The Drosophila ribosome protein S5 paralog RpS5b promotes germ cell and follicle cell differentiation during oogenesis

Ribosome availability places a limit on protein synthesis. As a result, cell growth and proliferation tightly correlate with ribosome biogenesis across species (Lempiainen and Shore, 2009). Emerging evidence suggests that ribosome biogenesis and the regulation of global protein synthesis may play previously underappreciated roles in regulating cell fate decisions and cell function (Buszczak et al., 2014; Slaidina and Lehmann, 2014; Zhang et al., 2014; Sanchez et al., 2016). Ribosome biogenesis mainly takes place in the nucleolus: a nuclear sub-domain that forms around rDNA loci. rRNA genes across eukaryotes tend to share the same basic structural elements, including the external transcribed spacer (ETS), the internal transcribed spacers (ITS), and the 28S, 18S and 5.8S rRNA coding sequences (Schneider, 2012). The 28S, 18S and 5.8S rRNAs are processed from a common 47S pre-rRNA transcribed by the multi-subunit RNA Polymerase I (Pol I) complex. A fourth rRNA, the 5S rRNA, is transcribed by RNA polymerase III. Many components of the Pol I complex share a high degree of amino acid sequence conservation from yeast to humans. However, the promoter elements and the transcription factors that drive Pol I transcription tend to diverge between species. Once transcribed, the pre-rRNA is extensively modified, which involves cleavage of the ETS and ITS, and various methylations and pseudouridylations (Granneman and Baserga, 2004; Klinge and Woolford, 2019). Many of these reactions are promoted by small nucleolar ribonucleoproteins (snoRNPs) and their associated small nucleolar RNAs (snoRNAs). Ribosomal proteins associate with rRNAs through a highly complex and coordinated process to form the 40S and 60S ribosome subunits. Each of these subunits is independently exported out of the nucleus and into the cytoplasm, where they participate in the translation of mRNA. Translation begins when a small subunit scans an mRNA in the 5′ to 3′ direction until it finds a start codon. At this point, a large subunit joins with the small subunit to form an 80S monosome. Monosomes then move along the mRNA to decode successive codons. Multiple monosomes can engage with a single mRNA to form polysomes.

Drosophila oogenesis has served as an outstanding model for studying the dynamic regulation of mRNA translation in space and time. Drosophila ovaries comprise tube-like structures called ovarioles that contain sequentially developing egg chambers, which go through 14 stages of development on the way to becoming a fertilizable egg. Egg chambers contain 16 interconnected germ cells, 15 polyploid nurse cells and one oocyte. This cyst is surrounded by a layer of somatically derived follicle cells. Follicle cells remain mitotically active until stage 6, when they begin to undergo endocycles and become polyploid. This transition is dependent on Notch signaling (Deng et al., 2001). Germ cells begin to express high levels of Delta, beginning at stage 5, and the presence of Delta promotes Notch signal activation in the surrounding follicle cells. Loss of Delta in the germline or downstream Notch pathway components in the follicle cells leads to follicle cell hyperplasia (Lopez-Schier and St Johnston, 2001).

Beyond the regulated transport and translation of individual mRNAs, dynamic increases and decreases of ribosome biogenesis and global protein synthesis help to drive various aspects of germ cell differentiation. The nucleoli of Drosophila female germline stem cells are large relative to their differentiating progeny (Neumuller et al., 2008), which coincides with high levels of rRNA transcription (Zhang et al., 2014). Different mechanisms ensure that GSCs display high levels of ribosome biogenesis (Fichelson and Huynh, 2009; Fichelson et al., 2009; Zhang et al., 2014). A subsequent study showed that changes in ribosomal protein levels and global protein synthesis help to regulate cell morphology and early germ cell differentiation (Sanchez et al., 2016). Beyond early germline cyst formation, ribosomal proteins have long been known to play important roles in supporting the growth and continued development of egg chambers. Mutations in ribosomal protein genes often result in dominant haploinsufficient Minute phenotypes characterized by small stature, thin bristles and fertility defects (Marygold et al., 2007). In addition, specific ribosomal protein gene mutants, such as string of pearls, which encodes RpS2, display egg chamber developmental arrest, fail to enter vitellogenesis and degenerate during mid-oogenesis (Cramton and Laski, 1994).

Emerging evidence from several model systems suggests that the population of ribosomes within a given cell may be heterogeneous (Xue and Barna, 2012; Barna, 2015; Shi and Barna, 2015). This proposed heterogeneity may contribute to distinct gene expression programs during development and in times of stress. For example, recent research suggests that the large subunit ribosomal protein RpL10A is not present in all actively translating ribosomes within mice embryonic stem cells (Shi et al., 2017). Ribosomes with RpL10A show selectivity in translation of mRNAs involved in blood vessel development. These mRNAs often have IRES-like elements within their 5′ untranslated region (5′ UTR). In a second example, loss of mouse RpL38 results in reduced translation of specific Hox mRNAs (Kondrashov et al., 2011). Again, these Hox mRNAs have elements with complex secondary structure within their 5′ UTRs, making them more sensitive to the presence or absence of RpL38 (Xue et al., 2015). Finally, in yeast, ribosomes that lack RpS26 tend to translate stress related mRNAs, which again carry IRES-like elements within their 5′ UTRs (Ferretti et al., 2017). Mutations in ribosomal protein genes and ribosome assembly factors often result in a group of human diseases known as ribosomopathies, which are often cited as further evidence for the functional importance of distinct ribosome populations. However, the extent to which ribosomopathies reflect a loss of specialized ribosomes or are caused by a reduction in overall ribosome levels remains unclear (Mills and Green, 2017).

The Drosophila genome encodes several ribosomal protein paralogs (Marygold et al., 2007). Interestingly, one of these paralogs tends to be ubiquitously expressed, while the other exhibits a much more restricted pattern of expression. For example, microarray analysis (Kai et al., 2005), confirmed by later RNA-seq experiments conducted by the modENCODE project (Graveley et al., 2011), showed that several of these ribosomal protein paralogs, including RpS19b and RpS10a, display enriched expression in gonads. Drosophila RpS5a and RpS5b also fall into this category. RpS5a is ubiquitously expressed whereas RpS5b exhibits enriched expression in the ovary. Both RpS5a and RpS5b have highly conserved C termini from yeast to humans but their N termini are divergent. The negatively charged N termini of RpS5 orthologs have been shown to bind to several translation factors and to be important for translation initiation and elongation in yeast (Valentin et al., 1992; Visweswaraiah et al., 2015; Visweswaraiah and Hinnebusch, 2017). In addition, N-terminally truncated yeast RpS5 shows an increased rate of frameshifting and stop codon read-through, resulting in reduced levels of translation fidelity (Visweswaraiah et al., 2015). These previous studies and the possibility that RpS5a and RpS5b define specialized heterogenous ribosomes prompted us and others (Kong et al., 2019) to examine RpS5a and RpS5b function during germ cell development.

