Localization of oskar mRNA by agglomeration in ribonucleoprotein granules

Localization of oskar mRNA to the posterior of the Drosophila oocyte is essential for abdominal patterning and germline development. oskar localization is a multi-step process involving temporally and mechanistically distinct transport modes. Numerous cis-acting elements and trans-acting factors have been identified that mediate earlier motor-dependent transport steps leading to an initial accumulation of oskar at the posterior. Little is known, however, about the requirements for the later localization phase, which depends on cytoplasmic flows and results in the accumulation of large oskar ribonucleoprotein granules, called founder granules, by the end of oogenesis. Using super-resolution microscopy, we show that founder granules are agglomerates of smaller oskar transport particles. In contrast to the earlier kinesin-dependent oskar transport, late-phase localization depends on the sequence as well as on the structure of the spliced oskar localization element (SOLE), but not on the adjacent exon junction complex deposition. Late-phase localization also requires the oskar 3′ untranslated region (3′ UTR), which targets oskar to founder granules. Together, our results show that 3′ UTR-mediated targeting together with SOLE-dependent agglomeration leads to accumulation of oskar in large founder granules at the posterior of the oocyte during late stages of oogenesis. In light of previous work showing that oskar transport particles are solid-like condensates, our findings indicate that founder granules form by a process distinct from that of well-characterized ribonucleoprotein granules like germ granules, P bodies, and stress granules. Additionally, they illustrate how an individual mRNA can be adapted to exploit different localization mechanisms depending on the cellular context.


Introduction
Localization of mRNAs to subcellular domains plays an important role in generating morphological and functional asymmetry through local protein production.Localization of oskar (osk) mRNA to the posterior of the Drosophila oocyte and its on-site translation are essential for formation of the germ plasm, the specialized embryonic cytoplasm required for abdominal patterning and germ cell formation during embryogenesis [1][2][3][4].The trafficking of mRNAs to different subcellular locations is controlled by cis-acting RNA sequences and/or structures that most commonly reside within 3 0 untranslated regions (3 0 UTRs), and proteins that recognize these localization elements.These proteins may interface with cellular transport machinery or they may facilitate association with cellular structures or organelles, including phase transitioned condensates [5,6].Multiple cis-acting elements along with a coterie of trans-acting factors contribute to the multi-step process of osk localization throughout oogenesis [7,8].
Drosophila oogenesis proceeds through 14 morphologically defined stages; during the first 10 stages the oocyte develops within an egg chamber, accompanied by 15 sister nurse cells that support oocyte growth and development by providing maternal mRNAs such as osk as well as proteins and organelles.osk is transported from the nurse cells to the oocyte by dynein; this transport requires sequence elements and structures in the 3 0 UTR [9,10].One of these, called the oocyte entry signal (OES), links osk mRNA to dynein by binding to the RNA-binding protein Egalitarian and the adaptor Bicaudal-D [10][11][12].Throughout stages 8 to 10, osk travels to the posterior of the oocyte by kinesin-mediated transport [13,14].This transport requires an unusual localization element, a stem-loop structure called the Spliced oskar Localization Element (SOLE) [15].The SOLE comprises the last 18 nucleotides of the first osk exon and the first 10 nucleotides of the second osk exon and therefore splicing of the first osk intron is crucial for localization.Mutational analysis showed that SOLE function depends on its structural integrity rather than on its sequence.Splicing of the first osk intron also plays a role in osk localization through recruitment of the exon junction complex (EJC) [15].The EJC/SOLE functions as a unit to activate kinesin motility, leading to accumulation of osk at the posterior [15,16].
In addition to its role in nurse cell-to-oocyte transport, the osk 3 0 UTR contributes in multiple ways to posterior transport within the oocyte [17].In the oocyte, osk multimerizes, forming ribonucleoprotein particles (RNPs) containing up to 4 osk mRNAs [18].The RNA-binding protein Bruno 1 (Bru1) binds to the osk 3 0 UTR and promotes osk oligomerization to form translationally repressed complexes [19,20].Bru1 drives assembly of osk transport RNPs through phase separation; these osk RNP condensates mature to a solid state, which is required for their transport, posterior accumulation, and translation [19].In addition, a short sequence at the tip of the OES promotes osk dimerization, allowing transgenic reporter mRNAs with this dimerization domain, but lacking the SOLE, to "hitchhike" to the posterior on endogenous wild-type osk mRNA [21,22].Whether hitchhiking contributes to localization of endogenous osk remains to be determined, however.
