e al., 2007). Pol III, like
Pol I and Pol II, is composed o mul iple subuni s o
which POLR3G (DNA-di ec ed RNA polyme ase III sub-
uni , RPC32, RPC7) is a Pol III-speci ic subuni wi h no
coun e pa in Pol I o Pol II. POLR3G subuni is needed
o ansc ip ional ini ia ion o Pol III, hus being impo -
an o he p ope unc ion o his polyme ase (Wang
and Roede , 1997).
In iguingly, consis en wi h p e ious s udies (Wong
e al., 2011; En e e al., 2005; Hau ie e al., 2010), ou
da a show ha POLR3G is highly exp essed in plu ipo-
en s em cells and is equi ed o main enance o he
plu ipo en s a e. Howe e , he mechanisms by which
how POLR3G con ibu es o main enance o s em cell
s a e and i s a ge s in s em cells a e no well known.
To explo e he unc ion o POLR3G in s em cells, we
ha e ca ied ou deep ansc ip ome analysis o mRNAs
and smallRNAs p esen in hESCs and egula ed in a
POLR3G-dependen manne . Ou da a e eal a special-
ized unc ion o POLR3G in he ansc ip ional main e-
nance o he hESC s a e and egula ion o de elop-
men al p og ams h ough a speci ic subse o coding
and non-coding ansc ip omes and egula ion o al e na-
i e splicing.
1442 S em Cell Repo s jVol. 8 j1442–1454 jMay 9, 2017 jª2017 The Au ho (s).
This is an open access a icle unde he CC BY license (h p://c ea i ecommons.o g/licenses/by/4.0/).
RESULTS
Deep PolyA
+
T ansc ip ome o hESCs
To explo e he egula ion o ansc ip ion by POLR3G in
hESCs, we i s es ablished a comp ehensi e e e ence an-
sc ip ome o hESCs. Fo his pu pose we explo ed how
la ge a p opo ion o he di e en ansc ip ypes a e p e-
sen in hESCs. Using mRNA sequencing (mRNA-seq),
om he o al numbe o 181–224 310
6
eads pe sample
we de ec ed al oge he 13,613 polyA
+
ansc ip s, which
we e p esen in all o he h ee hESC samples examined
wi h a minimum RPKM ( eads pe kilobase pe million
mapped eads) alue o 0.5 (Table S1A). This ep esen s
22% o he o al 60,679 ansc ip s p esen in he hg19
and miRBase e e ence genomes (Figu e 1A). The mos
abundan class o ansc ip s de ec ed in hESCs was p o-
ein-coding genes, which included 11,855 ansc ip s ep-
esen ing 58% o he p o ein-coding genes in he gencode
(hg19) anno a ions (Figu e 1A) and 87% o all he polyA
+
ansc ip s ha we e de ec ed in hESC ansc ip ome (Fig-
u e 1B). The h ee o he mos abundan classes o an-
sc ip s we e pseudogenes (n = 621, 5% o hESC ansc ip-
ome), long in e ening non-coding RNAs (lincRNAs)
(n = 428, 3% o hESC ansc ip ome), and an isense an-
sc ip s (n = 361, 3% o hESC ansc ip ome). Howe e , in
all hese o he classes less han 5% o he ansc ip s p e-
sen in he e e ence human genome we e de ec ed wi h
mRNA-seq (Figu e 1A).
SmallRNA T ansc ip ome o hESCs
Wi h smallRNA-seq we de ec ed al oge he 1,361 an-
sc ip s p esen in all hESC samples wi h a minimum ead
pe kilobase o 0.5 and a maximum leng h o 250 bp (Table
S1B). This ep esen ed 2% o all he ansc ip s in hg19 and
miRBase e e ence genomes (Figu e 1A). The mos abun-
dan class o smallRNA ansc ip s was ma u e miRNAs
(n = 864), which ep esen 33% o he miRNAs in he
miRBase e e ence anno a ions o 63% o all he smallRNAs
ha we e de ec ed in hESC small ansc ip ome (Figu e 1C).
The o he mos abundan classes o ansc ip s we e small
nucleola RNAs (snoRNAs) (n = 147 ansc ip s, 11% o
hESC smallRNA ansc ip ome) and p e-miRNAs (n = 81,
6% hESC small ansc ip ome), RNAs (n = 118, 9% o
hESC smallRNA ansc ip ome), and RNAs (n = 57, 4% o
hESC smallRNA ansc ip ome).