To see the evolutionary features of RpS5a and RpS5b, we examined the rates and mode of evolution of RpS5a and RpS5b genes using both divergence and polymorphism data (see Materials and Methods). For divergence-based analyses, gene sequences were retrieved from 13 species: D. melanogaster, D. simulans, D. sechellia, D. yakuba, D. erecta, D. takahashii, D. biarmipes, D. ficusphila, D. elegans, D. kikkawai, D. ananassae, D. rhopaloa (RpS5a only) and D. eugracilis (RpS5b only). Table S1 contains all the coding sequences and Figs S1 and S2 show RpS5a and RpS5b protein alignments. The results of sequence analysis with PAML (Table S1) indicate that RpS5b is evolving at a higher rate (ω_RpS5b=0.0284) compared with RpS5a (ω_RpS5a=0.0166). However, a model with two separate rates is not significantly better than a model with a single rate for both genes (ω=0.0257; χ²=2.5031; d.f.=1; P>0.05). In addition, no codons were detected to be under positive selection using either the PAML or FEL method from the HyPhy package (Table S2).

To complement divergence-based sequence analyses that might miss signals of positive selection, we applied divergence and polymorphism-based McDonald-Kreitman test (McDonald and Kreitman, 1991) to RpS5b sequences from D. melanogaster, D. simulans and D. yakuba (Table S3). This test compares polymorphism (Pol) and divergence (Div) data for synonymous (s) and nonsynonymous/replacement sites (n). Under neutral evolution, fixed differences represent a subset of neutral mutations that, at some point in time, were segregating in a population. In this case, the ratios of polymorphism to divergence for synonymous and replacement sites are expected to be equal. However, if the RpS5b gene is evolving under positive selection (accumulating adaptive substitutions), there should be an excess of fixed mutations at replacement sites. All three pairwise McDonald–Kreitman tests show significant excess of replacement substitutions in the RpS5b gene. In D. melanogaster to D. simulans comparison: Pol/Div (s)=51/9, Pol/Div(n)=7/10 (P=0.0006 Fisher's exact test). In D. melanogaster to D. yakuba comparison: Pol/Div (s)=45/31, Pol/Div(n)=4/11 (P=0.0255 Fisher's exact test). In D. simulans to D. yakuba comparison Pol/Div (s)=66/21, Pol/Div(n)=7/13 (P=0.0009 Fisher's exact test), supporting that RpS5b is under positive selection.

Sequence alignment of RpS5b genes from 12 species suggests faster evolution of the N-terminal region of the protein (Figs S1 and S2). Separate McDonald-Kreitman tests for N- and C-terminal parts of the protein in D. melanogaster, D. simulans and D. yakuba shows excess of replacement substitutions in the N terminus [D. melanogaster, D. simulans and D. yakuba comparison: Pol/Div (s)=9/4, Pol/Div(n)=8/19; P=0.0383 Fisher's exact test], suggesting that the target of positive selection might be restricted to the first 43 amino acids of the protein. The same region in other species includes multiple indels that might prevent proper orthologous codon alignment and the study of selection in divergence-based analyses (Fig. S2). It is possible that the N-terminus of the RpS5b gene is also under positive selection in other lineages, and indels might have been fixed as part of the functional turnover of the region. Indels have been previously observed in fast-evolving proteins under positive selection (Vacquier et al., 1997).

Next, to characterize the function of the Drosophila RpS5b within germ cells, we generated null mutations in the gene using CRISPR/Cas9-mediated techniques. A dsRed cassette was integrated into RpS5b locus (Fig. S3A) and relative mRNA expression of RpS5a and RpS5b in the mutant was tested by qRT-PCR (Fig. S3B). In Drosophila, mutations in ribosomal protein genes tend to result in Minute phenotypes (Marygold et al., 2007), but RpS5b mutants did not (Fig. S3C). The mutant homozygotes were viable, did not exhibit developmental delays, and displayed normal bristles. However, homozygous RpS5b mutant females were sterile (Fig. S3D), consistent with a recently published study (Kong et al., 2019). Further examination revealed that this sterility was caused by developmental arrest and degeneration of egg chambers around stage 7/8 (Fig. S3E). All examined stage 7/8 egg chambers of RpS5b mutants exhibited an abnormal nuclear structure, as shown by DAPI staining (n>50), and multiple layered follicle cells within the posterior region. The multiple layered follicle cell phenotype was also observed when we knocked down RpS5b using RNAi in combination with the germ cell driver vasa-gal4, but not with the somatic cell specific traffic jam (tj)-gal4 driver (Fig. S3F). These results indicate that the RpS5b mutant follicle cell defect is non-cell autonomous.

In yeast, most ribosomal protein paralogs are believed to be functionally redundant, with a few exceptions, especially in times of stress (Ghulam et al., 2020). To determine whether RpS5a and RpS5b can compensate for each other in germ cells, we introduced untagged RpS5a and RpS5b UAS-transgenes into the RpS5b mutant background and expressed them using the maternal α-tubulin (mat α-tub)-gal4 driver. Both RpS5a and RpS5b expression in germ cells could rescue the sterility of the RpS5b mutant (Fig. S3G), consistent with a previous report (Kong et al., 2019). The nuclear structure and multi-layered follicle cell phenotypes were also partially rescued by either RpS5a (45/56=80%) or RpS5b (43/50=86%) transgenic expression (Fig. S3H). To further test for functional differences between these paralogs, we next determined the relative fitness of progeny from mutant females expressing each transgene in a competitive assay. Female flies expressing RpS5a or RpS5b transgenes in a RpS5b mutant background were added to the same vial for 24 h. We then assayed the genotypes of the resulting adult progeny. Although we observed variability between different vials, no significant difference in the percentage of progeny derived from females expressing either transgene was detected (Fig. S3I). These results further suggest RpS5a and RpS5b containing ribosomes exhibit similar functionality under the standard laboratory conditions tested.

