Drosophila rhino Encodes a Female-Specific Chromo-domain Protein That Affects Chromosome Structure and Egg Polarity

Corresponding author: Celeste A. Berg, Department of Genetics, University of Washington, 1959 NE Pacific St., Box 357360, Seattle, WA 98195-7360., berg{at}genetics.washington.edu (E-mail)

Communicating editor: T. SCHÜPBACH

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Here we describe our analyses of Rhino, a novel member of the Heterochromatin Protein 1(HP1) subfamily of chromo box proteins. rhino (rhi) is expressed only in females and chiefly in the germline, thus providing a new tool to dissect the role of chromo-domain proteins in development. Mutations in rhi disrupt eggshell and embryonic patterning and arrest nurse cell nuclei during a stage-specific reorganization of their polyploid chromosomes, a mitotic-like state called the "five-blob" stage. These visible alterations in chromosome structure do not affect polarity by altering transcription of key patterning genes. Expression levels of gurken (grk), oskar (osk), bicoid (bcd), and decapentaplegic (dpp) transcripts are normal, with a slight delay in the appearance of bcd and dpp mRNAs. Mislocalization of grk and osk transcripts, however, suggests a defect in the microtubule reorganization that occurs during the middle stages of oogenesis and determines axial polarity. This defect likely results from aberrant Grk/Egfr signaling at earlier stages, since rhi mutations delay synthesis of Grk protein in germaria and early egg chambers. In addition, Grk protein accumulates in large, actin-caged vesicles near the endoplasmic reticulum of stages 6–10 egg chambers. We propose two hypotheses to explain these results. First, Rhi may play dual roles in oogenesis, independently regulating chromosome compaction in nurse cells at the end of the unique endoreplication cycle 5 and repressing transcription of genes that inhibit Grk synthesis. Thus, loss-of-function mutations arrest nurse cell chromosome reorganization at the five-blob stage and delay production or processing of Grk protein, leading to axial patterning defects. Second, Rhi may regulate chromosome compaction in both nurse cells and oocyte. Loss-of-function mutations block nurse cell nuclear transitions at the five-blob stage and activate checkpoint controls in the oocyte that arrest Grk synthesis and/or inhibit cytoskeletal functions. These functions may involve direct binding of Rhi to chromosomes or may involve indirect effects on pathways controlling these processes.

THE structure of chromatin powerfully influences gene expression and chromosome behavior in many systems (reviewed in SINGH and HUSKISSON 1998 ; PARO et al. 1998 ; STEGER et al. 1998 ). Drosophila oogenesis is an excellent model for examining the role of these processes in development. The egg chamber contains highly visible nurse cell chromosomes that undergo complex, dynamic changes throughout oogenesis. In addition, female-sterile mutations exist that alter nurse cell chromosome morphology, providing insight into the mechanisms that regulate these changes. Most such mutations cause egg chamber degeneration midway through oogenesis; a few, however, allow maturation of late-stage egg chambers that frequently exhibit specific eggshell patterning defects (reviewed in SPRADLING 1993 ). Here we describe rhino, a new female-sterile gene required for normal nurse cell chromosome structural reorganization and for establishing egg polarity.

The Drosophila ovary consists of ovarioles where egg chambers develop in an assembly-line-like process. Each egg chamber contains 16 interconnected germline cells surrounded by a layer of somatically derived follicle cells. The first germline-derived cell becomes the oocyte while the other 15 become nurse cells. Early in oogenesis, nurse cell chromosomes undergo endoreplication, increasing in ploidy. During this time, homologous chromosomes remain paired. By stage four (S4), when the DNA content is 32C, the polytene chromosomes show visible banding patterns. As endoreplication continues, homolog pairing weakens so that by S5, banding is lost but five distinct chromosome arms are still observable. This stage, called the five-blob stage, is transient. By S6, and for the rest of oogenesis, the chromosomes are diffuse and uniformly distributed throughout the nucleus. The nurse cell chromosomes continue endoreplicating until mid-oogenesis. By S10, the largest, most posterior nurse cells attain a DNA content of 1000C (KING 1970 ; HAMMOND and LAIRD 1985 ; DEJ and SPRADLING 1999 ).

Some female-sterile mutations may provide clues to the mechanisms that regulate these dynamic changes in chromosome morphology and link these events to other aspects of egg chamber development, such as cell-cycle regulation, meiosis, eggshell synthesis, and the establishment of egg polarity. Certain alleles of the genes ovarian tumor (otu), suppressor of Hairy wing (su[Hw]), string of pearls (sop), female sterile of Bridges (fs(2)B), fs(2)cup (cup), and morula (mr) produce egg chambers in which nurse cell polytene chromosomes persist well past S4 (KING 1970 ; HEINO 1989 ; SPRADLING 1993 ; CRAMTON and LASKI 1994 ; HEINO et al. 1995 ; REED and ORR-WEAVER 1997 ). For example, mutations in sop arrest oogenesis at S5, when pairing has begun to break down but five distinct chromosome arms are still visible in the nurse cells. Prior to degeneration, several consecutive S5 egg chambers are present within each ovariole, leading to the name string of pearls (CRAMTON and LASKI 1994 ). sop encodes a ribosomal protein, but how this gene product affects chromosome structure is unclear. Similarly, elegant genetic studies reveal an interaction between otu and cup (KEYES and SPRADLING 1997 ), but molecular analyses do not provide a clear mechanism for their effect on chromosome morphology, since both genes encode novel cytoplasmic proteins (STEINHAUER et al. 1989 ; KEYES and SPRADLING 1997 ). In contrast, studies on morula reveal a direct role in cell-cycle regulation and predict a distinct mitotic endo cell cycle at S5 (REED and ORR-WEAVER 1997 ), a prediction borne out by work of DEJ and SPRADLING 1999 . Mutations in morula alter cell-cycle controls, allowing a mitotic-like state with condensed chromosomes and the formation of spindles (REED and ORR-WEAVER 1997 ). Unlike mutations in otu and cup, however, these defects do not alter egg polarity (T. L. ORR-WEAVER, personal communication).

Other mutations that affect nurse cell development or number eventually lead to egg chamber degeneration in which abnormal nuclear morphology, including highly condensed chromatin, may be observed (reviewed by SPRADLING 1993 ). Some such degenerating mutations are not fully penetrant and allow production of a few late-stage egg chambers that exhibit specific patterning defects. For example, abnormal chromosome morphology and eggshell defects are sometimes seen in Bicaudal-C (Bic-C), pipsqueak (psq), and Dp mutants (SIEGEL et al. 1993 ; SPRADLING 1993 ; MYSTER et al. 2000 ). SAFFMAN et al. 1998 propose a model in which Bic-C affects translation through its activity as an RNA binding protein. In contrast, psq encodes a protein with a BTB/POZ domain, a motif found in transcriptional activators (HOROWITZ and BERG 1996 ; LEHMANN et al. 1998 ), while Dp encodes a subunit of the E2F transcription factor (reviewed by LA THANGUE 1994 ).

