Destabilisation of bam transcripts terminates the mitotic phase of Drosophila female germline differentiation

Adult stem cells divide repeatedly, producing new cells that differentiate to perform specialised functions, maintaining tissue homeostasis. The differentiation process of many adult stem cells includes a phase of transit-amplification, in which mitotic divisions increase the pool of differentiating cells, while minimising the number of divisions of the stem cells themselves (Hsu et al., 2014; Watt, 2001; Fuchs et al., 2004). Limiting stem cell divisions is thought to minimise the intrinsic risks associated with replication errors and uncontrolled proliferation. For the same reasons, the proliferative transit-amplifying phase of differentiation must be tightly controlled to protect against tumourigenesis.

Drosophila female germline stem cells (GSCs), which divide throughout adulthood to produce oocytes, are located in a structure called the germarium at the anterior of the ovary, where they are maintained in a stem cell niche (Fuller and Spradling, 2007). Upon exit from the niche, daughter cells (cystoblasts; CBs) enter differentiation and undergo precisely four mitotic divisions with incomplete cytokinesis to produce a 16-cell cyst (cc) connected by a structure called the fusome, before terminally differentiating into 16-cell egg chambers. The transition from GSC to CB is initiated by the transcriptional upregulation of Bag of marbles (Bam), which is repressed in GSCs by BMP/Dpp signalling from the stem cell niche (McKearin and Spradling, 1990; McKearin and Ohlstein, 1995; Chen and McKearin, 2003a,b; Song et al., 2004). When Bam expression was first observed in the mitotic cells of the germarium, Bam was immediately postulated as part of a mechanism to regulate mitotic division (McKearin and Ohlstein, 1995). However, the essential role of Bam in early differentiation means that bam mutant CBs do not enter differentiation and so an additional role for Bam in the exit from mitosis cannot be easily examined.

Despite this challenge, several studies have pointed to a role for Bam in promoting mitotic divisions. encore mutants, named due to the phenotype of an additional fifth mitotic division, exhibit an expanded domain of bam mRNA (Hawkins et al., 1996). Furthermore, while a bam^−/− mutant can be partially rescued with an inducible heat shock-driven bam construct, ∼50% of rescued egg chambers were found to contain only eight cells (i.e. from three mitotic divisions) (Ohlstein and McKearin, 1997). More recently, inducing overexpression of Bam in a wild type background was shown to generate 32-cell egg chambers, with a stabilised mutated Bam construct exhibiting a larger effect (Ji et al., 2017). Interestingly, overexpression of Cyclin A also leads to 32-cell egg chambers, and this phenotype is partially rescued by reducing the dosage of Bam, which usually stabilises Cyclin A (Lilly et al., 2000; Ji et al., 2017).

Here, we delve further into the regulation of Bam at the exit of mitosis. We show that depleting Bam in the mitotic region leads to the formation of eight-cell egg chambers. Using single molecule fluorescent in situ hybridisation (smFISH) and confocal imaging, we find that bam mRNA is rapidly cleared at the 8cc and early 16cc, and that the bam 3′ untranslated region (UTR) is sufficient for this selective destabilisation. Using a series of genetic tools, we show that Rbp9 is required for destabilising bam mRNA and restricting the domain of Bam protein expression. Given the central role of Rbp9, we examine its regulation during differentiation to reveal an intricate process involving both translation repression and the non-overlapping use of alternative transcription start sites, with the latter being driven downstream of the cytoplasmic isoform of Rbfox1. We suggest a model in which Bam expression upon exit from the niche initiates a self-limiting clock via Rbfox1 and Rbp9, that eventually leads to Bam downregulation and exit from the mitotic phase of differentiation.

It has been shown that ectopic Bam expression leads to an additional mitotic division during GSC differentiation (Ji et al., 2017), but it is not clear whether Bam is required for completing the normal four mitotic divisions. Depleting Bam with bam RNAi throughout germline development (e.g. with a nos-GAL4 driver) results in a complete block of differentiation and a tumourous accumulation of GSC-like cells (Blake et al., 2017). To deplete Bam from differentiating cells without impacting the initiation of differentiation, we used the bam-GAL4 driver, which is only activated upon niche exclusion. With bam-Gal4, bam RNAi is only induced where the bam promoter itself is active, in differentiating cells. In agreement with this, we dissected ovaries and observed no tumourous germaria, suggesting that differentiation is initiated normally. However, we found that 47% and 73% of ovarioles included at least one eight-cell egg chamber at 25°C and 29°C, respectively, which was very rarely observed in the mCherry RNAi controls (Fig. 1A,B). This result shows that lowering Bam expression reduces the number of mitotic divisions during differentiation, and in combination with the previously published results, we conclude that regulation of the Bam expression domain controls the number of mitotic divisions during female GSC differentiation.

