Copyright � 2007 by the Genetics Society of AmericaDOI: 10.1534/genetics.107.071472

The Molecular Chaperone Hsp90 Is Required for mRNA Localization inDrosophila melanogaster Embryos

Yan Song,*,1 Lanette Fee,* Tammy H. Lee* and Robin P. Wharton*,†,2

*Howard Hughes Medical Institute, Department of Molecular Genetics & Microbiology and †Department of Cell Biology,Duke University Medical School, Durham, North Carolina, 27710

Manuscript received January 29, 2007Accepted for publication May 25, 2007

ABSTRACT

Localization of maternal nanos mRNA to the posterior pole is essential for development of both theabdominal segments and primordial germ cells in the Drosophila embryo. Unlike maternal mRNAs such asbicoid and oskar that are localized by directed transport along microtubules, nanos is thought to be trapped asit swirls past the posterior pole during cytoplasmic streaming. Anchoring of nanos depends on integrity of theactin cytoskeleton and the pole plasm; other factors involved specifically in its localization have not beendescribed to date. Here we use genetic approaches to show that the Hsp90 chaperone (encoded by Hsp83 inDrosophila) is a localization factor for two mRNAs, nanos and pgc. Other components of the pole plasm arelocalized normally when Hsp90 function is partially compromised, suggesting a specific role for thechaperone in localization of nanos and pgc mRNAs. Although the mechanism by which Hsp90 acts is unclear,we find that levels of the LKB1 kinase are reduced in Hsp83 mutant egg chambers and that localization of pgc(but not nos) is rescued upon overexpression of LKB1 in such mutants. These observations suggest that LKB1is a primary Hsp90 target for pgc localization and that other Hsp90 partners mediate localization of nos.

SUBCELLULAR localization of mRNA is an efficientstrategy for spatially restricting the encoded pro-tein ( Johnstone and Lasko 2001; St. Johnston 2005).This strategy is common in highly polarized cell typessuch as neurons, as well as during the development ofoocytes or embryos when gene regulation is limitedto post-transcriptional mechanisms. Asymmetric RNAlocalization prior to mitosis also serves to distinguishdaughter cells in Saccharomyces cerevisiae. In this organ-ism, localization of ASH1 mRNA to buds just prior tocytokinesis currently provides the most completely un-derstood example of mRNA localization (Niessing et al.2004; Gonsalvez et al. 2005).

Localization of maternal mRNAs drives much of thepatterning along the anteroposterior axis of the Dro-sophila embryo. The anterior determinant, bicoid (bcd)mRNA, is localized during oogenesis as ribonucleopro-tein (RNP) cargo associated with molecular motors thattraverse the microtubule cytoskeleton (Riechmann et al.2002; Schnorrer et al. 2002; Snee et al. 2005; Weil et al.2006). Posterior patterning is nucleated by the localiza-tion of oskar (osk) mRNA to the posterior pole of theoocyte, also via directed movement along microtubules

(Cha et al. 2002; Braat et al. 2004; Huynh et al. 2004;Yano et al. 2004). Localization of both bcd and osk mRNAsis thought to occur via complex, multistep pathwayswith many components.

One role of localized Osk is to direct the subsequentlocalization of a fraction of the nanos (nos) mRNA late inoogenesis (Bergsten and Gavis 1999). Localized nosmRNA is the sole source of Nos protein in the earlyembryo (Gavis and Lehmann 1992) where it plays anumber of key roles in development. Nos is required inthe somatic cytoplasm of the early embryo to represstranslation of maternal hunchback mRNA, thereby gov-erning abdominal segmentation (Sonoda and Wharton1999). Nos is also required in the primordial germ cellsthat form at the posterior extreme of the embryo to de-lay proliferation, repress transcription, facilitate mi-gration into the somatic gonad, and promote survival(Kobayashi et al. 1996; Forbes and Lehmann 1998;Asaoka-Taguchiet al. 1999; Schaner et al. 2003; Hayashiet al. 2004; Kadyrova et al. 2007). The germ line func-tions of Nos appear to be conserved in many organisms(Subramaniam and Seydoux 1999; Tsuda et al. 2003).

The role of Osk in nos mRNA localization is indirect;Osk governs assembly of the pole plasm (specializedcytoplasm that specifies germ line identity) via anelaborate, genetically defined pathway in which recruit-ment of nos mRNA is one of the final steps (Ephrussiand Lehmann 1992; Kim-Ha et al. 1993). Recruitment isthought to involve trapping of nos RNPs as they swirlpast the posterior pole during cytoplasmic streaming

1Present address: Department of Pathology, Stanford University Schoolof Medicine, Stanford, CA.

2Corresponding author: Howard Hughes Medical Institute, Departmentof Molecular Genetics & Microbiology, Box 3657, Duke UniversityMedical School, Durham, NC 27710. E-mail: rwharton@duke.edu

Genetics 176: 2213–2222 (August 2007)

(Forrest and Gavis 2003; Serbus et al. 2005), ratherthan the directed movement that underlies localizationof osk or ASH1 RNPs. The factors involved specifically inlocalizing nos mRNA have yet to be identified.

In this report, we describe the results of a geneticscreen for nos mRNA localization factors that rely onthe abdomen-patterning role of Nos.

MATERIALS AND METHODS

Isolation and mapping of the 966 mutation: Homozygousw1118 ; e nosBN males were mutagenized with EMS and mateden masse with w1118 virgin females. Male progeny were individ-ually crossed to generate P½mini-nos1�, nosBN/e nosBN * females,which were screened for the ability to give rise to hatchingprogeny. The P½mini-nos1� transgene is described by Dahanukarand Wharton (1996), where it is named nos1(DBX). Candi-date mutant chromosomes (*) were recovered from siblingmales for further testing. Meiotic recombination with a rucucachromosome placed 966 between ru and h on the left arm ofthe third chromosome. Fine mapping by P element-inducedmale recombination further mapped 966 to a 57-kb intervalbetween P{SUPor-P}KG05210 and P{SUPor-P}KG00982. The966 mutation was definitively identified by sequencing theHsp90 coding region amplified from genomic DNA extractedfrom homozygous 966 larvae (identified by the absence of aGFP-marked balancer chromosome).

