eIF4A controls germline stem cell self-renewal bydirectly inhibiting BAM function in theDrosophila ovaryRun Shena,1, Changjiang Wenga,1, Junjing Yua,b, and Ting Xiea,c,2

aStowers Institute for Medical Research, 1000 East 50th Street, Kansas City, MO 64110; bInstitute of Biophysics, Chinese Academy of Sciences, 15 Da TunRoad, Beijing 100101, China; and cUniversity of Kansas School of Medicine, 3901 Rainbow Boulevard, Kansas City, KS 66160

Edited by Allan C. Spradling, Carnegie Institution of Washington, Baltimore, MD, and approved May 28, 2009 (received for review March 31, 2009)

Stem cell self-renewal is controlled by concerted actions of extrinsicniche signals and intrinsic factors in a variety of systems. Drosophilaovarian germline stem cells (GSCs) have been one of the mostproductive systems for identifying the factors controlling self-re-newal. The differentiation factor BAM is necessary and sufficient forGSC differentiation, but it still remains expressed in GSCs at lowlevels. However, it is unclear how its function is repressed in GSCs tomaintain self-renewal. Here, we report the identification of thetranslation initiation factor eIF4A for its essential role in self-renewalby directly inactivating BAM function. eIF4A can physically interactwith BAM in Drosophila S2 cells and yeast cells. eIF4A exhibitsdosage-specific interactions with bam in the regulation of GSC dif-ferentiation. It is required intrinsically for controlling GSC self-re-newal and proliferation but not survival. In addition, it is required formaintaining E-cadherin expression but not BMP signaling activity.Furthermore, BAM and BGCN together repress translation of E-cadherin through its 3� UTR in S2 cells. Therefore, we propose thatBAM functions as a translation repressor by interfering with trans-lation initiation and eIF4A maintains self-renewal by inhibiting BAMfunction and promoting E-cadherin expression.

BGCN � E-cadherin � niche � translation

S tem cells in adult animal tissues have the ability to self-renewand differentiate into functional cells that replenish lost cells.

Their self-renewal and differentiation are controlled by concertedactions of extrinsic factors and intrinsic factors (1, 2). Although aplethora of intrinsic factors has been identified for their roles instem cell regulation, it remains largely unclear how differentiationfactors are functionally repressed in stem cells. In this study, weshow that a translation initiation factor eIF4A maintains germlinestem cell (GSC) self-renewal in the Drosophila ovary by antago-nizing the differentiation factor BAM.

In the Drosophila ovary, 2 or 3 GSCs are located at the tip of thegermarium, where they are directly anchored to cap cells throughE-cadherin-mediated cell adhesion (3). In addition, GSCs are alsolaterally wrapped around by escort stem cells (4). After GSCdivision, the daughter attaching to cap cells/escort stem cells renewsas a stem cell, while the other daughter moving away from themdifferentiates (5). In addition, the differentiating GSC daughter,known as the cystoblast (CB), is enveloped by escort cells, which areproduced by escort stem cells (4). Genetic and cellular studies haveshown that cap cells, possibly along with escort stem cells, form afunctional niche for GSCs. Consistently, both Yb/Piwi and BMP,which are expressed in cap cells, maintain GSC self-renewal by repress-ing expression of differentiation-promoting genes such as bam (6–12).

bam and bgcn were identified for their specific roles in the regulationof GSC lineage differentiation. Mutations in either bam or bgcncompletely blocks GSC differentiation, causing the accumulation ofGSC-like cells (13–15). bam is transcriptionally repressed by BMPsignaling, leaving low levels of BAM expression in GSCs (16). We haverecently shown that low levels of BAM expression work along withBGCN to control GSC competition (16). Its expression is dramaticallyup-regulated in differentiated germ cells ranging from cystoblasts to

8-cell cysts (13). On the other hand, bgcn is expressed in GSCs as wellas in differentiated germ cells (14). Genetic studies have shown thatbam and bgcn require each other’s function to control germ celldifferentiation, suggesting that their protein products either form aprotein complex or function as a linear genetic pathway (14). AlthoughbgcnencodesaputativeDExHRNAbindingprotein(14),bamencodesa novel protein (15). However, their biochemical functions have re-mained a mystery.

Geneticandcellular studieshave indicated that translationregulationplays essential roles in maintaining GSC self-renewal in differentorganisms, including Drosophila (3, 17, 18). Translation initiation is therate-limiting step and thus a primary target for translational control(19). It is a complex process in which ribosomes are assembled byeukaryotic initiation factors (eIFs) to mRNA. eIF4F is a proteincomplex composed of eIF4A, eIF4E, and eIF4G, which respectivelyperform functions of an RNA helicase removing the secondary struc-ture at the 5� UTR, recognition of the mRNA 5� cap structure, andbridging the mRNA with the ribosome. In this study, we have providedexperimental evidence that eIF4A is a target of translation regulationin Drosophila ovarian GSCs. BAM and BGCN form a protein complexto repress E-cadherin expression at the translational level. AlthougheIF4A genetically antagonizes bam function in a dosage-dependentmanner, it is required for GSC self-renewal by regulating E-cadherinexpression. We propose that eIF4A regulates self-renewal by antago-nizing BAM-mediated translation repression and that BAM controlsGSC differentiation by interfering with translation initiation.