To explore how Bam protein expression is regulated, we examined bam mRNA expression using smFISH. This method is an advance on imaging methodologies previously used to visualise bam transcripts, because it can be combined with dyes and antibodies to clearly mark each stage of differentiation. We performed smFISH against bam, alongside staining the spectrosome and fusome (alpha-spectrin antibody), cell boundaries (Phalloidin to label f-actin) and DNA (Hoechst) (Fig. 1C, showing two z-sections of the same germarium, Fig. S1A). As expected, we observed very few bam transcripts in the GSCs, before bam was upregulated to the highest expression in the 4cc (Fig. 1D). This finding differs slightly from the earlier finding that bam is expressed most highly in CBs and 2cc (McKearin and Spradling, 1990), which is likely explained by our ability to simultaneously stain cell markers. At the 16cc stage very few bam transcripts were observed (Fig. 1D). We rarely observed nuclear bam mRNA spots in 16cc and early egg chambers, suggesting that transcription is switched off in late differentiation. Unfortunately, the bam introns are too small to design probes to examine transcription directly.

The region of peak bam mRNA expression matched the reported pattern of Bam protein expression (McKearin and Ohlstein, 1995). To examine Bam protein and mRNA together, we made use of a Bam::GFP transgene line from the FlyFOS collection (Sarov et al., 2016), which includes a GFP tag and all of the endogenous regulatory sequence (Fig. 1E). We performed smFISH against gfp alongside visualising the GFP protein (Fig. 1Fi) and found that Bam::GFP protein is highest in cells with the highest gfp smFISH signal, and is rapidly depleted at later stages of differentiation. Although we were unable to devise a reliable protocol to combine bam smFISH with anti-Bam antibody staining, the wild type samples we imaged were in agreement with the Bam::GFP line (Fig. S1B).

It has been previously reported that the Bam protein contains a highly destabilising PEST sequence (Rogers et al., 1986; McKearin and Spradling, 1990), so to examine the regulatory role of the coding sequence of Bam, we compared the Bam::GFP FlyFOS line with the widely used BamP-GFP transgene (Chen and McKearin, 2003b). BamP-GFP is driven by the bam promoter and uses the bam 3′ UTR, but the coding sequence (CDS) encodes GFP only (Fig. 1E). For this construct, smFISH showed a similar gfp mRNA expression domain as Bam::GFP FlyFOS, but the GFP protein persisted much longer, into the terminally differentiated 16cc (Fig. 1Fii). We conclude that the Bam::GFP fusion is less stable than GFP alone, likely due to the PEST sequence in Bam, though other differences in the protein fusions could play a role. The instability of Bam protein means that bam mRNA level is the primary determinant of Bam protein level. This finding is supported by our observation that Bam protein levels closely follow bam mRNA levels after induction by heat shock and subsequent decay (Samuels et al., 2024).

At the later stages of differentiation, bam mRNA is downregulated either through switching off transcription or increasing decay of bam transcripts. To distinguish between these possibilities, we took advantage of the Bam UTR sensor line (Pek et al., 2009), in which a ubiquitous tubulin promoter drives a transcript encoding GFP followed by the bam 3′ UTR (Fig. 2A, Fig. S2A). The 3′ UTRs contain regulatory sequences, with roles including controlling transcript stability. In ovaries of the Bam UTR sensor line, gfp transcripts were observed in the GSCs (where the tubulin promoter is active, but the bam promoter is not) and throughout differentiation, but with a distinctive absence of cytoplasmic gfp transcripts at 8cc (outlined light blue) and early 16cc (outlined dark blue). Nuclear gfp smFISH signal was observed in 8cc and 16cc (Fig. 2A, Fig. S2B, light blue arrowheads), confirming that the Bam UTR sensor transgene is transcriptionally active in these cells, though we could not determine whether the levels of transcription change in these stages. In later 16cc and early egg chambers, gfp transcripts were again observed in the cytoplasm. This result suggests that the gfp mRNA is specifically unstable in the 8cc and 16cc, via its bam 3′ UTR. The GFP protein persists into the 8cc and 16cc, likely because the GFP protein is more stable than Bam, as discussed above.

