Overexpression of BCL2 enhances survival of humanembryonic stem cells during stress and obviates therequirement for serum factorsReza Ardehalia,b,1,2, Matthew A. Inlaya,1, Shah R. Alia, Chad Tanga, Micha Drukkera, and Irving L. Weissmana,2

aInstitute for Stem Cell Biology and Regenerative Medicine, and bDepartment of Internal Medicine, Division of Cardiovascular Medicine, Stanford UniversitySchool of Medicine, Stanford, CA 94305

Contributed by Irving L. Weissman, December 23, 2010 (sent for review September 27, 2010)

The promise of pluripotent stem cells as a research and therapeutictool is partly undermined by the technical challenges of generat-ing and maintaining these cells in culture. Human embryonic stemcells (hESCs) are exquisitely sensitive to culture conditions, andrequire constant signaling by growth factors and cell–cell and cell–matrix interactions to prevent apoptosis, senescence, and differen-tiation. Previous work from our laboratory demonstrated thatoverexpression of the prosurvival gene BCL2 in mouse embryonicstem cells overrode the requirement of serum factors and feedercells to maintain mESCs in culture. To determine whether this pro-survival gene could similarly protect hESCs, we generated hESClines that constitutively or inducibly express BCL2. We find thatBCL2 overexpression significantly decreases dissociation-inducedapoptosis, resulting in enhanced colony formation from sortedsingle cells, and enhanced embryoid body formation. In addition,BCL2-hESCs exhibit normal growth in the absence of serum, butrequire basic fibroblast growth factor to remain undifferentiated.Furthermore, they maintain their pluripotency markers, form tera-tomas in vivo, and differentiate into all three germ layers. Ourdata suggest that the BCL2 signaling pathway plays an importantrole in inhibiting hESC apoptosis, such that its overexpression inhESCs offers both a survival benefit in conditions of stress byresisting apoptosis and obviates the requirement for serum ora feeder layer for maintenance.

embryonic stem cell survival | viability

Embryonic development proceeds through a series of tightlyregulated and precisely timed fate decisions to generate the

complex tissues and organs that comprise the developing or-ganism. The cells that form the early embryo are sensitive toimproper signals or conditions that could derail the develop-mental program. Embryonic stem cells, derived from the innercell mass of blastocysts, mirror this sensitivity and are prone toapoptosis, senescence, or differentiation when the in vitro cul-ture conditions are less than optimal. As a result, human em-bryonic stem cells (hESCs) and their progeny, including tissuestem cells, are technically difficult to maintain in culture, im-peding their use in studying intermediate developmental states.In particular, hESCs undergo significant cell death upon disso-ciation into single cells, which has been a major hurdle for theirdifferentiation, rapid expansion, and genetic manipulation (1, 2).Additionally, hESCs and their early progeny are vulnerable toapoptosis upon sorting through a flow cytometer, hinderingefforts to isolate rare populations. It is thought that a combina-tion of soluble growth factors and cell–cell or cell–matrix inter-actions acts in concert to maintain hESCs in a viable andpluripotent state, and these favorable interactions may be dis-rupted during single-cell dissociation or flow cytometry (3).Human ESC media has been carefully formulated to provide

a myriad of prosurvival signals. The standard media used formaintenance of hESCs on mouse embryonic fibroblasts (MEFs)is supplemented with basic fibroblast growth factor (bFGF) andknockout replacement serum (KOSR), a proprietary formulation

containing a variety of growth factors for hESC growth (4).Efforts have been made to develop defined culture systems toeliminate the need for animal-derived products in the media andthe supporting feeder-cell layer for the maintenance and ex-pansion of hESCs (5–7). Although several defined media haveoptimized the growth and maintenance condition for hESCswithout the requirement of a feeder layer, these media still relyon bFGF supplementation. FGF signaling plays a critical role inthe survival, self-renewal, proliferation, and maintenance of theundifferentiated hESC state in vitro (8–14).Despite the optimization of culture media for growth and

survival, apoptosis upon manipulation of hESCs remains a majorproblem. Hence, several recent studies have identified specificprosurvival factors to enhance hESC survival after dissociationand to facilitate cloning efficiency. A selective Rho-associatedkinase (ROCK) inhibitor (Y-27632) has been shown to markedlydiminish dissociation-induced apoptosis of hESCs (15). Othercompounds have been reported to enhance hESC survival bystabilizing E-cadherin signaling, one of the key signals disruptedduring single cell dissociation (3).Many of the prosurvival signals converge at the BCL2 path-

way. The BCL2 family of proteins contains three major sub-families: a proapoptotic family (e.g., BAX, BAK, BOK, and soforth), a prosurvival family (e.g., BCL2, BCL-XL, MCL1), andthe BH3-only subfamily (e.g., BAD, BID, BIP, PUMA, NOXA)(16). Growth factors continuously stimulate tyrosine kinase ac-tivity to suppress the proapoptotic and BH3-only family mem-bers, and thus prevent initiation of apoptosis (17, 18). Enforcedexpression of the 26-kDa BCL2 antiapoptotic protein has beenreported to result in enhanced survival and proliferation inseveral different systems (16, 19, 20). We demonstrated thatmouse embryonic stem-cell (mESC) lines expressing the humanBCL2 transgene could self-renew continuously under serum-freeand feeder cell-free conditions (21), and improved efforts toidentify and isolate mouse ES-derived HSC (22).Because signaling pathways that interact with the BCL2

pathway play a major role in the survival of hESCs, we evaluatedthe effect of overexpression of BCL2 on growth and survival ofhESCs. We established hESC clones that constitutively or tran-siently express human BCL2. After serial passages for more than30 generations, these BCL2-hESC lines continue to maintaintheir pluripotency, ability to differentiate in vitro, and potentialto form teratomas in vivo. We show that BCL2-hESCs display

Author contributions: R.A., M.A.I., S.R.A., and I.L.W. designed research; R.A., M.A.I., S.R.A.,and C.T. performed research; R.A., M.A.I., and S.R.A. contributed new reagents/analytictools; R.A., M.A.I., S.R.A., M.D., and I.L.W. analyzed data; and R.A. and M.A.I. wrote thepaper.