Fly strains and reagents: The following strains were fromthe Bloomington Stock Center: Hsp83 alleles scratch (08445),E317K (e6D), S529F (e6A), and j5C2A; transformants bearing the17.5 genomic Hsp83 rescue construct; flies with the P{SUPor-P}KG05210, P{SUPor-P}KG00982, P{SUPor-P}KG03657, andP{SUPor-P}KG07503 elements used in male recombination.Flies with the maternal tubulin-GAL4 driver as well as the GFP-LKB1 transgene (Huynh et al. 2001; Martin and St. Johnston2003) were from D. St. Johnston; flies with the P{GAL4-arm.S}11armadillo-GAL4 driver were from Bloomington. Antibodiesagainst various proteins were gifts of P. Macdonald (Hb), A.Nakamura (Nos), D. St. Johnston (Stau), A. Ephrussi (Osk), andK. Howard (Vas).

Immunohistochemistry: Egg chamber fixation and antigendetection were performed as described (Palacios and St.Johnston 2002). Primary antibodies were diluted as follows:chicken anti-Vas 1:2000, rabbit anti-Osk 1:2000, rabbit anti-Stau 1:2000, rat anti-Hb (1:500), and rabbit anti-Nos 1:1000.FITC-, rhodamine-, or Texas-red-conjugated secondary anti-bodies ( Jackson Laboratories) were used at 1:200. Nuclei were

stained with TOTO-3, oligreen, or TOPRO-3 (MolecularProbes). Samples were mounted in Vectashield and imagedon a Zeiss LSM510 confocal microscope. For embryo staining,primary antibodies were diluted as follows: rat anti-Hb (1:500),rabbit anti-Nos 1:1000, rabbit anti-Osk 1:2000. In situ hybrid-ization was by standard methods using digoxigenin-labeleddsDNA probes prepared from cDNA clones. The adducin-like/hts probe was from clone N4 (Ding et al. 1993).

Western and Northern blots: Homozygous 966 mutantlarvae (e.g., non-Tubby) were identified shortly after hatchingand grown under noncrowded conditions. Samples fromwhole larvae homogenized in SDS sample buffer were ana-lyzed following transfer to Immobilon P by standard methods.Hsp90 was detected with the 3E6-1.92 monoclonal antibody,a gift from R. Tanguay (Carbajal et al. 1990), and ECL Plus(Amersham). The loading control was a-tubulin, detectedwith DM 1A monoclonal antibody (Sigma F-2168) and ECL(Amersham). For Northern blots, 5-mg samples of total RNAprepared from 0- to 4.5-hr embryos were analyzed by standardmethods using radio-labeled probes to detect smaug mRNA(as a loading control) and mini-nos1 mRNA. Quantitationwas performed on a Typhoon phosphoimager.

RESULTS

A genetic screen for nos localization factors: Toidentify factors involved in nos mRNA localization, wechemically mutagenized flies and screened for domi-nant maternal-effect mutations that (further) compro-mise abdominal segmentation in a sensitized background.The rationale for the screen is shown in Figure 1.Inefficient localization and translation of a mini-nos1

mRNA generates only sufficient Nos activity to allowdevelopment of 5–6 abdominal segments if the endog-enous nos genes are mutant (Dahanukar and Wharton1996). In such a background, we reasoned that a mu-tation in one of the two alleles encoding a localizationfactor might compromise Nos activity sufficiently to pre-clude hatching (either because the allele is dominantnegative or the gene is haplo-insufficient). A similarrationale has been used extensively in screens based onchanges in morphology of the Drosophila eye (e.g., Rebayet al. 2000).

From a pilot screen of�6000 EMS-mutagenized thirdchromosomes, we isolated a number of mutations that

Figure 1.—A dominant modifier screen for nosmRNA localization factors. Shown schematicallyare the 39-UTRs of wild-type nos1 mRNA and amini-nos1 mRNA that lacks nt 185-849. The firstcolumn shows a qualitative assessment of the rela-tive efficiency of mRNA localization to the poste-rior pole (see Figure 2), the second columnshows the typical number of abdominal segmentsin embryos where the sole source of Nos is the in-dicated mRNA, and the third column is a drawingof the body plan after 24 hr of embryonic develop-ment. The rationale for the screen is that an EMS-induced mutation (*) would dominantly reducesegmentation such that embryos do not hatch(and thus remain within the vitelline membrane).

2214 Y. Song et al.

reduce or eliminate abdominal segmentation in thesensitized background. One mutation proved to be anallele of pumilio and another an allele of spindle-E; bothgenes are known to play roles in posterior specification(Lehmann and Nüsslein-Volhard 1987; Gillespieand Berg 1995; Martin et al. 2003; Cook et al. 2004),validating the premise of the screen. A third mutation,966, is the subject of this report.

The 966 mutation appears to affect the localizationbut not the synthesis or stability of mini-nos1 mRNA(Figure 2). Both the analysis of Northern blots (Figure2A) and examination of the unlocalized mini-nos1

mRNA in embryos hybridized with digoxigenin-labeledprobes (Figure 2B) support the idea that mini-nos1

mRNA stability is unaffected by the 966 mutation. Incontrast, localization of mini-nos1 mRNA is defective

from the beginning of embryonic development (stage1) through formation of the syncytial blastoderm (stage4) in embryos from heterozygous 966 mutant females(Figure 2B). Because Nos protein is generated exclu-sively from translation of localized mRNA, the embryosfrom 966 heterozygotes apparently have reduced Nosactivity, because Hunchback (Hb) accumulates in theposterior and they subsequently fail to develop abdom-inal segments (Figure 2B). The 966 allele is homozygouslethal and germ line clones are rudimentary, precludinganalysis of embryos derived from homozygous females.

We examined the distribution of other mRNAs inembryos from heterozygous 966 mutant females (here-after 966 mutant embryos) to determine whether thedefects in mini-nos1 mRNA localization are specific. Asshown in Figure 3, the localization of full-length nos1

Figure 2.—The 966 mutation dominantly inter-feres with localization of mini-nos1 mRNA. (A)Northern blot of samples from 0 to 4.5 hr nosBN/P½mini-nos1�, nosBN (lane 1) and 966 nosBN/P½mini-nos1�, nosBN (lane 2) embryos to detectsmaug and mini-nos1 mRNAs. The mobility of mo-lecular weight markers is indicated on the left. (B)The distributions of mini-nos1 mRNA during em-bryonic stages 2 (early cleavage, first row) and 4(syncytial blastoderm, second row), Hb proteinin nuclear cycle 10 (third row), and the resultingembryonicbodyplan(darkfieldmicrographsof se-creted cuticle in the fourth row) are shown. Hereand in all subsequent figures, embryos and eggchambers are oriented anterior to the left and dor-sal up. Note that the enzymatic reaction used todetect hybridized digoxigenin-labeled probe wasnearly twice as long for the embryos in row 2 (vs.row 1) due to the presence of reduced mRNA lev-els; control experiments suggest that an apprecia-

ble fraction of the signal in the somatic cytoplasm is due to background at the longer reaction time. In the cuticle photographs,abdominal segments are marked by the presenceof bands of large denticles that are white in darkfield; theeighth abdominal segmentis highlighted with an arrowhead on the left. The maternal genotypes are indicated above.