How do these mutations affect patterning? Most such mutations disrupt the synthesis or localization of Gurken (Grk), a TGF-like molecule that plays a key role in establishing anterior-posterior and dorsoventral polarity in the egg and embryo (reviewed in NILSON and SCHUPBACH 1999 ). Early in oogenesis grk mRNA is localized to the posterior of the oocyte. Signaling to posterior follicle cells through the epidermal growth factor receptor (Egfr) pathway determines posterior follicle cell fate. At S7, a reciprocal signal from posterior follicle cells to the oocyte induces a microtubule rearrangement that leads to the correct localization of axis determinants: oskar (osk) at the posterior, bicoid (bcd) at the anterior, and grk in a dorsal anterior cap above the oocyte nucleus. Grk then participates in a second signaling pathway to the dorsal follicle cells, inducing a patterning process that establishes dorsal cell fates in the eggshell and eventually leads to correct ventral cell fates in the embryo. The eggshell patterning process involves an autocrine mechanism in which Argos mediates inhibition of Egfr activity in a midline domain between two populations of dorsal follicle cells (reviewed in STEVENS 1998 ). Since the dorsal follicle cells respond to the signaling cascade by migrating and secreting the chorion that forms the dorsal respiratory appendages of the eggshell, defects in dorsal-ventral determination can be readily observed by examining eggshell phenotypes.

Here we describe rhino (rhi), a female-sterile mutant noted for its eggshell dorsal appendage defects. Our studies show that rhi mutants have an aberrant nurse cell chromosome structure and that grk and osk transcripts are mislocalized. Moreover, Grk protein first accumulates slowly but later is present in high amounts in the oocyte in large vesicles near the endoplasmic reticulum. We have cloned rhi and found that it encodes a protein with homology to members of the chromo-domain family of proteins. The best-characterized members of this family (e.g., HP1, reviewed by EISSENBERG and ELGIN 2000 ) interact with other proteins in the formation of a heritable chromatin domain that represses gene transcription. Some members of this family are important for centromere function and also anchor chromatin to the inner nuclear membrane (reviewed in CAVALLI and PARO 1998 ). rhi is required for a specific developmental transition in chromosome reorganization and coordinates these nuclear morphological changes with egg chamber maturation and axis formation.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Fly stocks:
Canton-S and cn; ry⁵⁰⁶ were used as wild-type controls. rhi¹ [originally called fs(2)ry1], was generated in a P[ry⁺] screen described in BERG and SPRADLING 1991 and is carried as a cn rhi¹/CyO, cn²; ry⁵⁰⁶ stock. rhi², formerly rhi²⁰⁸⁶, was generated in a P[lacZ, ry⁺] mutagenesis screen described in KARPEN and SPRADLING 1992 and is carried as a cn rhi²/CyO, cn²; ry⁵⁰⁶ stock. Fly stocks were maintained at 24° on standard cornmeal/yeast/molasses medium.

Excision screens:
Excision alleles were generated as described previously (HOROWITZ and BERG 1995 ) except that background mutations were eliminated by employing isogenized lines generated after outcrossing to cn; ry⁵⁰⁶ flies, the starting strain for both original mutagenesis screens. Several screens yielded 169 independent ry^- lines consisting of 7 fertile lines, 114 sterile lines, and 48 lethal or semilethal lines. This frequency of fertile revertants is unusually low and may be due to the insertion of both P elements in coding regions (see RESULTS). Presumably, perfect double-strand break repair must occur with the wild-type homolog to generate precise excisions and restore a valid open reading frame (ENGELS et al. 1990 ).

The extent of the deletions in nine female-sterile lines (four derived from rhi¹ and five derived from rhi²) was mapped at the molecular level. Six contained internal P-element deletions, including one line, rhi^2-S17, in which virtually all of the P element had been deleted, leaving only 50 bp of 5' P-element end. In three lines, the ends of the P element and the flanking DNA remained intact (data not shown); in these cases, a small internal deletion was inferred by the ry phenotype. Forty-eight lines were either lethal or semilethal and nine were chosen for complementation analysis. Four excision alleles of the PZ element (rhi^2-L12, rhi^2-L13, rhi^2-sL14, and rhi^2-sL15) belonged to the same complementation group. Southern blot analyses of DNA from rhi^2-L13 and rhi^2-sL15 revealed that these lines retain intact P-element ends, with no apparent loss of flanking DNA detected 12 kb 5' and 8.5 kb 3' to the insertion site. These results, along with transcript mapping and expression data (see below), suggest that the lethality is not due to disruption of the rhi gene. We speculate that a nearby essential gene contains a hotspot for P-element insertion and local hopping into this gene is responsible for the lethality.

4',6-Diamidino-2-phenylindole and Sytox green staining of ovaries:
Ovaries were dissected in phosphate-buffered saline (PBS, 130 mM NaCl, 7 mM Na₂HPO₄, 3 mM NaH₂PO₄, pH 7.0) and fixed for 20 min at room temperature in 4% paraformaldehyde in PBTE (PBS plus 0.2% Tween-20, 1 mM EDTA). Fixed ovaries were rinsed once with PBTE, permeabilized in PBTE plus 1% TRITON X-100 for 1 hr at room temperature and rinsed again. They were then stained with 0.2 µg/ml 4',6-diamidino-2-phenylindole (DAPI) in PBTE for at least 30 min at room temperature, rinsed several times, and mounted in PBTE plus 50% glycerol. Microscopy was carried out with a Nikon Microphot FXA microscope; photographs were digitally scanned and then manipulated in Adobe Photoshop.

For confocal microscopy, egg chambers were dissected in PBS and fixed as above in PBTE/4% paraformaldehyde. Fixed egg chambers were then washed three times for 5 min in TBST (25 mM Tris, 140 mM NaCl, 2.6 mM KCl, 0.2% Tween-20, pH 7.4). Sytox green (Molecular Probes, Eugene, OR) was added to a final concentration of 20 µM in TBST, incubated at room temperature for 1 hr, and washed five times 5 min with TBST. Egg chambers were mounted in 50% glycerol/TBST plus Vectashield antifade reagent (Vector, Burlingame, CA) and examined on a Bio-Rad (Hercules, CA) MRC600 laser scanning confocal microscope. Images were transferred to Adobe Photoshop and adjusted for optimal brightness and contrast.

Whole mount in situ hybridization:
cDNA in situ hybridizations were carried out as described previously (GILLESPIE and BERG 1995 ). Digoxigenin probes were prepared using a Boehringer Mannheim (Indianapolis) DNA labeling and detection kit. To prepare probes, we used a rhi cDNA (this article), a grk cDNA (provided by Trudi Schüpbach), a dpp cDNA (provided by Rick Fehon), an osk cDNA (provided by Ruth Lehmann), and a bcd cDNA (provided by Markus Noll). To allow direct comparison of expression levels in wild type and mutant, we set up equivalent conditions within these samples by optimizing several aspects of the protocol. Newly eclosed females were aged 2 days on wet yeast to ensure a distribution of stages similar to wild type. An equal volume of tissue was fixed and all samples were treated identically throughout the procedure. Equimolar amounts of probes were used, resulting in 5-min staining reactions for grk and osk, 15 min for bcd, and 60 min for dpp.

Immunohistochemistry:
Analysis of Grk protein levels and distribution was carried out as described by QUEENAN et al. 1999 . Monoclonal anti-Grk antibody 1D12 was provided by Trudi Schüpbach and used at a dilution of 1:10. Alexa-488 conjugated anti-mouse secondary antibody (Molecular Probes) was used at a dilution of 1:500. Ovaries were triple stained with rhodamine-phalloidin (2 units/ml) and DAPI (0.2 µg/ml) and examined using a Nikon Microphot FXA. Confocal images were collected using a Bio-Rad MRC600 microscope.

Molecular characterization of rhino:
Cloning DNA flanking the P elements: Genomic DNA flanking the ry11 P element in rhi¹ was cloned as follows: Total genomic DNA from rhi¹ adults was cut to completion with BamHI, ligated into -arms (Stratagene, La Jolla, CA), and packaged using the Gigapack Gold commercial extract from Stratagene. The resulting library was screened with a probe to the 5' P-element end and phage containing an insert composed of P-element sequences plus 1.8 kb of genomic DNA was isolated. The entire 5.6-kb insert was subcloned into pBluescript KS(+) (Stratagene) to create pCG5'-5.

