Plasticity of Drosophila germ granules during germ cell development

Compartmentalization of RNAs and proteins into membraneless structures called granules is a ubiquitous mechanism for organizing and regulating cohorts of RNAs. Germ granules are ribonucleoprotein (RNP) assemblies required for germline development across the animal kingdom, but their regulatory roles in germ cells are not fully understood. We show that after germ cell specification, Drosophila germ granules enlarge through fusion and this growth is accompanied by a shift in function. Whereas germ granules initially protect their constituent mRNAs from degradation, they subsequently target a subset of these mRNAs for degradation while maintaining protection of others. This functional shift occurs through the recruitment of decapping and degradation factors to the germ granules, which is promoted by decapping activators and renders these structures P body-like. Disrupting either the mRNA protection or degradation function results in germ cell migration defects. Our findings reveal plasticity in germ granule function that allows them to be repurposed at different stages of development to ensure population of the gonad by germ cells. Additionally, these results reveal an unexpected level of functional complexity whereby constituent RNAs within the same granule type can be differentially regulated.


Introduction
Ribonucleoprotein (RNP) granules are biomolecular condensates containing RNAs and RNAbinding proteins that create cytoplasmic compartments without the use of membranes. Compartmentalization in granules concentrates RNAs together with regulatory proteins, such as those involved in RNA localization, translational control, RNA processing, and control of mRNA stability [1][2][3]. Thus, RNP granules are hypothesized to be hubs of posttranscriptional regulation. Several types of RNP granules-such as processing bodies (P bodies), stress granules, and neuronal transport granules-are found in many different cell types and species, suggesting a common and highly conserved regulatory strategy [1,4].
One class of RNP granules, called germ granules, is a characteristic feature of germ cells across animal species [5]. Germ granules contain mRNAs and proteins that are necessary for germ cell development and are therefore thought to play important roles in germline development and function [6,7]. In some animals, like Drosophila, Xenopus, and zebrafish, germ granules form during oogenesis from maternally expressed proteins and RNAs. During early PLOS Biology | https://doi.org/10.1371/journal.pbio.3002069 April 13, 2023 1 / 28 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 embryogenesis, these maternally supplied granules are segregated to a subset of cells that will give rise to the germline. In other animals, including mammals, germ granules are not acquired maternally, but similar granules containing necessary germline determinants form de novo after the germ cells have been specified [5,8]. Despite differences in how germ granules arise, shared components like the RNA-binding protein Nanos (Nos), Tudor (Tud) domain proteins, and RNA helicases [5,9,10] suggest a conserved role in RNA metabolism.
In Drosophila, the germ granules form at the posterior of the oocyte within a specialized cytoplasm called the germ plasm. There, the germ granule proteins Oskar (Osk), Vasa (Vas), Tud, and Aubergine (Aub) form a scaffold [11][12][13] to recruit mRNAs that are necessary for germ cell development, including nanos (nos), polar granule component (pgc), and Cyclin B (CycB) [14][15][16]. Within the germ granules, these RNAs self-associate to form spatially distinct homotypic clusters [17][18][19]. The germ granules become anchored to the posterior cortex by the end of oogenesis and persist there into embryogenesis.
During embryogenesis, the germ plasm induces the formation of membrane buds around nuclei at the posterior of the syncytial embryo, which then pinch off to form the germ cell progenitors-called pole cells [20]. The germ granules accumulate around these nuclei and their associated centrosomes by dynein-dependent transport. Co-packaging of many different mRNAs into each granule ensures efficient trafficking and segregation of mRNAs to the pole cells as they bud and divide [21]. Germ granules also play a role in stabilizing constituent RNAs during the maternal to zygotic transition (MZT), when a majority of maternal mRNAs are degraded in the somatic region of the embryo. This degradation allows zygotic transcription to be activated, which is necessary for cellularization and differentiation of the soma and for body patterning [22,23]. In contrast, numerous transcripts, including important germline determinants, are stabilized in the germ plasm [23][24][25]. Although not all stabilized mRNAs are localized to the germ granules, sequestration within germ granules may be a mechanism to stabilize a subset of these RNAs by making them less accessible to mRNA decay factors.
After the pole cells have formed and their contents are physically separated from the soma, the need for protection from the somatic MZT is eliminated. Furthermore, once the pole cells enter mitotic quiescence in advance of gastrulation [26], the role of germ granules in distributing mRNAs to daughter cells is no longer required. However, the germ granules persist into later stages of embryogenesis [27]. Interestingly, germ granules increase in size in the pole cells [28][29][30]. The significance of this morphological change and what roles the germ granules play throughout the remainder of germ cell development have yet to be determined. Since germ granules are a conserved feature of differentiated germ cells, deciphering their regulatory functions at these stages is of particular interest.
Here, we have investigated a role for germ granules in regulating mRNA stability in pole cells. We found that during a period when germ granules gain the ability to fuse with each other, they sequentially recruit mRNA decay factors typically found in P bodies, and nos and pgc are destabilized. In contrast, CycB is maintained throughout embryogenesis, despite residing within the same granules. These findings suggest that in contrast to their protective role prior to pole cell formation, germ granules play a more complex role in pole cells, selectively protecting some mRNAs while promoting the degradation of others. Overexpression of an activating subunit of the decapping complex, Decapping protein 1 (DCP1), disrupts the protection of CycB, suggesting DCP1 levels are limiting for CycB degradation. We show that the decapping activators Edc3 and Patr-1 are necessary to localize the decapping complex to germ granules and disruption of decapping complex recruitment leads to aberrant stabilization of nos and pgc. Furthermore, disrupting the selective mRNA protection or degradation by germ granules leads to defects in pole cell migration, suggesting both of these germ granule functions are necessary for proper gonad formation. Overall, these findings reveal a shift in germ focused on the cytoplasmic granules, as only they contain mRNAs [30]. During nc9, these granules appear as diffraction limited spots that cluster around the budding nuclei ( Fig 1A). A few larger germ granules, with diameters between 500 and 700 nm, first begin to appear in the pole cells starting at nc12, approximately 10 min after pole cell formation (Fig 1A-1D). Over the next 90 min, there is a trend toward larger granules such that by the end of nc14, most granules appear much larger than those first segregated to the pole cells. The majority of these granules are at least 1 μm in diameter, with some reaching over 1.5 μm (Fig 1E and 1F).
At the end of nc14, gastrulation begins. Cellular movements carry the pole cells into the posterior midgut primordium, where they respond to chemotactic cues directing them to migrate through the midgut epithelium and attach to adjacent mesoderm cells. Once aligned, the germline and mesodermal cells migrate and coalesce to form the gonad [34]. In the gonad, the pole cells resume cell division and ultimately generate the germline stem cells capable of producing eggs or sperm. We visualized Osk-sfGFP throughout these migratory movements to determine how long germ granules persist. Both large (�1 μm diameter) and small granules are visible as the pole cells begin to migrate through the midgut and toward the presumptive mesoderm (Bownes stage 9). These granules persist throughout gastrulation and at least until the end of pole cell migration at stage 14 ( Fig 1G-1L). Similar results were obtained using another endogenously tagged germ granule marker, Vas-EGFP [35] (S1 Fig). Their persistence suggests that the larger granules are stable and that the germ granules could play a role in germ cell development throughout embryogenesis.