ResultseIF4A Is Identified as a BAM Interacting Protein. BAM is necessary andsufficient to induce GSC differentiation in the Drosophila ovary (13, 15,20). To understand how BAM controls GSC differentiation mecha-nistically, we performed a yeast 2-hybrid screen to identify BAMinteracting proteins using his3 as the selection marker by screening aGSC-enriched cDNA 2-hybrid library (Fig. 1A). In the 2-hybrid screen,we found that a C-terminal fragment of eIF4A (305–403 aa) fused withthe GAL4 activation domain (GAD) and activated his3 expression withGDB-BAM to support yeast growth on the His� medium, suggestingthat eIF4A is a putative BAM interacting protein. As a component ofthe translation initiation complex, eIF4A directly interacts with eIF4Eand eIF4G to control translation initiation (19). To confirm theinteraction between eIF4A and BAM, we made 4 different fusionproteins between eIF4A and GAD, named full-length GAD (FL-GAD), N terminus GAD (NT-GAD), middle portion GAD (MP-

Author contributions: R.S., C.W., and T.X. designed research; R.S., C.W., and J.Y. performedresearch; R.S. and C.W. contributed new reagents/analytic tools; R.S., C.W., J.Y., and T.X.analyzed data; and R.S. and T.X. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

1R.S. and C.W. contributed equally to this work.

2To whom correspondence should be addressed. E-mail: tgx@stowers-institute.org.

This article contains supporting information online at www.pnas.org/cgi/content/full/0903325106/DCSupplemental.

www.pnas.org�cgi�doi�10.1073�pnas.0903325106 PNAS � July 14, 2009 � vol. 106 � no. 28 � 11623–11628

DEV

ELO

PMEN

TAL

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

June

7, 2

020

GAD), and C terminus (also the same as the initially identifiedfragment) GAD (CT-GAD) (Fig. 1B). Interestingly, all of the 3 regions(NT, MP, and CT), but not GAD alone, interact with GDB-BAM,indicating that BAM contacts multiple regions of eIF4A (Fig. 1C). Thisfinding has been shown to be true for the interactions between thehuman tumor suppressor PDCD4 and eIF4A (21). However, thefull-length eIF4A and its N terminus had stronger interactions withBAM than the C-terminal part identified in our initial 2-hybrid screen(Fig. 1 B and C).

To further substantiate the yeast 2-hybrid results, we also per-formed coimmunoprecipitation (co-IP) experiments to confirm thephysical interaction between eIF4A and BAM in Drosophila S2cells. Flag-tagged BAM and Myc-tagged eIF4A could reciprocallybring down each other in the co-IP experiments (Fig. 1D and Fig.S1). Furthermore, the interaction between eIF4A and BAM re-mained unchanged after the removal of RNAs (Fig. 1D). Therefore,we conclude that BAM can physically interact with eIF4A in anRNA-independent manner, raising the interesting possibility thatBAM might control translation initiation.

eIF4A Genetically Interacts with bam. To establish a functionalconnection between eIF4A and bam, we used a sensitive bamgenetic background (bamZ3-2884/bam�86, referred to hereafter asbamZ/bam�86) to test genetic interactions between bam andeIF4A in the regulation of GSC differentiation. The ovaries ofdifferent genotypes were immunostained with the monoclonalanti-Hu-li tai shao (Hts) antibody, which labels spectrosomes inGSCs and CBs and fusomes in differentiated cysts (22). Asreported previously (14), each bamZ/bam�86 transheterozygousovariole had a tumorous germarium with zero or one normal eggchambers, which greatly contrasted with the wild-type ovariole,which had 7–10 egg chambers after the germarium (Fig. 2 A–C).In the bamZ/bam�86 mutant ovarioles, the tumorous germariacarried single spectrosome-containing GSC-like germ cells,whereas egg chambers also often contained single germ cells,which are indicated by spectrosomes (Fig. 2C). These bamZ/bam�86 mutant females were sterile. In contrast, most of thebamZ/bam�86; eIF4A/� ovarioles contained 4–6 normal eggchambers (Fig. 2D). Additionally, those mutant germaria con-tained fewer single germ cells than bamZ/bam�86 mutant ones(Fig. 2D). Four mutant eIF4A mutant alleles gave consistentsignificant suppression of the mutant bamZ/bam�86 phenotypes,

and those mutant females were fertile (Fig. 2B). However, all ofthe eIF4A alleles in the heterozygous state failed to suppress thebam null (bam�86) mutant tumorous phenotype (Fig. 2E),suggesting that eIF4A modulates BAM function in the regula-tion of GSC differentiation. We also tested mutations in vasa (aneIF4A-like gene specifically expressed in germ cells) and eIF4E(encoding a component of the translation initiation complex).Although a mutation in either vasa or eIF4E had significantsuppression of the bamZ/bam�86mutant phenotype, the effectwas much weaker than the mutations in eIF4A (Fig. 2F and G).These results indicate that eIF4A specifically antagonizes BAMfunction in the regulation of GSC differentiation.