The authors declare no conflict of interest.1R.A. and M.A.I. contributed equally to this work.2To whom correspondence may be addressed. E-mail: rezaa@stanford.edu or irv@stanford.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1019047108/-/DCSupplemental.

3282–3287 | PNAS | February 22, 2011 | vol. 108 | no. 8 www.pnas.org/cgi/doi/10.1073/pnas.1019047108

increased viability following single-cell dissociation, cell sorting,and in cultures lacking feeders and serum factors. Furthermore,the enhanced survival of the BCL2-expressing cells in the ab-sence of serum factors is partially a result of their resistance toapoptosis. Our results collectively demonstrate that the over-expression of BCL2 substantially promotes hESC survival with-out compromising their self-renewal and developmental potency.

ResultsExpression of BCL2 in hESCs. To determine whether overexpressionof human BCL2 could improve the viability of hESCs in con-ditions of stress, we generated hESC lines that either constitu-tively or inducibly express BCL2 (Figs. S1 and S2 ). ConstitutiveBCL2-expressing hESC lines were generated using lentiviralconstructs expressing BCL2 linked to GFP (C306) or CFP(C342), driven by the constitutively active EF1α promoter (Fig.S1A). Alternatively, Tet-inducible BCL2 hESC lines were gen-erated, in which the expression of BCL2 is induced by exogenousaddition of doxycycline (DOX) (Fig. S2). Expression of GFP/CFP was confirmed by microscopy, and BCL2 expression wasconfirmed by intracellular staining (Fig. S1 B and C). The pri-mary hESC line induced to constitutively express BCL2, andused for all subsequent experiments, was generated via the in-troduction of construct C306; it is hereafter denoted as “BCL2-hESC.” We estimated the copy number of the primary BCL2-hESC line by quantitative real-time PCR of genomic DNA usingprimers that amplify the endogenous locus and lentiviral con-struct equally (Fig. S1D). Based on this assay, we estimated that28 copies of the BCL2 construct were present in the BCL2-hESCline. Using quantitative real-time RT-PCR, we examined mRNAexpression of BCL2 in this line, and found an over 4,600-foldincrease in BCL2 transcripts compared with GFP-expressinghESC line, although this high fold-increase is partially because ofthe fact that hESCs normally express very low, if any, levels ofBCL2 (Fig. S1E). For BCL2-inducible lines, we found that themaximal BCL2 expression was exhibited in the first 24 h after

DOX addition (Fig. S2), and decreased gradually over the courseof 5 d following DOX withdrawal.BCL2-hESC lines have been serially passed for more than 30

generations yet still maintain the characteristic colony mor-phology and normal karyotype (Fig. S3A). Furthermore, BCL2-hESCs express the pluripotency markers Oct3/4, Nanog, andSSEA4 (Fig. S3B). When BCL2-hESCs were injected sub-cutaneously or under the kidney capsule of immunocompro-mised mice, they generated complex teratomas consisting of allthree germ layers (Fig. S3 C–E). Additionally, we observed nodefects in the ability of BCL2-hESCs to differentiate to eitherthe cardiomyocyte (Fig. S4 and Movie S1) or hematopoietic/endothelial lineages (Fig. S5). Thus, to the best of our knowl-edge, the constitutive overexpression of BCL2 in hESC linesaffects neither their pluripotency nor differentiation potential.

BCL2-hESC Cells Exhibit Improved Colony Formation Efficiency AfterSorting and Enhanced Survival During Embryoid Body Formation. Theefficiency of colony formation from a single hESC is notoriouslypoor, particularly after sorting via flow cytometer. The poorsurvival is attributed partially to disruption of favorable cell–celland cell–matrix interactions that prevent apoptosis. When singlewild-type hESCs were FACS-sorted directly into the wells ofa 96-well plate coated with Matrigel, we noted one colony out of∼100 plated cells (1 in 100), which improved to one colony per60 cells (1 in 60) in the presence of ROCK-inhibitor (Fig. 1 Aand B). However, when BCL2-hESCs were sorted onto matrigel-coated wells in the absence or presence of ROCK inhibitor, weobserved colony-formation efficiencies of 1 in 26 cells and 1 in 21cells, respectively, a three- to fourfold improvement over wild-type hESCs. Similarly, when sorted onto MEFs, wild-type hESCsyielded colony-forming efficiencies of 1 in 10, which improved to1 in 5 with ROCK inhibitor (Fig. 1 C and D). The efficiency ofcolony formation of BCL2-hESC on MEFs increased to 1 in 4.5and 1 in 2 cells plated in the absence or presence of ROCK-inhibitor, respectively. Taken together, BCL2 expression improved