Figure 3.—The 966 mutation does not signifi-cantly affect localization of other mRNAs in theearly embryo. The distributions of various mRNAs(indicated on the left) are shown in stage 2 em-bryos. Insets on rows 1 and 2 show the posteriorof stage 4 embryos. The maternal genotypes are in-dicated above. Note that other experiments showthat the distributions of pgc, bcd, and add-like mRNAsare unaffected by the nos genotype (not shown).

Hsp90 As an mRNA Localization Factor 2215

mRNA is essentially normal in early 966 mutant em-bryos, although the pole cells in slightly older embryosappear to retain somewhat reduced levels of mRNA. Incontrast, localization of pgc mRNA to the posterior orbicoid and adducin-like mRNAs to the anterior is in-distinguishable in wild-type and 966 mutant embryos.

We next wished to determine whether the 966 geneproduct acts late in the posterior pathway (when nosmRNA is localized) or early, perhaps acting indirectly togovern oocyte polarity and accumulation of Osk, forexample. Three observations suggest that the 966 geneproduct acts downstream of Osk. First, Osk accumula-tion is normal in 966 mutant embryos (Figure 4A). Sec-ond, Osk activity appears normal in 966 mutant embryosby the criterion that they have a similar number of polecells as wild-type embryos (data not shown). The forma-tion of pole cells is known to be a sensitive indicator ofOsk activity (Ephrussi and Lehmann 1992; Smith et al.1992). Third, we find that the 966 mutation interfereswith localization of mini-nos1 mRNA to the anterior ofosk-bcd embryos (Figure 4B). In osk-bcd embryos, a chi-meric mRNA (bearing the protein-coding region of oskand the localization signals of bcd) generates Osk activityat the anterior of the embryo independent of the up-stream factors that regulate localization and translationof native osk mRNA. This ectopic Osk directs efficientlocalization and translation of full-length nos1 mRNA,resulting in the suppression of head and thoracic seg-ments, which are replaced by a mirror symmetric dupli-cation of abdominal segments (Ephrussi and Lehmann1992). We find that mini-nos1 mRNA is inefficiently lo-calized to the anterior of osk-bcd embryos where it ap-parently is translated into only enough Nos to suppresshead segmentation (Figure 4B). If, in addition, the fe-males are heterozygous for 966, then recruitment ofmini-nos1 mRNA to the anterior is almost eliminatedand the suppression of anterior development by ectopicNos is largely relieved (Figure 4B).

Taken together, the results described above suggestthat the 966 mutation alters the function of a factor thatis required downstream of Osk to localize mini-nos1

mRNA.Hsp90 requirement for mRNA localization: We

mapped the 966 mutation (primarily scoring the asso-ciated homozygous lethality) using deficiencies, meioticrecombination, and P element-induced male recombi-nation. In the course of these experiments, we discov-ered that 966 is semilethal in trans to the scratch (stc)allele (Yue et al. 1999) of Hsp83, which encodes thehighly conserved Hsp90 chaperone. Two additionallines of evidence demonstrate that 966 is indeed anallele of Hsp83. First, the Hsp83 gene on the 966 chro-mosome bears a single nucleotide substitution thatresults in an alanine to aspartate substitution at a highlyconserved residue (133) in the N-terminal ATPasedomain (Figure 5A). Second, two independently iso-lated Hsp83 alleles that encode missense forms of the

Hsp90 protein (E317K and S592F) (Cutforth andRubin 1994; van der Straten et al. 1997) suppressabdominal segmentation in the mini-nos1 backgroundin a manner similar to 966 (Figure 5A). Thus, Hsp90function is required for normal localization of mini-nos1

mRNA.We next asked whether the A133D allele acts in a

dominant negative or haplo-insufficient manner tomodify mini-nos1-dependent segmentation. The E317Kand S592F alleles were identified in screens for domi-nant modifiers of kinase-dependent eye phenotypes;since they also modify mini-nos1 activity, we suspectedthat the A133D mutant and the other missense mutantsact as dominant negatives for mini-nos1 localization.Consistent with such an idea, a presumptive null alleleof Hsp83, j5C2A, which bears a P-element in the first(nonprotein-coding) exon, does not dominantly mod-ify the mini-nos1 segmentation phenotype. Moreover,rescue of the segmentation phenotype caused by theA133D allele is inefficient. A single wild-type Hsp83transgene rescues very poorly, yielding embryos with 1 to2 abdominal segments; two copies of the wild-type Hsp83transgene are required for full rescue (Figure 5B).Finally, we find that the A133D mutant protein accu-mulates to wild-type levels in extracts prepared fromlarvae �1 day before they die (Figure 5B). Takentogether, these results suggest that the A133D missenseprotein dominantly interferes with wild-type Hsp90 withrespect to the mini-nos1 segmentation phenotype.

To further investigate the role of Hsp90 in localiza-tion of maternal mRNAs in general and full-length nos1

Figure 4.—The 966 gene product acts downstream of Os-kar in the posterior pathway to localize mini-nos1 mRNA. (A)The distribution of Osk in stage 2 embryos, maternal geno-types as indicated. (B) The distribution of mini-nos1 mRNAin stage 2 embryos and anterior cuticle (phase-contrast micro-graphs) of mature embryos from females of the indicated ge-notype. Note that virtually the entire head skeleton is absentin the embryo on the left, with only a trace of sclerotized cu-ticle in evidence (arrowhead). With the exception of minordefects in the dorsal bridge (arrowhead), the head skeleton(arrowhead) in the embryo on the right is wild type.

2216 Y. Song et al.

mRNA in particular, it was necessary to bypass therequirement of Hsp90 function for viability. A compre-hensive study of Hsp83 alleles had shown that severaltrans-heterozygous allelic combinations are viable andfertile (Yue et al. 1999). We reinvestigated the issue andfound that Hsp83stc/Hsp83E317K females are reasonablyhealthy and produce large numbers of eggs, of which1–2% develop through late embryonic stages. We there-fore used this genetic background to investigate theconsequences of impairing maternal Hsp90 functionon mRNA localization.