Ribosomal protein production, rRNA transcription and processing must be coordinated to make a mature ribosome. Drosophila Minute mutations result in developmental delays and loss of fertility, presumably caused by lower levels of functional ribosomes and loss of protein production capacity (Marygold et al., 2007). We considered the possibility that loss of RpS5b causes rRNA transcription and/or processing defects in germ cells. Immunostaining for Fibrillarin revealed that all RpS5b mutant egg chambers exhibited nucleolar morphology defects, beginning at stage 5 (Fig. 1A). Again, the nucleolar defects were rescued by expression of either RpS5a or RpS5b transgenes driven by mat α-tub-gal4. Udd, which helps to drive rDNA transcription in Drosophila (Zhang et al., 2014), still localized to germ cell nucleoli in RpS5b mutant cells, but interestingly, its expression appeared higher and its localization within nucleoli more punctate. Western blot analysis comparing controls with RpS5b mutant samples confirmed that Udd expression increased in the absence of RpS5b (Fig. 1B,C). To determine whether loss of RpS5b resulted in changes in rRNA levels or rRNA-processing defects, we extracted RNAs from wild type and the RpS5b mutant ovaries, and analyzed these samples using northern blot analysis. In Drosophila, rRNA is transcribed as a long pre-rRNA that contains 18S, 5.8S, 2S and 28S, bridged by external transcribed spacers (ETS) and internal transcribed spacers (ITS) (Fig. 1D) (Schneider, 2012). The transcribed pre-rRNA is processed through multiple steps to generate mature rRNAs and using probes for specific regions of the pre-rRNA allows one to determine whether specific rRNA processing steps are affected in the mutant. When we normalized samples to total RNA levels, we detected significantly increased pre-rRNA and intermediate processed forms of rRNA in the mutant samples (Fig. 1E). These data suggest that rRNA transcription may increase in the absence of RpS5b.

Next, to directly compare rRNA transcription in control and mutant germ cells, we pulse-labeled nascent rRNA using bromouridine-triphosphate (Br-UTP). Br-UTP is an analog of UTP and incorporates with nascent RNA during transcription. Before Br-UTP labeling, we treated wild-type and mutant ovaries with α-amanitin to block RNA polymerases II and III, allowing us to specifically assay pre-rRNA transcription. We observed significantly increased Br-UTP incorporation in mutant nurse cell nucleoli (Fig. 1F,G), further suggesting that RpS5b mutant germ cells exhibit increased rRNA transcription.

Next, we performed ribosome profiling to determine whether loss of RpS5b changed the translation of specific genes. Ribosome profiling combines deep sequence ribosome footprint analysis with classic RNA-sequencing (RNA-seq) to reveal differences in both transcription and translation between samples (Fig. 2A) (Brar and Weissman, 2015). As RpS5b mutant ovaries do not carry egg chambers beyond stage 8, we sought to reduce the contribution of late egg chambers in control samples. Ovaries from young control and mutant female flies were hand dissected 48 h after eclosion and care was taken to remove any stage 10-14 egg chambers that were present. Biological duplicate control and mutant samples were processed for both RNA-seq and ribosome footprint-sequencing (ribo-seq), following established protocols (Greenblatt and Spradling, 2018). We obtained 33,955,476 and 43,040,528 reads from control RNA samples, and 48,218,060 and 38,989,829 reads from mutant samples for RNA-seq. We obtained 3,122,299 and 6,170,623 reads from control footprint samples and 13,942,184 and 8,714,570 reads from mutant footprint samples. These reads were trimmed and aligned to the Drosophila reference genome. Differential gene expression analysis and translation efficiency calculations were then performed. The two biological replicates showed high Pearson Correlation Coefficients for both RNA-seq and ribosome-seq analysis (Fig. S4A,B). The read length of the control and mutant footprint libraries typically fell between 29 and 31 nucleotides, as expected (Fig. S4C). More than half of the footprints were aligned within CDS regions (Fig. S4D).

We analyzed differences in translation efficiency between control and RpS5b mutant samples. 5′ UTR analysis based on sequence length, RNA fold and GC content suggests that shorter and less-complex 5′ UTR containing transcripts were translated more efficiently in the absence of RpS5b (Fig. S4E). Surprisingly, gene ontology (GO) analysis based on cellular components revealed that upregulated genes included those related to the ribosomal subunit (GO: 0044391, P.DE=1.03×10⁻¹⁵), ribosome (GO: 0005840, P.DE=1.72×10⁻¹⁰), small ribosomal subunit (GO: 0015935, P.DE=6.29×10⁻¹⁰) and large ribosomal subunit (GO: 0015934, P.DE=5.02×10⁻⁷) (Fig. 2B). Among the cytoplasmic ribosomal proteins, 14 genes showed upregulated translational efficiency and one gene showed downregulated translational efficiency (Fig. 2C,D).

Based on the increases in both rDNA transcription and the translation efficiency of various ribosome protein mRNAs, we tested whether levels of global translation changed upon loss of RpS5b. First, we performed a puromycin labeling assay. Briefly, puromycin resembles the 3′ end of an aminoacylated tRNA and enters into the ribosomal A site. Once there, it accepts the nascent polypeptide chain from the P-site peptidyl-tRNA. However, the resulting peptide bond cannot be cleaved by an incoming aa-tRNA, resulting in premature termination of translation and the formation of puromycin labeled peptides, which can be detected using anti-puromycin antibodies. Pulse-labeling experiments on control and mutant ovaries revealed that loss of RpS5b resulted in increased puromycin incorporation, suggesting that the mutants exhibit higher levels of protein synthesis (Fig. 3A). Next, we conducted a polysome fractionation experiment. This analysis showed that while control and mutant samples maintained similar levels of 40S and 60S subunits, the monosome peak appeared lower and polysome peaks appeared higher in the RpS5b mutants (Fig. 3B). Finally, we performed a metagene analysis on whole-transcriptome and ribosome footprint data to look for changes in translation efficiency. Consistent with increased puromycin incorporation and polysome peaks, transcripts from mutant samples generally displayed a higher translation efficiency (TE) score relative to control samples (Fig. 3C). Together, our data reveal an unexpected finding whereby disruption of RpS5b is accompanied by a global increase in protein synthesis.