POLR3G-Dependen PolyA
+
T ansc ip ome in hESCs
Consis en wi h p e ious indings (En e e al., 2005;
Wong e al., 2011; Hau ie e al., 2010), we ound bo h
POLR3G mRNA and p o ein o be exp essed a high le el
in hPSCs and o be apidly down egula ed du ing ea ly
di e en ia ion (Figu es 2A, S1A, and S1B). Exp ession o
Figu e 1. Deep SmallRNA and PolyA
+
RNA Re e ence T ansc ip ome o Human Emb yonic S em Cells
(A–C) PolyA
+
and smallRNAs we e isola ed om h ee independen expe imen s wi h wo di e en human emb yonic s em cells lines
(H9p38, HS360p62, HS360p66). The deep ansc ip ome was analyzed wi h smallRNA-seq and mRNA-seq. The sequence da a we e mapped
o he hg19 e e ence genome assembly and miRBase da abase (based on hg19 build). T ansc ip ome anno a ions we e aken om gencode
19 and RNA anno a ions om UCSC genome b owse (in addi ion o miRBase). The polyA
+
ansc ip s wi h a minimum RPKM alue o 0.5
and he smallRNA ansc ip s wi h a minimum RPM alue o 0.5 we e ex ac ed om he da a and classi ied in o di e en ansc ip ype
ca ego ies. In (A) he ba s ep esen he p opo ion o each ansc ip class de ec ed in hESCs in compa ison wi h he o al numbe
(indica ed in he igu e) o each ansc ip ype p esen in he e e ence genome. The p opo ions o di e en ansc ip classes p esen in
hESCs and de ec ed wi h mRNA-seq (B) and smallRNA-seq (C) a e isualized as pie cha s.
S em Cell Repo s jVol. 8 j1442–1454 jMay 9, 2017 1443
POLR3G p o ein was ecip ocal o al e na i e o m
POLR3GL, which was induced in esponse o di e en ia-
ion (Figu es S1A and S1B). In addi ion, POLR3G was ex-
p essed consis en ly in a POU5F1-dependen manne (Fig-
u e S1C). To s udy he unc ion and impo ance o POLR3G
in hESCs, we ca ied ou small in e e ing RNA (siRNA)-
media ed knockdown o POLR3G in wo di e en hESC
lines (H9 and HS360) wi h wo di e en siRNAs and
siRNA pool (see Supplemen al Expe imen al P ocedu es).
Silencing o POLR3G led o clea loss o mo phology ypical
o plu ipo en hESCs and dec eased he numbe o cells
p esen in he cul u es (Figu e S2A). In pa icula wi h
siRNA1, due o apid and s ong dec ease in cell numbe s,
no enough ma e ial was always ob ained o u he
Figu e 2. POLR3G-Dependen Human Emb yonic S em Cell T ansc ip ome
(A) No malized gene exp ession le els o POLR3G om ESTOOLS da a@hand da abase as measu ed wi h A yme ix a ays in 384 samples.
The combined numbe (n) o eplica ed samples om di e en s udies is indica ed in he igu e. A e age old changes and s a is ical
signi icance (unpai ed es ) be ween hPSCs in compa ison wi h la e emb yonic bodies (EB) and o he cell ypes is indica ed in he igu e.
See also Figu e S1.
(B) The p opo ions o di e en polyA
+
ansc ip ypes egula ed in POLR3G-dependen manne in hESCs wi h minimum old change o 1.5
and FDR %0.05 in h ee independen biological eplica es.
(C) The old changes and FDR alues o all POLR3G-dependen ansc ip s in mRNA-seq da a.
(D) The p opo ions o di e en smallRNA ansc ip ypes egula ed in POLR3G-dependen manne in hESCs wi h minimum old change o
1.5 and FDR %0.05 in ou siRNA samples in compa ison wi h h ee non- a ge ed siRNA con ols.
(E) The a e age old changes and FDR alues o all POLR3G-dependen ansc ip s wi h size below 250 bp in smallRNA-seq da a om ou
siRNA samples in compa ison wi h h ee non- a ge ed siRNA con ols.
See also Figu e S2.
1444 S em Cell Repo s jVol. 8 j1442–1454 jMay 9, 2017
analysis, indica ing he impo ance o POLR3G in hESC
main enance. We did no obse e an inc ease in he
numbe o dead cells in esponse o POLR3G knockdown,
al hough hese cells would ha e been washed o in he
daily media change (Figu e S2A).
We nex examined he unc ion o POLR3G in he egu-
la ion o hESC ansc ip ome. Compa ison o he polyA
+
ansc ip omes o hESCs be o e and a e POLR3G silencing
e ealed changes in exp ession o 718 ansc ip s ep esen -
ing 5% o he hESC ansc ip ome (Table S2A). The mos
abundan ansc ip ional class egula ed by POLR3G (Fig-
u e 2B) was p o ein-coding genes (n = 593), which ep e-
sen ed 83% o all he polyA
+
ansc ip s egula ed by
POLR3G and 5% o he p o ein-coding genes p esen in
hESC ansc ip ome. The o he less abundan polyA
+
an-
sc ip classes egula ed by POLR3G included 36 lincRNAs
(5%), 27 pseudogenes (4%), and 23 an isense ansc ip s
(6% o POLR3G-dependen ansc ip s). No inc ease in
he exp ession o al e na i e a ian POLR3GL was de-
ec ed a ei he mRNA o p o ein le el (da a no shown),
al hough his a ian is induced du ing ea ly di e en ia-
ion and is exp essed in ecip ocal manne o POLR3G (Fig-
u e S1). Consis en ly wi h ou obse a ions om cell cul-
u es and cell coun s (Figu e S2), among he genes wi h
he s onges dec ease in esponse o POLR3G was a ma ke
o p oli e a ion, MKI67 ( old change = 5.9, alse disco e y
a e [FDR] = 1.71 310
8
,Table S2), indica ing dec eased
p oli e a ion o he cells. Ou indings show ha POLR3G
egula es a speci ic subse o ansc ip s in hESCs consis ing
o mul iple ansc ip ypes. Howe e , he mos abundan
POLR3G-dependen ansc ip g oup in ou mRNA-seq
da a is clea ly he p o ein-coding genes. Mos o he
POLR3G-dependen polyA
+
ansc ip s we e dec eased
(n = 681, 94.8%) a he han inc eased (n = 37, 5.2%) in
esponse o POLR3G knockdown (Figu e 2C), indica ing
ha POLR3G has a unc ion p ima ily in main enance
a he han ep ession o ansc ip ion in hESCs.