To further investigate genetic interactions between bam andeIF4A, we also examined the effect of increasing eIF4A dosage inthe bam�86 heterozygous background. Normally, a wild-type ger-marium has 2 GSCs and 1–2 CBs (Fig. 2H). eIF4A overexpressiondid not significantly change the GSC and cystoblast numbers incomparison with the wild-type (Fig. 2 H and I). This lack of changemay be due to the abundance of eIF4A protein. Although the GSCnumber did not change much in the bam�86 heterozygous germaria,the CB number significantly increased, indicating that bam ishaploinsufficient in the regulation of GSC differentiation (Fig. 2 Hand J). Interestingly, eIF4A overexpression in the bam�86 heterozy-gous background caused the accumulation of significantly moreGSCs and cystoblasts than in the bam�86 heterozygote alone (Fig.2 H and K). These results further support that bam and eIF4Aexhibit dosage-dependent genetic interactions. Since BAM doesnot regulate eIF4A protein expression in germ cells (Fig. S2), theseresults also suggest that eIF4A directly antagonizes BAM functionto favor GSC self-renewal.

eIF4A Is Required for GSC Self-Renewal and Proliferation. BecauseeIF4A can bind to BAM and antagonize its function, we wouldexpect that it promotes GSC self-renewal. To directly investigatethe role of eIF4A in the regulation of GSC self-renewal, we usedFLP-mediated FRT recombination to generate marked eIF4Amutant GSCs and then studied their maintenance and prolifer-ation as we reported previously (10). The marked GSCs wereidentified by absence of armadillo-lacZ expression and presenceof a spectrosome close to cap cells, and the unmarked GSCs wereidentified by presence of armadillo-lacZ expression and a spec-trosome close to cap cells (Fig. 3 A–C and E–G). In the control,

Fig. 1. eIF4A interacts with BAM in yeast and Drosophila S2cells. (A) The yeast 2-hybrid system uses the his3 gene as theselection marker, supporting yeast growth on the mediumlacking histidine (His�). X-GAD and Y-GAD represent GADfused in frame with cDNA fragments generated from Dro-sophila ovarian GSC tumors. Y, but not X, interacts with BAMto activate his3 expression. (B) Four different regions of eIF4Afused with GAD: FL, NT, MP, and CT. (C) FL-GAD, NT-GAD,MP-GAD, and CT-GAD interact with GDB-BAM to supportyeast growth on the His� medium. (D) Flag-tagged BAM canbring down Myc-tagged eIF4A in the presence of RNase orBGCN in S2 cells.

11624 � www.pnas.org�cgi�doi�10.1073�pnas.0903325106 Shen et al.

Dow

nloa

ded

by g

uest

on

June

7, 2

020

76% of the marked GSCs detected 1 week after the cloneinduction (ACI) could be still detected 3 weeks ACI (Fig. 3A–D). In contrast, only �9% of the marked GSCs mutant foreIF4A1013 and eIF4A1006 detected 1 week ACI were maintained3 weeks ACI (Fig. 3D). Consequently, most of the marked eIF4Amutant GSCs were lost from the niche 2 weeks ACI (Fig. 3F),and most of the germaria carrying a marked mutant GSC 1 weekACI had only unmarked GSCs in the niche 3 weeks ACI (Fig.3G). The eIF4Ak01501 showed an intermediate GSC loss pheno-type, which may be due to the nature of a weak mutation (Fig.3D). Because eIF4A has been shown to control cell proliferationin a dosage-dependent manner in the Drosophila imaginal disc(23), we then examined the requirement of eIF4A for GSCdivision. The relative division rate is calculated by the number ofcysts produced by a marked GSC divided by the number of cystsproduced by a marked wild-type GSC. The relative division rateof the marked GSC mutant for eIF4A1006 was 0.45, indicating thateIF4A mutant GSCs divide slower than wild-type ones. These

results demonstrate that eIF4A is required intrinsically forcontrolling GSC maintenance and proliferation.