Despite elevated levels of ribosome biogenesis and global protein synthesis, the translation efficiencies of specific mRNAs were lower in the absence of RpS5b (Fig. 4A). As RpS5b mutant egg chambers do not complete stage 8, which is marked by the onset of vitellogenesis, we examined the expression of genes related to yolk production and uptake. Surprisingly, all three yolk proteins (Yps; Yp1/2/3) and one receptor, Yolkless (Yl) (Gelti-Douka et al., 1974; Barnett et al., 1980; Schonbaum et al., 2000), displayed decreased TE in RpS5b mutant samples. Yolkless mRNA levels appear similar in the wild-type and the mutant samples, but they displayed decreased translation efficiency (Fig. 4B,C). To see whether reduced translation efficiency levels of Yp1/2/3 and Yl affect yolk granule accumulation in the oocyte, we examined stage 8 egg chambers. We observed that RpS5b mutant egg chambers do not accumulate yolk granules, assayed by both differential interference contrast (DIC) imaging as well as yolk granule autofluorescence in the oocyte (100% penetrance, n>50 stage 8 wild-type egg chambers and n>20 stage 8 mutant egg chambers) (Fig. 4D,E). Vitellogenesis represents a well-characterized check point during Drosophila oogenesis, and disruption of this process causes egg chamber degeneration and reabsorption (McCall, 2004). Thus, data presented here suggest that although general protein production increases upon loss of RpS5b, reduced translation of specific mRNAs may underlie the observed egg chamber degeneration phenotypes.

Next, we attempted to better understand the multiple layered follicle cell phenotype exhibited by RpS5b mutants. Follicle cells are derived from follicle stem cells and encapsulate germ cells as a single layer. They continue to undergo mitotic cycles until stage 6, whereupon follicle cells enter an endocycle to become polyploid (Lilly and Duronio, 2005). To test whether follicle cells within the RpS5b mutant fail to undergo this transition, we immunostained for phospho-histone H3 (pH3), a marker of mitosis. Control egg chambers display pH3-positive cells before stage 6. By contrast, RpS5b mutant follicle cells were often pH3 positive during stage 7 (11/84 wild-type egg chambers and 42/76 mutant egg chambers) (Fig. 5A,B). This phenotype was only observed in the posterior region of the egg chamber, corresponding to where we observed multi-layered follicle cells.

Previous studies have shown that Notch-Delta signaling triggers follicle cells to transition from a mitotic cell cycle to an endocycle (Deng et al., 2001; Lopez-Schier and St Johnston, 2001; Schaeffer et al., 2004). To test whether loss of RpS5b disrupts Notch-Delta signaling, we characterized the expression of target genes downstream of Notch activation. For example, control follicle cells express Hnt, a positively regulated downstream factor of the Notch pathway (Sun and Deng, 2007), after stage 7. By contrast, RpS5b mutant egg chambers did not contain Hnt-positive cells at their posterior end (100% penetrance, n>100 ovarioles) (Fig. 5C). We also examined Cut expression as an independent marker of Notch signaling within follicle cells (Sun and Deng, 2005). Cut was not expressed in the anterior and lateral follicle cells after stage 7 in control egg chambers, but posterior follicle cells within RpS5b mutant egg chambers exhibited sustained Cut expression after stage 7 (100% penetrance, n>100 ovarioles) (Fig. 5D), further indicating these mutants fail to activate the Notch pathway in a stage-specific manner.

RNAi knockdown and cDNA rescue experiments showed the follicle cell phenotype in RpS5b mutants is non-cell autonomous (Fig. S3F-H). Based on these findings, we focused on analyzing the expression and localization of the Notch pathway ligand Delta within control and mutant egg chambers. Statistical analysis of the ribosome profiling data did not reveal any significant differences in the mRNA and translation efficiencies of Delta, or that of other components of the Delta-Notch signaling pathway (Fig. 6A). When we stained control and mutant ovaries with anti-Delta antibody, we observed a clear increase in Delta protein levels in mutant nurse cells, as well as Delta protein mislocalization within the oocyte (Fig. 6B). In control samples, Delta was expressed in the germ cells from stage 5 and localized at the nurse cell membranes as well as membranes between germ cells and follicle cells. By contrast, although modest levels of Delta protein accumulated in the oocytes of RpS5b mutant egg chambers, it did not localize to the membrane between the oocyte and the posterior follicle cells.

To further assay for Notch pathway activation, we stained ovaries for the Notch intracellular domain (NICD) and Notch extracellular domain (NECD). Consistent with the increased but mislocalized Delta protein we observe, RpS5b mutant ovaries also displayed increased NICD and NECD levels, suggesting full-length Notch receptor accumulates at the membrane, reflecting a loss of Notch signaling (Fig. 6C,D) (Yu et al., 2008). Delta protein internalization and turnover depends on ubiquitylation by the E3 ligases Mindbomb1 (Mib1) and Neuralized (Neur) (Lai et al., 2005; Wang and Struhl, 2005). Using anti-Mib1 antibodies, we observed lower levels of Mib1 in RpS5b oocytes relative to controls (Fig. 6E). All together, these data indicate that loss of RpS5b disrupts Delta trafficking and processing, leading to defects in Notch signal transduction in posterior follicle cells.

Microtubule-dependent transport of the essential mRNAs, proteins and the organelles from the nurse cells to the oocyte is crucial for the egg chamber development (Theurkauf and Hazelrigg, 1998). Based on the result that several cytoskeleton-related genes show decreased translation efficiency in RpS5b mutant ovaries [cytoskeleton organization (GO: 0007010, P.DE=3.35×10⁻¹²), microtubule-based process (GO: 0007017, P.DE=6.23×10⁻¹²) and microtubule cytoskeleton organization (GO: 0000226, P.DE=7.71×10⁻¹¹)] (Fig. S5A), we began to characterize the levels and polarity of microtubules in the absence of RpS5b. First, we stained ovaries using anti-α-Tubulin (αTub) antibody. Similar to a previous report (Kong et al., 2019), we observed an enrichment of αTub in the oocyte at stages 8 and 9 in wild-type egg chambers, while the levels of αTub in RpS5b oocytes appeared much lower (100% penetrance, n>20) (Fig. S5B). Consistent with these data, the microtubule minus-end marker khc::nod::lacZ displayed decreased levels of expression and an unpolarized pattern in RpS5b mutants during early and mid-oogenesis, suggesting the microtubule network across the nurse cells and oocytes was not polarized properly (95.7% penetrance, n>20) (Fig. S5C). We also observed mislocalization of Gurken protein during mid-oogenesis in RpS5b mutant oocytes (100% penetrance, n>10) (Fig. S5D), suggesting the possibility that loss of RpS5b results in microtubule polarity defects. However, Orb, an oocyte specification regulator (Pokrywka and Stephenson, 1995) is properly localized in the oocyte both in wild-type and mutant egg chambers (Fig. S5E). Furthermore, PFCs of wild type and the mutant express Midline, a downstream factor of Gurken/EGFR signaling pathway (Fregoso Lomas et al., 2013) (Fig. S5F), suggesting the microtubule network and A-P polarity of the oocyte are not entirely disrupted upon loss of RpS5b.