Genomic DNA flanking the PZ element in rhi² was cloned by plasmid rescue (ASHBURNER 1989 ) using the restriction enzymes XbaI and NheI. The resulting plasmid, pR1, contains 2.3 kb of genomic DNA flanking the 5' end of the PZ insertion.

The sites of P-element insertion for rhi¹ and rhi² were determined by sequencing pCG5'-5 and pR1 using the 5' P-element primer IRXb: 5'-GCTCTAGACGGGACCACCTTATGT-3' and comparing the resulting sequence to rhi cDNA and genomic sequence (see below).

Isolation and characterization of rhino cDNA and genomic clones: Seven rhi cDNAs of 1.6 kb length were isolated from an ovarian gt22 library (gift of Peter Tolias, Public Health Research Institute, New York; see STROUMBAKIS et al. 1994 ) by screening with the 1.8-kb HindIII genomic fragment from clone pCG5'-5. All had similar restriction maps; one was subcloned into pBluescript KS(+) to create pBSB2.

pBSB2 was sequenced by the dideoxy chain termination method using reagents supplied by United States Biochemical (Cleveland). Sequence data were compiled and organized using the IntelliGenetics program. BLASTp from National Institutes of Health (ALTSCHUL et al. 1990 ) and Genetics Computer Group (University of Wisconsin, Madison, WI) were used for database searches and alignments of homologous protein sequences. The rhi cDNA matches the Berkeley Drosophila Genome Project gene CG10683.

The cosmid 13H6, which maps to 54C, was obtained from the Drosophila Genome Mapping Project, courtesy of Dr. Inga Siden Kiamos (Foundation for Research and Technology–Hellas, Crete, Greece; see SIDEN-KIAMOS et al. 1990 and KAFATOS et al. 1991 ). Southern hybridization demonstrated that sequences homologous to the rhi cDNA were present within this genomic clone. A 6.5-kb EcoRI fragment that hybridized to the rhi cDNA was subcloned from cosmid 13H6 into pCaSpeR4 to create pHH76. Sequence analysis of this subclone permitted identification of the two introns within the rhi gene. This clone and cosmid 13H6 do not contain the final (third) exon of rhi. The absence of other introns within this final exon was established by its identification as one contiguous stretch of sequence within the bacterial artificial chromosome (BAC) clone, BACR12023 (accession no. 007697) of the Berkeley Drosophila Genome Project (http://flybase.bio.indiana.edu).

Analysis of deletions in rhino excision alleles: A genomic restriction map for the rhi locus was deduced from restriction digest analysis of the cosmid 13H6 and Southern analysis of total fly DNA isolated from wild-type and rhi mutant flies. DNA deletions present within the rhi excision alleles were mapped by Southern analysis of total fly DNA using rhi and P-element probes.

Northern blot analysis: RNA was prepared from adult flies, ovaries, and all developmental stages by the hot phenol method (JOWETT 1986 ). Hybridization was performed as described in GILLESPIE and BERG 1995 . All probes were labeled by the random hexamer-primed method (FEINBERG and VOGELSTEIN 1983 ).

The GenBank accession number for the Drosophila melanogaster rhino cDNA is AF411862.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Isolation of rhino mutations:
rhi¹ and rhi² are female-sterile mutations that were isolated in two independent, large-scale P-element mutagenesis screens. rhi¹ contains an insertion of a ry11 element and rhi² contains an insertion of a P[lacZ, ry⁺] (PZ) element (BERG and SPRADLING 1991 ; KARPEN and SPRADLING 1992 ). Both P elements map to the 54C/D border by in situ hybridization (data not shown; FlyBase places this locus at 54D5). Females homozygous for rhi¹ laid fewer eggs than wild-type flies do. Almost all laid eggs exhibited dorsal appendage defects (Fig 1; Table 1), and many had the single dorsal appendage for which rhino is named. rhi² homozygotes laid more eggs than did rhi¹ flies and almost half looked wild type. rhi² mutants also exhibited a mild dumpless phenotype; in some egg chambers, the nurse cells failed to transfer their contents into the oocyte at S11 (for review see COOLEY and THEURKAUF 1994 ). Finally, DAPI staining and other lines of evidence suggested that eggs laid by rhi mutant females either are not fertilized or arrest early in embryonic development (data not shown).

View larger version (66K): In this window In a new window Download PPT slide   Figure 1. Eggshells from rhino females display a range of dorsal-ventral defects. All alleles produce each phenotype although frequencies differ depending on allele strength. Anterior is to the left and dorsal is facing out of the page. (a) A wild-type S14 egg chamber has two dorsal appendages equidistant from the dorsal midline. (b) A weak rhino phenotype in which the appendages are fused at their bases only. (c) A forked dorsal appendage is typical of a ventralized eggshell. (d) Stronger allelic combinations produce partially dorsalized, short eggs; the appendages are broadly fused on the midline and carry extra spade-like material at the tip. (e) A large paddle of chorionic appendage material is observed here atop nurse cell remnants. (f) A strong dumpless phenotype in which shortened dorsal appendages (arrow) extend out over nurse cells, which failed to transfer their cytoplasmic contents into the oocyte and then undergo apoptosis.

We generated new rhi alleles through transposase-induced excision of each P element, using 2-3 transposase (ROBERTSON et al. 1988 ) and scoring for loss of the ry⁺ eye-color marker. This approach also allowed isolation of deficiencies for the region, since no deletions had been described. Multiple excision screens resulted in 169 independent ry^- lines. Precise excision of the P element produced 7 fertile revertant lines, indicating that the female sterility of the starting lines was indeed caused by the insertions. Most ry^- excision lines (114) were female sterile, however, and exhibited eggshell phenotypes similar to those of their respective starting lines. Molecular analyses of a subset of these lines revealed that imprecise excision had left a portion of the P element within the rhi gene. Two of these female-sterile excision lines, rhi^1-S6 and rhi^2-S18, are described in more detail here (Table 1). The remaining 48 lines were either lethal or semilethal. Complementation testing and Southern blot analyses of a subset of these lines, together with transcript mapping and expression data (see below), suggested that the lethality was not due to disruption of the rhi gene. One possibility is that a nearby essential gene contains a hotspot for P-element insertion and local hopping into this gene is responsible for the lethality. Finally, one lethal excision line, rhi^2-L12, contains a deletion of genomic DNA extending at least 11 kb from the insertion point, removing the presumptive start site for rhi transcription (see below) as well as a putative neighboring gene of unknown function, CG18186. Thus, rhi^2-L12 is deficient for the region and can be considered null for rhi function. We rename this deletion Df(2R)rhi^2-L12.