POLR3G-Dependen SmallRNA T ansc ip ome
in hESCs
The smallRNA-seq e ealed changes in 97 POLR3G-depen-
den smallRNAs ep esen ing 7% o he hESC smallRNA
ansc ip ome (Table S2B). The mos abundan ansc ip-
ional class egula ed by POLR3G (Figu e 2D) was ma u e
miRNAs (n = 59) ep esen ing 61% o all he smallRNAs
egula ed by POLR3G and 7% o he miRNAs p esen in
hESC smallRNA ansc ip ome. The o he abundan small-
RNA classes egula ed by POLR3G included p e-miRNAs
(n = 20) and snoRNAs (n = 10), ep esen ing 21% and
10% o all POLR3G-dependen smallRNAs, espec i ely.
Simila ly o polyA
+
ansc ip ome, mos o he POLR3G-
dependen smallRNAs (n = 84, 87%) we e down egula ed
and only a small p opo ion (n = 13, 13%) was up egula ed
in esponse o POLR3G knockdown (Figu e 2E and Table
S2B). In e es ingly, mos o he smallRNAs (n = 10, 77%)
induced in esponse o POLR3G silencing we e snoRNAs.
Hal o hese we e om imp in ed locus in he ch omo-
some 14q32. In compa ison wi h polyA
+
ansc ip s, he
impac o POLR3G deple ion on smallRNAs was s ong in
magni ude. Fo polyA
+
ansc ip s he a e age old change
o ep essed ansc ip s in esponse o POLR3G silencing
was 1.9, whe eas o smallRNAs i was 5.9. These esul s
indica e ha in addi ion o mul iple ypes o polyA
+
an-
sc ip s, POLR3G is equi ed o main enance o a speci ic
subse o miRNAs in hESCs. In addi ion, POLR3G ep esses
ansc ip ion o a small ac ion o snoRNAs.
Func ional En ichmen Analysis o
POLR3G-Regula ed T ansc ip s
To examine he unc ion o he POLR3G-dependen an-
sc ip s, we pe o med pa hway analysis wi h he Ingenui y
Pa hway Analysis Tool (Qiagen). Consis en wi h ou ind-
ings, he esul s linked he POLR3G- egula ed polyA
+
an-
sc ip s o impo an unc ions in he cellula main enance
and egula ion o ea ly de elopmen al p og ams (Table 1).
The unc ions o POLR3G-dependen non-coding an-
sc ip s, including smallRNAs, a e la gely unknown.
A o al o 114 molecules linked o plu ipo ency unc ion
we e ound in he Ingenui y Pa hway da abase. O hese,
eigh (DNMT1,APC,SMARCA4,NIPBL,RIF1,RTF1,TET1,
and LIN28A) showed down egula ion om 1.5- old o
2.9- old (FDR < 0.05) in esponse o POLR3G knockdown.
In addi ion, se e al o he miRNAs p edic ed o a ge hese
Table 1. Func ional En ichmen o he POLR3G-Dependen
PolyA
+
T ansc ip s
Func ional Ca ego y
No. o
Molecules
Fishe Exac
Tes p Value
Cellula and Molecula Func ions
Cellula assembly and o ganiza ion 174 0.0025
Cellula unc ion and main enance 150 0.0024
Cell cycle 109 0.0026
DNA eplica ion, ecombina ion, and epai 41 0.0023
Cell mo phology 148 0.0024
Physiological Sys em De elopmen and Func ion
Emb yonic de elopmen 122 0.0026
Ne ous sys em de elopmen and unc ion 145 0.0024
O gan de elopmen 79 0.0021
O ganismal de elopmen 172 0.0023
Tissue de elopmen 139 0.0024
S em Cell Repo s jVol. 8 j1442–1454 jMay 9, 2017 1445
ac o s we e egula ed in a POLR3G-dependen manne .
Howe e , mos o hem we e down egula ed a he han
up egula ed. Only wo o he miRNAs we e up egula ed
(hsa-miR-4472, 45.89- old, FDR = 0.04; and hsa-miR-4695-
5p, 24.63- old, FDR = 0.0005) and hei p edic ed a ge s
RTF1 (1.62, FDR = 0.0003) and RIF1 (1.55, FDR =
0.004), espec i ely, we e down egula ed. Among hedown-
egula ed plu ipo ency-associa ed genes was also LIN28A
(1.88, FDR = 2.12 310
4
), which has a well-known unc-
ion in he egula ion o mouse ESC sel - enewal h ough
supp ession o le -7 miRNA ma u a ion (Shyh-Chang and
Daley, 2013). Le -7 miRNAs again ep ess LIN28A du ing
s em cell di e en ia ion. The e o e, we looked in o he
le -7 le els in he da a and ound inc eased le els o hsa-
le -7b-5p (4.35- old, p = 0.02) and hsa-le -7c (5.21- old,
p = 0.01) wi h unadjus ed p alues in esponse o POLR3G
silencing. Noconsis en changes we eobse ed o heo he
le -7 amily membe s.