To test whether Notch activation in the posterior follicle cells can rescue the RpS5b follicle cell hyperplasia phenotype, we drove Notch intracellular domain (NICD) expression in posterior follicle cells using pnt-gal4. To overcome the lethality of NICD expression from early stages, we also introduced a GAL80ts transgene into the background. Raising the resulting flies at 18°C, which permits GAL80ts activity, results in the repression of GAL4-dependent NICD expression. After shifting female flies to the restrictive temperature (29°C) for 96 h, the multilayered follicle cell phenotype was partially rescued, indicating Notch pathway activation in the posterior follicle cells is sufficient to rescue this aspect of the RpS5b mutant phenotype (Fig. 7A,B). Overall, our data show that expression of RpS5b in germ cells regulates proper rDNA transcription and RP translation, as well as translation of microtubule-related genes to mediate Delta-Notch signaling pathway between germ cells and posterior follicle cells (Fig. 7C).

Here, we present data that indicate loss of the germ cell-enriched Drosophila ribosomal protein paralog RpS5b results in female sterility marked by egg chamber degeneration and posterior follicle cell hyperplasia. These results are consistent with the analysis of a second independent mutation within the locus (Kong et al., 2019). Both studies also documented defects in the microtubule cytoskeleton. New data presented here help to extend our understanding of the function of RpS5b. By using ribosome profiling approaches to characterize differences in transcription and ribosome footprints between control and RpS5b mutant ovaries on a global scale, we found that loss of RpS5b led to an unanticipated increase in ribosome biogenesis and protein synthesis. In addition, we found the follicle cell hyperplasia phenotypes of RpS5b mutants are caused by disruption of Delta trafficking and processing. Failure to properly activate Delta in oocytes results in loss of Notch signal transduction in the posterior follicle cells. We anticipate the disruption of normal Delta trafficking within RpS5b mutant oocytes is tightly linked with the observed defects in microtubule polarity.

Previous transcriptomic analysis showed that RpS5a is ubiquitously expressed during development and throughout adult tissues, while RpS5b appeared more restricted to the reproductive system (Kai et al., 2005; Graveley et al., 2011). RpS5b is expressed in both ovaries and testes, but disruption of the gene only effects female fertility (this study; Kong et al., 2019). The gonad-specific expression of RpS5b suggested that germ cells may form specialized ribosomes that contain either RpS5a and RpS5b. A reasonable extrapolation from this model is that RpS5b-containing ribosomes promote the translation of specific mRNAs within the germline. Indeed, Kong et al. (2019) provided evidence that mRNAs encoding genes linked with mitochondria function were enriched for RpS5b-containing ribosomes. However, our ribo-seq analysis did not reveal any specific changes in the translation efficiency of these same mRNAs in the absence of RpS5b. Moreover, expression of the ubiquitously expressed paralog RpS5a appears to rescue all the defects exhibited by RpS5b mutants tested to date (this study; Kong et al., 2019). Given the evidence that RpS5b is under positive selection among Drosophila species, future work may reveal a distinct requirement for RpS5b over RpS5a under certain environmental conditions or stresses. Alternatively, the germline may simply require higher expression levels of a RpS5 paralog. Interestingly, we observe an increase in RpS5a expression in the RpS5b mutant background, coincident with an increase of rRNA transcription and ribosomal protein production. However, the levels of this increase do not fully compensate for the reduced levels of RpS5b. Paralogs for several ribosomal proteins exist within the Drosophila genome, and there are multiple examples, including RpS5a/b, where the ubiquitously expressed gene is on the X chromosome, whereas the paralog that exhibits enriched expression in germ cells is located on an autosome. Perhaps differential regulation of gene expression on the X chromosome does not allow for the high levels of ribosomal protein expression required for germ cell differentiation.

The inability of the compensatory increases in endogenous RpS5a gene expression to suppress phenotypes caused by loss of RpS5b, along with the general increased ribosome biogenesis observed in the RpS5b mutant, suggests the unanticipated possibility that germ cells in RpS5b mutants have subpopulations of ribosomes that do not contain any RpS5 protein at all. Studies from yeast indicate that RpS5 plays a crucial role in controlling the rate and fidelity of translation (Visweswaraiah et al., 2015; Visweswaraiah and Hinnebusch, 2017). Previous work has shown that ribosome biogenesis is very sensitive to rates of translation initiation (Thoreen et al., 2012): higher rates of translation lead to more ribosome biogenesis, which feeds forward to support even higher levels of translation. One way loss of RpS5b could result in increased ribosome biogenesis is if RpS5a-containing ribosomes or ribosomes that do not contain any RpS5 paralog drive higher rates of translation, creating a feed-forward loop. Ribosomes that lack RpS5 may also exhibit lower levels of translation fidelity. Upsetting the balance of global translation during mid-oogenesis could have several detrimental effects leading to eventual egg chamber degeneration. Alternatively, the increase in rRNA transcription and ribosomal protein production could be the result of a feedback loop, whereby germ cells sense they do not have adequate translation capacity needed for the expression of proteins related to vitellogenesis, and try to compensate by increasing ribosome biogenesis. Previous results indicate that egg chambers experience reversible changes in protein localization and microtubule organization in response to nutrient stress (Shimada et al., 2011). Perhaps disruption of RpS5b induces a related, but irreversible, stress response. Further biochemical work will be needed to definitively assess whether ribosomes that lack any RpS5 are formed in RpS5b mutant germ cells and whether such ribosomes participate in translation. Despite the observations that transgenic expression of RpS5a can rescue a RpS5b mutant phenotype, the study of ribosomes in RpS5b mutant germ cells may still provide an additional platform, beyond those already studied in yeast and mice, for studying the function of sub-populations of ribosomes with an altered composition.

From the ribosome profiling analysis, we found that shorter and less-complex 5′ UTR containing transcripts are translated more efficiently in the RpS5b mutant ovaries. What accounts for this observed difference remains an open question. During translation initiation, the 48S preinitiation complex scans the 5′ UTR of a transcript in search of a start codon. This process requires eukaryotic initiation factor 4A (eIF4A), an RNA helicase that unwinds the secondary structure within 5′ UTRs (Andreou and Klostermeier, 2013). Inhibition of eIF4A causes a decrease in the translation of transcripts that contain complex 5′ UTRs (Rubio et al., 2014; Leppek et al., 2018). Although we do not observe differences in eIF4A transcription or translation between wild-type and RpS5b mutant samples, further studies are needed to test whether RpS5b affects eIF4A activity or scanning speed of the preinitiation complex to regulate initiation of transcripts with 5′ UTRs of different lengths.