To examine bam mRNA stability using an alternative approach, we performed a ‘pulse-chase’ inspired experiment using the heat-shock-bam line, in which the Bam CDS followed by the bam 3′ UTR is transcribed upon heat shock treatment (Ohlstein and McKearin, 1997) (Fig. 2B). We performed a 1 h heat shock (HS) at 37°C, returned the flies to 25°C and then examined bam mRNA after 20 min and after 2 h (Fig. 2B, Fig. S2C). Of note, given the strength of the heat-shock promoter, the exogenous hs-bam-derived transcripts are expressed at a much higher level than endogenous bam. bam was highly expressed in all cells 20 min after HS, and bright transcription foci were clearly observed in 8cc (outlined light blue, arrowhead). However, 2 h after HS, there was a distinct depletion of bam transcripts from the 8ccs (outlined light blue). In later 16cc and early egg chambers, bam transcripts were as stable as in the GSC to 4cc domain. This specific transcript depletion at the 8cc stage (outlined light blue) was not observed in a heat-shock-GFP control with an SV40 3′ UTR.

Collectively, these results suggest that bam mRNA is unstable at the 8cc and early 16cc stages, and this change in stability is mediated by the bam 3′ UTR. Interestingly, bam is stable in late 16cc and early egg chambers, but in wild type ovaries, bam transcripts do not accumulate at these stages. This finding suggests that bam transcription is eventually switched off upon terminal differentiation, supporting our earlier observation that nuclear bam transcription foci are rarely observed at late 16cc and early egg chambers (Fig. 1C).

As bam mRNA is selectively destabilised in the 8cc and early 16cc stages via its 3′ UTR, we reasoned that disrupting the RNA decay machinery would stabilise bam mRNA and expand the domain of bam mRNA expression. Knocking down pacman (pcm, also known as xrn1, an exoribonuclease that degrades decapped RNA) resulted in bam mRNA persistence into the 16cc (Fig. 3A,B,F), as well as cell death and defects in egg chamber cell number (Fig. S3Ai). Knocking down twin (the CCR4 deadenylase) also expanded the region of bam mRNA expression and caused egg chamber cell number defects (Fig. 3C,F, Fig. S3Aii) (Morris et al., 2005; Joly et al., 2013). However, although bam mRNA is specifically degraded at the 8cc and early 16cc stages, these core mRNA decay components are known to be pervasively expressed (Fig. S3B) (Samuels et al., 2024) and likely degrade many different transcripts in the germline. Therefore, it is most plausible that an intermediate protein acts to recognise bam mRNA and direct it to the RNA decay machinery at the correct stage of differentiation.

Rbp9 is a strong candidate for directing bam mRNA decay: Rbp9 depletion was previously shown to expand the domain of Bam protein expression, Rbp9 binds the bam 3′ UTR in vitro, and Rbp9 is upregulated at the 8cc stage (Kim-Ha et al., 1999; Jeong and Kim-Ha, 2004). To determine whether Rbp9 downregulates Bam via RNA stability or translation control, we performed smFISH for bam mRNA on ovaries of rbp9 RNAi. In these samples, we observed persistent bam mRNA in 8cc and 16cc, beyond the usual stage of bam decay (Fig. 3D,F). Together, this result suggests that Rbp9 destabilises bam mRNA, perhaps by recruiting the RNA decay machinery to the bam transcript. It is notable that even in the depletions of rbp9 and decay machinery, the persistent bam mRNA expression domain is eventually downregulated, which is likely explained because bam transcription is switched off in later differentiation.