Germ granules grow through fusion in the pole cells
During this period of germ granule growth, we observed a decrease in the number of granules that coincided with their increase in size (Fig 1). Therefore, we hypothesized that germ granules enlarge via fusion of smaller granules. To test this hypothesis, we performed time lapse confocal imaging of Osk-sfGFP during nc14, when the majority of germ granule growth occurs. Throughout nc14, we observed examples of apparent fusion between 2 granules, occurring over the course of 2 to 3 min (Fig 2A-2E, S1 Video). These events are slower than fusion of the more liquid-like germ granules in Caenorhabditis elegans, which occurs in seconds [36].
Furthermore, this analysis cannot distinguish true fusion from granules docking together without exchanging their materials. To confirm that fusion does occur, we endogenously tagged Osk with the photoconvertible fluorescent protein Dendra2 at its C terminus. Osk-Dendra2 was then photoconverted from green to red (shown here as green to magenta) within a small region of a pole cell to generate differentially labeled germ granules ( Fig 2F) that were tracked using time lapse imaging. During nc14, we observed almost entirely photoconverted (red) and unconverted green granules that appear to "stick" together for a period of about 1 to 3 min without mixing (Fig 2F-2H, S2 and S3 Videos). After this initial period, the 2 colors begin to mix, resulting in a granule with green and red signals distributed uniformly throughout (Fig 2I-2K, S2 and S3 Videos). Together, these data suggest that germ granule growth at nc14 occurs at least in part by the slow fusion of smaller granules. Since fusion has not been observed at earlier stages [30,37], these results suggest that the germ granules are more dynamic at nc14.

nos and pgc become destabilized in the pole cells while CycB is protected
To determine if enlargement of granules is accompanied by changes in their composition, we performed single-molecule fluorescence in situ hybridization (smFISH) analysis, which revealed that some of the larger granules at nc14 lack nos and pgc mRNA in contrast to earlier nuclear cycles when granules almost always contain at least 1 of these 2 mRNAs (Fig 3A). Since nos and pgc do not appear outside of germ granules at nc14, these seemingly "empty" granules suggest that the enlargement of germ granules in nc12-14 may coincide with a loss of granule mRNAs. To test this hypothesis, we measured total mRNA levels per pole cell of nos, pgc, and another abundant germ granule RNA, CycB [17,38], during each nuclear cycle, starting with pole cell budding at nc9. mRNAs were detected by smFISH and their levels were normalized to either endogenously tagged Osk-sfGFP or Vas-EGFP ( Fig 3B) to account for the overall decrease in the amount of germ plasm per pole cell that occurs as pole cells divide prior to gastrulation.
Both nos and pgc levels decline modestly (25% and 19%, respectively) between nc9 and nc10 (Fig 3C and 3D). This decrease could be due to incomplete protection from the somatic MZT since the pole buds and somatic nuclei are still in a common cytoplasm until the end of nc10 when the pole cells cellularize. After this initial decrease, nos and pgc levels remain relatively constant until nc14, when they drop dramatically, by 80% and 54%, respectively ( Fig  3B-3D). In contrast, CycB levels remain constant throughout this period (Fig 3B and 3E).
Because CycB is more abundant in the germ plasm than nos or pgc (approximately 1.8-fold and 6.5-fold, respectively) [38], CycB might appear to be stable if a constant rate of mRNA decay depletes the larger pool of CycB more slowly. Decreasing CycB levels by 25% at nc10-11 using a chromosomal deficiency (S2A Fig) has no effect on protection of CycB (S2B Fig), however. Conversely, we evaluated the effect of increasing nos levels using a nos-egfp transgene [39]. Whereas the total nos RNA level in nc10-11 nos-egfp embryos is 1.6-fold higher than in the wild type (Fig 3F), the fraction of nos remaining by nc14 is indistinguishable (Fig 3G). Together, these results support the conclusion that CycB is selectively stabilized.
To determine how long CycB remains stable, we quantified mRNA levels in the pole cells throughout their migration, until the pole cells coalesce in the gonad at stage 14. Although CycB levels do decrease significantly at stage 8, the magnitude of this decrease is much less than that of nos and pgc (34% for CycB versus 85% for nos and pgc) (Fig 3H-3J). After stage 8, nos and pgc levels continue to decline to 3% and 0.25% of their stage 4 levels, respectively ( Fig  3H and 3I), while CycB levels remain steady until stage 12 ( Fig 3J). The small but significant increase in CycB at stage 12 may result from zygotic transcription, suggesting that zygotic transcripts can accumulate in germ granules. Due to the selective protection of CycB throughout embryogenesis, a greater fraction of germ granules at stage 14 contain CycB compared to nos and pgc, despite these mRNAs occupying the same fraction of germ granules when the pole cells initially form (Fig 3K and 3L).