We and others (Kong et al., 2019) observe changes in the microtubule levels and polarity within RpS5b mutant oocytes. Early in oogenesis, Orb protein and oskar RNA are trafficked correctly to the oocyte, indicating that loss of RpS5b does not completely disrupt the microtubule network. However, Gurken protein localization is disrupted during mid-oogenesis. The correct targeting of molecules such as Gurken to specific locations within the oocyte depends on the repolarization of microtubules starting at stage 6 of oogenesis. In RpS5b mutants, lack of microtubule repolarization results in their plus ends being incorrectly oriented towards the interior of the oocyte, as is also observed in other mutants. Ribosome profiling analysis indicates that a number of proteins linked with microtubule assembly and a number of microtubule motors exhibit decreased expression in the absence of RpS5b. Loss of any number of these factors could account for the observed microtubule defects in RpS5b mutants.

In addition to the degeneration and microtubule phenotypes, RpS5b mutant egg chambers also exhibit follicle cell hyperplasia at their posterior ends. Previous work has shown that Delta-Notch signaling controls when follicle cells exit the cell cycle (Deng et al., 2001; Lopez-Schier and St Johnston, 2001). We find that the normal expression patterns of various genes that respond to Notch pathway activation in the posterior follicle cells are disrupted in RpS5b mutants. Germ cell-specific RNAi knockdown and transgenic rescue experiments indicate that this follicle cell phenotype is caused by defects in the germline. These results led us to investigate Delta expression and trafficking within the oocyte. Delta transcript levels and translation efficiency appeared similar in control and RpS5b mutant samples, based on ribosome profiling. However, immunofluorescence revealed a dramatic increase in total Delta protein levels within nurse cells. Significantly, however, we observed less Delta protein localized to the interface between the oocyte and the follicle cells within the RpS5b mutants. Enhanced NICD and NECD staining further suggests that loss of RpS5b interferes with the normal trafficking, activation and turnover of the Delta ligand. Previous work has shown that ubiquitylation of Delta by the ubiquitin ligases Mindbomb1 and Neuralized is essential for proper Notch signaling (Lai et al., 2005; Wang and Struhl, 2005). In wild-type egg chambers, we find Mindbomb1 becomes enriched at the posterior end of the oocyte during mid-oogenesis. Loss of RpS5b disrupts this localization. Expression of activated Notch specifically in the posterior follicle cells circumvents these defects and rescues the follicle cell hyperplasia. Thus, our data are consistent with a model whereby disruption of microtubule re-polarization in the RpS5b mutants results in a depletion of Delta and Mindbomb1 at the posterior pole of the oocyte, ultimately compromising Notch signaling in the posterior follicle cells.

Fly stocks were maintained at 25°C on standard cornmeal-agar-yeast food, except where noted.

y¹v¹; P{TRiP.GL01502}attP2/TM3, Sb¹(RpS5a RNAi),

y¹v¹; P{TRiP.GL01864}attP2 (RpS5b RNAi),

w; P{UASp-khc::nod::lacZ}3/TM3, Sb¹,

P{hsFLP}1, y¹w; P{UAS-N.intra.GS}2/Cyo; MKRS/TM2,

w; P{GMR45D11-GAL4}attP2 (pnt-gal4),

w; P{tubP-GAL80[ts]}20; TM2/TM6B, Tb¹ and

w; P{matalpha4-GAL4-VP16}V2H were obtained from Bloomington stock center.

Other lines used include udd¹ (Zhang et al., 2014), vasa-gal4 (a gift from Yukiko Yamashita, Whitehead Institute, USA) and tj-gal4 (Panchal et al., 2017). UAS-RpS5a and RpS5b were inserted into attP40w landing site using phiC31 integrase (Rainbow Transgenics). For NICD expression in RpS5b mutants, flies were incubated in 18°C until eclosion and then shifted to 29°C for the designated time.

RNA was extracted from w¹¹¹⁸ ovaries using TRIzol reagent (Thermo Fisher Scientific), following the manufacturer's protocol. cDNAs were made using a SuperScript III First-Strand Synthesis System (Invitrogen), followed by PCR using RpS5a and RpS5b specific primers (RpS5a forward, 5′-CACCATGGCCGAAGTTGCTGAAAACG; RpS5a reverse, 5′-TTAACGGTTGGACTTGGCGAC; RpS5b forward, 5′-CACCATGTCCGAGGAAGTGGTGG; and RpS5b reverse, 5′-TTAACGATTCGACTTGGCCACG). PCR products were purified, cloned into pENTR vector and swapped into pPW (Drosophila Gateway Vector Collection) using an LR reaction.

To generate the RpS5b mutant, guide RNAs were designed using http://tools.flycrispr.molbio.wisc.edu/targetFinder and synthesized as 5-unphosphorylated oligonucleotides, annealed, phosphorylated and ligated into the BbsI sites of the pU6-BbsI-chiRNA plasmid (gRNA1 sense, 5′-CTTCGTGTAAGTCGGCATCAATTCA; gRNA1 antisense, 5′-AAACTGAATTGATGCCGACTTACAC; gRNA2 sense, 5′-CTTCGAATTCGAGCTAGATCTCAAA; and gRNA2 antisense, 5′-AAACTTTGAGATCTAGCTCGAATTC) (Gratz et al., 2013). Homology arms were synthesized (IDT) and cloned into pHD-dsRed-attP (Gratz et al., 2014) (Addgene). Guide RNAs and the donor vector were co-injected into nosP Cas9 attP embryos at the following concentrations: 250 ng/ml pHD-dsRed-attP donor vector and 20 ng/ml of each of the pU6-BbsI-chiRNA plasmids containing the guide RNAs (Rainbow Transgenics).