Mutations in rhino disrupt dorsal appendage structures of the eggshell:
Eggs laid by rhi mutants displayed a range of late-stage eggshell phenotypes (Fig 1; Table 1). A wild-type S14 egg has two dorsal anterior respiratory appendages that are equidistant from the dorsal midline (Fig 1A). rhi¹ females laid a few wild-type eggs (5%) but most eggs had a single dorsal appendage fused on the dorsal midline. Some eggs carried two appendages that emanated from one base; such phenotypes are produced by weak ventralizing mutations. Other eggs had a single dorsal appendage with extra appendage material; the eggs themselves were shorter than wild-type eggs. These characteristics are produced by dorsalizing mutations. This variability in patterning defects typifies mutations that disrupt the synthesis or distribution of Grk; small changes in the concentration of morphogen lead to dramatic differences in eggshell structures (ROTH and SCHUPBACH 1994 ; GHABRIAL et al. 1998 ). Eggs laid by rhi² mothers also displayed diverse eggshell phenotypes but the range of phenotypes clearly differed from that caused by the rhi¹ mutation. Fifty percent of eggs laid by rhi² mothers were indistinguishable from wild-type eggs and 11% exhibited some degree of incomplete cytoplasmic transfer from the nurse cells. Dorsal appendage defects observed in rhi² eggs again varied greatly, resembling those of rhi¹ eggs. Excision alleles and heteroallelic combinations displayed highly variable eggshell phenotypes similar to those seen in the starting lines; examples are shown in Fig 1. Most alleles also produced a variable number of dumpless egg chambers (Table 1). Fig 1B shows a weakly ventralized eggshell with two appendages fused at their bases; forked dorsal appendages were also common (Fig 1C). Appendages that were fused along their entire length usually displayed a spade-like tip (Fig 1D), unusual for most ventralizing mutants but typical for rhi and similar to defects produced by gain-of-function alleles of psq (HOROWITZ and BERG 1996 ). Excess chorion material on a single broad dorsal appendage coupled with a truncated egg shape was also observed and represents a weakly dorsalizing phenotype (Fig 1E). Some eggs failed to transfer their nurse cell contents into the oocyte, due either to a direct dumping defect or to premature migration of centripetal follicle cells. These eggs exhibited remnants of degenerating nurse cells (Fig 1E and Fig F). Rarely, we observed egg chambers with two micropyles (data not shown). Finally, females carrying any allele in trans to Df(2R)rhi^2-L12 laid few or no eggs. Dissected ovaries revealed a higher incidence of degenerating egg chambers and a higher frequency of strong eggshell phenotypes compared with phenotypes produced by homozygotes or mutants carrying various heteroallelic combinations (data not shown). The increased severity of phenotypes in hemizygous animals demonstrates that these P-element mutations are not null.

These data suggest an allelic series in which rhi² and sterile rhi² excision alleles are weaker than rhi¹ and its sterile derivatives. Further, all P alleles provide more function than the deficiency chromosome. Interestingly, all P alleles also produce a range of eggshell phenotypes. This variability could be due to the molecular mechanism that allows some function from these P alleles or to a partial redundancy in the process in which Rhi functions.

Molecular structure and expression of the rhino gene:
To characterize the rhi gene (Fig 2A), we cloned DNA sequences flanking the 5' ends of the two P-element insertions. We used the 1.8 kb of DNA adjacent to ry11 in rhi¹ to probe a Northern blot of RNA from wild-type females, wild-type males, and rhi¹ females. In wild-type females, a highly abundant 1.6-kb transcript was present and enriched in ovaries. This same transcript was not detected in wild-type males nor in rhi¹ females (data not shown). We refer to this 1.6-kb transcript as the rhi transcript. We observed a less abundant >9-kb transcript in RNA from all three sources. Using this same genomic probe, we isolated a 1.6-kb cDNA from an ovarian cDNA library (STROUMBAKIS et al. 1994 ). When this cDNA was used to probe Northern blots (Fig 2B), it hybridized to the same 1.6-kb ovary-enriched transcript and to a higher molecular weight transcript of >9 kb present in wild-type female and male RNA. We detected this large message in all mutants tested; the P elements do not appear to disrupt this transcript. Although the 1.6-kb transcript was highly enriched in ovaries, some rhi transcript was also observed in RNA prepared from osk³⁰¹ flies, which lack a germline (EPHRUSSI et al. 1991 ). By this criterion, rhi is expressed in both germline and somatic cells.

View larger version (49K): In this window In a new window Download PPT slide   Figure 2. Molecular and expression analyses of rhino. (A) Molecular map of rhi locus based on genomic blots of rhi1 and rhi2 flies and DNA from cosmid 13H6 (EDGP). The rhi transcript is indicated by boxes connected with thin lines; the hatched regions encode protein, the gray box represents the chromo domain, the black box represents the chromo-shadow domain. The P elements inserted in rhi1 and rhi2, shown above the line that indicates genomic DNA, are not diagrammed to scale. pCG5'-5 denotes a plasmid containing DNA flanking rhi1; pR1 contains DNA flanking rhi2. pHH76 is a subclone made from cosmid 13H6 (not shown). The deletion in Df(2R)rhi2-L12 is indicated by thin lines; a notch marks the 3' deletion endpoint; the 5' endpoint is unknown but resides at least 11 kb 5' to the region shown here. Restriction sites correspond only to the genomic map: E, EcoRI; B, BamHI; N, NheI. (B) Northern blot probed with the 1.6-kb rhi cDNA. Lanes contain 35 µg of total RNA. The 1.6-kb rhi transcript is female specific and is highly enriched in the ovary (solid arrow). osk301 females have no germline and produce very low levels of the 1.6-kb transcript. This transcript is apparently absent in RNA from rhi homozygous females. Low levels of truncated transcripts are visible in rhi2 and rhi2-sl15 females (open arrows). These truncated transcripts also hybridize to sequences present in the l(3)S12 gene (data not shown). A >9-kb transcript is not affected in any rhi mutants (gray arrow). (C) The same blot reprobed with the ribosomal gene rpS5 as a loading control. (D) rhi cDNA used as a probe to wild-type ovaries. The rhi transcript is present at low levels in the germarium (not shown) and begins to accumulate in the oocyte at S5 (arrowhead). rhi mRNA is localized to the posterior of the oocyte at S9 (arrow) but is no longer localized by S10. At S10B, rhi transcription is strongly induced in the nurse cells (open arrowhead). (E) Developmental Northern blot probed with the rhi cDNA. The 1.6-kb transcript is indicated with a solid arrow. It is abundant in ovaries and its level gradually tapers off throughout embryogenesis. The >9-kb transcript is predominant in late embryogenesis (open arrow). (F) The same blot reprobed with the ribosomal gene rp49 as a loading control.

The rhi transcript was not detected in RNA prepared from rhi¹ and rhi² homozygous females (Fig 2B) nor in RNA from flies carrying any of the sterile excision alleles. rhi² and rhi^2-sL15 females produced two lower molecular weight transcripts of 1.3 and 0.5 kb; these messages were observed with long exposures and appeared somewhat heterogeneous in length. These messages also hybridized to a probe for the putative 1(3)S12 gene (data not shown), the 3' end of which is contained in the PZ element (HOROWITZ and BERG 1995 ). These RNAs most likely are truncated rhi transcripts that terminate in and/or splice into the l(3) S12 gene present in the PZ element. Consistent with this hypothesis, the rhi^2-S17 mutation deletes the l(3)S12 sequences and RNA from flies carrying this allele does not exhibit the 1.3- and 0.5-kb transcripts.

To determine when and where rhi is expressed during oogenesis, we performed in situ hybridizations to ovaries using a digoxigenin-labeled rhi cDNA probe (Fig 2D). rhi transcript is first detectable at low levels in region I of the germarium, where it is localized perinuclearly in the germ cells (data not shown). By S5, rhi transcript accumulates in the oocyte and is later found at the posterior of S8–10A egg chambers (Fig 2D). By S10B, this posterior localization disappears. Finally, transcript levels increase dramatically in the nurse cells at S10 (Fig 2D) and rhi mRNA is loaded into the oocyte as maternal message (Fig 2E, 0–2 hr lane). We also performed Northern analysis on RNA isolated from embryonic stages (Fig 2E). The 1.6-kb rhi transcript was present in high levels in ovaries and in 0- to 2-hr embryos and then tapered off to low but detectable levels later in embryogenesis. This expression profile is typical for maternal transcripts that are deposited into the embryo.

rhino (CG10683) encodes a novel member of the chromo-domain family:
Sequence analysis of the 1.6-kb cDNA revealed a single open reading frame that on conceptual translation encodes a protein of 418 amino acids with a predicted molecular weight of 46 kD (Fig 3). The cDNA possesses a short 5' untranslated region (UTR) containing an in-frame stop codon just upstream of the predicted translational start site. A nuclear localization signal is present in the C-terminal portion of the protein. The rhi cDNA matches the Berkeley Drosophila Genome Project gene CG10683.