Taken oge he , based on unc ional en ichmen anal-
ysis, POLR3G is equi ed o main enance o ansc ip s
impo an in egula ion o cellula main enance, p oli e a-
ion, and ea ly de elopmen al p og ams. Impo an ly, he
unc ions o he POLR3G-dependen non-coding RNAs
a e la gely unknown, highligh ing he lack o knowledge
in he impo ance o non-coding ansc ip ome in he
egula ion o plu ipo ency.
Regula ion o POLR3G-Dependen T ansc ip s du ing
Ea ly Di e en ia ion
To s udy whe he he POLR3G-dependen ansc ip s a e
egula ed du ing di e en ia ion o hESCs, we examined
egula ion o hese ansc ip s in he deeply sequenced
RNA-seqda a a ailable ondi e en ia ion o hESCs in o h ee
di e en ge m cell lineages (endode m, mesode m, and ec o-
de m) by Gi o d e al. (2013). O ou 718 POLR3G- egula ed
polyA
+
ansc ip s, 537 we e p esen in he da a by Gi o d
e al. (2013). O hese, 70% (n = 377) showed a leas 1.5-
old change (p alue cu o 0.05) in esponse o di e en ia-
ion (Table S3). This ep esen ed 3.8% o he all 10,010 di e -
en ially egula ed genes in he da a by Gi o d e al. (2013).
The POLR3G- egula ed ansc ip s included bo h induced
and ep essed genes and we e associa ed wi h all h ee
ge m cell lineages endode m (98 down, 120 up), ec ode m
(108 down, 116 up), and mesode m (151 down, 104 up).
F om hese obse a ions we can conclude ha POLR3G eg-
ula es a speci ic subse o ansc ip s in hESCs, mos o which
a e di e en ially egula ed du ing ea ly di e en ia ion o
hESCs o all he di e en ge m cell lineages.
Impac o POLR3G Silencing on Al e na i e Splicing o
T ansc ip s
We nex analyzed he impac o POLR3G silencing on he
al e na i e splicing o ansc ip s. Using he MATS me hod
( eplica e mul i a ia e analysis o ansc ip splicing) (Shen
e al., 2014), he numbe o splicing e en s de ec ed in he
samples anged om housands ( e ained in on) o ens o
housands (skipped exon) based on coun s om eads map-
ping o exon junc ions and al e na i e exons (and simila
numbe s based on jus junc ion coun s), whe eas he com-
bined numbe o al e na i e splicing e en s (ASEs) om
bo h junc ion coun s and coun s om junc ion and al e -
na i e exon eads (FDR < 0.05, DJ > 0.2) o he ollowing
subclasses we e: 109 (71) skipped exon (SE), 25 (16) al e -
na i e 50splice si e (A5SS), 15 (7) al e na i e 30splice si e
(A3SS), 62 (49) e ained in on (RI), and 22 (11) mu ually
exclusi e exons (MXE) (Table S4). The numbe s in pa en-
heses indica e he numbe o ASEs whe ein he sign o
DJ was consis en among all h ee case-con ol pai s (in
con as o ASEs ha had oo ew eads in some samples
o hose whe ein he sign o DJ a ied among case-con ol
pai s).
Splicing analysis wi h MISO (mix u e o iso o ms) (Ka z
e al., 2010) was pe o med independen ly o each o
he h ee case-con ol pai eplica es and splicing e en s
mee ing he il e ing c i e ia (Bayes ac o R5, DJ >0.2,
numbe o eads suppo ing inclusion/exclusion iso o m
o case/con ol samples, o ice e sa, R10) o all he
h ee eplica e pai s and wi h consis en sign o DJ we e
conside ed signi ican . Using hese ad hoc il e ing c i e ia,
only six ASEs we e de ec ed (Table 2).
Gi en he ela i ely small numbe o s a is ically signi i-
can splicing e en s om MISO analysis and disc epancies
in he splicing e en anno a ions, only one ASE, skipped
exon o HDAC7 gene (exon 9 o Re Seq NM_015401), was
de ec ed by bo h MATS (FDR = 8.82 310
22
,DJ = 0.29)
and MISO (Bayes ac o = 1.00E+12, DJ = 0.21–0.37) (Fig-
u e S3). This ASE was subjec ed o u he expe imen al
alida ion wi h qRT-PCR, which con i med inc eased
exp ession ( old change 1.52–7.03, p = 0.09) o he a ian
lacking he exon a e silencing o POLR3G in all ou bio-
logical eplica es examined (da a no shown).