We have shown that Rbp9 is required for the decay of the bam mRNA, so we asked how Rbp9 itself is regulated during differentiation. Rbp9 is part of a cascade of RNA-binding proteins (RBPs) that are temporally activated during differentiation (Tastan et al., 2010). Once Bam is expressed in the CB, cytoplasmic Rbfox1 (previously A2BP1) is upregulated, followed by Rbp9. Both rbp9 and rbfox1 mutants exhibit a range of phenotypes from germline cystic tumours to egg chambers with 32 cells (Kim-Ha et al., 1999; Tastan et al., 2010) (Fig. S3C). Interestingly, Rbp9 is lost in rbfox1 mutants and the Bam protein expression domain is expanded (Tastan et al., 2010). Similarly, when Rbfox1 was depleted with a germline-specific CRISPR knock out, we found that the region of bam mRNA expression was greatly expanded (Fig. 3E,F). The effect of Rbfox1 germline knockout was greater than any of the RNAi depletions, likely because the CRISPR leads to null mutations compared to incomplete depletion with RNAi. These findings are consistent with a model in which Rbfox1 regulates bam mRNA via Rbp9.

To examine Rbp9 regulation during differentiation, we performed smFISH and found that rbp9 mRNA is slightly upregulated during differentiation (Fig. 4A), with a moderate number of transcripts being reproducibly observed in GSCs and early stages of differentiation when Rbp9 protein is not expressed (Tastan et al., 2010). When Rbfox1 was depleted, the accumulation of rbp9 mRNA was dramatically reduced during differentiation, but the level of rbp9 transcripts was unchanged in the GSCs (Fig. 4A, Fig. S4A).

The rbp9 gene is annotated to have three different transcription start sites (TSSs), each producing a transcript with the same CDS, but a different 5′ UTR (Gramates et al., 2022). We previously used a GSC synchronised differentiation system to show that the middle TTS, TTS #2, of rbp9 is activated only in late differentiation (Samuels et al., 2024). TSS #2 has been shown to be repressed by the insulator-binding protein Su(Hw) during egg chamber formation (Soshnev et al., 2013). To test rbp9 TSS usage in wild type differentiation, we designed hybridisation chain reaction (HCR) probes against each 5′ UTR and observed a near complete switch from TSS #1 to TTS #2 during germline differentiation (Fig. 4B, Fig. S4B). Remarkably, the timing of the switch to TSS #2 corresponds to the appearance of Rbp9 protein (Tastan et al., 2010), suggesting that only the 5′ UTR from TSS #2 allows the translation of Rbp9. TSS #3 did not show cytoplasmic germline transcripts, with some expression in somatic and muscle cells, and some transcription foci in the germline (Fig. S4B). Notably, in this experimental design, signal from a downstream TSS probe may also be picked up in the pre-spliced introns from an upstream TSS.

We asked how loss of Rbfox1 affects rbp9 transcription by testing the TSS usage in the Rbfox1 germline-specific knockout. Despite the observation of branched fusomes identifying 8cc, TSS #2 was never activated during differentiation (Fig. 4B). It is likely that the loss of the TSS #2 isoform upon Rbfox1 depletion prevents the upregulation of the Rbp9 protein (Tastan et al., 2010). Interestingly, in the rbfox1 depletion, TSS #1 was largely switched off in cysts with branched fusomes (light blue outline), suggesting that the two TSSs are regulated independently and the switching off of TSS #1 does not require Rbfox1.

One step upstream, Rbfox1 itself is upregulated during differentiation via alternative splicing, which leads to the production of an isoform that lacks the nuclear localisation signal (Tastan et al., 2010; Carreira-Rosario et al., 2016). We previously showed that the introduction of Bam protein in a bam^−/− mutant is sufficient to drive this switch (Samuels et al., 2024). Altogether, our data, combined with the previously published data, supports our proposed model (Fig. 4C) in which induction of Bam expression initiates a series of events that eventually terminate the Bam expression domain and the mitotic phase of differentiation. At the onset of differentiation in the CB, BMP-mediated repression of bam transcription is released. Bam drives an alternative splicing event to upregulate cytoplasmic Rbfox1, which then upregulates Rbp9 via switching to a downstream TSS, including an alternative 5′ UTR allowing Rbp9 translation. Rbp9 then binds to the bam mRNA via its 3′ UTR to initiate RNA decay, and the intrinsic instability of Bam protein via the PEST sequence limits the domain of Bam protein expression, resulting in cells exiting the transit-amplifying mitotic phase.