To help maintain the GSC, eIF4A could act by either regu-lating survival or self-renewal. To investigate whether eIF4A isrequired for GSC survival, we examined the apoptosis of themarked mutant eIF4A1006 GSCs by using an ApopTag labelingkit (Chemicon). After examining 27 marked mutant eIF4A1006

mutant GSCs, we did not find that any of them were positive forApopTag labeling (Fig. 3H). Consistently, we often observed theexistence of one or a few mutant cysts in the germaria that hadrecently lost their marked mutant GSCs (Fig. 3F and Fig. S3 Aand B). These results show that eIF4A maintains GSCs bypreventing their differentiation and is also required for germ cellgrowth and oocyte development in egg chambers.

eIF4A Regulates E-Cadherin Expression, but Not BMP Signaling, in GSCs.BMP signaling is required for maintaining GSCs by preventingGSC differentiation, at least partly, through repressing bam expres-

A B

C D E

F G

H

I

J

K

Fig. 2. eIF4A genetically interactswith bam in a dosage-dependent man-ner. (A) A wild-type ovariole carries agermarium (g) and 5 egg chambers(ec). (B) Barograph showing that eIF4A,and vasa or eIF4E to much less extent,can make the bamZ/bam�86 transhet-erozygous ovaries produce dramati-cally more normal egg chambers (n, to-tal ovarioles examined). P value for agiven phenotype is compared with thebam transheterozygotes. (C) A bamZ/bam�86 ovariolecarriesatumorousger-marium (Inset shows individual spec-trosomes) and 2 egg chambers, one ofwhich (arrowhead) also carries manysingle germ cells (broken lines). (D) AbamZ/bam�86;eIF4AS02439/� ovariolecarries a tumorous germarium and 5normal egg chambers. (E) A bam�86/bam�86;eIF4AS02439/� ovariole carries atumorous germarium and no nor-mal egg chambers. (F) A bamZ/bam�86;vasaRJ36/� ovariole carries a tu-morous germarium, 1 normal eggchamber, and 3 tumorous cham-bers (arrowheads). (G) A bamZ/bam�86;vasaRJ36/� ovariole carries a tu-morous germarium, 1 normal eggchamber, and 1 egg chamber (arrow-head) with single germ cells. (H)Barograph showing that germ cell-specific eIF4A overexpression can en-hance theGSCdifferentiationdefectofthe bam�86 heterozygous mutant (n,totalgermaniaexamined).Pvalueforagiven phenotype is compared withwild type. (I) A nos-gal4 UASp-eIF4A/�germarium carries 2 GSCs (arrows) and1 CB (arrowhead). (J) A bam�86/� ger-marium carries 2 GSCs (arrows) and2CBs (arrowheads). (K) A nos-gal4UASp-eIF4A/�; bam�86/� carries 2 GSCs(arrows) and 5 CBs (arrowheads). (Scalebars: A–G, 50 �m; I–K, 10 �m.)

Shen et al. PNAS � July 14, 2009 � vol. 106 � no. 28 � 11625

DEV

ELO

PMEN

TAL

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

June

7, 2

020

sion (11, 12). Recently, eIF4A has been shown to negatively regulateBMP signaling during early Drosophila development (24). Toinvestigate if eIF4A modulates BMP signaling in GSCs, we exam-ined the expression of Dad-lacZ and bam-GFP in the markedmutant GSCs. Dad is a direct BMP target gene, and Dad-LacZ isa reliable Dad transcriptional reporter line (25). bam-GFP has beenshown to be a reliable transcriptional reporter and is repressed inthe GSC (26). Surprisingly, both Dad-lacZ (total 15 marked GSCsexamined) and bam-GFP (total 19 marked GSCs examined)showed no difference in expression between the marked eIF4A1006

mutant GSCs and the unmarked control GSCs in the same ger-maria (Fig. 3 I–L). This finding strongly argues against the role ofeIF4A in the regulation of BMP signaling in the GSC.

Another factor essential for GSC maintenance is E-cadherin. Loss ofE-cadherin expression in GSCs results in the detachment of GSCs fromthe niche and their subsequent loss (27). To investigate whether eIF4Aregulates E-cadherin expression in the GSC, we used reconstructed 3-Dconfocal images to quantitatively measure E-cadherin expression in thestem cell-niche junction as previously reported (16). The markedeIF4A1006 mutant GSCs expressed significantly less E-cadherin in thestem cell-niche junction than the unmarked control ones in the samegermaria (Fig. 3 M–P). This finding indicates that eIF4A controls GSCmaintenance, at least partly, through regulating E-cadherin expressionin the stem cell-niche junction.