The osk 3 0 UTR also contains binding sites for the double-stranded RNA-binding protein Staufen (Stau) [23,24].Stau associates with osk RNPs upon their entry to the oocyte and remains colocalized with osk at the posterior of the oocyte [1,2,18,25,26].Stau facilitates kinesin-dependent transport [13,23,25] and is also required for osk mRNA translation and anchoring at the posterior [26].Additionally, Osk protein itself is required to maintain osk mRNA at the posterior of the oocyte [4,27].
At the posterior, Osk initiates assembly of the germ plasm and nucleates formation of RNP condensates called germ granules [8].Numerous mRNAs become enriched in the germ plasm through their incorporation into germ granules, which ensures their inheritance by the germ cell progenitors, called pole cells, during embryogenesis.Among these is nanos (nos), whose translation depends on Osk and is crucial for both abdominal patterning and pole cell development [28][29][30][31].For many germ granule mRNAs including nos, elements contained within 3 0 UTRs are sufficient to direct their germ granule localization [32][33][34].
At the end of stage 10 of oogenesis, the nurse cells extrude their contents en masse into the oocyte and degenerate.Just prior to this nurse cell dumping, a bulk cytoplasmic flow called ooplasmic streaming is initiated that mixes the oocyte and incoming nurse cell contents [35,36].Notably, additional osk mRNA enters the oocyte and accumulates at the posterior during this later period, propelled by nurse cell dumping and ooplasmic streaming rather than by motor-dependent transport [37,38].This late phase of osk accumulation amplifies the germ plasm, allowing production of additional Osk protein and enlargement of germ granules [38][39][40].Since the quantity of germ plasm formed, and consequently the quantities of the abdominal determinant Nos and germ cells produced in the embryo, depend directly on the amount of osk mRNA localized during oogenesis [30,41,42], this amplification is important for abdominal patterning and reproduction.Failure to accumulate osk mRNA and protein at late stages of oogenesis results in loss of abdominal segments and pole cells [38,39].
By the end of oogenesis, posteriorly localized osk resides in large granules containing up to 250 osk mRNAs [18].These granules ultimately mediate the degradation of osk mRNA in the embryo, where it is toxic to pole cells, and we have referred to them as founder granules to distinguish them from germ granules [43].How osk transport RNPs are organized into larger founder granules has not been investigated.Moreover, which features of osk mediate the late phase of localization and whether they are distinct from the signals that mediate the earlier microtubule-dependent steps remain unknown.Using a combination of quantitative confocal imaging and qualitative stimulated emission depletion (STED) microscopy, we show that founder granules are agglomerations of smaller osk transport RNPs.By analyzing the behavior of osk transgenes with mutations or deletions that affect the osk SOLE and 3 0 UTR, we show that, in contrast to localization during stages 8 to 10, both the structure and the sequence of the SOLE, but not the nearby exon junction complex (EJC) deposition, are necessary for latephase localization.Late-phase localization also requires the osk 3 0 UTR and by swapping the nos and osk 3 0 UTRs, we show that each is sufficient to target mRNA to the appropriate granule.Together, our results demonstrate that 3 0 UTR-mediated targeting together with SOLEdependent agglomeration result in accumulation of osk in large founder granules at the posterior of the oocyte during late stages of oogenesis.

Founder granules in late-stage oocytes are agglomerates of osk RNPs
To examine how osk accumulates in founder granules, we performed a spatiotemporal analysis of wild-type osk mRNA, visualizing osk in both nurse cells and anterior region of the oocyte from stage 8 to stage 9 (S1  and 1).Because we were limited to 2D STED, the analysis is qualitative rather than quantitative.From stage 8 to stage 10, diffraction-limited confocal spots resolve into individual, largely round puncta consistent with osk transport RNPs.These particles accumulate at high density at the anterior oocyte cortex at stage 8 and the posterior cortex at stages 8 to 10 but remain distinct from each other in both confocal and STED images (Figs S1 and 1).By stage 12, as the late phase of localization progresses, individual confocal spots often resolve to multiple puncta and by stage 13, they appear to comprise agglomerates of many individual osk RNPs, independent of probe concentration (Figs 1 and S2A).This is congruent with the dramatic increase in mRNA content of localized osk granules that occurs during the later stages of oogenesis [18].By contrast, lipid droplets, which are similar in size to founder granules, appear homogeneous (S2B Fig).