View larger version (69K): In this window In a new window Download PPT slide   Figure 3. Sequence of rhino cDNA and predicted sequence of Rhino protein. Numbers correspond to nucleotide sequence. Inverted triangles indicate the insertion points of the two P elements. The chromo domain is boxed and the chromo-shadow domain is underlined. The intron positions are indicated by T-bars. A putative nuclear localization signal is shown in boldface. The polyadenylation signal is both underlined and in boldface type.

Database searches performed with the BLASTp and GCG programs revealed that Rhi contains a chromo domain (chromatin organization modifier; PARO and HOGNESS 1991 ), a motif found in proteins involved in transcriptional repression via gross chromatin reorganization (reviewed in CAVALLI and PARO 1998 ). At least 70 members of the chromo-domain family have been discovered to date in many organisms, including yeast and humans. Two general types of chromo-domain motif subdivide this family into two classes, HP1-like and Pc-like, named after two Drosophila family members Heterochromatin Protein 1 (HP1) and Polycomb (Pc). HP1 binds to heterochromatin and is involved in position-effect variegation while Pc heritably silences transcription of the homeotic genes (LEWIS 1978 ; JAMES and ELGIN 1986 ; PARO and HOGNESS 1991 ). Rhi is a member of the HP1-like class (Fig 4). Across the 40-amino-acid chromo domain, Rhi is 48% identical and 73% similar at the amino acid level to its closest homologs, the murine and human HP1 proteins (also called CBX5; SAUNDERS et al. 1993 ; LE DOUARIN et al. 1996 ). Slightly less similarity in this region (68%) is found when Rhi is aligned with a human protein, p25ß (also called CBX1; SINGH et al. 1991 ). Rhino's chromo domain shares less homology with analogous regions in Polycomb and Polycomb-homologous proteins.

View larger version (48K): In this window In a new window Download PPT slide   Figure 4. Alignments of the Rhino chromo domain and chromo-shadow domain with the closest chromo-domain-family homologs. (A) Chromo domain. (B) Chromo-shadow domain. Numbers correspond to amino acid sequence. Identical amino acids are shaded in dark gray; similarities are indicated in light gray. Homologs depicted are: murine HP1 (CBX5), human HP1 (CBX5), human p25ß (CBX1), murine MoMOD1 and MoMOD2 (CBX3), human HPIhs, Drosophila HP1, human Y chromosome CDY, and Tetrahymena Pddp1. Dm, D. melanogaster; Mm, Mus musculus; Hs, Homo sapiens; Tt, Tetrahymena thermophila.

Rhino also shares a region of homology at its carboxyl terminus with a subset of the HP1-like chromo-domain proteins. This C-terminal motif has been named the "chromo shadow domain," since it has a low level of homology to the chromo domain (AASLAND and STEWART 1995 ; KOONIN et al. 1995 ). The shadow domain is found at the C-terminal ends of Drosophila melanogaster and D. virilis HP1 and in two human, two murine, one Xenopus, and one chicken HP1 homologs (SINGH et al. 1991 ; SAUNDERS et al. 1993 ). Although significant homology between Rhi and other subfamily members exists in the chromo-shadow domain, Rhi sequence diverges to a greater extent (Fig 4B). Finally, the central region of Rhi contains weak homology to mouse and human HP1 chromo-domain proteins (CBX5), suggesting a more recent evolutionary relationship with these proteins.

We determined the positions of the two P elements relative to the rhi coding region by sequencing the flanking DNA cloned from the rhi¹ and rhi² flies (Fig 2A and Fig 3). The P element in rhi¹ is inserted within the coding region at nucleotide 55 of the cDNA, 10 amino acids downstream from the conceptual start site of the protein. Since rhi¹ hemizygotes produce a phenotype more severe than that of rhi¹ homozygotes, we speculate that rare splicing events produce low quantities of a functional Rhi protein. Thus, rhi¹ is a strong hypomorphic mutation of the gene. In contrast, the PZ element in rhi² is inserted 81 amino acids from the protein start site, potentially permitting translation of the entire chromo domain and thereby producing a partially functional protein. Chromo domains are involved in protein:protein interactions and are included in large multiprotein complexes (SINGH et al. 1991 ; COWELL and AUSTIN 1997 ; HUANG et al. 1998 ; YAMADA et al. 1999 ); thus, a truncated Rhino protein potentially present in rhi² flies could induce a mutant phenotype by interfering with proper complex formation. Since rhi² alleles overall produce phenotypes weaker than those of rhi¹ alleles, however, we favor an alternative hypothesis, that the Rhi chromo domain alone provides some aspects of normal Rhi function.

Mutations in rhino affect chromosome structure or organization in the nurse cell nucleus:
rhi encodes a putative chromatin-binding protein. We therefore stained ovaries with the DNA dyes DAPI or Sytox green to ask if mutations in rhi produce visible effects on chromosome structure. We used females hemizygous for each rhi allele to decrease the gene dose and to eliminate the contributions of potential background mutations. Although some ovariole degeneration was seen in these heteroallelic combinations, including loss of germline stem cells and their derivatives, the most striking feature of the egg chambers was their aberrant nurse cell chromosome configuration (Table 2; Fig 5). Normally, a specific progression of changes in higher-order chromosome structure occurs during oogenesis (Fig 5A; KING 1970 ; HAMMOND and LAIRD 1985 ; DEJ and SPRADLING 1999 ). Early endoreplicating nurse cell chromosomes are paired and as the DNA content increases, a polytene banding pattern becomes visible. At S5, pairing breaks down and the chromosomes assume a stereotypic five-lobed or wagon-wheel morphology (Fig 5D). This morphology is transient, however, and after S5 the chromosomes assume a fairly uniform distribution throughout the nucleus (Fig 5G and Fig J). The nurse cell chromosomes then retain this morphology for the rest of oogenesis.

View larger version (135K): In this window In a new window Download PPT slide   Figure 5. rhino mutants exhibit altered chromosomal morphology. (a, d, g, and j) Wild-type ovarioles containing a normal five-lobed S5 egg chamber (arrow, a, enlarged in d), and later-staged egg chambers with uniformly distributed chromatin (a and g). A magnified view of early S10 nurse cell nuclei is shown in j. (b, e, h, and k) Similarly staged egg chambers from rhi1/Df(2R)rhi2L12 females, exhibiting a five-lobed chromosome morphology in S5 and persisting into later stages (arrows, b, enlarged in e). Note the close association of the chromatin with the nuclear periphery in S10 nurse cells (k), leaving a hole in the center of the nucleus. (c and f) The five-lobed phenotype also persists in rhi2/Df(2R)rhi2L21 egg chambers (arrows), but the chromatin lobes are slightly larger and more evenly distributed in the later stages compared to rhi1/Df(2R)rhi2L12 egg chambers. (i) rhi2-SL15/Df(2R)rhi2L12 ovarioles contain egg chambers with abnormal numbers of germline-derived cells (arrow); similar egg chambers were observed in other rhi allelic combinations. (l) Nurse cell chromosomes in rhi1-S6/Df(2R)rhi2L1 females break into fragments before degeneration. Regions enclosed by boxes in a–c are expanded in d–f. Bars, 25 µm.