As shown in Figure 6, localization of full-length nos1

mRNA is defective in 90–95% of embryos when mater-nal Hsp90 function is compromised. The phenotype isheterogeneous and includes nearly normal localizationof a reduced amount of nos RNA to a crescent along theposterior cortex, localization of a ‘‘ball’’ of nos mRNAonly partially in contact with the posterior cortex,diffuse localization of a cloud near the posterior, andno detectable localization. These defects in mRNAlocalization have predictable effects on subsequentdevelopment: essentially all Hsp83stc/Hsp83E317K embryoshave reduced levels of Nos protein (Figure 6). Very fewof these embryos cellularize, presumably due to someother requirement(s) for maternal Hsp90 function. Butthe 1–2% of embryos that eventually secrete cuticle haveon average 3 abdominal segments rather than thenormal complement of 8 (Figure 6). Thus, maternalHsp90 is critical for posterior localization of nos1 mRNA.

We next wished to determine whether localization ofother maternal mRNAs also relies on Hsp90. Accord-

ingly, we examined the distributions of polar granulecomponent (pgc), germcell-less (gcl), CyclinB (CycB) (Lehnerand O’Farrell 1990; Jongens et al. 1992; Nakamuraet al. 1996), and bcd mRNAs in Hsp83stc/Hsp83E317K em-bryos. Of these, only pgc mRNA localization is abnormal(Figure 6 and not shown). Localization of pgc mRNA isaffected to a greater extent than is localization of nosmRNA. Because Hsp90 activity is only partially ablated inthe Hsp83stc/Hsp83E317K egg chambers, we do not knowwhether the posterior localization of gcl and CycB mRNAsand the anterior localization of bcd mRNA rely on lowerHsp90 activity or whether they are Hsp90 independent.However, the significant observation is that proper de-ployment of only a subset of the late-localizing posteriormRNAs (nos and pgc) requires normal Hsp90 activity.

Hsp90 is thought to interact with hundreds of pro-teins in most cells (Millson et al. 2005; Zhao et al. 2005)and Hsp83 mutants are highly pleiotropic (Yue et al.1999). We therefore considered whether the defects innos and pgc localization might be quite indirectly causedby earlier defects in the assembly of pole plasm com-ponents at the posterior of the egg chamber. To thisend, we examined the localization of upstream compo-nents of the posterior pathway in Hsp83stc/Hsp83E317K eggchambers where Hsp90 activity is compromised.

Reduction of Hsp90 activity has a relatively specificeffect on the posterior localization of nos and pgc mRNAs.The initial localization of three key factors that act up-stream of nos is essentially normal in Hsp83stc/Hsp83E317K

egg chambers (Figure 7). These include: (1) Stau protein½an effective proxy for osk mRNA (Martin et al. 2003)�,

Figure 5.—966 is a dominantnegative allele of Hsp83. (A) Atthe top is a schematic representa-tion of the domains of Hsp90and the positions of missense sub-stitutions encoded by the 966 al-lele and the two other allelesused in these studies. Abdominalsegmentation is visualized in dark-field photographs of cuticle se-creted by embryos from femalesheterozygous for the indicatedHsp83 allele (except in row 1,which is from an Hsp831/Hsp831

female). The females here and inB are also P½mini-nos1�, nosBN/nosBN,such that the sole source of Nosin embryos is from the mini- nos1

transgene. (B) The A133D proteinacts in a dominant negative fashion.Cuticle of embryos from femalesheterozygous for the indicatedHsp83 allele, as in A. Embryos inrows 2 and 3 are from females that

also bear oneand two extra copies of an Hsp831 transgene, respectively. Note that onecopy of the transgene does not efficiently rescuethe lethalityofHsp83A133D/Hsp83stc animals, consistentwith theidea that thisallele, likeothermissenseallelesofHsp83(CutforthandRubin 1994; van der Straten et al. 1997), is dominant-negative. Western blot of extracts (each containing 5 mg of protein) preparedfrom w1118 (lane 1), nosBN/TM3 (lane 2), and 966 nosBN/966 nosBN (lane 3) larvae 72–74 hr after egg laying. Following transfer, themembrane was cut in half and probed separately with antibodies against Hsp90 and a-tubulin.

Hsp90 As an mRNA Localization Factor 2217

(2) Osk protein, and (3) Vasa protein. Since accumula-tion of Osk is acutely interdependent with localizationof osk mRNA, Vasa, and the Par-1 kinase (Breitwieseret al. 1996; Riechmann et al. 2002; Benton and St.Johnston 2003; Johnstone and Lasko 2004), the datapresented in Figure 7 suggest that pole plasm assemblyis essentially normal until the late recruitment of nos andpgc. Oogenesis appears grossly normal in the Hsp83stc/Hsp83E317K background, which suggests that polarizationof the microtubule cytoskeleton is likely to be normal.This idea is further supported by the normal initial lo-calization of Stau to the posterior and the maintenanceof bcd mRNA at the anterior, both dependent criticallyon microtubule integrity (Pokrywka and Stephenson1991; Brendza et al. 2000; Weil et al. 2006). Not onlyis the initial formation of pole plasm normal, but also itsintegrity is maintained in Hsp83stc/Hsp83E317K embryos,based on the normal posterior localization of Osk andVasa (Figure 7). Distribution of the latter protein was

detected in double-staining experiments, which showthat Hsp83stc/Hsp83E317K embryos with very little or nodetectable Nos have a normal crescent of Vasa at theposterior pole. Thus, both the initial formation andmaintenance of the pole plasm appear unperturbed inHsp83stc/Hsp83E317K flies.

Taken together, the observations outlined above leadus to conclude that, despite its general pleiotropy,Hsp90 plays a relatively specific role for the localizationof nos and pgc mRNAs to the posterior of the embryo.