En ichmen o Plu ipo ency Fac o s in he P oximi y
o POLR3G-Dependen T ansc ip s
We nex examined he binding o ansc ip ion ac o s,
impo an o plu ipo ency, in he p omo e s o he
POLR3G- egula ed ansc ip s. Fo he compa ison we
used exis ing ch oma in immunop ecipi a ion sequencing
(ChIP-seq) da a by Lis e e al. (2009) on POU5F1, NANOG,
SOX2, and KLF4 binding. Addi ionally EP300, a his one
ace yl ans e ase and ansc ip ional co-ac i a o impo -
an o s em cell di e en ia ion, was included in he
analysis. The p omo e was de ined as a egion including
1,000 bp ups eam om he ansc ip ion s a si e (TSS)
o he ansc ip and included i s exon and in on, i p e-
sen . We ound signi ican en ichmen o NANOG, SOX2,
1446 S em Cell Repo s jVol. 8 j1442–1454 jMay 9, 2017

and KLF4, bu no POU5F1, in he p omo e s o he
POLR3G- egula ed polyA
+
ansc ip s in compa ison wi h
simila en ichmen in he all he genes exp essed in hESCs
(Table 3). In e es ingly, en ichmen o he subuni o
Pol III complex, POLR3A, was no obse ed, al hough
his subuni has been p e iously epo ed o co-localize
wi h he plu ipo ency ac o s (Alla and Cai ns, 2014). A
o al o 311 (30.5%) o all he POLR3G-dependen polyA
+
ansc ip s exp essed in hESCs had a binding si e o
one o mo e plu ipo ency ac o s in he p omo e egion
(Figu e 3A).
POLR3G Binding Si es in Genome
ChIP-seq was ca ied ou o iden i y he genomic binding
si es o POLR3G and di ec a ge s o ansc ip ional
egula ion. A e se e al ailed expe imen s wi h hESC
lines, good-quali y ChIP-seq da a we e ob ained wi h one
o he wo eplica es o plu ipo en NT2D1 line. Al oge he
836 peaks (p = 1.00 310
4
) we e de ec ed in he POLR3G
ChIP-seq da a. Fi s we compa ed he genomic egions
bound by POLR3G wi h he da a a ailable on POLR3A
binding si es in H1 hESCs by Alla and Cai ns (2014).O
he 330 egions bound by POLR3A, 225 we e o e lapping
wi h genomic egions bound by POLR3G. These egions
we e chosen o u he analysis o ensu e iden i ica ion
o POLR3G- egula ed a ge s wi h high con idence. The
majo i y o he POLR3G binding si es we e localized as
clea ly de ined peaks o e lapping wi h 245 di e en
RNA genes (Table S5). In addi ion, 13 peaks we e o e lap-
ping wi h non-coding RNAs encoding o componen s
o he ibonucleop o eins (RMRP,RN7SK,-7SL1,-7SL2,
RNU6-1,-2,-8,-9,-ATAC,RNY1,-3,-4,-5). Two peaks o e -
lapped miRNAs (p edic ed AC008738.2,MIR3676) and one
aul RNA (VTRNA1-3).
To iden i y he di ec a ge s o he POLR3G pu a i ely
impo an in he main enance o hESC s a us, we iden i ied
he o e laps in he genomic binding si es o POLR3G/
POLR3A and changes in he ansc ip ion o coding o
non-coding genes in esponse o POLR3G knockdown.
Acco ding o ou esul s, m DNA polyme ase POLG was
he only p o ein-coding gene, down egula ed (1.5- old,
FDR = 2.00 310
2
) in esponse o POLR3G knockdown
wi h POLR3G/POLR3A binding si e in he p oximal p o-
mo e . The exac binding si e o POLR3G in POLG p o-
mo e o e lapped wi h he RNA-A g-TCG-1-1 gene, unc-
ional genomic elemen wi h ac i e o weak p omo e
his one ma ks, and nume ous ansc ip ion ac o binding
si es based on ENCODE da a (ENCODE P ojec Con-
so ium, 2012)(Figu e 3B). Consis en ly, RNA-A g-CGA
ansc ip exp ession was inc eased on a e age 5.0- old
(p = 4.00 310
2
) in smallRNA-seq da a ac oss he biological
eplica es. In addi ion, se e al o he RNAs bound showed
al e ed exp ession le els in esponse o POLR3G knock-
down. These included RNA-Th -ACY ( old change = 4.5,
FDR = 3.10 310
2
), RNA-Leu-TTA(m) ( old change =
7.15, FDR = 8.00 310
3
), RNA-Ala-GCA ( old change =
3.1, p = 4.90 310
2
) in he smallRNA-seq da a, and
RNA-Gly-GGA ( old change = 2.2, p = 2.32 310
2
) and
RNA-Me ( old change = 2.68, p = 2.10 310
2
) in he
mRNA-seq da a. Compa ison o he ansc ip ome da a
Table 2. Al e na i e Splicing E en s De ec ed by MISO S a is ical Model in Response o Silencing o POLR3G in hESCs
Splicing E en (Genomic Loci) Ca ego y Ensembl ID Gene ID Bayes Fac o Dc
ch 12:48189990:48190081:-
@ch 12:48189689:48189799:-
@ch 12:48189370:48189550:-
SE ENSG00000061273 HDAC7 R1.00E+12 R0.21
ch 17:74087224:74087316:-
@ch 17:74086410:74086478:-
@ch 17:74085256:74085401:-
SE ENSG00000182473 EXOC7 R76.48 %0.20
ch 19:2226182:2227126:-
@ch 19:2227736:2228381:-
@ch 19:2229784:2232577:-
SE ENSG00000104885 DOT1L R100E+12 %0.24
ch 15:137356720:137356886:-
@ch 15:137354644:137354835:-
@ch 15:137353991:137354203:-
SE ENSG00000031003 FAM13B R6.95 %0.27
ch 4:119459016:119459112I 119459163:+
@ch 4:119461374:119461544:+
A5SS ENSG00000154608 CEP170P1 R11.89 %0.3
ch 7:23562051-23561740:-
@ ch 7:23561459-23561326:-
RI ENSG00000164548 TRA2A R610.64 %0.24
See also Figu e S3.