The mRNA decay machinery is sequentially recruited to germ granules
The loss of nos and pgc could result from their selective degradation within the germ granules, or from their selective release and subsequent degradation in the cytoplasm. To distinguish between these two possibilities, we visualized proteins involved in each major step of the 5 0 to 3 0 mRNA decay pathway by immunofluorescence during the period when nos and pgc degradation begins. CCR4, a component of the CCR4-NOT deadenylation complex, forms puncta that do not colocalize with the germ granules at any point during nc9 to nc14 (S3 Fig), suggesting that deadenylation is not occurring in the germ granules at these stages and may have preceded pole cell formation. At nc10, DCP1, an activating subunit of the mRNA decapping Quantification of the number of nos, pgc, and CycB puncta relative to the number of germ granules in nc10 and stage 14 embryos, n = 5-6 embryos per stage. One-way ANOVA with Tukey-Kramer post hoc tests were performed at each time point to compare the 3 mRNAs. Individual data points and mean (C-E, H-J) or mean ± SD (F, G, L) are shown. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, and n.s., not significant. Source data for the graphs in Fig 3C-3J and 3L are provided in S1 Data. Scale bars: 10 μm. smFISH, single-molecule fluorescence in situ hybridization.
https://doi.org/10.1371/journal.pbio.3002069.g003 complex, and Pacman (Pcm), the Drosophila homolog of the 5 0 to 3 0 XrnI exonuclease, form puncta in the pole cells that do not overlap with germ granules (Fig 4A and 4B). However, colocalization of germ granules with DCP1 can be detected beginning at nc12. At this time, 1 to 2 germ granules per pole cell appear to colocalize with DCP1 ( Fig 4A). Interestingly, this initial colocalization occurs at the same nuclear cycle when larger germ granules first appear ( Fig  1D). The overlap between DCP1 and germ granules further increases in nc13 and nc14 such that 30% of granules associate with DCP1 (Fig 4A and 4C). Pcm follows a similar pattern, but its recruitment to germ granules is delayed by 1 nuclear cycle relative to DCP1. Pcm is first detected in a few germ granules per pole cell at nc13 (Fig 4B). Colocalization increases at nc14, when approximately 30% of granules are associated with Pcm (Fig 4B and 4D). The finding that a decapping co-factor and the Pcm exoribonuclease associate with germ granules just before nos and pgc levels decrease suggests that germ granules play a role in promoting mRNA degradation in pole cells, which contrasts with their stabilizing role in early embryos.

DCP1 is not recruited to homotypic clusters within germ granules
The presence of mRNA decay pathway proteins in germ granules raises the question of how nos and pgc can be targeted for degradation, while CycB RNA in the same granules is not. The organization of granule mRNAs in homotypic clusters suggests that proteins involved in the mRNA decay pathway may be selectively recruited to clusters of some RNAs, but not to others. We therefore performed stimulated emission depletion (STED) microscopy to visualize the distribution of DCP1 in relation to RNAs within germ granules before and after germ granules increase in size. Before pole cell formation, we detected nos, pgc, and CycB in spatially distinct homotypic clusters (S4A Fig), as has been previously described in the germ granules of early embryos [18,19]. By contrast, no separation of probes was detected when 2 differentially labeled probes for nos were used simultaneously. Occasionally, up to 2 homotypic clusters of nos and/or pgc were observed in the same granule (S4A Fig). In the larger granules at nc14, nos, pgc, and CycB remain confined to non-overlapping puncta (Figs 5A and 5B, S4B). On average, these puncta are brighter than the single transcripts detected in the bulk cytoplasm (S4C Fig), as has been found for clusters in earlier germ granules [18,19], suggesting that they contain multiple copies of the same transcript and are therefore homotypic clusters. Individual granules contain multiple clusters of CycB, pgc, and nos, with the number of distinct puncta per granule ranging from 2 to 12. This increase in the number of clusters in larger granules is consistent with granule growth through fusion. Surprisingly, in both large and small granules, most DCP1 puncta are spatially separated from nos and CycB during nc13 and nc14 (Figs 5C and S4D). Therefore, DCP1 localization to homotypic clusters is not required for degradation of germ granule mRNAs.
An alternate hypothesis is that the levels or activity of decapping proteins are limiting, causing the decapping complex to preferentially target mRNAs with higher binding affinity for the complex or a decapping regulatory factor. To test if DCP1 levels are limiting, we overexpressed DCP1. To avoid potential effects of excess DCP1 on germ granule assembly during oogenesis, we fused the DCP1 coding region to the smaug (smg) 3 0 UTR, which inhibits smg translation until egg activation [40] (S5A- S5C Fig). DCP1-smg3 0 UTR RNA was expressed under GAL4/UAS control using a strong maternally active GAL4 driver, matα-GAL4. STED analysis of embryos overexpressing DCP1 showed that within individual germ granules, DCP1 still accumulates in distinct puncta. While many of these puncta are separated from nos and CycB, there is greater overlap between DCP1 puncta and CycB as compared to the wild type (Fig 5C and 5D).
To determine the effect of DCP1 overexpression on CycB stability, we analyzed CycB levels in the pole cells within the gonad by smFISH. In order to maximize the difference in DCP1 levels, we compared embryos expressing DCP1-smg3 0 UTR to embryos heterozygous for a null DCP1 allele [41]. In DCP1-smg3 0 UTR embryos, the initial level of CycB in the pole cells is unchanged ( S5E Fig). However, there is a small but significant reduction in CycB levels at nc14 (S5F Fig), and a greater decrease in the gonad (Fig 5E and 5F) when compared to DCP1 . The percent of cytoplasmic Vas puncta that colocalize with DCP1 (C) and Pcm (D) puncta was quantified at each nc, n = 5-7 embryos per nc. Nuclear Vas puncta were masked using Imaris software. Values for individual embryos and means are shown. *p < 0.05, **p < 0.01, ***p < 0.001 as determined by Welch ANOVA with Dunnett's T3 post hoc test. Source data for the graphs in Fig 4C and 4D are provided in S1 Data. Scale bar: 5 μm.
https://doi.org/10.1371/journal.pbio.3002069.g004 Notably, DCP1 overexpression causes pole cell migration defects, with an increased number of "lost" pole cells that fail to reach the gonad (Fig 5G and 5H). Since DCP1-smg3 0 UTR RNA is translated throughout the embryo (S5A Fig), this effect on pole cell migration could be due to excess DCP1 in the soma or in the pole cells. To distinguish between these possibilities, we fused DCP1 to the nos3 0 UTR, which targets DCP1 synthesis to the germ plasm (S5A- S5C  Fig). Similarly to DCP1-smg3 0 UTR embryos, DCP1 is more diffuse within the granules of DCP1-nos3 0 UTR embryos, CycB is aberrantly degraded after the pole cells form, and there is an increased frequency of lost pole cells (S5D, S5E and S5J- S5L Fig). This migration defect indicates that the protection of one or more germ granule mRNAs is likely important for proper pole cell function, although an indirect effect on pole cell migration due to excess DCP1 in the posterior soma cannot be completely ruled out.