BAM and BGCN Repress E-Cadherin Translation Through Its 3� UTR inS2 Cells. Genetic studies have also shown that bam and bgcn requireeach other to regulate GSC differentiation, suggesting that theymay physically interact (14). Both BAM and BGCN have beenrecently shown to repress E-cadherin expression in the GSC (16).To further understand how BAM and BGCN regulate E-cadherinexpression at the molecular level, we used yeast 2-hybrid and co-IPexperiments to investigate whether BAM and BGCN can physicallyinteract. In the yeast 2-hybrid system, GDB-BAM and BGCN-GAD or GDB-BGCN and BAM-GAD, but not BAM-GAD orBGCN-GAD alone with GDB, could activate his3 expression,suggesting that BAM and BGCN can interact with each other (Fig.

4A). In S2 cells, the Flag-tagged BAM could bring down theMyc-tagged BGCN in the presence or absence of RNase (Fig. 4B),whereas the Myc-tagged BGCN could precipitate the Flag-taggedBAM (Fig. S4). Interestingly, the Flag-tagged BAM could not bringdown the Myc-tagged version of other known RNA binding pro-teins VASA, Rm62, and Me31B, indicating that BAM specificallyinteracts with BGCN to form a protein complex (Fig. S5). Based onthe observation that BAM can interact with eIF4A in the presenceof BGCN and with BGCN in the presence of eIF4A (Figs. 1D and4B), we propose that BAM, BGCN, and eIF4A form a ternaryprotein complex in a RNA-independent manner.

Because BGCN contains a putative DEXH RNA binding domain(14), we then tested if BAM and BGCN can repress E-cadherinexpression at the posttranscriptional level in S2 cells. Because the 3�UTR is frequently the target sequence for mRNA degradation ortranslation regulation, the shg (encoding E-cadherin) 3� UTR was fusedwith the Firefly luciferase reporter and, as the internal control, theactin5C 3� UTR was fused with the Renilla luciferase reporter. Bothreporter gene constructs were expressed under the control of theactin5C promoter. Without BAM and BGCN expression, the expres-sion ratio of the Firefly luciferase versus Renilla luciferase was about 8,which represents their relative protein expression levels (Fig. 4C). In thepresence of either BAM or BGCN alone, the ratios were still similar tothe absence of both BAM and BGCN (Fig. 4C). Note that S2 cellsexpress very low levels of bam and bgcn mRNAs (Fig. S6). However, inthe presence of both BAM and BGCN, the ratio was reduced approx-imately 4-fold in comparison to neither BAM nor BGCN or with eitherBAM or BGCN, indicating that BAM and BGCN function together tonegatively regulate mRNA expression through the shg 3� UTR. SuchE-cadherin expression could either be due to mRNA degradation ortranslational repression. We did not observe any obvious changes inmRNA levels between BAM/BGCM co-expression and BAM orBGCN alone based on the RT-PCR results, suggesting that BAM andBGCN do not regulate reporter mRNA stability through the shg 3�UTR (Fig. 4D). Based on the physical interaction between BAM andeIF4A, we propose that BAM and BGCN repress E-cadherin at thetranslational level as a protein complex.

Fig. 3. eIF4A maintains GSCs by regulating E-cadherinexpression but not BMP signaling. Solid circles indicateunmarked wild-type control GSCs, whereas broken circlesdenote marked eIF4A mutant GSCs. (A–C) A marked wild-type GSC is detected 1 week (A), 2 weeks (B), and 3 weeks(C)ACI. (D)ThemarkedeIF4AmutantGSCsare lostat fasterrates than the marked wild-type control GSCs. The initia-tion percentages 1 week ACI are normalized to 100% forcomparison, and the percentages at the subsequent timepoints are calculated by the actual percentages of thegermaria carrying a marked GSC divided by the actualpercentages, 1 week ACI. (E) A marked eIF4A mutant GSCand an unmarked wild-type GSC in the germarium, 1 weekACI. (F) Two unmarked wild-type GSCs in the germariumwith 1 marked eIF4A mutant cyst (broken lines) developedfromtherecently lostmarkedmutantGSC,2weeksACI. (G)Two unmarked wild-type GSCs in the germarium with 1lost marked mutant GSC exiting the germarium, 3 weeksACI. (H) A marked eIF4A mutant GSC which is negative forApopTag labeling. (I and J) The marked eIF4A mutant GSCand the unmarked wild-type GSC in the germarium havesimilar levels of Dad-lacZ expression. (K and L) Both themarked eIF4A mutant GSC and the unmarked wild-typeGSC in the germarium show no bam-GFP up-regulation.(M–P) Confocal images (M–O) and quantitative data (P)show that the marked eIF4A mutant GSC has less E-cadherin expression than the unmarked wild-type GSC inthe stem cell-niche junction (highlighted by red and yellowlines in N). The germaria in panels A–C and E–L are shownat the same scale. (Scale bars: A and M, 10 �m.)

11626 � www.pnas.org�cgi�doi�10.1073�pnas.0903325106 Shen et al.