S em Cell Repo s jVol. 8 j1442–1454 jMay 9, 2017 1447
and he ChIP-seq da a e ealed o e laps also in a ew o he
non-coding smallRNAs ha had al e ed exp ession le els,
al hough he s a is ical signi icance was no high. The p e-
dic ed miRNA AC008738.2 was bound by POLR3G and
exp ession was dec eased 1.6- o 9.3- old (p = 1.00 3
10
2
) in esponse o knockdown. In addi ion, he small-
RNAs RNY1 (ENSG00000201098.1, old change = 1.5–5.0,
p = 4.60 310
2
), RNY4 (ENSG00000252316.1, old
change = 1.5–5.9, p = 4.00 310
2
), and RNY5
(ENSG00000252310.1, old change = 1.4–2.6, p = 3.40 3
10
2
), bound by POLR3G, had inc eased exp ession
in esponse o POLR3G knockdown in all o he ou
biological eplica es. Also, exp ession o he Vaul RNA
(VTRNA1–3), bound by POLR3G, had al e ed exp ession
(1.5- o 1.8- old) in all o he ou biological eplica es,
al hough s a is ical signi icance was low (p = 2.01 310
1
).
In conclusion, ou da a indica e ha main enance o
s em cell s a us by POLR3G may be media ed h ough
egula ion o POLG gene. The egula ion o s em cell s a us
may also in ol e ac i i y o AC008738.2,VTRNA1–3,a
subse o speci ic RNA genes, and componen s o ibonu-
cleop o eins RNY1,RNY4, and RNY5.
DISCUSSION
POLR3G is a key ac o equi ed o main enance o hESC
s a e, as silencing o POLR3G leads o di e en ia ion and
a enua es p oli e a ion o hESCs (Wong e al., 2011; En-
e e al., 2005; Hau ie e al., 2010). Howe e , he mecha-
nisms by which POLR3G sus ains undi e en ia ed s a us
o hESCs and p e en s di e en ia ion ha e no been
cha ac e ized in de ail. Ou esul s show ha POLR3G
is equi ed o he main enance o se e al ypes o
polyA
+
and smallRNA ansc ip s, including p o ein-cod-
ing genes, pseudogenes, lincRNAs, an isense ansc ip s,
and miRNAs in hESCs. In e es ingly, he POLR3G-depen-
den ansc ip ome is a he speci ic, as only 5% o he
polyA
+
ansc ip s and 7% o he smallRNA ansc ip s
in hESCs we e POLR3G dependen . This indica es ha a
la ge p opo ion o he ansc ip s in hESCs a e also main-
ained h ough POLR3G-independen mechanisms. How-
e e , he se e e pheno ypic and p oli e a ion changes
induced in esponse o POLR3G silencing and accompa-
nied by hese ansc ip ional al e a ions indica e ha
POLR3G-dependen egula ion o hese speci ic an-
sc ip s is c ucial o he main enance o s em cell s a e
and p oli e a ion.
We also ound ha silencing o POLR3G leads o
changes in he exp ession le els o al e na i ely spliced
ansc ip a ian s. O hese we u he alida ed he
s onges obse a ion, skipping o exon 9 o he HDAC7
Re Seq a ian 1. Elucida ion o he unc ional signi i-
cance o his ASE equi es u he s udies. The impo ance
o HDAC7 in he egula ion o ea ly s em cell di e en ia-
ion and de elopmen is suppo ed by a p e ious s udy
on a mouse s em cell model, which showed ha Hdac7
is equi ed o di e en ia ion o smoo h muscle cells
and o unde go al e na i e splicing du ing di e en ia ion,
albei a di e en si es a ec ing usage o exon 1 (Ma ga i i
e al., 2009).
The POLR3G-dependen ansc ip s a e unc ionally
linked o emb yonic de elopmen and key cellula unc-
ions impo an o cellula main enance and p oli e a ion.
The known plu ipo ency egula o s a e en iched in he
TSSs o hese genes, and he majo i y o he polyA
+
an-
sc ip s a e egula ed du ing he di e en ia ion o hESCs
o h ee di e en ge m cell lineages. This egula ion o
POLR3G-media ed ansc ip s mus be impo an , enabling
lineage speci ica ion and u he de elopmen o he cells
and issues, as p e ious s udies ha e shown ha o e ex-
p ession o POLR3G in he cells leads o esis ance o di e -
en ia ion (Wong e al., 2011).
The e ec s o POLR3G knockdown on smallRNA exp es-
sion we e s onge han on polyA
+
ansc ip ome, indi-
ca ing ha POLR3G is likely o ha e an impo an unc ion
in he main enance o s em cell s a e and egula ion o plu-
ipo ency h ough speci ic subse s o smallRNAs.