Edc3 and Patr-1 promote recruitment of the decapping complex to germ granules
The observation that mRNA decay factors are recruited to the germ granules, beginning with the decapping complex at nc12 raises the question of why and how they are recruited after pole cell formation. To better understand the mechanism of this recruitment, we knocked down the expression of known regulators of mRNA stability by RNAi expressed using matα-GAL4 and evaluated the effect on DCP1 localization to germ granules marked with Osk-sfGFP. The efficacy of the RNAi was confirmed by RT-qPCR ( By contrast, knockdown of two decapping regulators, edc3 and patr-1 (also called hpat), leads to a decrease in colocalization of germ granules with DCP1 at nc13-14 by 29% and 19%, respectively (Fig 6A and 6B). DCP1 levels are unaffected (S6C Fig), suggesting impaired recruitment of the decapping complex. Since Edc3 and Patr-1 share several binding partners [42,43], we hypothesized that they play partially redundant roles in recruiting the decapping complex to germ granules. Consistent with this hypothesis, simultaneous knockdown of edc3 and patr-1 decreases the frequency of germ granule colocalization with DCP1 by 46% (Fig 6A  and 6B), without affecting DCP1 levels (S6C Fig). Interestingly, edc3+patr-1 knockdown also appears to reduce granule size at nc14 (Fig 6A), suggesting that decapping activity may be required for germ granule growth.
Since Edc3 and Patr-1 can each bind to several components of the decapping complex [42,43], we sought to determine if Edc3 and Patr-1 recruit DCP1 to the germ granules as part of the decapping complex. Therefore, we asked whether Edc3 and Patr-1 localize to germ granules during the period of recruitment. Colocalization of germ granules with Edc3 puncta was not detected (Fig 6C), suggesting Edc3 is not a component of the decapping complexes being recruited to the germ granules. Therefore, its effect on recruitment is likely indirect. By contrast, a few germ granules per cell colocalize with Patr-1 at nc12, followed by further increases at nc13 and nc14 such that 30% of granules become associated with Patr-1 (Fig 6C and 6D). This pattern of localization closely mirrors that of DCP1 (Fig 4A and 4C), supporting the RNAi, patr-1 RNAi, and edc3 and patr-1 double RNAi embryos expressing an osk-sfgfp transgene. Osk-sfGFP was detected by direct fluorescence (green) together with anti-DCP1 immunofluorescence (magenta). Enlargements of the boxed regions show germ granules that recruit DCP1 in control embryos and granules that fail to recruit DCP1 in RNAi embryos. Yellow arrows indicate germ granules in the RNAi embryos that recruit DCP1. (B) Quantification of the percent of cytoplasmic Osk-GFP puncta that colocalize with DCP1 in control and RNAi embryos. Genotypes were compared by Ordinary one-way ANOVA and Dunnett's multiple comparison test, n = 10-18 embryos per genotype. (C) Single confocal sections of the pole cells at nc14 in embryos expressing Vas-GFP (green). Vas-GFP (green) was detected by direct fluorescence and Edc3 and Patr-1 (magenta) were detected immunofluorescence. Enlargements of the boxed regions show Patr-1 (cyan box), but not Edc3 (yellow box) puncta, overlap with germ granules. (D) The percent of germ granules, marked by cytoplasmic Vas-GFP, that colocalize with Patr-1 was quantified from nc10 to nc14. Nuclear cycles were compared by Ordinary oneway ANOVA and Dunnett's multiple comparison test, n = 5-6 embryos per nuclear cycle. Nuclear puncta of Osk (B) or Vas (D) were masked using Imaris software. (E) The proportion of nos and pgc remaining in the pole cells at nc14 was quantified in control and double RNAi embryos. nos and pgc were detected by smFISH and their total intensities at nc14 were normalized to their average intensities during nc9-13. Genotypes were compared by Mann-Whitney test, n = 11-12 embryos. (F) Pole cells were detected by Vas immunohistochemistry in control and double RNAi embryos. The lost pole cells at Bownes stages 13 and later were counted, n = 103-106 embryos per genotype. Genotypes were compared by Mann-Whitney test. Individual data points and mean values (D, F) or mean ± SD (B, E) are shown; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Source data for the graphs Fig 6B and 6D-6F are provided in S1 Data. Scale bars: 5 μm. smFISH, single-molecule fluorescence in situ hybridization. https://doi.org/10.1371/journal.pbio.3002069.g006

PLOS BIOLOGY
Developmental plasticity of germ granules hypothesis that Patr-1 functions as part of the decapping complex to promote DCP1 recruitment to germ granules.
One known binding partner of Patr-1, the conserved DEAD-box helicase Me31B [42], is a component of the germ granules in the oocyte and newly fertilized embryo [44,45]. To determine if Patr-1 could promote granule localization through its interactions with Me31B, we investigated whether Me31B localizes to germ granules in the pole cells. Prior to and during pole cell budding, Me31B is present throughout the posterior of the embryo, but does not appear to be associated with germ granules (Fig 7). At nc11, Me31B accumulates at pole cell nuclei in a pattern similar to but more diffuse than that of Vas, consistent with enrichment in

PLOS BIOLOGY
Developmental plasticity of germ granules germ granules. This enrichment persists into nc14 (Fig 7). Although the diffuse signal precludes quantification, Me31B appears to accumulate in germ granules before Patr-1 (Fig 6C  and 6D) and DCP1 (Fig 4A and 4C) and throughout their recruitment period. Thus, Me31B could act upstream of Patr-1 to promote DCP1 recruitment. However, because oogenesis is arrested in me31B mutants [46], this role could not be directly tested.