Dow

nloa

ded

by g

uest

on

June

7, 2

020

Because BGCN is a putative RNA binding protein (14), it isplausible that the role of BGCN in the BAM/BGCN complex is tobring BAM to its target mRNAs. To test this idea, we generated afusion protein between BAM and �N, which could bind to theFirefly luciferase reporter with the 3� UTR containing a �N bindingsequence known as the 5B box (28). To show that the �N-5B systemworks properly in S2 cells, we tested it using eIF4G and GW182.eIF4G is a m7G cap-binding component of the translation initiationcomplex, which activates translation when loaded onto mRNA (29),whereas GW182 protein is a potent translational repressor forwhichever mRNA it is tethered to (30). Indeed, �N-GW182 re-pressed the translation of the mRNA carrying the 5B at the 3� end,whereas �N-eIF4G dramatically enhanced translation of the samemRNA, indicating that the �N-5B system works properly in S2 cells(Fig. 4E). Interestingly, the �N-BAM fusion protein was sufficientto repress its target mRNA expression in the absence of BGCN(Fig. 4E). This result suggests that the major role of BGCN in thecomplex is to help load BAM to its target mRNAs.

DiscussionThis study has revealed the biochemical function of the BAM/BGCN complex as a translational repressor. We have also showneIF4A in the regulation of GSC self-renewal to be a direct antag-onist of BAM function in the Drosophila ovary. Here, we proposea model explaining how GSC self-renewal is controlled by con-certed actions of intrinsic factors and the extrinsic BMP signal (Fig.

4F). BMP signaling directly represses bam expression, yet leaves lowlevels of BAM protein expression in the GSC (10–12, 16). eIF4Aand other unidentified germline factors in the GSC can effectivelydismantle BAM/BGCN�s repression of GSC maintenance factors,including E-cadherin, through physical interactions, leading to highexpression of maintenance factors in the GSC. In the CB, high levelsof BAM along with BGCN can keep eIF4A proteins out of theactive pool and thus effectively repress GSC maintenance factors,promoting CB differentiation. Therefore, this study has signifi-cantly advanced our current understanding of how GSC self-renewal and differentiation are regulated by translation factors.

BAM and BCGN Form a Protein Complex Involved in TranslationRepression. bam and bgcn genetically require each other’s functionto control CB differentiation (13, 14). Although they are expressedat low levels in GSCs, they have an important role in regulating GSCcompetition (16). However, their biochemical functions remainedunclear until this study. We showed that BAM specifically interactswith BGCN, but not other RNA-binding proteins VASA, Rm62,and Me31B, to form a protein complex (Fig. 4). In addition, we havealso shown that BAM and BGCN together, but not BAM or BGCNalone, are capable of suppressing the expression of the reportercontaining the shg 3� UTR (Fig. 4). Furthermore, BAM and BGCNdo not affect the stability of the reporter mRNA, further supportingthat they regulate mRNA translation but not stability. To reveal therole of BGCN in the function of the BAM/BCGN complex, we

Fig. 4. BAM and BGCN form a complex for repressingE-cadherin translation. (A) BAM and BGCN interactwith each other in the yeast 2-hybrid assay. (B) Flag-BAM can bring down Myc-BGCN in the presence ofRNase or eIF4A in S2 cells. (C) BAM and BGCN togethercan significantly repress the expression of the fireflyluciferase reporter carrying a shg 3� UTR in comparisonwith BAM or BGCN alone. (D) BAM, BGCN, or BAM andBGCN together cannot affect the mRNA stability of thefirefly luciferase reporter carrying a shg 3�UTR. (E)Tethering BAM to the 3� UTR through the �N-5B boxinteraction can repress the firefly luciferase reporter inabsence of BGCN. �N-tagged GW182 (�N-GW) andeIF4G (�N-4G) function as positive controls becausethey are known to repress and activate translation,respectively. (F) A working model explaining howeIF4A controls GSC self-renewal by preventingdifferentiation.

Shen et al. PNAS � July 14, 2009 � vol. 106 � no. 28 � 11627

DEV

ELO

PMEN

TAL

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

June

7, 2

020

showed that direct tethering of BAM to the 3� UTR of the targetmRNA can bypass the requirement of BGCN and sufficientlysuppress the expression of the reporter. Based on the fact thatBGCN contains a putative DEXH RNA binding domain, wepropose that BGCN helps bring BAM to its target mRNAs torepress their translation. Therefore, this study has revealed thebiochemical functions of BAM and BGCN.