The ability to resolve individual osk RNPs within founder granules indicates a different organization from germ granules.In contrast to osk, germ granule mRNAs travel within the oocyte in RNPs containing only a single transcript.Within the germ plasm, these RNPs become incorporated into phase-separated protein scaffolds nucleated by Osk.Upon incorporation, mRNAs self-associate to form homotypic clusters containing many like transcripts.Distinct clusters of different mRNAs (e.g., nos versus Cyclin B) can be resolved by super-resolution microscopy and although an individual cluster can contain tens of transcripts, each cluster appears as a single puncta [40,44].For direct comparison to osk in founder granules, we visualized nos using STED microscopy.By confocal microscopy, germ granule associated nos appears as bright spots, each corresponding to a germ granule as previously shown [18] andin contrast to founder granules-these resolve to single puncta with STED (Fig 1).These data suggest that founder granules form by a different mechanism than germ granules, through the agglomeration of RNPs that remain physically distinguishable.

Late phase osk mRNA localization does not appear to require the EJC
We next sought to determine whether elements involved in the active transport of osk to the posterior during stages 8 to 10 also function in osk localization during late oogenesis.We generated flies expressing a genomic osk transgene tagged with a superfolder gfp (sfgfp) sequence (osk-sfgfp), as well as a version with a deletion of the first two introns (oskΔi1,2-sfgfp) (Fig 2A).Deletion of intron 1 does not affect SOLE formation but prevents EJC deposition near the exon-exon junction [15].The sfgfp sequence tag allows for detection of the transgenic mRNA by smFISH in the context of endogenous osk mRNA.Since the oskΔi1,2-sfgfp transgenic mRNA is not expected to localize on its own by kinesin-dependent transport, we reasoned that the presence of endogenous osk mRNA and consequent production of Osk protein would be necessary for the initial establishment of germ plasm and for the retention of osk mRNA during late stages of oogenesis.The osk-sfgfp and oskΔi1,2-sfgfp were expressed at comparable levels as determined by RT-qPCR (S3 Fig).
Quantification of the total localized sfgfp fluorescence intensity in confocal sections through the entire germ plasm at stage 10 showed that in the presence of wild-type osk, both the osksfgfp and oskΔi1,2-sfgfp mRNAs can localize during mid-oogenesis (Fig 2B -2E and 2H).However, localization of the oskΔi1,2-sfgfp mRNA is less efficient than the osk-sfgfp mRNA (Fig 2D ,  2E and 2H).This is consistent with the finding that without EJC deposition on osk mRNA, transport efficiency of osk RNPs is reduced [15] and suggests that the oskΔi1,2-sfgfp mRNA only localizes by hitchhiking with endogenous osk using the dimerization domain in the 3 0 UTR [22].By stage 13, the localized fluorescence intensities of the osk-sfgfp and oskΔi1,2-sfgfp mRNAs increase by 34% and 43%, respectively (Fig 2F -2H).Therefore, deposition of the EJC near the first exon-exon junction is not required for late-phase osk localization.

Late-phase osk localization requires the SOLE UA-rich proximal stem sequence
Although the oskΔi1,2-sfgfp mRNA is not bound by the EJC near the first exon-exon junction, it can form the SOLE.To test whether the SOLE influences late-phase osk localization independently of the EJC, we generated a version of the osk-sfgfp transgene with the SOLE proximal stem sequence substituted by lacZ sequence (SOLE PS-Lz ; Fig 3A).This mutation was previously shown to disrupt SOLE localization activity during mid-oogenesis without affecting EJC deposition [15].As expected, the oskSOLE PS-Lz -sfgfp mRNA localizes to the posterior of the oocyte by stage 10 in the presence of endogenous osk, although with reduced efficiency compared to the osk-sfgfp control mRNA (Fig 3B,3D and 3J).This is consistent with the reduced localization efficiency at stage 10 of the oskΔi1,2-sfgfp mRNA (Fig 2D , 2E and 2H).However, in contrast to the oskΔi1,2-sfgfp mRNA, the oskSOLE PS-Lz -sfgfp mRNA is not further enriched by stage 13 (Fig 3C , 3E and 3J).Therefore, the SOLE is required for late-phase osk localization, and this role is independent of adjacent EJC deposition.