Recruitment of decapping factors to germ granules requires translation
Our results suggest that there is a temporally regulated, ordered recruitment of the mRNA decay machinery to germ granules. We therefore sought to determine the nature of the trigger that initiates this recruitment. Because the pole cells are transcriptionally quiescent during this time [47], the initiation of degradation in germ granules must be triggered by a protein translated during this period and/or posttranslational modification of a protein in the pole cells. To distinguish between these possibilities, we tested the dependence of DCP1 recruitment to the germ granules on translation, by injecting the translational inhibitor cycloheximide (CHX) into the posterior of the embryos prior to nc12 and monitoring DCP1 distribution. Although CHX can cause dissociation of P bodies, a previous study showed that they are resistant to CHX at this stage [48] and similarly, we detected DCP1 puncta in both the soma and pole cells after CHX injection ( Fig 8A). Moreover, CHX injection does not affect the overall DCP1 level ( Fig 8B). However, DCP1 fails to localize to germ granules in CHX injected embryos (Fig 8A  and 8C). Therefore, recruitment of DCP1 to germ granules depends on translation.
One possible explanation for this observation is that translation of germ granule mRNAs makes them vulnerable to DCP1 binding and degradation, such as through gradual shortening of the poly(A) tail. This hypothesis is consistent with the protection of CycB RNA, as CycB is translationally repressed during this period [49,50]. However, in nos mutant embryos, where CycB is aberrantly translated in the pole cells during nc14 [49,50], CycB levels remain stable (S7A and S7B Fig). Furthermore, a nos RNA that cannot bind translational repressors and produces excess Nos protein [51] is slightly more stable than wild-type nos (S7C Fig). Therefore, translational activity is not sufficient to target germ granule RNAs for degradation. Alternatively, translation may be required to generate a critical threshold of a regulatory protein, in much the same way that synthesis of Smg has been proposed to time events in the soma at the MZT [52]. patr-1 and edc3 transcripts are present in pole cells [24] suggesting that accumulation of Edc3 and/or Patr-1 could trigger the recruitment of DCP1 to the germ granules. Immunofluorescence analysis showed that there is no significant difference in Edc3 or Patr-1 in the pole cells before and after nc12, however (S8A and S8B Fig). Similarly, Me31B levels are unchanged (S8C Fig). Therefore, the timing of decapping complex recruitment does not appear to be regulated by synthesis of Edc3, Patr-1, or Me31B. Rather Edc3, Patr-1, and potentially Me31B, act downstream of the trigger to promote efficient recruitment.

Edc3 and Patr-1 are necessary for proper pole cell development
Finally, we investigated the functional importance of DCP1 recruitment to germ granules in pole cells. Because oogenesis is arrested in DCP1 mutants [41], we could not test a requirement for DCP1 directly. Therefore, we took advantage of the defect in DCP1 recruitment in edc3 +patr-1 double-knockdown embryos to ask whether DCP1 recruitment is required for germ granule mRNA degradation in pole cells. Quantitative smFISH analysis showed that edc3 +patr-1 RNAi does not affect germ plasm assembly prior to pole cell formation (S6D and S6E  Fig) or nos and pgc levels in the pole cells prior to nc14 (S6F and S6G Fig). At nc14, nos and pgc drop to 28% and 40% of their pre-nc14 levels, respectively, in control embryos. By contrast, 67% of nos and 77% of pgc remain intact at nc14 in double RNAi embryos (Fig 6E), suggesting that DCP1 recruitment to germ granules is necessary for nos and pgc degradation. As with DCP1 overexpression, hsp83 was unaffected (S6H Fig), suggesting that depletion of edc-3 and patr-1 does not prevent RNA degradation globally in pole cells. Additionally, the stabilization of RNAs in the double-knockdown embryos allowed us to test if germ granule mRNA degradation is necessary for germline development. We found that upon edc3+patr-1 RNAi, there is a significant increase in the frequency of lost pole cells compared to controls (Fig 6F), suggesting that recruitment of DCP1 and degradation of at least a subset of germ granule mRNAs in the pole cells is necessary for proper pole cell migration to populate the gonad. to mark the germ plasm after water or CHX injection. Vas-EGFP was detected by direct fluorescence together with anti-DCP1 immunofluorescence. Enlargements of the boxed regions show DCP1 localization to germ granules in water injected (control) embryos (yellow) and the lack of colocalization in CHX injected embryos (cyan). (B) Western blot analysis of DCP1 levels in CHX and water injected embryos. Khc is used as a loading control. For the unprocessed data see S1 Raw Images. (C) The percent of cytoplasmic Vas-EGFP puncta that colocalize with DCP1 was quantified in both injection conditions, n = 18 embryos per condition. Nuclear Vas puncta were masked using Imaris software. Individual data points and mean ± SD are shown. ****p < 0.0001 as determined by Welch's t test. Source data for the graph in Fig 7C are

Discussion
Despite the ubiquity of germ granules among animal germ cells [5,6], their role in RNA metabolism is not well understood. We show that after their segregation to the pole cells, Drosophila germ granules undergo coordinated changes in size and function. During the period when germ granules grow by fusion, proteins involved in decapping and 5 0 to 3 0 degradation, but not deadenylation, are sequentially recruited to the germ granules. Therefore, deadenylation likely occurs before pole cell formation. Together, our data suggest that, in contrast to their broadly protective role in early embryos, germ granules become sites of selective mRNA decapping and degradation in pole cells and this plasticity is necessary for robust germline development.

Role of decapping activators in DCP1 recruitment
The shift in germ granule function is promoted by two decapping activators, Patr-1 and Edc3. Patr-1 and its homologs interact with multiple proteins in the decapping complex, including the catalytic enzyme DCP2 and Me31B [42,[53][54][55] and contribute to RNA binding [53,56]. Patr-1 localizes to germ granules 1 nuclear cycle after Me31B and concurrently with DCP1, suggesting it may serve as a direct link between the decapping complex and Me31B or mRNAs in the germ granules for recruitment of the complex. In contrast, Edc3 does not localize to germ granules, suggesting it promotes decapping complex recruitment indirectly, possibly by regulating the stability or translation of an unidentified RNA. Despite their distinct behaviors, Edc3 and Patr-1 are partially redundant as has been observed for decapping activators in other systems [55,57]. Structural and biochemical studies in yeast and Drosophila have shown that decapping activators such as Patr-1, Edc3, Trailer Hitch (TraI), and their homologs compete for the same binding surfaces on Me31B (Dhh1 in yeast) [42,43,55,58] and DCP2 [59], which could underlie this functional redundancy. Such redundancy would ensure effective decapping complex recruitment and mRNA degradation in pole cells.
We previously showed that founder granules, a distinct class of germ plasm RNP granules that house osk mRNA, also recruit mRNA decay factors [60]. This process begins much earlier, at nc5, to degrade osk and minimize its uptake by pole cells. During this period when germ granules and founder granules are intermingled within the germ plasm, DCP1 and Pcm associate selectively with founder granules. What makes germ granules "invisible" to the decay machinery at this time but susceptible later is an important question for future work. However, our results suggest that this susceptibility of germ granules requires production of an unidentified factor that triggers recruitment.