Table 3. Binding and En ichmen o Regula o y Fac o s in he P omo e s o POLR3G-Dependen PolyA
+
Genes in Compa ison wi h
All De ec ed Genes
DNA Binding
Fac o
No. o Bound Genes
in he Genome
No. o Bound Genes
in hESCs
No. o Bound
POLR3G-Dependen Genes
Hype geome ic
Tes p Value
Adjus ed p Value
(Benjamini-Hochbe g)
NANOG 25,071 1,668 (12.2%) 166 (16.3%) 3.64 310
5
2.19 310
4
KLF4 3,793 1,116 (8.2%) 105 (10.3%) 4.55 310
4
1.36 310
3
SOX2 5,682 467 (3.4%) 49 (4.8%) 6.57 310
3
1.31 310
2
EP300 3,093 280 (2.1%) 37 (3.6%) 9.08 310
3
1.36 310
2
POU5F1 3,889 1,524 (11.1%) 123 (12.1%) 1.74 310
1
2.09 310
1
POLR3A 389 40 (0.3%) 3 (0.3%) 5.81 310
1
5.81 310
1
1448 S em Cell Repo s jVol. 8 j1442–1454 jMay 9, 2017
In e es ingly, al hough based on ansc ip ome analysis,
POLR3G is equi ed o main ain exp ession o mul iple
ypes o ansc ip s in hESCs, only ew a e di ec a ge s
o POLR3G. We ound m DNA polyme ase POLG o
be he only p o ein-coding gene di ec ly egula ed by
POLR3G. The e o e, POLG is a s ong candida e o
Figu e 3. O e lap o he Genomic Binding Si es o Known Plu ipo ency Regula o s and POLR3G wi h he POLR3G-Regula ed
T ansc ip ome
(A) En ichmen s o POU5F1, NANOG, SOX2, and KLF4 in he p omo e egion o POLR3G-dependen ansc ip s in he genome we e
examined by compa ing he binding o he ac o s o he p oximi y (1,000 om TSS + i s exon and in on, i p esen ) o POLR3G-
egula ed polyA
+
genes in compa ison wi h all he polyA
+
genes exp essed by hESCs be o e o a e POLR3G knockdown. The o e lap o he
binding si es o di e en ac o s is illus a ed in he igu e.
(B) Binding o POLR3G in o he genome o emb yonal ca cinoma-de i ed plu ipo en s em cells (NT2D1) was s udied wi h ChIP-seq.
Binding si es we e compa ed wi h he ansc ip ional changes obse ed in esponse o POLR3G knockdown and in co ela ion o ENCODE
da a a ailable on ch oma in s a us o plu ipo en s em cells. The igu e shows a snapsho om in eg a i e analysis a he ansc ip ional
s a si e o POLG gene showing genomic binding and ansc ip ional egula ion o he locus by POLR3G.
S em Cell Repo s jVol. 8 j1442–1454 jMay 9, 2017 1449
media ing POLR3G-dependen main enance o s em cell
s a us. POLG encodes he ca aly ic subuni o he m DNA
polyme ase, and is hus equi ed o he p ope unc ion
and genomic in eg i y o mi ochond ia and is essen ial
o ea ly emb yonic de elopmen . Mi ochond ial unc ion
has been shown o be c ucial o he plu ipo ency and di -
e en ia ion o emb yonic s em cells (Hance e al., 2005;
Facucho-Oli ei a e al., 2007; Xu e al., 2013). Silencing
o POLG in mouse emb yonic s em cells leads o loss o
POU5F1 and induc ion o B achyu y p o ein exp ession
(Hance e al., 2005; Facucho-Oli ei a e al., 2007; Xu
e al., 2013), demons a ing he impo ance o he gene
in main enance o plu ipo ency.
In addi ion, binding and POLR3G-dependen exp ession
was de ec ed o a subse o RNA genes, p edic ed miRNA
AC008738.2,VTRNA1–3, and ibonucleop o eins RNY1,
RNY4, and RNY5, wi h cu en ly unclea unc ion in he
egula ion o plu ipo ency. S ong binding o POLR3G
o nume ous RNAs wi hou ansc ip ional changes in
esponse o knockdown hin s a a po en ial unc ion o he
han di ec egula ion o ansc ip ion. Fu he s udies a e
needed o cla i y, o example, whe he POLR3G has a
unc ion in con olling accessibili y o ch oma in o egula-
o y ac o s o media es egula ion o gene ansc ip ion
h ough long- ange in e ac ions o he ch oma in.
In summa y, ou esul s p o ide insigh s in o he molec-
ula mechanisms by which he s em cell-speci ic subuni o
Pol III complex, POLR3G, egula es sel - enewal and plu ip-
o ency. Fu he mo e, since Pol III is esponsible o an-
sc ibing many o he co e RNA componen s o he cy osolic
ansla ion machine y, ou indings sugges a p e iously
un epo ed mechanism o coo dina ed egula ion o p o-
ein syn hesis and mi ochond ial biogenesis.
EXPERIMENTAL PROCEDURES
Full expe imen al p ocedu es a e p o ided in he Supplemen al
In o ma ion.