The earlier phase of osk localization relies on the structure, but not the sequence of the SOLE proximal stem [15].To determine whether this is also the case for the late phase of osk localization, we generated a new mutation that disrupts base-pairing of the proximal stem and is thus predicted to disrupt the regular helical structure of the SOLE (SOLE UA-mut ) [45] as well as a compensatory mutation (SOLE UA-mut-comp ) that would restore the proximal stem basepairing, but not the sequence (Fig 3A).Like the oskSOLE PS-Lz -sfgfp mRNA, oskSOLE UA-mutsfgfp mRNA shows reduced localization efficiency at stage 10 in the context of endogenous osk and fails to enrich further at stage 13 (Fig 3F , 3G and 3J).The compensatory SOLE UA-mut-comp mutation restores localization efficiency at stage 10, but does not restore late phase localization (Fig 3H -3J).This is in contrast to the earlier phase of osk localization [15].In all cases, RT-qPCR analysis confirmed that loss of enrichment is not due to low expression of the transgenic mRNA (S3 Fig) .Quantification of the number of founder granules containing transgenic mRNA (detected particles containing �4 mRNAs, see Materials and Methods) and their size (i.e., the number of mRNAs per granule) in stage 13 oocytes showed that although none of the mutations significantly affect the number of osk-sfgfp containing granules, both the average and maximum number of osk-sfgfp in each granule are reduced (Fig 3K -3M).Thus, these data show that both the sequence and the structure of the SOLE are required for late-phase osk accumulation.

SOLE function is required for accumulation of osk in founder granules
Our results suggest that the SOLE mutations impair the ability of osk to accumulate in founder granules after stage 10.To investigate this possibility, we performed STED microscopy.Similarly to endogenous osk at stage 13, osk-sfgfp mRNA resides in large granules containing multiple osk-sfgfp puncta (Fig 4).By contrast, oskSOLE PS-Lz -sfgfp mRNA, oskSOLE UA-mut -sfgfp, and oskSOLE UA-mut-comp -sfgfp appear largely as individual or small groups of puncta, similar to endogenous osk prior to stage 12 (Figs 1, 4 and S1) and consistent with the quantitative analysis of granule size (Fig 3K and 3L).Because wild-type endogenous osk is also present and forms founder granules in these oocytes, this pattern suggests that although the SOLE mutant transcripts can localize along with endogenous osk prior to nurse cell dumping, they are unable to accumulate in founder granules during late stages of oogenesis.Together, both the quantitative and qualitative data lead us to conclude that SOLE function during late oogenesis is necessary for enrichment of osk in the germ plasm by promoting its accumulation in founder granules.

Late phase osk localization requires the osk 3 0 UTR
In addition to the SOLE, the osk 3 0 UTR is required for osk localization during stages 8 to 10 [17].Furthermore, the osk 3 0 UTR of an mRNA lacking the SOLE can facilitate localization by dimerizing with the 3 0 UTR of wild-type endogenous osk, allowing the SOLE-less mRNA to hitchhike to the posterior [21,22].The finding that none of the SOLE mutants tested above support late-phase accumulation indicates that hitchhiking is not occurring at these later stages, however.To test whether the 3 0 UTR plays any role in late-phase localization, we generated an osk-sfgfp transgene (including the SOLE) in which the osk 3 0 UTR was replaced by sequences from the fs(1)K10 3 0 UTR that promote transport of mRNA from the nurse cells to the oocyte, but not localization within the oocyte [46] (Fig 5A).osk-sfgfp-K10_3 0 UTR mRNA is readily detectable by smFISH when expressed under GAL4/UAS control.As expected, osksfgfp-K10_3 0 UTR mRNA does not accumulate at the posterior of the oocyte prior to stage 10 even in the presence of wild-type osk (Fig 5B).Furthermore, the osk-sfgfp-K10_3 0 UTR does not localize by stage 13 either (Fig 5C ), indicating that the osk 3 0 UTR is required in addition to the SOLE for both earlier and later phases of osk localization.