Our previous genetic study showed that BAM and BGCNnegatively regulate E-cadherin expression in GSCs to control GSCcompetition (16), but the underlying molecular mechanism remainsdefined. In this study, we showed that in Drosophila S2 cells BAMand BGCN could repress E-cadherin expression through its 3� UTRat the translational level. Along with our previous observation thatBAM and BGCN negatively regulate E-cadherin expression inGSCs in vivo, we propose that BAM and BGCN likely repressE-cadherin expression in GSCs at the translational level. In thefuture, it will be important to show if BAM and BGCN directly bindto the shg 3� UTR to repress E-cadherin expression in the GSC.

eIF4A Directly Antagonizes BAM Function to Maintain E-CadherinExpression in GSCs. eIF4A, an RNA helicase, is one component ofthe translation initiation complex eIF4F, which is required forloading the small 40S ribosome subunit onto the target mRNA toinitiate its translation (30). The helicase activity of eIF4A itself isweak but is enhanced upon binding to eIF4G, another componentof eIF4F. Such helicase activity is important to remove the sec-ondary structure of the 5� UTR, facilitating the ribosome scanningalong mRNA to find the initiation codon ATG. To reveal howBAM and BGCN confer translation repression, we used the yeast2-hybrid screen to identify eIF4A as a BAM interacting protein(Fig. 1). Then, we have provided 2 pieces of genetic of evidencesupporting the idea that eIF4A and bam function together to controlthe balance between GSC self-renewal and differentiation. First,one copy of the mutations in eIF4A can dramatically promote germcell differentiation in the hypomorphic bamZ/bam�86 transheterozy-gous ovaries. However, a mutation in eIF4A cannot suppress thetumorous phenotype of the bam�86 homozygous ovaries (no bamfunction at all), suggesting that the reduction of eIF4A dosage helpsenhance the remaining BAM function. Second, overexpression ofeIF4A can enhance the differentiation defect in the bam�86 het-erozygote. These genetic results support the antagonizing relation-ship between bam and eIF4A (Fig. 4F).

The antagonizing genetic relationship between bam and eIF4Asuggests that eIF4A favors GSC maintenance over differentiation. Ourgenetic analysis of the marked eIF4A mutant GSC clones shows thateIF4A is indeed required in GSCs for their self-renewal and division. Touncover the genetic mechanism underlying the function of eIF4A inmaintaining GSCs, we have also shown that the marked eIF4A mutantGSC has normal BMP signaling activities in comparison with itsneighboring wild-type GSC based on expression results from 2 BMPresponses genes, bam and Dad, but has significantly reduced E-cadherinexpression in comparison with its neighboring wild-type GSC. Thesegenetic and cell biological results demonstrate that eIF4A controls GSCmaintenance at least partly by maintaining E-cadherin expression. Inmammalian cells, overexpression of translation initiation factors, suchas eIF4A, 4G, and 4E, is implicated in different kinds of cancer due totheirability to increasecellproliferation(19). In theDrosophila imaginaldisc, the block in cell proliferation caused by mutations in eIF4A can bebypassed by E2F overexpression, indicating that eIF4A regulates cellcycle progression and consequently cell proliferation (23). In this study,we show that eIF4A is also required for controlling GSC division.Therefore, we propose that eIF4A controls GSC proliferation byregulating cell cycle progression like in Drosophila imaginal tissues.

Materials and MethodsDrosophila Stocks and Genetic Clonal Analysis. Information about the Drosophilastrains is described in the flybase (http://www.flybase.bio.indiana.edu), andthe strains are specified in SI Text.

Yeast 2-Hybrid System. The Drosophila ovarian GSC cDNA pool was con-structed by using SMART technology (Clontech), and the mRNAs were isolatedfrom Dpp overexpression-induced GSC tumors according to the instructionmanuals. Additional details are described in SI Text.

Immunohistochemistry. Multiple antisera were used, and a complete list isgiven in SI Text. The immunostaining protocol and the TUNEL assay using theApopTag kit from Chemicon have been described in Results.

For the details about plasmid constructions, immunoprecipitation, and lucif-erase assay, please see SI Text.

ACKNOWLEDGMENTS. We thank B. Edgar (University of Washington, Seattle),P. Lasko, and the Drosophila stock center for reagents, D. McKearin for sharingunpublished results, and H. Li for advice on statistical analysis. We also thank theXie laboratory members for stimulating discussions and C. Lee and C. Tanzie forhelp with manuscript preparation. This work is supported by Grant R01GM064428 from the National Institute of General Medical Sciences and StowersInstitute for Medical Research.

1. Li L, Xie T (2005) Stem cell niche: Structure and function. Annu Rev Cell Dev Biol 21:605–631.2. MorrisonSJ, SpradlingAC(2008)Stemcellsandniches:Mechanismsthatpromotestemcell

maintenance throughout life. Cell 132:598–611.3. Kirilly D, Xie T (2007) The Drosophila ovary: An active stem cell community. Cell Res 17:15–25.4. Decotto E, Spradling AC (2005) The Drosophila ovarian and testis stem cell niches: Similar

somatic stem cells and signals. Dev Cell 9:501–510.5. XieT,SpradlingAC(2000)Anichemaintaininggermlinestemcells intheDrosophilaovary.