0

10

20

30

40

50

0.1 1 10 100 1000 100000

10

20

30

40

50

0.1 1 10 100 1000 10000

no

. co

lon

ies

per

wel

l

no

. co

lon

ies

per

wel

l

no. cells plated per well no. cells plated per well

H9GFPBCL2

H9GFPBCL2

H9GFPBCL2

0

10

20

30

40

50

0.1 1 10 100 1000

H9GFPBCL2

0

10

20

30

40

50

0.1 1 10 100 1000

no

. co

lon

ies

per

wel

l

no

. co

lon

ies

per

wel

l

no. cells plated per well

w/o Rock-i with Rock-i

on

Mat

rig

elo

n M

EFS

A B

C D

E

F

RFP

BCL2-GFP

no. cells plated per well

100% BCL275% BCL2 : 25% RFP50% BCL2 : 50% RFP25% BCL2 : 75% RFP5% BCL2 : 95% RFP100% RFP

Days0 1 2 3 4

100%

75%

50%

25%

0%% B

CL2

cel

ls in

EB

Days0 1 2 3 4

No

. cel

ls p

er E

B

1000

0

5000

4000

3000

2000

GFPConstitutive BCL2Inducible BCL2 + DOXInducible BCL2 - DOX

G H

Fig. 1. Improvement in colony and EB formation with BCL2-hESCs. (A–D) Colony formation from FACS-sorted wild-type (red line), GFP- (green line), and BCL2-(blue line) hESCs when plated onto matrigel (A and B) or MEFs (C and D) in the absence (A and C) or presence (B and D) of ROCK inhibitor. Colony formationefficiency was plotted versus cells plated per well. Cell counts were capped at 50 colonies per well. (E and F) EBs were generated from RFP-hESCs (E) and BCL2-hESCs (F) after single-cell dissociation. Shown are brightfield (Left) and RFP/GFP fluorescence (Right) of hESC clusters in microwells. (G) Chimeric EBs generatedfrom BCL2-hESCs and RFP-hESCs, mixed at multiple ratios. Plotted is the percentage of BCL2-hESCs within the EBs at 0, 1, and 4 d after generation, determinedby FACS. (H) Absolute number of live hESC-derived cells per EB of GFP-hESCs (orange), constitutively expressed BCL2-hESCs (red), and inducible BCL2-hESCs inthe absence (green) or presence (blue) of DOX. Two-thousand hESCs were seeded per well (40 wells per line) at day 0, then pooled at day 1. Half the EBs wereharvested and counted at day 1, and the other half at day 4. Error bars are SD (n = 3) of three independent experiments.

Ardehali et al. PNAS | February 22, 2011 | vol. 108 | no. 8 | 3283

DEV

ELOPM

ENTA

LBIOLO

GY

colony-formation from sorted single cells between two- and five-fold, depending on the precise plating conditions.A large number of cells also undergo cell death in the initial

stages of embryoid body (EB) formation, particularly when EBsare formed from single-cell aggregation. To determine whetherBCL2-hESCs exhibit improved survival during EB formation,BCL2-hESCs were mixed with wild-type hESCs that constitu-tively express red fluorescent protein (RFP). BCL2-hESCs andRFP-hESCs were dissociated into single cells and were mixed atdifferent ratios to form chimeric EBs. EBs formed from 100%BCL2-hESC cells exhibited greater uniformity, were symmetri-cally spherical with distinct borders, and relatively few detachedsingle cells were noted in the wells (Fig. 1 E and F). Conversely,RFP-hESC EBs were less uniform and more vulnerable to dis-ruption upon transfer. Furthermore, when chimeric EBs gener-ated from different mixtures of BCL2- and RFP-hESC wereevaluated 24 h after EB formation, we found a significant in-crease in the ratio of BCL2- to RFP-hESCs within the EBs (Fig.1G). This result was found to be caused by a combination offactors, including continued growth of BCL2-hESC cells andpoor survival of RFP-hESCs, especially in the first 24 h of EBformation (Fig. 1H). In EBs formed from control hESCs alone,the total number of cells decreased by 45% after 24 h andreturned to baseline after 4 d. However, the total number of cellsin EBs formed from constitutive BCL2-hESCs increased by 13%and nearly doubled after 4 d. A similar trend was noted whenEBs were formed from inducible BCL2-hESCs, but only whenBCL2 expression was induced by DOX addition 24 h before EBformation (Fig. 1H). These results indicate that during the first24 h after EB formation from single cells, massive cell deathoccurs that can be largely rescued by expression of BCL2.

BCL2-hESC Cells Exhibit Normal Growth Under Serum-Free Conditions.Conventional hESC media contains factors that provide pro-survival signals to hESCs. We have previously shown that mESCsthat express BCL2 are able to survive in media lacking serum orfeeders, the primary sources of these prosurvival signals (21). Todetermine whether BCL2 overexpression can similarly compen-sate for the absence of prosurvival signals in hESCs, parentaland BCL2-hESC cells were cultured in feeder-free conditions(Matrigel) and with decreasing concentrations of KOSR. Thegrowth kinetics were determined over the course of two pas-sages. Unmanipulated hESCs required 20% KOSR for normalgrowth and failed to survive when KOSR was decreased to aconcentration of 1% or removed entirely (Fig. 2A). However,BCL2-expressing hESCs grew normally in the absence of KOSR,and displayed increased growth kinetics in 20 and 1% KOSR.Thus, overexpression of BCL2 appears to override the require-ment for KOSR in hESC culture.BCL2-hESC colonies appeared morphologically normal,

whether grown in “complete” media—containing the standardconcentrations of KOSR (20%) and bFGF (10 ng/mL)—or inmedia lacking KOSR (0%) but containing normal bFGF (Fig.2B). BCL2-hESCs expanded independently of KOSR and feed-ers but required bFGF to form tightly packed colonies with in-distinct cell boundaries, a hallmark of undifferentiated hESCs.When bFGF was withdrawn from the serum-free media, thehESC colonies survived for a couple of passages before extensivedifferentiation was noted (Fig. 2C). Additionally, BCL2-hESCscultured in decreasing KOSR and bFGF concentrations afterone passage exhibited alkaline phosphatase activity (a marker ofpluripotent cells of embryonic origin), although parental hESCcells lost alkaline phosphatase activity upon KOSR withdrawal(Fig. 2D).To determine whether the increased growth kinetics of BCL2-

hESCs is because of increased cycling, we cultured GFP- andBCL2-hESCs in the presence of a BrdU analog, EdU, andmeasured the rate of EdU incorporation and DNA content after