Rescue of pgc mRNA localization in Hsp83 mutantembryos by overexpression of LKB1: Hsp90 is a molec-ular chaperone that activates and stabilizes a wide varietyof client regulatory and signaling proteins (Pearl andProdromou 2001); a priori, it seemed unlikely that Hsp90interacts directly with nos and pgc mRNAs. Therefore, oneapproach to further understanding its role in mRNAlocalization would be to identify molecules whose activ-ity is dependent on Hsp90. To date, none of the othergenetic modifiers identified in the screen that yieldedthe 966 allele of Hsp83 encodes an obvious Hsp90 client.Therefore, we turned to a candidate gene approach, fo-cusing on protein kinases previously implicated in vari-ous aspects of posterior patterning. Two such proteins arePar-1 and LKB1, which are required for polarization ofthe oocyte microtubule cytoskeleton and the proper de-position of osk mRNA at the posterior (Shulman et al.2000; Tomancak et al. 2000; Martin and St. Johnston2003; Doerflinger et al. 2006).

Two lines of evidence suggest that LKB1 is a signifi-cant Hsp90 client for the localization of pgc mRNA. First,the level of GFP-LKB1 is significantly reduced whenHsp90 activity is compromised in essentially all Hsp83stc/Hsp83E317K egg chambers (Figure 8A). In contrast, noconsistent effect of reducing Hsp90 activity is seen onlevels of GFP-Par-1 (not shown). Second, overexpres-sion of LKB1 in Hsp83stc/Hsp83E317K females significantlyrescues the localization of pgc mRNA (Figure 8B). Forthis experiment, we used the armadillo-GAL4 driver toachieve low-level overexpression of GFP-LKB1 in the ova-ries, as previously described (Martin and St. Johnston2003). The rescue of pgc localization does not appear tobe due to an indirect elevation of posterior Osk levels,which are normal during both oogenesis and early em-bryogenesis (Figure 8B); these observations are consistentwith the previous finding that up to 10-fold overexpres-sion of LKB1 has no significant effect on localization of thepole plasm component Stau (Martin and St. Johnston2003). In contrast to the rescue of pgc, no significant res-cue of nos localization was observed (Figure 8B). Similarnegative results were obtained using nos-GAL4-VP16 todrive higher level expression of LKB1 in the germ line(Van Doren et al. 1998) (not shown). Taken together,these observations suggest that, for localization of pgcmRNA, a major function of Hsp90 is to stabilize LKB1.Presumably other Hsp90 partners or targets mediate lo-calization of nos mRNA.

Figure 6.—Hsp90 is required for localization of nos and pgcmRNAs. The distributions of various mRNAs and proteins(except in row 3, which shows cuticle secreted by mature em-bryos) in stage 2 embryos from wild-type (w1118) and Hsp83stc/Hsp83E317K females. Many of the Hsp83stc/Hsp83E317K embryoshave terminal defects not readily visible in the photograph,consistent with previous reports that Hsp90 is a cofactor ofthe Raf kinase (van der Straten et al. 1997), which actsdownstream of Torso to specify terminal identity in the em-bryo. The embryos in row 5 were hybridized simultaneouslywith probes to bcd (the prominent signal at the anterior)and pgc (the low-level unlocalized signal as well as the poste-rior cap that is present in wild type but not Hsp83 mutant em-bryos). Note that many Hsp83 mutant embryos have anabnormal shape, with elongated anterior ends to which thevitelline membrane and micropyle frequently remain at-tached (particularly evident in rows 4 and 5).

2218 Y. Song et al.

DISCUSSION

The specific role that maternal Hsp90 plays inlocalization of a subset of mRNAs to the pole plasm issomewhat surprising, given the number of proteins thatare thought to require the activity of this chaperone.Although ubiquitously distributed, Hsp90 is enrichedin the testes and ovaries and the male germ line is par-ticularly sensitive to a reduction in Hsp90 activity (Yueet al. 1999). The experiments reported here define a rolefor maternal Hsp90 in the localization of nos and pgcmRNAs. We do not yet know the mechanism by whichHsp90 acts. However, the apparent integrity of the poleplasm in Hsp83stc/Hsp83E317K ovaries and embryos (Fig-ures 6 and 7) and the rescue of pgc localization uponoverexpression of a single kinase (Figure 8) are consis-tent with the idea that the defects in pgc (and perhapsalso nos) localization arise from the reduction in activityof a few discrete Hsp90 clients. Hsp83 mRNA is itselfconcentrated in the pole plasm (Ding et al. 1993) andHsp90 is found at particularly high levels in the germline precursors of other organisms (Vanmuylder et al.2002; Inoue et al. 2003), which may reflect a conservedrole in mRNA localization.

We do not know whether Hsp90 acts directly orindirectly to stabilize LKB1. Mammalian LKB1 bindsdirectly to Hsp90 and Cdc37, a cochaperone for kinaseclients (Boudeau et al. 2003). However, we have notobserved a direct interaction between Drosophila Hsp90and LKB1, either by co-immunoprecipitation or in yeast

interaction experiments in which the DNA-bindingdomain was fused to the Hsp90 C terminus to avoidinterfering with dimerization, as described (Millsonet al. 2005). We do not currently know whether Hsp90binding to LKB1 is ephemeral (and thus difficult todetect) or whether Hsp90 acts indirectly to stabilizeLKB1.

How might LKB1 act to localize pgc mRNA? Despiteits conserved role in regulation of cellular polarity (seeAlessi et al. 2006 and references therein), no LKB1substrate that plays a direct role in mRNA localizationhas been described, to our knowledge. Loss of LKB1function in germ line clones of presumptive null allelesprevents the reorganization of the oocyte microtubulenetwork at stage 7 that is required for posterior lo-calization of osk mRNA and affects epithelial polarity inthe ovarian follicle cells (Martin and St. Johnston2003). LKB1 colocalizes with cortical actin in the oocyte,integrity of which is required for anchoring of poleplasm components and nos mRNA (Lantz et al. 1999;Forrest and Gavis 2003). It is therefore attractive tospeculate that LKB1 might act at the cortex, where actinand microtubule filaments meet, phosphorylating acurrently unknown substrate to promote the trappingof pgc-containing RNPs. Our results suggest that thelevel of LKB1 in Hsp83stc/Hsp83E317K flies is insufficientfor pgc localization but sufficient for viability as well asproper polarization of microtubules during oogenesisand localization of osk. According to this idea, LKB1

Figure 7.—Assembly and maintenance of thepole plasm in Hsp83 mutants. The distributionsof various proteins in stage 10B egg chambersor stage 2 embryos from wild-type (w1118) andHsp83stc/Hsp83E317K females. Nuclei in rows 1–3are in blue, green, and red, respectively. Notethat the initial accumulation of Osk at the poste-rior is slightly defective in some stage Hsp83 mu-tant egg chambers at stage 7–8; however, by stage9–10, the level of posterior Osk in all mutant eggchambers is indistinguishable from that in wildtype. Also note that the level of Osk in manyHsp83 mutant embryos is higher for reasons wedo not understand. As a result, the protein is de-tectable further toward the anterior than in wild-type embryos. In row 5, distributions of Vasa(red) and Nos (green) at the posterior of stage2 embryos are shown, with both channels overlaidin the third image for each genotype.