The osk 3 0 UTR is sufficient for RNP accumulation in founder granules
Founder granules occupy the germ plasm together with germ granules, but remain physically and functionally distinct.Whereas germ granules are actively incorporated into the pole cells during embryogenesis, founder granules are not and instead recruit degradation machinery to stages 10 and 13; n = 6-8 oocytes each.(K-M) Quantification of the number of founder granules (particles �4 mRNAs) containing the indicated mRNAs (K), or the average (L) and maximum (M) number of mRNAs in those granules; n = 8-11 oocytes each.Similar results were obtained using a requirement of >2 (the minimum for osk transport particles [18]).eliminate osk mRNA [18,43].As osk RNA is toxic to pole cells, its segregation from germ granules is crucial for fertility [43].Studies of several germ granule mRNAs, including nos, polar granule component, and germ cell-less, have shown that their 3 0 UTRs are sufficient for their accumulation in germ granules.To determine if the osk 3 0 UTR is sufficient to direct mRNA to founder granules, we generated tagged, genomic osk and nos transgenes with their 3 0 UTRs exchanged (Fig 6A) and asked whether the hybrid mRNAs were associated with founder granules or germ granules.The sfgfp and egfp sequences allowed the respective transgenic mRNAs to be distinguished from endogenous osk and nos by smFISH.Stau was used as a founder granule marker [43] and Cyclin B (CycB), an abundant germ granule transcript [18], was used as a germ granule marker.We performed dual smFISH and immunofluorescence to detect the transgenic mRNAs with sfgfp or egfp probes coupled to one of two far-red fluorophores; germ granules with CycB probes coupled to a 565 fluorophore; and founder granules with anti-Stau and a secondary antibody coupled to a 488 fluorophore.The experiments were performed using early embryos, due to the impenetrability of late-stage oocytes to antibodies.We monitored germ granule-association as far-red fluorescence signal (transgenic mRNA) colocalized with 565 fluorescence signal (CycB).Because the transgenes encode GFP-labeled proteins, to eliminate any contribution from residual GFP fluorescence we identified founder granules by the criteria they contain mRNA (far-red) and Stau (488) but not CycB (565).Based on the previous finding that fewer than 25% of osk particles in the germ plasm colocalize with germ granules [18], we set a threshold for granule association at 25% of detected sfgfp or egfp particles colocalized with the germ granule or founder granule marker as described above.The control osk-sfgfp and egfp-nos mRNAs behave as expected.43% of osk-sfgfp particles colocalize with Stau (Fig 6B and 6J), whereas colocalization with CycB is 22%, below the threshold (Fig 6C and 6J).Conversely, 50% of egfp-nos particles colocalize with CycB (Fig 6E and 6J), similarly to the behavior of nos [18].In contrast, colocalization of egfp-nos with Stau is 6% (Fig 6D and  6J).Similarly to egfp-nos, 49% of osk-sfgfp-nos3 0 UTR particles colocalize with CycB (Fig 6G and  6J), whereas 23% colocalize with Stau (Fig 6F and 6J).Furthermore, similarly to osk-sfgfp, 56% of egfp-nos-osk3 0 UTR particles colocalize with Stau (Fig 6H and 6J), whereas only 10% colocalize with CycB (Fig 6I and 6J).Thus, the granule preference of the transgenic mRNAs is determined by the 3 0 UTRs.Together with the analysis of osk-sfgfp-K10_3 0 UTR mRNA, these results indicate that the SOLE is not sufficient to target mRNAs to founder granules, rather the osk 3 0 UTR imparts this function just as the nos 3 0 UTR directs mRNAs to germ granules.

The osk coding sequences and/or 5 0 UTR, but not the SOLE, are required to maintain osk mRNA accumulation in late-stage oocytes
Although the osk 3 0 UTR is sufficient to target nos to founder granules, egfp-nos-osk3 0 UTR mRNA is only weakly enriched in the embryonic germ plasm (Fig 7B).To determine whether this results from a defect in the initial localization, presumably by hitchhiking, or a defect in maintenance of egfp-nos-osk3 0 UTR in founder granules over time, we monitored localization over the course of late oogenesis, from stages 10 to 13.In contrast to osk-sfgfp mRNA, egfpnos-osk3 0 UTR does not continue to accumulate at the posterior after stage 10 (Fig 7C -7H and  7L).This behavior is consistent with early phase localization of egfp-nos-osk3 0 UTR by 3 0 UTRmediated hitchhiking but failure to accumulate further due to lack of the SOLE.Moreover, the amount of egfp-nos-osk3 0 UTR mRNA localized at the posterior of the oocyte decreases from stage 12 to stage 13, whereas osk-sfgfp mRNA is largely unchanged (Fig 7C -7H and 7L).Both the number of mRNAs per founder granule and the number of founder granules that contain the transgenic mRNA are reduced for egfp-nos-osk3 0 UTR as compared to osk-sfgfp mRNA (Fig 3K -3M).Total transgenic mRNA levels remain constant between these stages, however, suggesting that the loss of egfp-nos-osk3 0 UTR is unlikely a result of degradation (S4 Fig) .Together, these data are most consistent with a failure to maintain egfp-nos-osk3 0 UTR in founder granules.