Ovaries were dissected in PBS and fixed for 10 min with gentle rocking in 4% formaldehyde in PBS. After fixation, ovaries were washed in PBT (1×PBS, 0.5% BSA and 0.3% Triton-X 100) followed by primary antibody incubation overnight at 4°C. Ovaries were washed and incubated for four hours with secondary antibodies at room temperature. Ovaries were then washed and mounted in VectaShield Mounting medium with DAPI (Vector Laboratories). The following antibodies were used: rat anti-HA (1:200) (Roche, 11867423001), mouse anti-Fibrillarin (1:1000, Abcam, ab4566), mouse anti-αTubulin (1:100) (Abcam, ab7291), rabbit anti-Mib1 (1:100) (Abcam, ab124929), rat anti-BrdU (1:50) (Abcam, ab6326), rabbit anti-pH3 (1:2000) (Upstate, 06-570), mouse anti-β-galactosidase (1:1000) (Promega, Z3781), guinea pig anti-Midline (1:100) (a gift from Laura A. Nilson, McGill University, Canada), mouse anti-Hts (1:20) [Developmental Studies Hybridoma Bank (DSHB), 1B1], mouse anti-Orb (1:20) (DSHB, 4H8), mouse anti-Hnt (1:20) (DSHB, 1G9), mouse anti-Cut (1:20) (DSHB, 2B10), mouse anti-Delta (1:20) (DSHB, C594.9B), mouse anti-NECD (1:20) (DSHB, C458.2H), mouse anti-NICD (1:20) (DSHB, C17.9C6), mouse anti-Gurken (1:10) (DSHB, 1D12), Alexa Fluor 488 goat anti-mouse IgG (H+L) (1:200, Invitrogen, A11029), Alexa Fluor 568 goat anti-mouse IgG (H+L) (1:200, Invitrogen, A11031), Cy3-conjugated AffiniPure donkey anti-rabbit IgG (H+L) (1:200, Jackson ImmunoResearch, 711-165-152), Cy3-conjugated AffiniPure goat anti-rat IgM (H+L) (1:200, Jackson ImmunoResearch, 112-165-075) and Fluorescein (FITC)-conjugated AffiniPure donkey anti-guinea pig IgG (H+L) (1:200, Jackson ImmunoResearch, 706-095-148).

Ten ovaries of wild type and 20 ovaries of the mutant were dissected and RNAs extracted using TRIzol reagent (Thermo Fisher Scientific), following the manufacturer's protocol. RNAs were then reverse transcribed with the SuperScript III First-Strand System (Thermo Fisher Scientific) and subjected to the real-time PCR reactions with SYBR Green PCR Master Mix (Thermo Fisher Scientific). Three replicates were used to determine relative changes in transcript levels and unpaired Student's t-test was used for statistical analysis. Primers were: RpS5a specific forward, 5′-GTGGAGCGTTTGACCTGCTC; RpS5a specific reverse, 5′-GTTGACGCGACGCAGGG; RpS5b specific forward, 5′-GCGGCCAAGAGATTCCGC; and RpS5b specific reverse, 5′-CCTGAAGCGGATTCTCCGAG.

50 pairs of ovaries were dissected in ice-cold PBS and lysed in polysome lysis buffer [20 mM Tris (pH 7.4), 5 mM MgCl₂, 100 mM NaCl, 200 μg/ml emetine, 0.1% NP-40 and Complete Protease Inhibitor Cocktail (Roche)]. After centrifugation, the supernatants were subjected to a ∼10-50% sucrose column and centrifuged at 154,005 g for 3 h at 4°C. The samples were fractionated into 20 fractions and absorbance was measured at 254 nm using a Biologic LP system (Biorad).

Ovaries were dissected in PBS and lysed in protein lysis buffer [20 mM Tris (pH 7.4), 100 mM NaCl, 5 mM MgCl₂, 1% Triton X-100 and Complete Protease Inhibitor Cocktail (Roche)]. Lysates were clarified by centrifugation at 16,363 g at 4°C for 15 min and supernatants were subjected to the western blot analysis. The following antibodies were used: mouse anti-β-Tubulin (1:4000, Developmental Studies Hybridoma Ban, E7), guinea pig anti-Udd (Zhang et al., 2014), mouse anti-Puromycin (1:10000, Sigma, 12D10), peroxidase-conjugated AffiniPure donkey anti-mouse IgG (H+L) (1:1000, Jackson ImmunoResarch, 715-035-150) and peroxidase-conjugated AffiniPure goat anti-guinea pig IgG (H+L), 106-035-003).

Virgin females were collected and aged for 2-3 days. Single females were crossed with wild-type males for 24 h and transferred to a new vial. Eggs laid over a 24 h period and their subsequent hatching rates were counted.

Ovaries were dissected in Schneider's media and permeabilized with 350 ng/μl digitonin in permeabilization buffer (PB) (2 mM MgCl₂, 1 mM DTT and protease inhibitor in PBS) for 5 min on ice. Samples were washed with PB once and incubated with 250 ng/μl α-amanitin in PB on ice. After 10 min, transcription mix (final concentration: 2 mM ATP, 0.5 mM CTP, 0.5 mM GTP and 2 mM Br-UTP) was added to the samples and run-on transcription was carried out at 25°C for 5 min. The reaction was terminated by washing with PBS. Fixation and staining procedures were the same as described earlier.

Ovaries were dissected in Schneider's media and then incubated with 5 µg/ml of puromycin for 40 min (Deliu et al., 2017). Ovaries were washed and lysed for western blot analysis.

Northern blot analysis was performed using NorthernMax Kit (Ambion) as described previously (Zhang et al., 2014). Probes were synthesized using the DIG Oligonucleotide 3′-End Labeling Kit, 2nd generation (Roche): ETS, 5′-CGAACAATGCGAGGTCGGCAACCACTGCCTACC-3′; ITS1, 5′-GGTTGTTGCATTAGCCAACGTATGCCCATAACTAAGATG-3′; ITS2, 5′-AGAAAATATTTCTCTTCGTTTTTCACATTCAAATGTGAGATAATG-3′; 5.8S, 5′-GTCGATGTTCATGTGTCCTGCAGTTCACACGATGACGCACAG-3′; and 28Sb, 5′-GTAACTAGCGCGGCATCAGGTGATCGAAGATCCTCCC-3′.