Expression of Bam in a synchronised GSC differentiation system drives two sequential waves of gene expression, with the first wave being enriched for genes involved in DNA replication and the cell cycle (Samuels et al., 2024). Previously, we suggested that the loss of Bam is crucial for allowing cells to enter into the second wave of gene expression changes. This idea is supported by the findings presented here: Bam must be depleted to complete the mitotic phase of differentiation in the wild-type female germline. This is in contrast to the findings in males, in which a threshold level of Bam accumulated is required for entry to terminal differentiation (Gönczy et al., 1997; Insco et al., 2009).

It is unclear from our experiments whether the proposed self-limiting Bam regulatory mechanism is tuned to ‘count’ division numbers or to ‘time’ the proliferative phase. These possibilities might be distinguished with experiments to manipulate the cell division rate. The proposed self-limiting model for Bam expression involves the regulation of gene expression at multiple levels, including transcription, splicing, mRNA stability, translation control and protein stability. In the ‘timing’ model, we speculate that the different regulatory mechanisms could influence the speed of the clock: for example, the pace by which transcriptional upregulation or alternative splicing may result in functional changes is probably slower than that of the regulatory layer involved in RNA or protein decay.

Our model most likely overlooks various unknown intermediate regulators – for example the cytoplasmic RBP Rbfox1 likely mediates the nuclear rbp9 TSS change indirectly, perhaps via downregulation of Su(Hw) at the 8cc/16cc stage (Soshnev et al., 2013). The model is further simplified in its omission of additional downstream targets of each regulator, which may have effects beyond the regulation of Bam. For example, Rbfox1 binds to the translational repressor Bruno (Sugimura and Lilly, 2006; Wang and Lin, 2007; Tastan et al., 2010), and destabilises pumilio mRNA (Carreira-Rosario et al., 2016).

It is unclear how the interactions between regulators may differ between tissues: while Bam is germline-specific, both Rbfox1 and Rbp9 are highly expressed in the Drosophila nervous system (modENCODE; Brown et al., 2014). Rbp9 is closely regulated to the Drosophila neuronal RBPs Found in neurons (Fne) and Elav, as well as the human ELAVL/Hu family of proteins, which play diverse roles in RNA regulation including splicing, 3′ UTR lengthening and RNA stability (Mulligan and Bicknell, 2023). In both fly and human, this family of RBPs is widely involved in differentiation, emphasising the importance of RNA regulation in stem cells and their progeny (Yao et al., 1993; Grassi et al., 2019; Kota et al., 2021).

It is intriguing that the protein which initiates differentiation must be removed to end the transit-amplifying phase and allow entry into terminal differentiation. We speculate that the self-limiting mechanism regulating the Bam expression domain is protective, preventing the tumourous growth downstream of an uncontrolled proliferative phase caused by unlimited Bam expression. The system is made further robust through a layer of transcriptional regulation of bam. Other master differentiation factors are also switched off in mature terminally differentiated cell types, such as Prospero in differentiating Drosophila neurons (Spana and Doe, 1995) and MyoD in differentiating muscle (Hinterberger et al., 1991), suggesting that this could be a widespread mechanism ensuring unidirectional differentiation and preventing tumourigenesis.

Unless otherwise stated, stocks and crosses were maintained on standard propionic food at 25°C for experiments. For heat shocks, flies were incubated in pre-warmed vials for 1 h at 37°C and then transferred to fresh vials at 25°C for either 20 min or 2 h before dissection. The Drosophila melanogaster stocks used can be found in Table S2.