These results also suggest that the osk coding sequences or 5 0 UTR contains an element important for maintenance of osk mRNA in founder granules for the duration of oogenesis.Since the oskSOLE UA-mut -sfgfp mRNA is less enriched in the germ plasm of late oocytes than osk-sfgfp mRNA, we hypothesized that the SOLE could be such a maintenance element.In contrast to egfp-nos-osk3'UTR mRNA, however, oskSOLE UA-mut -sfgfp mRNA is not lost from the posterior after stage 12 (Fig 7I -7L).Thus, the SOLE and 3 0 UTR most likely function in the association of osk RNPs to form founder granules whereas long-term persistence requires additional coding region or 5 0 UTR sequences.

Discussion
Our finding that founder granules appear to be agglomerates of osk RNPs provides new insight into the process by which osk accumulates at the posterior of the oocyte during late stages of oogenesis.osk is transported in RNPs containing 2-4 transcripts [18].Recent work has shown that these RNPs are initially liquid-like condensates but they rapidly mature to a non-dynamic, solid state that prevents incorporation of additional mRNA molecules.Inducing a more liquid-like state results in formation of large, dynamic condensates at the posterior of late-stage oocytes that subsequently detach [19] indicating that the solid state is necessary for proper founder granule assembly and anchoring.Our observation that founder granules contain multiple physically distinct osk RNPs packed together is consistent with the solid-like properties of these RNPs and indicates that they do not form through the collapse of transport RNPs into larger condensates but rather through an aggregative process.This mechanism contrasts with Drosophila germ granules, whereby pre-formed protein condensates are populated by RNPs containing single transcripts, which then self-assemble within the granules to form homotypic clusters [40].What limits the agglomeration of osk RNPs into founder granules to the posterior of the oocyte remains unclear.RNP-RNP associations may be fostered by the high posterior concentration of osk RNPs achieved previously by kinesin-dependent transport.Since proteins can partition into osk RNPs after their transition to a solid-like state [19], proteins recruited to the germ plasm, perhaps by Osk protein itself, could mediate this behavior.Intriguingly, zebrafish germ plasm mRNAs form homotypic RNPs that aggregate into compact structures while retaining their distinct spherical appearance [47].This similarity with founder granules suggests that agglomeration may be a more generalized mechanism for mRNA compartmentalization.
We interrogated the function of the earlier acting EJC/SOLE complex in late-phase osk localization and found that the SOLE, but not the adjacent EJC, is required.Whereas the function of the SOLE in the earlier localization phase relies only on the structure of the proximal stem [15], both the sequence of the proximal stem and its structure are important for latephase localization.How the SOLE collaborates with the EJC to promote kinesin-dependent osk motility and what, if any regulatory factor interacts with it are not yet known.The sequence-dependence of the SOLE and lack of requirement for the EJC in late-phase localization suggests a different mode of action, possibly through the binding of a different protein to the proximal stem and recruitment of new RNP components or through RNA-RNA interactions.The change in ovarian physiology with the onset of nurse cell dumping could lead to an exchange of proteins associated with osk, to inhibit kinesin-dependent motility and promote posterior agglomeration.