2 h (Fig. 2E). We found that both hESC lines incorporated EdUat an equivalent rate, indicating that overexpression of BCL2does not cause hESCs to cycle faster. We also examined 11different BCL2-hESC clones, bearing a range of copy numbersfrom 1 to 10, to determine the relationship between BCL2 copynumber and expression level to viability in minimal media, andfound a general trend of enhanced survival in clones with highercopy numbers (Fig. S6). Taken together, these results suggestthat BCL2-hESCs can grow and expand in the absence of KOSRsupplementation in the medium, but require bFGF to preventdifferentiation. Furthermore, the increase in the growth kinet-ics noted in BCL2-hESCs is not due to an augmented cellcycle activity.To determine whether culture conditions devoid of serum

affected hESC pluripotency, we examined pluripotency markerexpression in BCL2-hESCs cultured for three passages in theabsence of knockout serum. We found that BCL2-hESCsmaintained high expression of Oct3/4, Nanog, and SSEA4 (Fig.3A), similar to BCL2-hESCs cultured in complete media (Fig.S3B). Conversely, control hESCs lost pluripotency marker ex-pression after three passages in minimal media. To determinewhether BCL2-hESCs remain pluripotent after long-term cul-ture without serum, we cultured control and BCL2-hESCs incomplete media, and in media lacking serum for 12 passages(67 d). Control hESCs did not survive beyond a few passages inthe absence of serum, whereas BCL2-hESCs grew normally forseveral passages, although at later passages their growth kineticsbegan to slow and their colonies remained small (Fig. 3B). De-spite their reduced growth rate, BCL2-hESCs continued to ex-press high levels of Oct3/4 and Nanog proteins (Fig. 3C). Inaddition, they rapidly returned to normal colony size and growthkinetics when the same colonies were cultured back onto com-plete media. These data indicate that BCL2-hESCs retain theirpluripotency expression profiles even when deprived of serumfactors for over 2 mo.

0 1 2 3 4 5 6 7

20

1

0

10 5 1 010 5 1 0

20

1

0

GFP-hESC BCL2-hESC

B

C

0% KOSR10ng/mL bFGFComplete media

0% KOSR and no bFGFDay 3 Day 7

G1 S newG2/M

oldG2/M

GFP

BCL2

Perc

ent

of c

ells

0%

10%

20%

30%

40%

0

1

2

3

Cel

l co

un

ts (x

106 )

BCL2 - 20% KOSR

BCL2 - 1% KOSR

BCL2 - 0% KOSR

GFP - 20% KOSR

GFP - 1% KOSR

GFP - 0% KOSR

Days

A

D

E

0 5000 10K 15K 20K 25K

0

10 3

10 4

10 5

35

21

9.7

32

DNA content0 5000 10K 15K 20K 25K

0

10 3

10 4

10 5

32

22

12

30S new G2/M

G1 old G2/M

GFP-hESC BCL2-hESC

EdU

DNA contentKOSR (%)

bFGF(ng/mL)

Fig. 2. Enhanced survival of BCL2-hESCs in serum-free media. (A) Growthkinetics of hESCs at different KOSR concentrations. Wild-type and BCL2-hESCswere plated on Matrigel and cultured in media with KOSR concentrations of20, 1, and 0%, and cells were counted daily. Error bars are SD (n = 3). (B)Representative BCL2-hESC colonies grown inmedia with 20%KOSR (Left) andin 0% KOSR (Right). (C) Representative colony of BCL2-hESCs grown in medialacking both KOSR and bFGF. The same colony is shown at day 3 (Left) and day7 (Right) after plating. (D) Alkaline phosphatase activity of BCL2-hESCs (Left)and wild-type hESCs (Right) cultured for two passages under different con-centrations of KOSR and bFGF. (E) Cell-cycle analysis of GFP- (Upper Left plotand Lower graph, white bars) or BCL2-hESCs (Upper Right plot and Lowergraph, black bars) cultured in complete hESC media. EdU was added 2 h be-fore analysis. Shown are the percentages of cells at each cell cycle stage.

3284 | www.pnas.org/cgi/doi/10.1073/pnas.1019047108 Ardehali et al.

BCL2 Partially Blocks Apoptosis in hESCs Cultured in Serum-FreeMedia. We hypothesized that the enhanced survival of BCL2-hESCs in serum-free media was caused by increased resistance toapoptosis. To determine whether BCL2 overexpression protectshESCs from apoptosis, we passaged BCL2- and control hESCsonto Matrigel-coated dishes in complete media (20% KOSR,10 ng/mL bFGF), then after 2 to 3 d switched them into a min-imal media with 0% KSR and 1 ng/mL of bFGF, and measuredthe frequency of apoptotic cells by Annexin V staining after 24and 48 h in minimal media (Fig. 4). We found that controlhESCs cultured in minimal media for 24 or 48 h rapidly un-derwent apoptosis and anoikis compared with cells cultured incomplete media for the duration of the experiment (Fig. 4 B–D).However, BCL2-hESCs exhibited minimal apoptosis despite thelack of survival factors in minimal media, such that both thefrequency and absolute number of live nonapoptotic cells re-mained near that of BCL2-hESCs cultured in complete media(Fig. 4 B–E).To determine if the apoptosis that results from the removal of

KOSR and reduction of bFGF is caspase-dependent, we pre-treated wild-type hESCs and BCL2-hESCs with the caspase-inhibitor ZVAD-FMK at 25 μM 1 h before changing to minimalmedia, and measured apoptosis after 48 h (Fig. 4F). We foundthat wild-type hESCs cultured in minimal media containingZVAD-FMK were more resistant to apoptosis than untreatedcells, indicating that the growth factor starvation-induced apo-ptosis is caspase-dependent. However, treatment with ZVAD-FMK only reduced apoptosis in wild-type hESCs, as BCL2-hESCs appeared unaffected by ZVAD-FMK. BCL2 hESCsshowed higher viability than the ZVAD-FMK treated H9 cells.To eliminate the possibility that the apoptosis resistance is an

integration artifact of this particular BCL2-hESC line, we re-peated these experiments using an inducible BCL2 hESC line(Fig. S7). BCL2 expression was induced by addition of DOX 24 hbefore changing to minimal media, and thus we could comparethe effects of BCL2 expression within the same cell line (Fig.S7A). Like their constitutively-expressed counterparts, BCL2-induced cells were more resistant to apoptosis in minimal mediacompared with uninduced cells (Fig. S7 B–E). Taken together,

these results indicate that the increased growth kinetics of BCL2-expressing hESCs is likely in large part the result of an increasedresistance to apoptosis when critical survival factors contained inKOSR are reduced or removed.