In rhi mutants, chromosome structure was normal up to S5. In egg chambers from each of the rhi alleles, however, the wagon-wheel morphology persisted well after this stage; the chromosomes did not assume a uniform distribution. Nevertheless, DNA replication apparently continued in these nuclei since the intensity of DAPI staining increased in older egg chambers (Fig 5B and Fig C).

Females carrying alleles derived from the rhi¹ P element (rhi¹ and rhi^1-S6) produced egg chambers with a distinctive five-lobed phenotype (Fig 5B and Fig E). In S10 egg chambers from rhi¹/Df(2R)rhi^2L12 females, the lobes of chromosomes were at the nuclear periphery, leaving a central "hole" that lacked DAPI staining (Fig 5H and Fig K). Alleles generated from the rhi² P-element insertion (rhi² and rhi^2-SL15) also resulted in the persistent wagon-wheel phenotype (Fig 5C, Fig F, and Fig I). The chromosome lobes were larger and more broadly distributed throughout the nucleus, however, compared to chromosome defects produced by the rhi¹-based alleles (Fig 5B and Fig C).

Females hemizygous for each of the rhi alleles occasionally produced egg chambers with improper numbers of nurse cells (Table 2 and Fig 5I). These egg chambers usually contained 32 cells, indicative of either an extra round of mitosis or a follicle-cell encapsulation event involving two germline-derived cysts. Less frequently, egg chambers were produced that had too few nurse cells. In older females (10 days), germaria usually contained fewer numbers of germline cysts. Ultimately, most egg chambers hemizygous for the various rhi alleles degenerated before completing oogenesis. Notably, chromosomes in rhi^1-S6/Df(2R)rhi^2L12 nuclei broke up into unusually small and numerous fragments before degenerating completely (Fig 5L). The morphology of the chromosomes in the oocyte nucleus and follicle-cell nuclei was not visibly altered in any of the rhi mutants (data not shown). These data reveal the need for Rhi in restructuring or reorganizing nurse cell chromosomes during the unique S5 endoreplication cycle. It is possible that rhi mutations affect chromosome structure in other ovarian cell types but defects are not detectable in the mutants due to the lower ploidy of these cells.

rhino mutations affect transcripts that are important for early axis establishment:
The similarity of Rhi to chromo-domain proteins and the aberrant nurse cell chromosome morphology and eggshell phenotypes produced by rhi alleles suggested three hypotheses for the role of rhi during oogenesis.

First, Rhino could bind specific sites on the chromosome as part of a multiprotein chromatin-binding complex, regulating expression of a variety of genes including key genes required for patterning. In this scenario, loss-of-function mutations would disrupt chromatin conformation and alter expression of the target genes. If Rhi binds a large number of sites on these highly polyploid chromosomes, visible changes in chromosome structure might be observed in rhi mutants. Alternatively, Rhi might be required at the end of endocycle 5 to repress expression of specific genes whose products inhibit chromosome dispersal; loss of Rhi would therefore prevent the normal chromosome structural transitions. At the same time, eggshell defects might result if, for example, Rhino normally represses transcription of gurken in the nurse cells; lack of Rhino would allow overexpression and generate dorsalized egg chambers.

In the second hypothesis, Rhino does not act as a transcriptional regulator but rather participates in chromosome structural changes during specific transitions of the cell cycle, such as the chromosome reorganizational events that take place at S5 of oogenesis. The ordered progression of chromosome modifications that occur during the cell cycle would be monitored and this information then integrated with patterning processes to ensure coordinated egg chamber maturation. Schüpbach and colleagues have described a link between cell-cycle checkpoint functions that occur in the oocyte and pathways that regulate grk mRNA translation (GHABRIAL et al. 1998 ; GHABRIAL and SCHUPBACH 1999 ). This hypothesis would extend that connection by suggesting that cell cycle events in the nurse cells are also integrated with patterning mechanisms. Thus, in this scenario, rhi mutations would lead to defects in S5 chromosome structural transitions and these defects would then affect Grk protein synthesis through putative checkpoint controls. Transcription of most genes, however, would be normal.

Finally, the third hypothesis states that Rhino regulates the synthesis or activity of cytoskeletal or transport proteins that control the distribution of chromatin modifying proteins, nuclear import/export functions and/or microtubule organizing center components. Loss-of-function mutations in rhi would then indirectly lead to defects in nurse cell chromosome conformation and egg chamber polarity. A similar scenario is proposed for cup and otu, which encode cytoplasmic proteins and interact genetically to regulate nurse cell chromosome conformation and overall egg chamber maturation (KEYES and SPRADLING 1997 ). Since rhi mutants exhibit only a subset of the phenotypes produced by mutations in otu or cup, rhi may affect only a specific branch of the overall regulatory process. Nevertheless, this third hypothesis predicts that rhi mutants would exhibit defects in the subcellular distribution of proteins that regulate chromosome structure and mRNAs that determine axial patterning.

To distinguish between these hypotheses, we tested if rhi mutations resulted in gross changes in gene expression for four genes whose expression levels and mRNA localization patterns are required for and/or reveal the establishment of egg polarity.

We first examined grk mRNA because of its primary role in dorsal-ventral patterning. In wild-type ovaries, grk message is localized to the posterior of the oocyte early in oogenesis and then becomes concentrated in a dorsal anterior cap over the oocyte nucleus at S9 and S10 (Fig 6A and NEUMAN-SILBERBERG and SCHUPBACH 1993 ). We observed this pattern of expression in 93.6% of Canton-S egg chambers (n = 218). In contrast, only about one-third of rhi¹/Df(2R)rhi^2-L12 and half of rhi²/Df(2R)rhi^2-L12 egg chambers exhibited this pattern (32.5%, n = 77 and 50.7%, n = 67, respectively). The remaining egg chambers mislocalized the mRNA laterally in early stages (Fig 6B) or in a ring at late stages (Fig 6C) or lacked detectable grk transcripts (not shown). We did not see excess grk mRNA, as expected if Rhi acted to repress grk transcription. In some S13 egg chambers, we did see residual grk mRNA in nurse cell remnants; this late expression could contribute to the variety of unusual dorsal appendage shapes of rhi mutants but is not consistent with a general role in grk transcriptional repression. Finally, we observed defects in grk mRNA localization before S6, the time at which we detected aberrant nurse cell chromosome morphology. Normally, grk mRNA is localized at the posterior of the oocyte in early stages; we observed grk mRNA in lateral or anterior regions of the oocyte (arrows, Fig 6B). Taken together, these results suggest that Rhi does not act as a direct regulator of grk gene expression but rather affects eggshell patterning at least in part by disrupting the processes that localize grk transcripts.

View larger version (115K): In this window In a new window Download PPT slide   Figure 6. Expression patterns in rhino mutants reveal normal transcript levels but some mislocalized mRNAs. Wild-type (a, d, g, and j) and rhi/Df(2R)rhi2-L12 (b, d, e, f, h, i, k, and l) egg chambers probed with grk (a–c), osk (d–f), bcd (g–i), and dpp (j–l) cDNAs. (a) In the wild type, grk mRNA (arrowheads) is localized to the posterior of the oocyte in early egg chambers (first in line), transiently in a ring at S7 or S8 (middle egg chamber), and then in a dorsal anterior cap above the oocyte nucleus by S9 (last in line). (b) In rhi mutants, grk mRNA is frequently mislocalized to the side of the egg chamber in early stages (arrows; compare first egg chamber in a to indicated egg chambers in b). (c) In S10 rhi egg chambers, grk remains localized in a ring at the anterior of the oocyte (arrow). (d) In wild type, osk mRNA is localized to the posterior of the oocyte in early egg chambers, diffusely throughout the oocyte at S7, and then again at the posterior by S9 (arrowhead). (e and f) osk mRNA (arrows) may be found at anterior and posterior (e) or diffuse throughout the oocyte (f) in rhi mutant egg chambers. (g) bcd message (arrowhead) is localized to the anterior cortex by S9 in wild-type egg chambers. (h) In some S9 rhi egg chambers, bcd expression is absent (arrow). (i) By S10, however, rhi egg chambers express bcd normally (arrow). (j) In wild type, dpp (arrowheads) is expressed in anterior follicle cells at S8 and then more dramatically in a ring of centripetally migrating cells at S10. (k and l) In some rhi egg chambers, dpp expression is delayed (arrows, k, indicate lack of expression in early egg chamber yet normal expression in late egg chamber). In most mutant egg chambers, however, dpp expression is normal (arrows, l, highlight expression in both egg chambers).