Science 290:328–330.6. Szakmary A, Cox DN, Wang Z, Lin H (2005) Regulatory relationship among piwi, pumilio,

and bag-of-marbles in Drosophila germline stem cell self-renewal and differentiation.Curr Biol 15:171–178.

7. Chen D, McKearin D (2005) Gene circuitry controlling a stem cell niche. Curr Biol 15:179–184.8. Cox DN, et al. (1998) A novel class of evolutionarily conserved genes defined by piwi are

essential for stem cell self-renewal. Genes Dev 12:3715–3727.9. King FJ, Szakmary A, Cox DN, Lin H (2001) Yb modulates the divisions of both germline and

somatic stem cells through piwi- and hh-mediated mechanisms in the Drosophila ovary.Mol Cell 7:497–508.

10. Xie T, Spradling AC (1998) decapentaplegic is essential for the maintenance and divisionof germline stem cells in the Drosophila ovary. Cell 94:251–260.

11. Song X, et al. (2004) Bmp signals from niche cells directly repress transcription of adifferentiation-promoting gene, bag of marbles, in germline stem cells in the Drosophilaovary. Development 131:1353–1364.

12. Chen D, McKearin D (2003) Dpp signaling silences bam transcription directly to establishasymmetric divisions of germline stem cells. Curr Biol 13:1786–1791.

13. McKearin D, Ohlstein B (1995) A role for the Drosophila bag-of-marbles protein in thedifferentiation of cystoblasts from germline stem cells. Development 121:2937–2947.

14. Ohlstein B, Lavoie CA, Vef O, Gateff E, McKearin DM (2000) The Drosophila cystoblastdifferentiation factor, benign gonial cell neoplasm, is related to DExH-box proteins andinteracts genetically with bag-of-marbles. Genetics 155:1809–1819.

15. McKearin DM, Spradling AC (1990) bag-of-marbles: A Drosophila gene required to initiateboth male and female gametogenesis. Genes Dev 4:2242–2251.

16. Jin Z, et al. (2008) Differentiation-defective stem cells outcompete normal stem cells forniche occupancy in the Drosophila ovary. Cell Stem Cell 2:39–49.

17. Kimble J, Crittenden SL (2007) Controls of germline stem cells, entry into meiosis, and thesperm/oocyte decision in Caenorhabditis elegans. Annu Rev Cell Dev Biol 23:405–433.

18. Tsuda M, et al. (2003) Conserved role of nanos proteins in germ cell development. Science301:1239–1241.

19. Sonenberg N, Hinnebusch AG (2009) Regulation of translation initiation in eukaryotes:Mechanisms and biological targets. Cell 136:731–745.

20. OhlsteinB,McKearinD(1997)EctopicexpressionoftheDrosophilaBamproteineliminatesoogenic germline stem cells. Development 124:3651–3662.

21. Suzuki C, et al. (2008) PDCD4 inhibits translation initiation by binding to eIF4A using bothits MA3 domains. Proc Natl Acad Sci USA 105:3274–3279.

22. Lin H, Yue L, Spradling AC (1994) The Drosophila fusome, a germline-specific organelle,contains membrane skeletal proteins and functions in cyst formation. Development120:947–956.

23. Galloni M, Edgar BA (1999) Cell-autonomous and non-autonomous growth-defectivemutants of Drosophila melanogaster. Development 126:2365–2375.

24. Li J, Li WX (2006) A novel function of Drosophila eIF4A as a negative regulator of Dpp/BMPsignalling that mediates SMAD degradation. Nat Cell Biol 8:1407–1414.

25. Tsuneizumi K, et al. (1997) Daughters against dpp modulates dpp organizing activity inDrosophila wing development. Nature 389:627–631.

26. Chen D, McKearin DM (2003) A discrete transcriptional silencer in the bam gene deter-mines asymmetric division of the Drosophila germline stem cell. Development 130:1159–1170.

27. Song X, Zhu CH, Doan C, Xie T (2002) Germline stem cells anchored by adherens junctionsin the Drosophila ovary niches. Science 296:1855–1857.

28. Baron-Benhamou J, Gehring NH, Kulozik AE, Hentze MW (2004) Using the lambdaNpeptide to tether proteins to RNAs. Methods Mol Biol 257:135–154.

29. De Gregorio E, Baron J, Preiss T, Hentze MW (2001) Tethered-function analysis reveals thatelF4E can recruit ribosomes independent of its binding to the cap structure. RNA 7:106–113.

30. Ding L, Han M (2007) GW182 family proteins are crucial for microRNA-mediated genesilencing. Trends Cell Biol 17:411–416.

11628 � www.pnas.org�cgi�doi�10.1073�pnas.0903325106 Shen et al.

Dow

nloa

ded

by g

uest

on

June

7, 2

020