Selectivity of degradation within germ granules
Our data suggest that there are two distinct class of mRNAs within the germ granules: one that is selectively targeted for decapping and degradation and one that is protected. DCP1 forms puncta in germ granules but surprisingly, DCP1 puncta do not colocalize with homotypic clusters of either class. Phase separation has been shown to repress the activity of the yeast DCP1/ DCP2 complex in the absence of activators such as Edc3 [61]. Therefore, within the phase separated germ granules, these puncta could contain an inactive population of DCP1/DCP2 that cannot initiate decapping without activation. In this scenario, individual DCP1/DCP2 complexes, which would be undetectable by immunofluorescence, must exit these puncta to interact with target RNAs and become activated. Given the transient nature of enzyme-substrate interactions, DCP1/DCP2 may not accumulate at clusters. Decapping activators such as Patr-1, whose homologs recruit the decapping complex to mRNAs and catalyze cap removal [61,62] likely play a role in activating DCP1/DCP2 within the granules. Recent work in yeast has shown that decapping activators can regulate substrate specificity. Binding of different combinations of decapping activators, including Edc3 and Pat1, to distinct cis-regulatory sites within the C-terminal domain of DCP2 generates distinct decapping complexes that target different sets of transcripts [57,63]. Although the specific binding sites are likely different in Drosophila due low conservation of the DCP2 C-terminal domain [59,64], similar interaction networks within the decapping complex occur in Drosophila [42,43,58]. Therefore, similar mechanisms could regulate the activity and specificity of decapping in Drosophila, suggesting that the selective targeting of mRNAs for degradation could be achieved by the same decapping activators that recruit DCP1 to the granules. Whether the homologs of other decapping activators that regulate substrate specificity in yeast [57,63], such as Trailer Hitch and Upf1, regulate RNA degradation in germ granules remains to be determined. By contrast, DCP1 is a general decapping activator that does not influence RNA binding [65,66]. Therefore, increasing the effective concentration of DCP1 in the granules by overexpression may be sufficient to activate decapping, but would not confer substrate specificity, leading to the loss of CycB in addition to nos and pgc.

Functional significance of differential germ granule mRNA stability
Knockdown of decapping activators compromises nos and pgc RNA degradation and overexpression of DCP1 compromises CycB RNA protection, and both result in defective pole cell migration. This phenotype likely results from improper regulation of many germ granule mRNAs and speaks to the need for differential regulation of RNA stability. We envision that overexpression of DCP1 causes untoward decapping and degradation of mRNAs that encode proteins needed for further germ cell development and function, whereas knockdown of edc3 and patr1 expression results in stabilization of mRNAs and production of proteins whose functions are no longer required and may inhibit further development.
For example, CycB encodes a cyclin that promotes mitosis in the pole cells [26,49]. Prior to gastrulation, pole cells arrest the cell cycle in G2 through repression of CycB translation and maintain mitotic quiescence until after they reach the gonad [26,49,50]. CycB must be re-expressed in pole cells in the gonad for their proliferation [14]. Maintaining a pool of repressed CycB RNA would allow a rapid off-to-on switch, ensuring efficient entry into mitosis when the pole cells need to resume divisions in the gonad. On the other hand, nos and pgc encode proteins that delay zygotic genome activation in the pole cells [15,67] until the onset of gastrulation [24]. Pgc protein is cleared from pole cells by Bownes stage 6, which lifts global repression of transcription [15]. We find that pgc RNA degradation precedes protein degradation, which would limit translation and allow for effective Pgc clearance. Therefore, pgc degradation may promote the MZT. By contrast, Nos protein is detected in the pole cells throughout activation of zygotic transcription [47,68], suggesting repression of transcription can be lifted without Nos degradation. Therefore, the purpose of nos RNA degradation is unclear.
Interestingly, a recent study found that in the absence of maternal Pgc, premature miRNA transcription leads to precocious degradation of several germ granule and non-granule mRNAs in the pole cells. This zygotically driven degradation, which likely targets different RNAs than the maternal pathway described here, leads to loss of Nos, derepression of CycB translation, and ultimately pole cell apoptosis [69]. Therefore, the specificity and timing of mRNA degradation is likely important for pole cell development. For example, delaying degradation until nc14 could ensure global transcriptional repression is not lifted before the onset of another mechanism to silence somatic genes, such as chromatin remodeling.

Plasticity of germ granules allows their repurposing throughout development
Our findings uncover functional plasticity of germ granules during development, with their role in localization and stabilization of maternal mRNAs in the early embryonic germ plasm supplanted by roles during pole cell formation, and then during subsequent germline development. As pole cells become less dependent on maternal mRNAs, germ granules sequentially recruit mRNA degradation proteins, which makes them more P body-like and leads to turnover of select maternal mRNAs. Recent work revealed that C. elegans germ granules colocalize with pre-formed P bodies and become sites of deadenylation in the germline founder cells [70]. Thus, although these species use different mechanisms to become more P body-like, the shift in function from mRNA protection to degradation may be a conserved feature of germ granules. Recruitment of new proteins provides a mechanism to regulate the function of these long-lived granules, repurposing them as needed at different developmental stages.
Interestingly, the change in germ granule function coincides with enlargement of granules by fusion. Whether this growth is a cause or effect of the functional change remains an outstanding question. An intriguing hypothesis is that germ granules are restructured to facilitate new functions. The large number of small granules facilitates distribution of the germ granule material evenly among the pole cells as they bud and divide. Once the pole cells cease division, consolidation of germ granules into fewer, large granules might be favorable by concentrating mRNAs into fewer reaction sites. Association of limiting amounts of DCP1 and Pcm with larger granules would effectively bring them into contact with a larger number of transcripts, allowing them to degrade more RNAs without having to disassociate and then reassociate with other granules. Additionally, neuronal Me31B/Imp granules increase in size and recruit new RNAs during aging, which leads to translational repression [71]. Therefore, the modulation of granule size may be a conserved mechanism to alter granule function throughout the life span of an organism.
The gRNA and homology-directed repair plasmids were injected into a nos-Cas9 line (BDSC 78781) by BestGene. Following isolation of lines with successful insertion events and removal of the Cas9 by outcrossing, flies were crossed to a nos-PBac line [77] to remove the DsRed cassette. Individual F1 flies that had lost DsRed expression were used to generate independent lines and the insertion was confirmed by sequencing.