To test the generality of this result, we examined the expression of three other genes, oskar (osk), bicoid (bcd), and decapentaplegic (dpp). Like grk, osk mRNA is localized to the posterior of the oocyte in early egg chambers. During microtubule reorganization at S7, osk transcripts are uniformly distributed within the oocyte or are transiently localized in an anterior ring. By late S8, however, osk mRNA once again is localized to the posterior (Fig 6D and EPHRUSSI et al. 1991 ; Canton-S 94.5%, n = 71). In ovaries from either rhi¹ or rhi² hemizygous females, only half the egg chambers exhibited this pattern (rhi¹ 49.2%, n = 130; rhi² 52.6%, n = 213). Mislocalization in early stages and diffuse localization into late stages were common phenotypes (Fig 6E and Fig F). Virtually all S8–11 egg chambers exhibited mislocalized osk mRNA. No defects in level of expression were detected. These results reveal a defect in the establishment of posterior polarity within the oocyte and support the hypothesis that rhi mutations disrupt a mRNA localization process common to grk and osk transcripts.

To test if genes normally expressed only after S5 are regulated correctly and to ask if the mislocalization defects were a general phenomenon, we examined expression of bcd, whose mRNA is localized to the anterior of the oocyte beginning at S8 (Fig 6G and ST. JOHNSTON et al. 1989 ; Canton-S 98.2%, n = 218). We observed no defects in early bcd expression in rhi¹ and rhi² hemizygotes but did find some S8 and S9 egg chambers with little or no bcd message, suggesting a delay in the onset of bcd transcription or transport. Later-stage egg chambers either exhibited normal bcd transcript levels and distribution or were degenerating. Overall 69.7% of rhi¹ hemizygous (n = 98) and 82.2% of rhi² hemizygous (n = 135) egg chambers exhibited mid-oogenesis defects. The normal distribution of bcd RNA in late-stage egg chambers demonstrates that some aspects of RNA localization function properly in rhi mutant ovaries. The delay in bcd mRNA expression and localization, however, suggests a defect in the coordination between patterning processes and other aspects of egg chamber differentiation, such as yolk uptake and follicle-cell movements that define particular stages of egg chamber development. Finally, we tested if rhi mutants misregulate expression of dpp. dpp is normally expressed only in the follicle cells and affects eggshell structures by determining anterior follicle cell fate (TWOMBLY et al. 1996 ; DENG and BOWNES 1997 ; XIE and SPRADLING 1998 ; PERI and ROTH 2000 ). Ectopic expression of dpp in the germline or misexpression in the follicle cells could cause altered eggshell morphology in rhi mutants. In wild-type egg chambers, dpp is expressed at very low levels in the germarium and then at more readily detectable levels in anterior follicle cells beginning at S8. This expression resolves into a ring of staining in centripetally migrating cells at S9 and S10 (Fig 6J and TWOMBLY et al. 1996 ; Canton-S 98.4%, n = 126). Almost all rhi¹ and rhi² hemizygous egg chambers exhibited normal dpp expression (Fig 6L; rhi¹ 91.5%, n = 130; rhi² 95.9%, n = 271). A few S8 and S9 egg chambers produced no detectable dpp transcript (S9 egg chamber in Fig 6K). Thus, the eggshell phenotypes we observe are likely due to defects in grk mRNA localization and not to defects in dpp expression.

In summary, these results suggest that Rhi does not affect eggshell patterning by regulating transcription of grk or dpp, key genes required for determining the fate of follicle cells that synthesize eggshell structures. Moreover, it is likely that Rhi does not act as a general transcriptional repressor in oogenesis. Severe hypomorphic alleles still allow development of most egg chambers to S10 and some egg chambers to S14. Although our survey of genes was limited, none of the genes we assayed exhibited altered levels of gene expression. Rather, mutations in rhi disrupt grk and osk mRNA localization in a large fraction of S8–10 egg chambers. Thus, Rhi likely acts upstream in a pathway controlling the microtubule reorganizations prerequisite for establishing axial polarity. In addition, our rhi results contrast with those of Heino and colleagues, who found that mutations in otu disrupt chromosome conformation and lead to aberrant accumulation of bcd and other mRNAs in nurse cell nuclei. They found no defect in osk mRNA localization or in the localization of other transcripts involved in egg and embryonic patterning (HEINO et al. 1995 ). Thus, rhi and otu share some common phenotypes but differ in their mechanism of action.

Gurken protein accumulates slowly and associates with vesicles in rhino mutants:
The grk and osk mRNA localization defects observed in rhi mutants suggested that the S6–S8 oocyte microtubule rearrangement that ensures proper mRNA distribution was defective. This cytoskeletal reorganization is directed by a signal from posterior follicle cells, which rely on earlier Grk signaling to determine their posterior fate (reviewed by NILSON and SCHUPBACH 1999 ). rhi mutations could disrupt this earlier Grk patterning process, leading to microtubule defects as a downstream consequence, or they could affect a S5/S6 regulatory process that integrates nurse cell chromosome transitions with microtubule functions. If acting early in oogenesis, Rhi might be required for proper expression of any gene involved in Grk signaling (e.g., homeless, vasa, cornichon, etc., but not grk) or it might regulate meiotic chromosome structural changes that occur in the oocyte and are then linked to Grk translation (GHABRIAL et al. 1998 ; GHABRIAL and SCHUPBACH 1999 ). To begin to address these possibilities, we asked if Grk protein is expressed correctly in rhi mutant egg chambers.

In wild-type egg chambers, Grk protein is present in a punctate pattern in region II of the germarium and is localized to the oocyte in subsequent stages (Fig 7A; ROTH et al. 1995 ; NEUMAN-SILBERBERG and SCHUPBACH 1996 ). S1 egg chambers frequently exhibit -Grk staining mainly in the oocyte with cortical staining in neighboring nurse cells (Fig 7A). During S2–S5, Grk protein is tightly localized to the posterior of the oocyte (Fig 7A) but becomes diffuse throughout the oocyte during microtubule reorganization at S6 and S7 (Fig 7B). Transient localization of Grk protein in a ring at the anterior of the oocyte may occur in early S8 egg chambers but by S9, all Grk protein is found in a dorsal anterior cap outlining the oocyte nucleus (Fig 7C). Secretion of Grk protein may be monitored by visualization of punctate dots in the overlying follicular epithelium (e.g., Fig 7B; QUEENAN et al. 1999 ). We observed this precise pattern of staining in 89.2% (n = 158) of +/Df(2R)rhi^2-L12 egg chambers (Fig 7, A–C). The remaining egg chambers exhibited subtle defects in the timing at which Grk localization changed, such as S2 egg chambers with remaining nurse cell expression, S6 egg chambers with continued posterior localization and two S9 egg chambers with a diffuse oocyte cortical pattern (data not shown).