DiscussionWe have previously shown that transgenic mESCs expressinghuman BCL2 continue to self-renew in serum- and feeder-freeconditions when supplemented with leukemia inhibitory factor(21). These studies suggest that leukemia inhibitory factor andBCL2 overexpression are sufficient to maintain and expandmESCs in culture. The data summarized here is a logical ex-tension of the previous study, where we show that BCL2 over-expression, in the presence of bFGF, is sufficient to maintainhESCs in an undifferentiated state for prolonged periods andoffers a survival benefit in conditions of stress by resisting apo-ptosis. Although differentiation to only the mesoderm-derivedcardiomyocyte, endothelial, and hematopoietic lineages werespecifically tested, BCL2-hESCs can form all three germ layersupon teratoma formation.It is known that growth factors and cytokines present in KOSR

can stimulate multiple receptor tyrosine kinase signaling, leadingto activation of a variety of tyrosine kinases, such as AKT and

H9 BCL2

Complete day 67

Complete day 67

Minimal day 67

Minimal day 61

Completeday 67

BCL2-hESC GFP-hESC

Oct3/4

Nanog

SSEA4

Sox17

A B

C

Nanog0 10 2 10 3 10 4 10 5

Oct3/40 10 2 10 3 10 4 10 5

H9 iso

H9 C67

BCL2 iso

BCL2 C67

BCL2 M67

BCL2 M61-C67

Fig. 3. Human ESCs overexpressing BCL2 maintain their pluripotencymarkers when cultured for short-term (3 passages) or long-term (12 pas-sages) in serum-free media. (A) Immunofluorescence staining of BCL2-hESCsafter three passages in minimal media containing no KOSR with anti-Oct3/4,anti-Nanog, anti-SSEA4, and anti-Sox 17. Compared with control GFP-hESCs,BCL2 expressing hESCs had normal expression of Oct3/4, Nanog, and SSEA4.Absence of Sox 17 indicates lack of differentiation in the BCL2-hESC colonies.(B) Representative colonies of parental H9 and BCL2 expressing hESCs cul-tured on MEFs for 67 d (12 passages) in complete media, in minimal medialacking serum, or in minimal media for 61 d, then passaged into completemedia for additional 6 d. Unmanipulated H9 cells in minimal media did notsurvive beyond three passages, and therefore were not analyzed. (C) In-tracellular FACS profile of Oct3/4 (Left) and Nanog (Right) expression of hESCcultures at day 67. H9 and BCL2 isotype controls are shown in red and greenlines, respectively. For ease of comparison, the crest of Oct3/4 and Nanogexpression curves for the H9 culture in complete media are indicated by thedotted line.

H9

BC

L2

Complete Minimal 24hr Minimal 48hr

C0

20000

40000

60000

80000

100000

120000

140000

Complete Minimal24hr

Minimal48hr

H9BCL2

Ave

rag

e n

um

ber

of l

ive

cells p < 0.02

p < 0.03

Day 0 Day 2 Day 3 Day 4Change to

minimal media(48hr)

Change tominimal media

(24hr)