Embryo collection and fixation
Embryos were collected on yeasted apple juice agar plates at 25˚C and staged by nuclear density for nuclear cycles 9 to 14 or morphological features for Bownes stages 6 to 15. Prior to imaging, embryos were dechorionated, fixed, and devitellinized as previously described [78] and stored in methanol at −20˚C for up to 2 weeks. For western blotting or RT-qPCR, embryos were dechorionated, flash frozen in liquid nitrogen, and stored at −80˚C.

Live imaging and photoconversion
The 2-to 3-hour-old embryos were prepared for live imaging as described in [21]. Briefly, the posterior poles of embryos were stuck to a #1.5 coverslip (VWR) using heptane glue and embryos were covered in Halocarbon-95 oil (PolySciences). Time lapse imaging was then performed on a Nikon A1R laser scanning microscope with a resonant scanner. An approximately 8 μm z-series encompassing the pole cells was acquired using a 500 nm step size. Imaging of Osk-sfGFP was performed with a 60 × 1.4 NA oil immersion objective with a 2× optical zoom, 8× line averaging, and a capture rate of 1 z-stack per minute. Imaging of Osk-Dendra was performed with a 40 × 1.3 NA oil immersion objective with a 4× optical zoom, 4× line averaging, and a capture rate of 1 stack every 20 s. Osk-Dendra2 granules were photoconverted by 1-s stimulation with a 405 nm laser. Granules were manually tracked to identify fusion events.
smFISH was performed as previously described [78], followed by DAPI staining and mounting as described above. For embryos Bownes stage 7 and later, an additional 1 h incubation in 2% Triton-X 100 (Sigma Aldrich) was added prior to the pre-hybridization step to improve probe penetration. After staining, the late-stage embryos were cleared in RapiClear 1.47 (SUNJin Lab) overnight and mounted in 1:1 RapiClear:Vectashield.

Confocal imaging and quantification of fluorescence intensity
Confocal imaging was performed on a Nikon A1 laser scanning microscope with a 60 × 1.4 NA oil immersion objective and GaAsp detectors, with the exception of Figs 5E and S5J, for which a 20 × 0.75 NA air objective was used. Imaging parameters were kept identical for all samples within each experiment. To quantify fluorescence intensity per pole cell (Fig 3C-3E and 3H-3J), z-series images were captured with a 300 nm step size; to quantify total fluorescence intensity of the pole cell population, a 1 μm step size was used. Images were quantified in FIJI using the "sum slices function" to capture the entire volume of a pole cell of interest and the entire pole cell population, respectively. Total fluorescence intensity (integrated density in FIJI) of the signal in the pole cell(s) and of the background signal of a region outside the embryo were then measured. Background subtracted intensities were calculated for each pole cell or embryo.
The number and intensity of puncta were analyzed using the Spots function in Imaris. Spots were detected using consistent quality thresholds within each experiment. The number of puncta and the Intensity Sum of each spot were recorded. To quantify the normalized number of nos, pgc, and CycB puncta (Fig 3L), Vas, nos, and pgc or Osk and CycB spots were detected in the same nc10 and stage 14 images used to quantify total fluorescence intensity (Fig 3C-3E and 3H-3J). The number of RNA spots was normalized to the number of Osk or Vas spots in each embryo. Average CycB intensity per puncta (Figs 5F and S5K) was quantified from confocal z-series images of the whole embryo captured with 1 μm steps. Background signal was removed in FIJI using the Subtract Background function with a 2 pixel rolling ball radius and a sliding paraboloid. CycB signal that did not overlap with Vas (i.e., outside of the pole cells) was masked using the "Mask Channel" function in Imaris. The sum intensity of all spots was normalized to the number of spots to get an average intensity per spot for each embryo.

Quantification of colocalization
Confocal imaging was performed using a Nikon A1 laser scanning microscope with a 60 × 1.4 NA oil immersion objective and GaAsp detectors. A 2× optical zoom was applied to all images. An approximately 5 μm z-series encompassing the pole cells was acquired using a 300 nm step size. Prior to colocalization analysis, nuclear Vas or Osk granules were masked using the "Mask Channel" function in Imaris. The nuclear volume used for masking was defined based on DAPI signal using the surfaces function. Colocalization of Vas with DCP1 and Pcm, and of Osk and DCP1, was then analyzed as previously described [18] using a threshold of 300 nm to be considered colocalized in all analyses. Imaging conditions and thresholds were kept identical within each experiment.

STED microscopy
For STED imaging, 1:250 goat anti-mouse STAR RED secondary antibody was used for immunofluorescence, and nos, pgc, or CycB probes conjugated to atto594 or atto647N were used for smFISH. Embryos were mounted in Prolong Diamond Antifade Mountant, and 2D STED imaging was performed on a Nikon A1R laser scanning microscope coupled to a STEDYCON STED module with a 100 × 1.45 NA oil immersion objective. Images were acquired in the STE-DYCON software. A confocal image of the posterior region or pole cells was used to identify individual germ granules as the ROI for STED. STED images of a ROI within the bulk cytoplasm of early embryos were acquired from the same slides. After acquisition, STED images were deconvolved using the NIS Elements software.
To determine if germ granule puncta contain multiple transcripts, we compared the intensities of nos, CycB, and pgc puncta in the germ granules to the intensities of puncta in the bulk cytoplasm of early embryos, which represent single transcripts [17]. The sum intensity of each puncta was measured from deconvolved STED images using Imaris as described above. For each transcript, the intensity of each germ granule puncta was normalized to the average intensity of puncta in the bulk cytoplasm.

Immunohistochemistry
Pole cells were detected in fixed 10-to 14-hour-old embryos using anti-Vas immunohistochemistry as previously described [60]. The following antibodies were used: 1:500 rabbit anti-Vas and 1:500 biotin goat anti-rabbit (Jackson Immuno Research Laboratories). Embryos were mounted in 80% glycerol and analyzed using a Zeiss compound microscope. Images of representative embryos were captured on a Nikon DsQi-2 microscope using a 20 × 0.75 NA air objective and DIC optics.