View larger version (41K): In this window In a new window Download PPT slide   Figure 7. Gurken protein accumulation is aberrant in rhino mutants. +/Df(2R)rhi2-L12 (A, B, and C), rhi1/Df(2R)rhi2-L12 (D, E, and F), and rhi2/Df(2R)rhi2-L12 (G, H, and I) egg chambers probed with monoclonal antibody 1D12 against Gurken protein (green) and counter-stained with rhodamine-phalloidin to highlight the actin cytoskeleton (red). In all panels, anterior is to the upper left. (A) In the germarium (bracket), Grk protein is detectable in region II, accumulating in the oocyte and two neighboring nurse cells by S2. Young egg chambers contain higher quantities of Grk protein, especially in a cortical strip at the posterior of the oocyte. (B) Upon a signal from posterior follicle cells at S6, microtubule reorganization leads to redistribution of Grk protein; a transient uniform localization throughout the oocyte is followed by accumulation at the nurse cell/oocyte boundary. (C) In early S9 egg chambers,Grk is present in a dorsal anterior cap above the oocyte nucleus. (D) rhi1 hemizygous egg chambers lack detectable Grk expression in the germarium (bracket) but contain approximately normal levels of protein by S4. (E) As early as S6, Grk protein accumulates in large vesicles caged by actin (arrow). (F) In this S9 egg chamber, most Grk protein resides in vesicles (arrow). (G) In rhi2 hemizygous egg chambers, a low level of Grk protein is present in the germarium (bracket). (H) The S5 egg chamber lacks Grk protein (arrowhead) while the S7 egg chamber contains large, Grk-filled vesicles (arrow). (I) A late S8/early S9 rhi2 hemizygous egg chamber in which Grk protein is present in a diffuse band at the anterior of the oocyte. In all panels, punctate dots over the follicular epithelium indicate secreted Grk protein.

In egg chambers produced by rhi¹ and rhi² hemizygous females, we observed two major defects in Grk expression. Virtually all germaria (n = 34) and one-third of S2–S4 rhi¹/Df egg chambers (32%, n = 37) lacked detectable -Grk staining (Fig 7D). By S5, however, almost all egg chambers produced normal levels of Grk protein (Fig 7D). Thus, rhi egg chambers exhibit a delay in Grk protein accumulation. rhi²/Df egg chambers exhibited a weaker phenotype; half the germaria and S2–S4 egg chambers produced normal levels of Grk protein (52%, n = 54; Fig 7G). The remaining samples lacked detectable expression (37%) or exhibited only weak staining (11%; data not shown). These results suggest that rhi mutations affect a process required early in oogenesis for the translation or stability of Grk protein.

The second defect apparent in rhi mutant egg chambers was the presence of large, Grk-containing vesicles in S6–S10 egg chambers (Fig 7E, Fig F, and Fig H). Rhodamine-phalloidin staining demonstrated that actin co-localized with these vesicles, giving the appearance of a cage surrounding Grk protein (yellow overlap in Fig 7E, Fig F, and Fig H). These structures were found in 54.3% (n = 70) of S6–S10 rhi¹/Df egg chambers and 29.6% (n = 88) of S6–S10 rhi²/Df egg chambers. We never observed such vesicles in wild type (n = 68). We speculate that rhi mutants are defective in Grk translation or processing such that Grk protein is delayed in the endoplasmic reticulum, yielding large vesicles. Although a delay may exist in Grk synthesis, some product is secreted and taken up by the overlying follicle cells (Fig 7F, Fig H, and Fig I). Our results suggest that Rhi is necessary for the synthesis or maturation of Grk protein. Lack of Rhi function leads to a delay in Grk production early in oogenesis and the accumulation of Grk protein in vesicles in S6–S10 egg chambers. Taken together, our results suggest that Rhi is not acting to repress grk transcription nor is it required specifically to coordinate S5 nurse cell chromosome reorganization with axis-determining events. The effects on eggshell patterning may involve Rhi function as a transcriptional regulator or as a mediator of chromosome structural reorganization, but at least one role must occur before the obvious S6–S10 chromosomal and D/V polarity defects observed in rhi mutant egg chambers.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We describe the genetic and molecular characterization of a new Drosophila female-sterile gene, rhi, which encodes a protein molecularly similar to the HP1-like subfamily of chromo-domain proteins (for review see CAVALLI and PARO 1998 ). rhi is required for proper chromosome conformation in the nurse cells during oogenesis and for patterning events that are reflected in eggshell structures. We generated two independent P-element alleles and many excision lines, all of which disrupt the major ovarian transcript of 1.6 kb, encoding a protein of 46 kD. rhi expression is not detectable in males but is found at low levels in somatic tissue in adult females and at higher levels in the germline; thus, rhi likely plays a specific role in oogenesis. Here we discuss Rhi's homology to a subfamily of chromo-domain proteins. We then consider the chromosome, eggshell, mRNA-mislocalization, and Grk-protein defects in rhi mutants and present hypotheses for rhi function during oogenesis.

Rhino is a member of the chromo-domain protein family:
The chromo domain was first recognized as a conserved protein motif through homologies shared by Drosophila proteins Heterochromatin-associated Protein 1 (HP1) and Polycomb (Pc) (PARO and HOGNESS 1991 ). HP1 was identified genetically as a suppressor of position-effect variegation, an epigenetic phenomenon in which euchromatic genes are silenced when they are next to heterochromatin (EISSENBERG et al. 1990 , reviewed in WEILER and WAKIMOTO 1995 ). Pc was first described as a repressor of homeotic genes and is now known to be part of a large protein complex that binds 100 sites at specific times during embryogenesis and maintains transcriptional quiescence (LEWIS 1978 ; PARO and HARTE 1996 ). As new members of the chromo-domain family were isolated through their homology to HP1 and Pc, it became clear that two distinct classes exist within the family. The HP1-like proteins contain a carboxyl-terminal motif, the chromo-shadow domain, that shares homology with the amino-terminal chromo domain (AASLAND and STEWART 1995 ). Rhi belongs to the HP1-like class of proteins containing the chromo-shadow domain. Although this class of proteins is believed to function in heterochromatin formation, repressing transcription of euchromatic genes, disrupting these proteins may also activate transcription (HEARN et al. 1991 ; GORMAN et al. 1995 ).

Insight into how chromo-domain proteins can function as transcriptional repressors has been provided by the three-dimensional structure of the chromo domain from the HP1-like MoMOD1 (BALL et al. 1997 ) and by structural and biochemical studies of the chromo-shadow domains from fission yeast Swi6 and Drosophila HP1a, HP1b, and HP1c (COWIESON et al. 2000 ; DELATTRE et al. 2000 ; SMOTHERS and HENIKOFF 2000 , SMOTHERS and HENIKOFF 2001 ). These authors postulate that the chromo-domain and chromo-shadow domain are modules that facilitate interaction with proteins at target sites designated for compaction. This work provides a mechanism to explain the localization of HP1-like proteins (PLATERO et al. 1995 ; COWELL and AUSTIN 1997 ), since few specific DNA sequences bound by HP1-like molecules are known (SUGIMOTO et al. 1996 ). Indeed, the common feature of the HP1 subfamily of chromo-domain proteins is their association with heterochromatin found near centromeres. Thus, loss-of-function mutations not only affect transcription and position-effect variegation but chromosome segregation as well. For example, the Schizosaccharomyces pombe chromo-domain-containing Swi6 protein is required for silencing at the silent-mating-type loci and is also an essential centromere component (EKWALL et al. 1995 ; ALLSHIRE 1996 ). swi6 mutants have a high rate of chromosome loss, as do HP1 Drosophila mutants (EK