Cell Cul u e and Di e en ia ion Assays
Human ESC lines we e main ained on human o eskin ib oblas
eede s o in eede - ee cul u e condi ions on Ma igel (BD Biosci-
ences) in mTeSR1 medium (STEMCELL Technologies) as p e iously
desc ibed (Na a e al., 2012; Konki e al., 2016). In eede - ee cul-
u e condi ions he cells we e main ained on Ma igel (BD Biosci-
ences) in mTeSR1 medium (STEMCELL). Di e en ia ion o hESCs
was pe o med as desc ibed by Na a e al. (2012). In b ie , o
spon aneous emb yonic body di e en ia ion he cells we e pla ed
wi hou eede s and we e g own in suspension in s anda d hESC
medium wi hou ib oblas g ow h ac o 2. Fo e inoic acid-
induced di e en ia ion, cells we e pla ed in eede - ee condi-
ions and medium was supplemen ed wi h 13.7 mM e inoic acid
(Sigma). The ka yo ypes o he lines we e ou inely moni o ed
wi h G-banding and/o Ka yoLi e BoBs assay (Lund e al., 2012).
RNA In e e ence and T ans ec ions o hESC Lines
hESC cul u es om wo di e en cell lines (H9p38, HS360p62,
HS360p63, HS360p66) we e used in POLR3G siRNA expe imen s
as indica ed. HS360 was de i ed in and ob ained om he Ka olin-
ska Ins i u e . The cells we e i s expanded on Ma igel in mTeSR1
o wo o h ee passages o emo e eede cells. Silencing expe i-
men s we e pe o med wi h wo di e en siRNA oligonucleo ides
(Sigma) and a pool o siRNAs (San a C uz Bio echnology) and
wi h non- a ge ing siRNA con ol, which we e ans ec ed in o
he cells wi h Lipo ec amine RNAi Max o Lipo ec amine 2000
(In i ogen) acco ding o manu ac u e ’s p o ocols. The sequence
in o ma ion o he siRNAs can be ound in Supplemen al Expe i-
men al P ocedu es. Double ans ec ions o siRNAs we e pe -
o med a 24 and 48 h a e pla ing o cells on eede - ee condi-
ions. The samples we e ha es ed on day 3 o day 4 a e he
i s ans ec ion o he expe imen s and analysis. The Cedex XS
(Inno a is) sys em was used o calcula e he amoun o iable
and dead cells in he cul u es based on ypan blue s aining and
cell mo phology. F om he collec ed samples o al RNA, including
smallRNAs, DNA and p o eins we e ex ac ed simul aneously wi h
a Qiagen Allp ep ki .
Silencing o POU5F1, L1TD1, SOX2, and NANOG was ca ied ou
as p e iously desc ibed (Na a e al., 2012). In b ie , Lipo ec -
amine2000 (In i ogen) eagen was used o double ans ec ions
a e 2 o 3 days o pla ing, and cells we e collec ed o expe imen s
1–5 days a e he second ans ec ion.
RT-PCR
Taqman eal- ime qRT-PCR was un as p e iously desc ibed (Lund
e al., 2013). RNA was ea ed wi h DNase I (Qiagen) du ing
column pu i ica ion, and a second ound o DNase ea men
was ca ied ou o 500 ng o o al RNA wi h DNase I Ampli ica ion
G ade (In i ogen). To e i y ha no genomic DNA was p esen ,
we pe o med a nega i e RT-PCR con ol measu emen wi h
housekeeping gene EF1a om o al RNA. cDNA was syn hesized
using a Supe sc ip II ki (Gibco). The le els o he indica ed genes
o in e es we e measu ed wi h he 7900HT Fas Real-Time PCR
Sys em (Applied Biosys ems). The cycles o h eshold alues (C )
we e compa ed wi h hose o housekeeping gene o ob ain
no malized log
2
exp ession le els o he ansc ip s (DC ). The
p ime s and p obes we e designed using a Uni e sal P obe Lib a y
Assay Design Cen e (Roche). Analysis o HDAC7 a ian s was ca -
ied ou wi h RT
2
SYBR G een qPCR Mas e mix (Qiagen). P ime
and p obe sequences a e lis ed in Supplemen al Expe imen al
P ocedu es.
Wes e n Blo ing
P o ein le el analysis was pe o med as p e iously desc ibed
(Na a e al., 2012). In b ie , he cells we e lysed in bu e wi h
50 mM T is-HCl (pH 7.5), 150 mM NaCl, 0.5% T i on X-100, 5%
glyce ol, 1% SDS, 1 mM Na
3
VO
4
, 10 mM NaF, and 1 mM PMSF.
A e sonica ion, p o ein concen a ions we e measu ed wi h DC
P o ein Assay (Bio-Rad), a e which 63SDS bu e (0.5 M T is-
HCl [pH 6.8], 28% glyce ol, 9% SDS, 5% 2-me cap oe hanol,
0.01% b omophenol blue) was added. A e boiling o 5 min,
lysa es we e un in elec opho esis using 10% SDS-PAGE gel and
ans e ed in o a ni ocellulose memb ane. Memb anes we e
1450 S em Cell Repo s jVol. 8 j1442–1454 jMay 9, 2017