Cycloheximide injections
Dechorionated 45-to 75-minute-old embryos were arranged end to end with the posterior poles facing the same direction and stuck to a coverslip using heptane glue. After 12 min of desiccation with Drierite (Fisher Scientific), embryos were covered in Halocarbon 200-oil (PolySciences). Embryos between 60 min and 90 min old were injected with either 1 mg/mL cycloheximide or water, at a lateral site near the posterior pole. Following injection, embryos were returned to a 25˚C incubator for 70 min to develop to nc14. The embryos were removed from the coverslip by washing with heptane to dissolve the glue and rinsed thoroughly with water to remove the remaining heptane.
For western blot analysis, embryos were blotted dry, flash frozen in liquid nitrogen, and stored at −80˚C. For immunofluorescence, embryos were fixed in a glass crystallization dish for 30 min. The fixative was prepared by combining equal parts heptane and a formaldehyde solution (0.5× PBS, 18.5% formaldehyde) and shaking vigorously for 1 min prior to use to saturate the heptane with formaldehyde. Several drops of the heptane and 1 drop of the aqueous phase were added to the embryos for fixation. Following fixation, embryos were hand devitellinized in PBST. Devitellinized embryos were rinsed 4× in PBST and stored in 1 mL PBST at 4˚C for up to 4 days. Immunofluorescence was performed as described above.

Western blotting
Frozen embryos were homogenized in 2 μl of sample buffer (0.125 M Tris-HCl (pH 6.8), 4% SDS, 20% glycerol, 5 M urea, 0.01% bromophenol blue, and 0.1 M DTT) per mg tissue and boiled for 4 min. Samples were then used for immunoblotting as previously described [60] using the following antibodies: 1:1,500 mouse anti-DCP1, 1:10,000 rabbit anti-Kinesin heavy chain (Cytoskeleton), 1:2,000 HRP Sheep anti-mouse, and 1:2,000 HRP donkey anti-rabbit. Blots were imaged using an iBright FL1000 Imaging System (Invitrogen). Embryos were staged by nuclear cycle or Bownes stage according to nuclear density or morphological features, respectively. Osk-sfGFP (green) was visualized by anti-GFP immunofluorescence; Vas-EGFP (green) was detected by direct fluorescence; nuclei were stained with DAPI (blue). The brightness and contrast were adjusted individually for each image to best show the features of the germ granules at that stage. Scale bar: 10 μm.  Fig. nos, pgc, and CycB localize to homotypic clusters in large and small germ granules. (A) The 2D STED images of nos, CycB, and pgc RNA detected pairwise by smFISH in granules in pre-pole bud stage embryos. Individual granules were selected for STED imaging from confocal images as shown in the example on the left (white box). (B) The 2D STED images (indicated by the white boxes on the confocal sections shown in the left-most panels) from pole cells at nc14. pgc (green) was detected together with CycB or nos (magenta) by smFISH. Fluorescence intensity was measured along the paths marked with white lines and intensity profiles of each channel, normalized to the maximum value, are plotted. (C) The sum intensity of nos and pgc puncta were measured from STED images of the bulk cytoplasm of early embryos (black data points) and of clusters in single germ granules (magenta data points). Values were normalized to the average intensity of a puncta in the bulk cytoplasm. Individual data points and means are displayed. ****p < 0.0001 by Mann-Whitney test. Source data for the graphs in per genotype. (C) Western blot analysis of DCP1 levels in embryos overexpressing DCP1 compared to DCP1 heterozygotes. Khc was used as a loading control. For the unprocessed data, see S1 Raw Images. (D) The 2D STED images of DCP1 (green) relative to CycB or nos (magenta) in DCP1-nos3 0 UTR overexpressing embryos. (E, F) CycB was detected by smFISH and the total CycB intensity in pole cells was quantified at nc10-13, n = 4-10 embryos per genotype (E) and nc14, n = 10-11 embryos per genotype (F). Total CycB intensity at nc14 was normalized to the average intensity at nc10-11 (F). (G) hsp83 (green) detected together with CycB (magenta) by smFISH in a wild-type nc14 embryo showing enrichment of hsp83 in pole cells, but not in germ granules. (H) Total hsp83 intensity in pole cells at nc10-11 (H). (I) Total hsp83 intensity at nc14 was normalized to the average intensity at nc10-11 for the same genotype, n = 10-11 embryos per genotype. (J) Maximum intensity confocal z-projections of CycB (gray) and Vas (red) in the gonads of DCP1 heterozygous and DCP1-nos3 0 UTR overexpressing embryos. CycB was detected by smFISH and Vas by immunofluorescence. (K) The average intensity of CycB puncta per embryo in DCP1 heterozygous, DCP1-nos3 0 UTR, and DCP1-smg3 0 UTR (also shown in Fig 5) overexpressing embryos was measured using Imaris, n = 13-15 embryos. (L) Quantification of the number of lost pole cells, detected by anti-Vas immunohistochemistry in DCP1 heterozygotes, DCP1-nos3 0 UTR, and DCP1-smg3 0 UTR (also shown in Fig 5) overexpression embryos, n = 73-223 embryos per genotype. Graphs display individual data points and means (E, H, K, L) or mean ± SD. (B, F, I). *p < 0.05, **p < 0.01, ***p < 0.001, ****p CycB was detected by smFISH during nc9-13 and at nc14. Representative embryos at nc10 (WT), nc11 (nos -), and nc14 (both) are shown. (B) Quantification of the sum fluorescence intensity of CycB in the germ plasm, n = 10-12 embryos per time point. CycB levels were compared by Kruskal-Wallis ANOVA. n.s., not significant. (C) nos was detected by smFISH in wild-type and gnosSREs -GRHembryos at nc10-11 and nc14. The gnosSREs -GRHtransgene produces nos mRNA with mutations in binding sites for the Smaug (SREs) and Glorund (GRH) repressors (51). Total nos fluorescence intensity in the germ plasm was quantified and the intensity at nc14 was normalized to the average intensity at nc10-11, n = 10-11 embryos per genotype. *p < 0.05 by Student's t test. Individual data points and means ± SD are shown. Source data for the graphs in S7B and S7C