Harvest/Analyze

0 10 3 10 4 10 5

0

10 3

10 4

10 5 14

5.3

80

0 10 3 10 4 10 5

0

10 3

10 4

10 5 6.53.5

89

0 10 3 10 4 10 5

0

10 3

10 4

10 5 5313

29

0 10 3 10 4 10 5

0

10 3

10 4

10 5 8.73.0

86

0 10 3 10 4 10 5

0

10 3

10 4

10 5 5819

21

0 10 3 10 4 10 5

0

10 3

10 4

10 5 157.6

75

Complete Minimal 24hr Minimal 48hr

H9

BC

L2

Live

Dead

Dying

7-AAD

An

nex

in V

0%

25%

50%

75%

100%

H9 BCL2 H9 BCL2 H9 BCL2

LiveDyingDead

Complete Minimal 24hr Minimal 48hr

Ave

rag

e p

erce

nt

p < 0.002 p < 0.004A

B

D

E

F Complete Minimal48 hr

Minimal 48 hr +25µM ZVAD-FMK

0 10 3 10 4 10 5

0

10 3

10 4

10 5 7.73.9

870 10 3 10 4 10 5

0

10 3

10 4

10 5 4114

390 10 3 10 4 10 5

0

10 3

10 4

10 5 105.9

80

0 10 3 10 4 10 5

0

10 3

10 4

10 5 3.62.2

930 10 3 10 4 10 5

0

10 3

10 4

10 5 8.12.4

870 10 3 10 4 10 5

0

10 3

10 4

10 5 6.42.7

87

H9

BC

L2

7-AAD

An

nex

in V

Plate hESCin completemedia

Fig. 4. Apoptosis analysis of BCL2-hESCs in minimal media. (A) Experimentaldesign. Control and BCL2 hESCs were passaged from feeder plates ontoMatrigel-coated wells in complete media. At 48 (day 2) or 24 (day 3) h beforeanalysis, wells were changed into minimal (0% KOSR, 1 ng/mL bFGF) mediafor the 48 and 24 h time points, respectively. At day 4, all floating and at-tached cells were harvested and analyzed. (B) Photographs of hESC coloniestaken at day 4 from control (H9, Upper) and BCL2 (Lower) wells in completemedia (Left), or in minimal media 24 h (Center) or 48 h (Right) before anal-ysis. (C) Apoptosis analysis of control and BCL2 hESCs, using Annexin V (y axis)and 7-AAD (x axis). Percentages of live cells (Annexin V−, 7-AAD−), dying cells(Annexin V+, 7AAD−), and dead cells (Annexin V+, 7AAD+) are shown. (D)Average percent of live (white), dying (gray), and dead (cells) from controland BCL2 hESCs in complete and minimal media. Error bars are SD (n = 3) andP values are calculated by the t test. (E) Average absolute number of live cellsfrom control (white) and BCL2 (black) hESCs. Error bars are SD (n = 3) and Pvalues are calculated by the t test. (F) Analysis of apoptosis of H9- and BCL2-hESCs treated with the caspase inhibitor ZVAD-FMK in minimal media. H9-and BCL2-hESCs were plated as described in A, but some wells were pre-treated for 1 h with 25 μM ZVAD-FMK before changing to minimal media,then retreated with an additional 25 μM ZVAD-FMK after each media change.Human ESCs were harvested 48 h after changing to minimal media, and an-alyzed for Annexin V binding (y axis) and 7-AAD incorporation (x axis). Valuesshown are the percent of live, dying, and dead cells.

Ardehali et al. PNAS | February 22, 2011 | vol. 108 | no. 8 | 3285

DEV

ELOPM

ENTA

LBIOLO

GY

ERK, which can directly inhibit proapoptotic BCL2 familymembers (23). In the absence of serum factors, those proa-poptotic proteins may accumulate and override the activity ofBCL2 and the other prosurvival family members. The over-expression of BCL2 proteins may increase the threshold levelof proapoptotic signaling required for initiating apoptosis, andthereby prevent apoptosis when other survival signals are re-moved. This theory may explain our finding that the protectiveeffect of BCL2 correlates with its expression level. That BCL2overexpression alone may obviate the requirement for KOSRsuggests that the primary function of the myriad growth factorscontained in KOSR is to provide survival signals, and that theBCL2 pathway lies at the hub of those signals.Although we found KOSR dispensable in maintaining BCL2-

hESC lines, bFGF was not. When bFGF was removed from theculture media, BCL2-hESCs continued to proliferate but grad-ually differentiated. Accumulating evidence suggests that FGFsignaling plays a central role in sustaining hESCs in an undiffer-entiated state by providing both survival and antidifferentiationsignals (8, 10, 11, 24). Our results suggest that although BCL2can replace the prosurvival function of bFGF, it cannot supplantthe antidifferentiation signal that bFGF provides. This findingmay partially explain why BCL2-hESCs can grow in low concen-trations of bFGF (1 ng/mL) but not in its complete absence.Analogous to the requirement for leukemia inhibitory factor in theserum-free and feeder-free culture of BCL2-expressing mESCs,bFGF appears to be required for maintaining the pluripotency ofhESCs. Additionally, signaling from TGF-β family members hasbeen demonstrated to play an integral role in hESC self-renewal,maintenance of pluripotency, and regulation of differentiation(25). Although activities of FGF and TGF-β superfamily may par-tially overlap, each is necessary to sustain hESCs in an undiffer-entiated state. Our data suggest that short-term culture conditionsin which KOSR is lacking, overexpression of BCL2 in conjunctionwith the presence of bFGF in the medium may partially overrideTGF-β signaling for maintenance of pluripotency.Human ESCs are especially vulnerable to apoptosis when

dissociated into single cells. The molecular mechanisms thatunderlie this sensitivity are not completely understood. Recently,Xu et al. reported the essential role of E-cadherin signaling forhESC survival: E-cadherin-mediated cell–cell interactions con-stitute an essential regulatory mechanism that controls hESCsurvival and self-renewal (3). Dissociation of hESCs into singlecells disrupts this signal, perhaps leading to cell death. Here, weshow that BCL2 expression offers a significant survival benefitfor hESCs upon single-cell dissociation and cell sorting. Addi-tionally, we have shown that EB formation from single-cell sus-pensions and their subsequent growth is considerably moreefficient when BCL2 is overexpressed. As the survival benefitconferred by BCL2 is most obvious during the first day of EBformation, when individual cells must aggregate and join, it ispossible that BCL2’s benefit in EB formation is caused by itsprotection of hESCs when they exist as single cells.Human ESC lines that overexpress BCL2, either constitutively

or transiently, have several practical uses: first, BCL2 improvesthe viability of hESCs in protocols that require single cell dis-sociation, such as (i) clonal isolation following genetic modifi-cation, (ii) EB formation, (iii) single-cell differentiation proto-cols, and (iv) culture of populations of hESCs after sorting bymagnetic-activated cell sorting or flow cytometry. Second, BCL2overexpression allows for the efficient differentiation of hESCsin protocols that require the removal of survival factors. In sup-port of this notion, BCL2 overexpression was shown to enhancemESC-derived hematopoietic differentiation (22). Finally, BCL2-hESC lines can be cultured in serum- and feeder-free conditions toassess the effect of individual growth factors, cytokines, and smallmolecules on the proliferation and differentiation of hESCs, with-out concern for survival.