Ribosome profiling was performed as described previously (Greenblatt and Spradling, 2018), with some modifications. 200 pairs of ovaries from young (day 1-2) female were dissected in PBS and flash frozen in liquid nitrogen every 20 min. Ovaries were lysed and homogenized with a motorized pestle in 300 µl of ribosome lysis buffer [150 mM NaCl, 5 mM MgCl₂, 50 mM Tris (pH 7.5), 1 mM DTT, 0.5% Triton X-100, 20 µg/ml emetine, 20 U/ml SUPERaseIn and 50 µM GMP-PNP] followed by trituration through a 27.5-gauge needle. Samples were centrifuged at 3000 g for 5 min at 4°C. The aqueous phase was transferred to pre-chilled RNAse-free tube and centrifuged at 20,000 g for 10 min at 4°C. The supernatant was transferred to a new tube. For mRNA sequencing, RNAs from 10 µl of the sample were extracted using TRIzol reagent (Thermo Fisher Scientific), following the manufacturer's protocol. mRNA-seq libraries were prepared as poly(A) selected mRNA according to the manufacturer's instructions using the Illumina TrueSeq Stranded RNA LT Kit. For ribosome footprint sequencing, samples were treated with Micrococcal Nuclease S7 (Roche) and monosomes were enriched through a 34% sucrose cushion. RNA was extracted from the pellet using TRIzol reagent and 28-34mer footprints were selected on a 15% TBE-Urea gel. Dephosphorylation of the footprints was performed with shrimp alkaline phosphatase (rSAP, NEB) and ligated with 3′ preadenylated adapter (rAppTGGAATTCTCGGGTGCCAAGG) by T4 RNA ligase 2, truncated (NEB). The 49-55mer ligated products were selected on a 10% TBE-Urea gel. Reverse transcription was performed with SuperScript III (company) using the primer (/5phos/GATCGTCGGACTGTAGAACTCTGAACGTGTAGATCTCGGTGGTCGC/isp18/CACTCA/isp18/CCTTGGCACCCGAGAATTCCA) following rRNA depletion. Samples were circularized by two sequential reactions with CircLigase (Epicentre) and amplified by 9-12 PCR cycles with primers compatible with Illumina flow cells.

Raw fastq files were quality checked using FastQC v0.11.5 to determine usability and then trimmed using TrimGalore v0.4.1 for RNA-seq and FastX for footprint-seq. After trimming, rRNAs were removed from the reads using bowtie v1.0.0. To accomplish this, rRNA information was extracted from the Drosophila reference BDGP6.22 and was used to generate the bowtie reference. The resulting fastq files were aligned to the full Drosophila reference using TopHat v2.1.2 in conjunction with bowtie and its built-in ‘bowtie1’ flag. Samtools v0.1.19 was used to index the bam files and bedtools v2.26.0 along with UCSC to generate the bed files and BigWig files. In order to calculate p-site offset, Plastid was used and then the metagenome analysis was preformed using Plastid as well. Count tables were then generated from the aligned reads using Plastid and differential expression analysis was performed using the DESeq2 package in R.

For gene-level analysis, the trimmed fastq files were aligned using HISAT2 v2.1.0 instead of bowtie. This was carried out to allow the use of StringTie v1.3.2d and its prepDE.py script. This allows for differential expression analysis on a gene-wide level and a generation of read count tables. Differential expression analysis can now be performed using DESeq2 package in R. Translational efficiency was determined using a combination of RNA-seq and footprint-seq count tables, whereby RNA-seq counts were divided by the footprint-seq counts.

We used FlyBase to retrieve nucleotide sequences of the RpS5a and RpS5b genes for 13 species (Table S1): D. melanogaster, D. simulans, D. sechellia, D. yakuba, D. erecta, D. takahashii, D. biarmipes, D. ficusphila, D. elegans, D. kikkawai, D. ananassae, D. rhopaloa (RpS5a only) and D. eugracilis (RpS5b only) (Celniker et al., 2002; Kaminker et al., 2002; Misra et al., 2002; Drosophila 12 Genomes et al., 2007; Hoskins et al., 2007; Chen et al., 2014). For genomes where the gene was not annotated, sequences were retrieved after recovering the tBlastN hit of the D. melanogaster proteins sequences to the assembled genome.

Polymorphism data for D. melanogaster, D. simulans and D. yakuba for RpS5b were also obtained from previously sequenced genomes. Polymorphism data for D. melanogaster ribosomal protein RpS5b (CG7014) were obtained from 167 Zambia lines available for download from PopFly Drosophila population genomics browser (https://popfly.uab.cat). Polymorphism data for RpS5b orthologs in D. simulans (GD20391) and D. yakuba (GE26420) were obtained using next-generation sequencing data from 20 lines of each species (https://www.ncbi.nlm.nih.gov/sra accessions: PRJNA215876 and PRJNA215932). Paired-end reads were aligned to D. simulans and D. yakuba reference genomes with BWA, version 0.7.12 (http://bio-bwa.sourceforge.net). Regions of alignments with properly paired reads mapped to genes of interest were used to call variants with BCFtools, version 1.9 (http://www.htslib.org/doc/bcftools.html). Whenever heterozygous sites were called, an allele representing a new polymorphism in a population was retained to reflect the sequence of a given line. Thus, although reconstructed sequences are suitable for counting polymorphic sites, they cannot be used to obtain frequency spectrum of mutations. Aligned sequences are provided in Table S3.

Protein-coding sequences of RpS5a and RpS5b were aligned after translation using ClustalW software (Thompson et al., 1994) in MEGA (Kumar et al., 2016). PAML (Drosophila 12 Genomes et al., 2007; http://abacus.gene.ucl.ac.uk/software/paml.html) and HyPhy (http://www.hyphy.org) packages were used to analyze the rates and mode of evolution of both genes. Branch models from PAML were used to test whether the two genes evolve at different rates (Table S1). A likelihood ratio test was used to compare the model with a single rate (ω, dn/ds) with a model with a separate rate for each gene. Next, site models (NSsites) from PAML were used to test whether any sites in the RpS5b gene evolve under positive selection (Table S2; Nielsen and Yang, 1998; Yang et al., 2000). A likelihood ratio test was used to compare Model M7 (which does not allow for sites under positive selection) with model M8 (which does allow for sites under positive selection). The sequences of RpS5b were also analyzed using the HyPhy package. Fixed effects likelihood (FEL) method (Kosakovsky Pond and Frost, 2005; http://www.datamonkey.org) was performed to detect positively selected codons in RpS5b phylogenies. These codon analyses are presumed to be more realistic than the PAML site models because they allow for synonymous rate variation across sites. Species pairwise McDonald-Kreitman tests (McDonald and Kreitman, 1991) were performed using DnaSP software (http://www.ub.edu/dnasp) (Rozas et al., 2017). Divergence and polymorphism for N- and C-terminal parts of the protein in D. melanogaster, D. simulans and D. yakuba was exported from DnaSP and scored manually for the three-way McDonald-Kreitman tests.

We thank Y. Yamashita, D. Godt, A. Ephrussi, L. A. Nilson, the Developmental Hybridoma Bank and the Bloomington Stock Center for reagents. We thank P. Lasko for sharing data prior to publication. We thank members of the Buszczak lab and M. Osterfield for comments and advice, and O. Cheng with assistance with the figures. We thank E. Greenblatt for advice on the ribosome profiling experiment. We also thank S. Barnes from the UT Southwestern Bioinformatics Core Facility, funded by the Cancer Prevention and Research Institute of Texas (CPRIT, RP150596) for ribosome profiling analysis.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199511