Hsp90 As an mRNA Localization Factor 2219

hypomorphs might exhibit many of the defects weobserve in flies with reduced Hsp90 function.

Two other kinases have been implicated in mRNAlocalization in Drosophila, but as is the case for the roleof LKB1 in pgc localization, the critical direct sub-strate(s) for each have yet to be identified. Proteinkinase A (PKA) is required for the microtubule re-organization described above that leads to posteriorlocalization of osk mRNA (Lane and Kalderon 1994).Although PKA has been shown to phosphorylate LKB1at residue 535 in vitro (Martin and St. Johnston 2003),overexpression of LKB1 bearing a phosphomimeticS535E substitution does not rescue the microtubule de-fects in PKA mutant ovaries, suggesting that some otherprotein is the major target for PKA during microtubulereorganization (Steinhauer and Kalderon 2005). Asecond kinase, IkB kinase-like2 (Ik2), and its bindingpartner, Spindle-F (Spn-F), have recently been shown toregulate both microtubule and actin filament distribu-tions in the female germ line (Abdu et al. 2006; Shapiroand Anderson 2006). The authors proposed that Ik2/Spn-F facilitates the connection of a subset of micro-tubules to cortical actin, although the mechanism of theiraction is unknown. For each of these kinases, LKB1,PKA, and Ik2, further biochemical and genetic experi-ments will be required to determine how they act tolocalize mRNA.

The differential effects we observe on localization ofnos, pgc, and CycB (Figures 6 and 8) suggest that eachmRNA is localized by a somewhat different mechanism.CycB mRNA localization is normal in Hsp83stc/Hsp83E317K

embryos and thus appears to be relatively Hsp90 in-dependent; pgc mRNA localization requires normallevels of Hsp90 activity, primarily to stabilize LKB1; andnos mRNA localization requires normal levels of Hsp90activity, presumably to stabilize or activate other (cur-rently unknown) factors. Studies of the polar granulecomponent Tudor support the idea that different mech-anisms underlie localization of nos, pgc, and gcl mRNAs,each of which is concentrated at the posterior pole latein oogenesis (Thomson and Lasko 2004; Arkov et al.2006). Among this class of mRNAs, only nos has beenstudied in detail (Forrest and Gavis 2003). Similardetailed studies of pgc, gcl, and CycB localization mightreveal some of the mechanistic differences.

The genetic screen we employed to identify Hsp90appears to constitute a promising approach for theidentification of additional nos mRNA localization factors.One key aspect of the screen was reliance on the identifi-cation of dominant modifier mutations. Such mutationsmay be relatively easy to isolate in the case of Hsp83, asthe encoded protein is a multidomain dimer that formslarge complexes with other factors, including its cocha-perones (Pearl and Prodromou 2001). Nevertheless,

Figure 8.—Overexpression ofthe LKB1 kinase rescues pgcmRNA localization in Hsp83 mu-tant embryos. (A) The distribu-tion of GFP-LKB1 in sibling wildtype (Hsp83�/1) or Hsp83stc/Hsp83E317K stage 10B egg cham-bers. (B) The distributions ofpgc mRNA, Osk, and nos mRNAare shown in stage 2 embryos orstage 10B egg chambers for threematernal genotypes: wild type(w1118), Hsp83stc/Hsp83E317K, andarmadillo-GAL4/UAS-GFP-LKB1 ;Hsp83stc/Hsp83E317K .

2220 Y. Song et al.

we are optimistic that a scaled-up version of the screenthat surveys the entire genome and characterization ofresulting mutants will yield additional components ofthe nos mRNA localization machinery.

We especially thank Daniel St. Johnston for his generosity insupplying strains and reagents; Anne Ephrussi, Akira Nakamura, PaulLasko, Paul Macdonald, and Ken Howard for reagents; Joe Heitman,Chris Nicchitta, and Danny Lew for comments on the manuscript andsuggestions; Sandy Curlee and Estelle Tsalik for assistance in the officeand lab, respectively; and Jian Chen for media preparation. We are alsoindebted to the Bloomington Stock Center. Y. S. was a PredoctoralFellow of the HHMI and Robin Wharton is an Investigator of theHoward Hughes Medical Institute.

LITERATURE CITED

Abdu, U., D. Bar and T. Schupbach, 2006 spn-F encodes a novelprotein that affects oocyte patterning and bristle morphologyin Drosophila. Development 133: 1477–1484.

Alessi, D. R., K. Sakamoto and J. R. Bayascas, 2006 Lkb1-dependentsignaling pathways. Annu. Rev. Biochem. 75: 137–163.

Arkov, A. L., J. Y. Wang, A. Ramos and R. Lehmann, 2006 The roleof Tudor domains in germline development and polar granulearchitecture. Development 133: 4053–4062.

Asaoka-Taguchi, M., M. Yamada, A. Nakamura, K. Hanyu and S.Kobayashi, 1999 Maternal Pumilio acts together with Nanosin germline development in Drosophila embryos. Nat. Cell Biol.1: 431–437.

Benton, R., and D. St. Johnston, 2003 Drosophila PAR-1 and 14–3-3 inhibit Bazooka/PAR-3 to establish complementary corticaldomains in polarized cells. Cell 115: 691–704.

Bergsten, S. E., and E. R. Gavis, 1999 Role for mRNA localizationin translational activation but not spatial restriction of nanosRNA. Development 126: 659–669.

Boudeau, J., M. Deak, M. A. Lawlor, N. A. Morrice and D. R. Alessi,2003 Heat-shock protein 90 and Cdc37 interact with LKB1 andregulate its stability. Biochem. J. 370: 849–857.

Braat, A. K., N. Yan, E. Arn, D. Harrison and P. M. Macdonald,2004 Localization-dependent oskar protein accumulation; con-trol after the initiation of translation. Dev. Cell 7: 125–131.

Breitwieser, W., F. H. Markussen, H. Horstmann and A. Ephrussi,1996 Oskar protein interaction with Vasa represents an essen-tial step in polar granule assembly. Genes Dev. 10: 2179–2188.