Although increased expression of BCL2, whether by transgeneor translocation, appears to predispose some cell types to neo-plasia, overexpression alone is not oncogenic (26). In fact, wepreviously showed that chimeric mice generated from mouse EScells expressing human BCL2 appear normal and remain tumor-free despite ubiquitous expression of BCL2 (21). Nonetheless,the constitutive expression of an antiapoptotic gene in an hESCline may gradually alter its epigenetic state over time. Althoughthe constitutive BCL2-expressing hESC lines used in this studyhave remained healthy and pluripotent throughout the manypassages they have undergone, there remains the possibility thatgradual genomic alterations that usually signal apoptosis mayaccumulate over time and compromise the potential of theselines. To circumvent this limitation, we generated an induciblesystem that allows for temporal expression of BCL2 only duringthe brief periods where viability is most challenged. Specifically,the addition of DOX the day before challenge ensures maximalBCL2 protection when needed most. BCL2 expression returnsto baseline ∼5 d after removal of DOX (Fig. S2D), therebypreventing any long-term effects that may occur. Although BCL2overexpression in pluripotent stem cells has a variety of appli-cations, the genetic modification required for expression andthe uncertainty of genomic instability from permanent inhibitionof apoptosis may preclude their use in clinical applications. Tothat end, we have found that the caspase inhibitor ZVAD-FMKcan partially protect hESCs from apoptosis when survival factorsare removed. Although this protection does not appear to beas robust as BCL2 overexpression, this inhibitor can be safelyadded to cultures without genetic modification or use of ani-mal products.In summary, overexpression of BCL2 enables hESCs to ex-

pand and proliferate under serum- and feeder-free conditions,and this effect was mediated primarily by increased resistanceto apoptosis. Additionally, we have demonstrated a significantsurvival benefit when BCL2-hESCs are dissociated into singlecells or formed into EBs. Furthermore, BCL2-hESCs maintaintheir pluripotency, are able to differentiate to all three germlayers, and display a normal karyotype. We find that BCL2overexpression can be used to enhance hESC viability in anyassay performed under harsh conditions. Our data suggest thatnumerous survival signals are required for hESC culture, butthese signals intersect at the BCL2 pathway and thus over-expression of BCL2 can obviate the need for many of thesesignals, such as those provided by KOSR. Further studies arewarranted to delineate the molecular mechanism by which BCL2exerts its effect.

Materials and MethodsFlow Cytometry. All flow cytometry was performed using a FACSAria 2(BD Bioscience).Intracellular staining. Human ESCs were dissociated into single cells, then fixedin 2% paraformaldehyde, and permeabilized with 3% saponin. Human ESCswere stained with PE-conjugated mouse antihuman BCL2 antibody (Cat. no.MHBCL04, clone 100; Invitrogen) at 1:100 dilution. For intracellular Oct3/4 andNanog stains, cells were first treated with Fixable Viability Dye eFluor450according to themanufacturer’s protocol (eBioscience; Cat. no. 65–0863), thenstained with αOct3/4-PE, αNanog-PE, or isotype control PE at 1:10 dilutions.Surface stains. Single-cell suspensions of hESCs or derived products werestained in staining media (PBS with 2% FCS), on ice, for 15 min.Cell cycle analysis. The BrdU analog EdU was added to healthy hESC cultures at10 μM for 2 h. Human ESCs were then dissociated into single cells and the cellcycle status analyzed using the Click-iT EdU flow cytometry assay kit (Cat. no.A10202; Invitrogen) according to the manufacturer’s protocol.Antibodies. Clone numbers, catalog numbers, and conjugates of all antibodiesused in this study are listed in Table S1.

Immunostaining and Immunohistochemistry. Cells were fixed in 4% para-formaldehyde (Electron Microscopy Sciences) for 15 min, permeabilized with0.1% Triton X-100 (Fisher Scientific) in PBS for 5 min, and then blocked with1% goat serum for 15 min. Cells were incubated overnight with primary

3286 | www.pnas.org/cgi/doi/10.1073/pnas.1019047108 Ardehali et al.

antibodies against Oct3/4, Nanog, SSEA4, Sox 17, troponin I, and β-myosinheavy-chain isoform, washed, and then incubated with goat anti-mouseAlexa 594, goat anti-mouse Alexa 488, or donkey anti-goat Alexa 594 sec-ondary antibody for 1 h. Teratomas formed under the kidney capsule werefixed as described in SI Materials and Methods, and 6-μm sections wereprepared for immunohistochemical analysis.

Cell Proliferation and Viability Assays. To assess the effect of KOSR and bFGFon the growth of BCL2-expressing cells, the concentration of bFGF and KOSRwere varied, respectively, to 5 ng/mL, 1 ng/mL, 0 ng/mL, and 1%, 0%. Wild-type and BCL2 hESCs were passaged onto Matrigel-coated wells in the dif-ferent media formulations, and maintained in those media for two passagesbefore their growth kinetics were evaluated. Cumulative cell counts atdesignated days were determined by dissociating the colonies into single cellsusing Trypsin 0.05% and manually counting. Cells were passaged at day 4.

Apoptosis Assays. Human ESCs were passaged onto Matrigel-coated dishes incomplete hESC media. At day 2 to 3 after plating, wells were washed in PBS,and then either complete (20% KSR, 10 ng/mL bFGF) or minimal (0% KSR,1 ng/mL bFGF) hESC media was added. After 24 to 48 h, supernatants wereaspirated, then hESCs were harvested with accutase and pooled with thesupernatants. Human ESCs were then stained with Annexin-APC (Cat. no.550474; BD Biosciences) and 7-AAD (Cat. no. 00–6993-50; eBioscience)

according to the manufacturer’s protocol (BD Bioscience). For all cultures,media was changed daily, whether to complete or minimal media. For theinducible BCL2 hESCs, DOX (1 μg/mL) was added to the media of selectedwells the day before changing to minimal media, and was re-added at eachmedia change until analysis. To count the absolute number of cells in eachwell, a set number of accudrop beads (Cat. no. 345249; BD Biosciences) wasadded to samples before FACS analysis. After FACS analysis, the percent ofsample analyzed could be estimated based on the amount of accudropbeads counted in relation to the amount added. The caspase inhibitor ZVAD-FMK was used at 25 μM according to the manufacturer’s instructions (R&D;Cat. no. FMK001).