smFISH probes against bam, egfp, sfgfp and rbp9 were designed with the Stellaris Probe Designer (Biosearch Technologies) (Table S1). Oligos were labelled with ATTO 633 as described by Gaspar et al. (2017). smFISH and antibody staining was performed as in Samuels et al. (2024): ovaries were dissected in PBS, fixed for 25 min at room temperature (RT) in 4% formaldehyde in 0.3% PBSTX (0.3% Triton-X in 1× PBS) and then washed 3×15 min in 0.3% PBSTX. Samples were transferred to Wash buffer [2× saline sodium citrate (SSC), 10% deionised formamide in nuclease-free water] for 10 min at RT. Hybridisation buffer [2× SSC, 10% deionised formamide, 20 mM vanadyl ribonucleoside complex, 0.1 mg/ml bovine serum albumin (BSA), competitor (1:50 dilution of 5 mg/ml E. coli tRNA and 5 mg/ml salmon sperm ssDNA) in nuclease-free water] was prepared and smFISH probes, primary antibodies and Phalloidin (Alexa Fluor 405 or 488 Phalloidin, Thermo Fisher Scientific) were added. Ovaries were incubated in this Hybridisation buffer at 37°C overnight in the dark, then washed 3×15 min in Wash buffer. Samples were incubated with secondary antibodies in Wash buffer for 2 h at RT, then finally washed in Wash buffer, with the addition of Hoechst in one wash step. Finally, samples were mounted in VectaShield mounting media (Vector Laboratories).

For bam smFISH with anti-Bam antibody, the protocol was adapted by performing the smFISH without any primary antibodies first in an overnight step, then transferring samples to 0.3% PBSTX. Ovaries were incubated with primary antibody in Block (0.1 mg/ml BSA in 0.3% PBSTX) overnight at 4°C, washed and then incubated with secondary antibody in Block for 2 h at RT before final washes and mounting. The antibodies used were: alpha-spectrin (mouse) [1:100, Developmental Studies Hybridoma Bank (DSHB), #3A9(323 or M10-2)], Bam (mouse) (1:20, DSHB, bam), GFP Booster ATTO 488 (1:500, Chromotek, gba488-100) and Cy3-AffiniPure Donkey Anti-Mouse IgG (H+L; 1:500, Stratech, 715-165-150-JIR).

Split-initiator probes for version 3.0 HCR (Molecular Instruments) were designed according to the parameters described by (Choi et al., 2018) (Table S1). Ovaries were dissected in PBS and fixed for 25 min at RT in 4% formaldehyde in 0.2% PBSTX (0.2% Triton-X in 1× PBS). Ovaries were washed 3×15 min in 0.2% PBSTX, then samples were transferred to Hybridisation buffer (Molecular Instruments) for 30 min at 37°C. We added 1.6 μl of 1 μM probes stock for each probe to 100 μl of Hybridisation buffer and this was incubated with ovaries at 37°C overnight. On the second day, ovaries were washed with pre-warmed Probe Wash Buffer for 4×15 min at 37°C. Samples were then washed 3×5 min in 5× SSCT (5× SSC in 0.1% Tween in water) at RT and then left in 5× SSCT until the evening. Samples were incubated in pre-amplification buffer for 30 min at RT and the hairpins (B2-488, B2-594, B3-647) were prepared according to manufacturer's instructions (heating to 95°C and then cooling to RT; Molecular Instruments). Samples were incubated with hairpins in amplification buffer overnight in the dark at RT. Ovaries were finally washed in 3×15 min in 5× SSCT, with the addition of Hoechst in one wash step. Finally, samples were mounted in VectaShield mounting media (Vector Laboratories).

Images were acquired on a Leica SP8 confocal microscope with a 20× dry objective or 40× oil objective. All imaging experiments were performed for at least three replicates. Image processing was performed using Fiji (Schindelin et al., 2012). Cell counting of egg chambers (Fig. 1B) was done visually on the microscope based on DAPI staining. smFISH quantitation for Figs 1D and 3F was performed by taking average intensity measurements at the midpoint of each cyst based on Phalloidin-labelled cell boundaries. Background measurements were taken from each image and removed from intensity values. Values were then normalised to intensity in the GSC for each sample. smFISH quantitation of rbp9 in the GSCs (Fig. S4A) was performed by counting spots at the midpoint of each GSC, based on Phalloidin-labelled cell boundaries. Statistical significance was calculated using ANOVA or t-tests as specified in figure legends.

We thank Toshie Kai, the Vienna Drosophila Resource Center, and the Bloomington Drosophila Stock Center for fly reagents, and Eleanor Blake for assistance with the HCR protocol. We also gratefully acknowledge the help of the following University of Cambridge facilities: the Department of Genetics Fly Facility, and Dr Ian Clark, Dr Antonina J. Kruppa, Dr Ben Sutcliffe and Dr Jonathan D. Howe from the Department of Genetics Imaging Facility.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.204405.reviewer-comments.pdf