Brendza, R. P., L. R. Serbus, J. B. Duffy and W. M. Saxton, 2000 Afunction for kinesin I in the posterior transport of oskar mRNAand Staufen protein. Science 289: 2120–2122.

Carbajal, M. E., J. P. Valet, P. M. Charest and R. M. Tanguay,1990 Purification of Drosophila hsp 83 and immunoelectronmicroscopic localization. Eur. J. Cell Biol. 52: 147–156.

Cha, B. J., L. R. Serbus, B. S. Koppetsch and W. E. Theurkauf,2002 Kinesin I-dependent cortical exclusion restricts pole plasmto the oocyte posterior. Nat. Cell Biol. 4: 592–598.

Cook, H. A., B. S. Koppetsch, J. Wu and W. E. Theurkauf, 2004 TheDrosophila SDE3 homolog armitage is required for oskar mRNAsilencing and embryonic axis specification. Cell 116: 817–829.

Cutforth, T., and G. M. Rubin, 1994 Mutations in Hsp83 andcdc37 impair signaling by the sevenless receptor tyrosine kinasein Drosophila. Cell 77: 1027–1036.

Dahanukar, A., and R. P. Wharton, 1996 The Nanos gradient inDrosophila embryos is generated by translational regulation.Genes Dev. 10: 2610–2620.

Ding, D., S. M. Parkhurst, S. R. Halsell and H. D. Lipshitz,1993 Dynamic Hsp83 RNA localization during Drosophila oo-genesis and embryogenesis. Mol. Cell. Biol. 13: 3773–3781.

Doerflinger, H., R. Benton, I. L. Torres, M. F. Zwart and D. St.Johnston, 2006 Drosophila anterior-posterior polarity requiresactin-dependent PAR-1 recruitment to the oocyte posterior. Curr.Biol. 16: 1090–1095.

Ephrussi, A., and R. Lehmann, 1992 Induction of germ cell forma-tion by oskar. Nature 358: 387–392.

Forbes, A., and R. Lehmann, 1998 Nanos and Pumilio have criticalroles in the development and function of Drosophila germlinestem cells. Development 125: 679–690.

Forrest, K. M., and E. R. Gavis, 2003 Live imaging of endogenousRNA reveals a diffusion and entrapment mechanism for nanosmRNA localization in Drosophila. Curr. Biol. 13: 1159–1168.

Gavis, E. R., and R. Lehmann, 1992 Localization of nanos RNA con-trols embryonic polarity. Cell 71: 301–313.

Gillespie, D. E., and C. A. Berg, 1995 Homeless is required forRNA localization in Drosophila oogenesis and encodes a newmember of the DE-H family of RNA-dependent ATPases. GenesDev. 9: 2495–2508.

Gonsalvez, G. B., C. R. Urbinati and R. M. Long, 2005 RNA local-ization in yeast: moving towards a mechanism. Biol. Cell 97: 75–86.

Hayashi, Y., M. Hayashi and S. Kobayashi, 2004 Nanos suppressessomatic cell fate in Drosophila germ line. Proc. Natl. Acad. Sci.USA 101: 10338–10342.

Huynh, J. R., J. M. Shulman, R. Benton and D. St. Johnston,2001 PAR-1 is required for the maintenance of oocyte fate inDrosophila. Development 128: 1201–1209.

Huynh, J. R., T. P. Munro, K. Smith-Litiere, J. A. Lepesant andD. St. Johnston, 2004 The Drosophila hnRNPA/B homolog,Hrp48, is specifically required for a distinct step in osk mRNAlocalization. Dev. Cell 6: 625–635.

Inoue, T., K. Takamura, H. Yamae, N. Ise, M. Kawakami et al.,2003 Caenorhabditis elegans DAF-21 (HSP90) is characteristi-cally and predominantly expressed in germline cells: spatialand temporal analysis. Dev. Growth Differ. 45: 369–376.

Johnstone, O., and P. Lasko, 2001 Translational regulation andRNA localization in Drosophila oocytes and embryos. Annu.Rev. Genet. 35: 365–406.

Johnstone, O., and P. Lasko, 2004 Interaction with eIF5B is essen-tial for Vasa function during development. Development 131:4167–4178.

Jongens, T. A., B. Hay, L. Y. Jan and Y. N. Jan, 1992 The germ cell-less gene product: a posteriorly localized component necessaryfor germ cell development in Drosophila. Cell 70: 569–584.

Kadyrova, L. Y., Y. Habara, T. H. Lee and R. P. Wharton,2007 Translational control of maternal Cyclin B mRNA by Nanosin the Drosophila germline. Development 134: 1519–1527.

Kim-Ha, J., P. J. Webster, J. L. Smith and P. M. Macdonald,1993 Multiple RNA regulatory elements mediate distinct stepsin localization of oskar mRNA. Development 119: 169–178.

Kobayashi, S., M. Yamada, M. Asaoka and T. Kitamura, 1996 Es-sential role of the posterior morphogen nanos for germline de-velopment in Drosophila. Nature 380: 708–711.

Lane, M. E., and D. Kalderon, 1994 RNA localization along the an-teroposterior axis of the Drosophila oocyte requires PKA-mediatedsignal transduction to direct normal microtubule organization.Genes Dev. 8: 2986–2995.

Lantz, V. A., S. E. Clemens and K. G. Miller, 1999 The actin cyto-skeleton is required for maintenance of posterior pole plasmcomponents in the Drosophila embryo. Mech. Dev. 85: 111–122.

Lehmann, R., and C. Nüsslein-Volhard, 1987 Involvement of thepumilio gene in the transport of an abdominal signal in the Dro-sophila embryo. Nature 329: 167–170.

Lehner, C. F., and P. H. O’Farrell, 1990 The roles of Drosophilacyclins A and B in mitotic control. Cell 61: 535–547.

Martin, S. G., and D. St. Johnston, 2003 A role for DrosophilaLKB1 in anterior-posterior axis formation and epithelial polarity.Nature 421: 379–384.

Martin, S. G., V. Leclerc, K. Smith-Litiere and D. St. Johnston,2003 The identification of novel genes required for Drosophilaanteroposterior axis formation in a germline clone screen usingGFP-Staufen. Development 130: 4201–4215.

Millson, S. H., A. W. Truman, V. King, C. Prodromou, L. H. Pearlet al., 2005 A two-hybrid screen of the yeast proteome for Hsp90interactors uncovers a novel Hsp90 chaperone requirement inthe activity of a stress-activated mitogen-activated protein kinase,Slt2p (Mpk1p). Eukaryot. Cell 4: 849–860.