ACKNOWLEDGMENTS. The authors thank Stephen Willingham for helpfuldiscussion, Libuse Jerabek for excellent laboratory management, andmembers of the I.L.W. laboratory for helpful advice and assistance. Thiswork was supported in part by Grant RCI 00354 from the California Institutefor Regenerative Medicine; Fellowship TG2-01159 from the CaliforniaInstitute for Regenerative Medicine Training Program (to R.A.); fellowshipsfrom the National Cancer Institute (Public Health Service Grant CA09151),the California Institute for Regenerative Medicine (T1-00001), and theNational Institutes of Health (5 T32 AI07290, Molecular and CellularImmunobiology) (to M.A.I.); fellowships from the Stanford Medical ScholarsResearch Program and the Howard Hughes Medical Institute (to S.R.A. andC.T.); and the Soros Fellowships for New Americans (to S.R.A.).

1. Hasegawa K, Fujioka T, Nakamura Y, Nakatsuji N, Suemori H (2006) A method for theselection of human embryonic stem cell sublines with high replating efficiency aftersingle-cell dissociation. Stem Cells 24:2649–2660.

2. Pyle AD, Lock LF, Donovan PJ (2006) Neurotrophins mediate human embryonic stemcell survival. Nat Biotechnol 24:344–350.

3. Xu Y, et al. (2010) Revealing a core signaling regulatory mechanism for pluripotentstem cell survival and self-renewal by small molecules. Proc Natl Acad Sci USA 107:8129–8134.

4. Skottman H, Narkilahti S, Hovatta O (2007) Challenges and approaches to the cultureof pluripotent human embryonic stem cells. Regen Med 2:265–273.

5. Chin AC, Padmanabhan J, Oh SK, Choo AB (2010) Defined and serum-free mediasupport undifferentiated human embryonic stem cell growth. Stem Cells Dev 19:753–761.

6. Ludwig TE, et al. (2006) Feeder-independent culture of human embryonic stem cells.Nat Methods 3:637–646.

7. Ludwig TE, et al. (2006) Derivation of human embryonic stem cells in definedconditions. Nat Biotechnol 24:185–187.

8. Dvorak P, et al. (2005) Expression and potential role of fibroblast growth factor 2 andits receptors in human embryonic stem cells. Stem Cells 23:1200–1211.

9. Eiselleova L, et al. (2009) A complex role for FGF-2 in self-renewal, survival, andadhesion of human embryonic stem cells. Stem Cells 27:1847–1857.

10. James D, Levine AJ, Besser D, Hemmati-Brivanlou A (2005) TGFbeta/activin/nodalsignaling is necessary for the maintenance of pluripotency in human embryonic stemcells. Development 132:1273–1282.

11. Levenstein ME, et al. (2006) Basic fibroblast growth factor support of humanembryonic stem cell self-renewal. Stem Cells 24:568–574.

12. Li J, et al. (2007) MEK/ERK signaling contributes to the maintenance of humanembryonic stem cell self-renewal. Differentiation 75:299–307.

13. Wang X, Lin G, Martins-Taylor K, Zeng H, Xu RH (2009) Inhibition of caspase-mediatedanoikis is critical for basic fibroblast growth factor-sustained culture of humanpluripotent stem cells. J Biol Chem 284:34054–34064.

14. Xu RH, et al. (2008) NANOG is a direct target of TGFbeta/activin-mediated SMADsignaling in human ESCs. Cell Stem Cell 3:196–206.

15. Watanabe K, et al. (2007) A ROCK inhibitor permits survival of dissociated humanembryonic stem cells. Nat Biotechnol 25:681–686.

16. Gross A, McDonnell JM, Korsmeyer SJ (1999) BCL-2 family members and themitochondria in apoptosis. Genes Dev 13:1899–1911.

17. Cory S, Adams JM (2002) The Bcl2 family: Regulators of the cellular life-or-deathswitch. Nat Rev Cancer 2:647–656.

18. Opferman JT, Korsmeyer SJ (2003) Apoptosis in the development and maintenanceof the immune system. Nat Immunol 4:410–415.

19. Kutschka I, et al. (2006) Adenoviral human BCL-2 transgene expression attenuatesearly donor cell death after cardiomyoblast transplantation into ischemic rat hearts.Circulation 114(1, Suppl)I174–I180.

20. Sasaki T, et al. (2006) Bcl2 enhances survival of newborn neurons in the normal andischemic hippocampus. J Neurosci Res 84:1187–1196.

21. Yamane T, Dylla SJ, Muijtjens M, Weissman IL (2005) Enforced Bcl-2 expressionoverrides serum and feeder cell requirements for mouse embryonic stem cell self-renewal. Proc Natl Acad Sci USA 102:3312–3317.

22. Yamane T, Hosen N, Yamazaki H, Weissman IL (2009) Expression of AA4.1 markslymphohematopoietic progenitors in early mouse development. Proc Natl Acad SciUSA 106:8953–8958.

23. Wang L, et al. (2007) Self-renewal of human embryonic stem cells requires insulin-likegrowth factor-1 receptor and ERBB2 receptor signaling. Blood 110:4111–4119.

24. Vallier L, et al. (2009) Activin/Nodal signalling maintains pluripotency by controllingNanog expression. Development 136:1339–1349.

25. Vallier L, Alexander M, Pedersen RA (2005) Activin/Nodal and FGF pathwayscooperate to maintain pluripotency of human embryonic stem cells. J Cell Sci 118:4495–4509.

26. Jamieson CHM, Passegue E, Weissman IL (2004) Stem Cells in the Nervous System:Functional and Clinical Implications, ed Gage R (Springer, Berlin), pp 157–182.

Ardehali et al. PNAS | February 22, 2011 | vol. 108 | no. 8 | 3287

DEV

ELOPM

ENTA

LBIOLO

GY