Nakamura, A., R. Amikura, M. Mukai, S. Kobayashi and P. F. Lasko,1996 Requirement for a noncoding RNA in Drosophila polargranules for germ cell establishment. Science 274: 2075–2079.

Niessing, D., S. Huttelmaier, D. Zenklusen, R. H. Singer and S. K.Burley, 2004 She2p is a novel RNA binding protein with a ba-sic helical hairpin motif. Cell 119: 491–502.

Palacios, I. M., and D. St. Johnston, 2002 Kinesin light chain-independent function of the Kinesin heavy chain in cytoplasmic

Hsp90 As an mRNA Localization Factor 2221

streaming and posterior localisation in the Drosophila oocyte.Development 129: 5473–5485.

Pearl, L. H., and C. Prodromou, 2001 Structure, function, andmechanism of the Hsp90 molecular chaperone. Adv. ProteinChem. 59: 157–186.

Pokrywka, N. J., and E. C. Stephenson, 1991 Microtubules mediatethe localization of bicoid RNA during Drosophila oogenesis.Development 113: 55–66.

Rebay, I., F. Chen, F. Hsiao, P. A. Kolodziej, B. H. Kuang et al.,2000 A genetic screen for novel components of the Ras/Mitogen-activated protein kinase signaling pathway that interact with the yangeneofDrosophilaidentifiessplitends,anewRNArecognitionmotif-containing protein. Genetics 154: 695–712.

Riechmann, V., G. J. Gutierrez, P. Filardo, A. R. Nebreda and A.Ephrussi, 2002 Par-1 regulates stability of the posterior deter-minant Oskar by phosphorylation. Nat. Cell Biol. 4: 337–342.

Schaner, C. E., G. Deshpande, P. D. Schedl and W. G. Kelly,2003 A conserved chromatin architecture marks and maintainsthe restricted germ cell lineage in worms and flies. Dev. Cell 5:747–757.

Schnorrer, F., S. Luschnig, I. Koch and C. Nusslein-Volhard,2002 Gamma-tubulin37C and gamma-tubulin ring complex pro-tein 75 are essential for bicoid RNA localization during drosophilaoogenesis. Dev. Cell 3: 685–696.

Serbus, L. R., B. J. Cha, W. E. Theurkauf and W. M. Saxton,2005 Dynein and the actin cytoskeleton control kinesin-drivencytoplasmic streaming in Drosophila oocytes. Development 132:3743–3752.

Shapiro, R. S., and K. V. Anderson, 2006 Drosophila Ik2, a memberof the I kappa B kinase family, is required for mRNA localizationduring oogenesis. Development 133: 1467–1475.

Shulman, J. M., R. Benton and D. St. Johnston, 2000 The Dro-sophila homolog of C. elegans PAR-1 organizes the oocyte cyto-skeleton and directs oskar mRNA localization to the posteriorpole. Cell 101: 377–388.

Smith, J. L., J. E. Wilson and P. M. Macdonald, 1992 Over-expression of oskar directs ectopic activation of nanos and pre-sumptive pole cell formation in Drosophila embryos. Cell 70:849–859.

Snee, M. J., E. A. Arn, S. L. Bullock and P. M. Macdonald,2005 Recognition of the bcd mRNA localization signal in Dro-sophila embryos and ovaries. Mol. Cell Biol. 25: 1501–1510.

Sonoda, J., and R. P. Wharton, 1999 Recruitment of Nanos tohunchback mRNA by Pumilio. Genes Dev. 13: 2704–2712.

Steinhauer, J., and D. Kalderon, 2005 The RNA-binding proteinSquid is required for the establishment of anteroposterior polar-ity in the Drosophila oocyte. Development 132: 5515–5525.

St. Johnston, D., 2005 Moving messages: the intracellular localiza-tion of mRNAs. Nat. Rev. Mol. Cell Biol. 6: 363–375.

Subramaniam, K., and G. Seydoux, 1999 nos-1 and nos-2, two genesrelated to Drosophila nanos, regulate primordial germ cell devel-opment and survival in Caenorhabditis elegans. Development126: 4861–4871.

Thomson, T., and P. Lasko, 2004 Drosophila tudor is essential forpolar granule assembly and pole cell specification, but not forposterior patterning. Genesis 40: 164–170.

Tomancak, P., F. Piano, V. Riechmann, K.C. Gunsalus, K. J. Kemphueset al., 2000 A Drosophila melanogaster homologue of Caeno-rhabditis elegans par-1 acts at an early step in embryonic-axis for-mation. Nat. Cell Biol. 2: 458–460.

Tsuda, M., Y. Sasaoka, M. Kiso, K. Abe, S. Haraguchi et al.,2003 Conserved role of nanos proteins in germ cell develop-ment. Science 301: 1239–1241.

van der Straten, A., C. Rommel, B. Dickson and E. Hafen,1997 The heat shock protein 83 (Hsp83) is required for Raf-mediated signalling in Drosophila. EMBO J. 16: 1961–1969.

Van Doren, M., A. L. Williamson and R. Lehmann, 1998 Regu-lation of zygotic gene expression in Drosophila primordial germcells. Curr. Biol. 8: 243–246.

Vanmuylder, N., A. Werry-Huet, M. Rooze and S. Louryan,2002 Heat shock protein HSP86 expression during mouse em-bryo development, especially in the germ-line. Anat. Embryol.205: 301–306.

Weil, T. T., K. M. Forrest and E. R. Gavis, 2006 Localization of bi-coid mRNA in late oocytes is maintained by continual activetransport. Dev. Cell 11: 251–262.

Yano, T., S. Lopez de Quinto, Y. Matsui, A. Shevchenko and A.Ephrussi, 2004 Hrp48, a Drosophila hnRNPA/B homolog, bindsand regulates translation of oskar mRNA. Dev. Cell 6: 637–648.

Yue, L., T. L. Karr, D. F. Nathan, H. Swift, S. Srinivasan et al.,1999 Genetic analysis of viable Hsp90 alleles reveals a criticalrole in Drosophila spermatogenesis. Genetics 151: 1065–1079.

Zhao, R., M. Davey, Y. C. Hsu, P. Kaplanek, A. Tong et al.,2005 Navigating the chaperone network: an integrative mapof physical and genetic interactions mediated by the hsp90 chap-erone. Cell 120: 715–727.

Communicating editor: W. M. Gelbart

2222 Y. Song et al.