The failure of osk-K10_3 0 UTR mRNA to localize at late stages of oogenesis despite the presence of the SOLE indicates that similarly to the earlier phase, late-phase localization depends on both the SOLE and the 3 0 UTR.This dependence on the 3 0 UTR for the late accumulation of osk is not for the purpose of hitchhiking, and by swapping the osk and nos 3 0 UTRs we showed that the osk 3 0 UTR specifies association of osk RNPs in founder granules independently of the SOLE.This function of the osk 3 0 UTR may be conferred by the same 3 0 UTRbinding proteins that control the formation and/or initial localization of osk transport RNPs and remain associated with osk at the posterior pole, such as Bru1, Stau, or Hrp48.For example, a prion-like domain in Bru1 required for formation of osk transport RNPs [19] could also mediate self-association of these RNPs when they come in contact at the posterior pole.Likewise, mammalian Stau has the propensity to form cytoplasmic aggregates [48].The requirements for Bru1 and Stau in the earlier phase of osk localization [1,2,19] make it difficult to test this idea, however.Additionally, the osk 3 0 UTR may function to prevent co-condensation of osk RNPs with germ granules through the recruitment of proteins like Hrp48, which maintains the solid-like properties of osk RNPs [19].
Results from swapping the osk and nos 3 0 UTRs also suggest that either osk 5 0 UTR and/or coding sequences other than the SOLE contribute to maintaining osk RNPs in founder granules.Since multivalent interactions are typically required for inclusion of components in phase separated condensates [49], it is not surprising that binding of founder granule components to multiple sites within osk would be required for the integrity of these granules.Further dissection of the sequence requirements and identification of interacting factors will be necessary to define the mechanisms by which the various osk elements accomplish the different tasks.
The process by which osk mRNA achieves its posterior localization is remarkably complex and labor intensive, involving distinct machineries for transport into the oocyte, movement to the posterior pole during stages 8 to 10, and further accumulation during late stages of oogenesis.Given the dependence of embryonic abdominal patterning and germ cell formation on the amount of osk mRNA localized during oogenesis [30,41,42], the reliance on numerous distinct contributions to osk localization likely provides robustness to processes governing the targeting of osk RNPs to the right location and the accumulation of sufficient osk there.
Moreover, the distinct process of assembling founder granules ensures that osk mRNA remains separated from germ granules to promote its degradation in the embryonic germ plasm and minimize its inheritance by pole cells.

Construction of transgenes and transgenic lines
The osk-sfgfp and oskΔi1,2-sfgfp transgenes and transgenic lines were previously described and contain sfgfp sequences inserted just before the osk stop codon in an 8 kb genomic osk rescue fragment in the pattB vector [43].oskΔi1,2-sfgfp lacks the first and second osk introns [43].The egfp-nos transgenes and transgenic lines were previously described and contain egfp sequences inserted just after the nos start codon in a 4.3 kb nos genomic rescue fragment [50].To generate osk-sfgfp-nos3 0 UTR, the osk 3 0 UTR and 3 0 genomic DNA were removed from the plasmid pattB-osk-sfgfp [43] and replaced with a 1.3 kb fragment containing the nos 3 0 UTR and 451 bp of 3 0 genomic nos DNA.To generate egfp-nos-osk3 0 UTR, a fragment containing the nos promoter and 5 0 UTR, egfp, and nos coding region (including introns) was removed from pCaS-peR-Pnos-gfp-nos [50], fused to the osk 3 0 UTR and 3.2 kb of osk 3 0 genomic sequences, and cloned into pattB.To generate the osk-sfgfp SOLE mutant transgenes, a 774 bp SphI-SacI fragment from pattB-osk-sfgfp [43] was replaced with a 774 bp SphI-SacI fragment synthesized by Genewiz with either the PS-Lz, UA mut , or UA mut-comp mutation shown in Fig 3A .UASp-osksfgfp-K10_3 0 UTR was generated by inserting a fragment from pattB-osk-sfgfp [43] spanning a naturally occurring BamHI site just before the osk transcription start site to an engineered BamHI site immediately following the osk stop codon into the BamHI site of pattB-UASp.Transgenes were integrated into the attP40 site by phiC31-mediated recombination.

Tissue collection and fixation
Ovaries: Females were fed on yeast paste at 25˚C for 3 days.Ovaries were dissected into PBS, lightly teased apart, and fixed and stepped into methanol as previously described [51].Ovaries were stored in methanol at -20˚C for �1 month.Ovaries used for RNA extraction and RT-qPCR were transferred to 1.5 mL Eppendorf tubes after dissection, flash frozen in liquid nitrogen, and stored at -80˚C.Embryos: Embryos were collected on apple juice agar plates at room temperature, then dechorionated, fixed, and devitellinized as described [51].Embryos were stored in methanol at -20˚C for �1 month.Embryos used for RNA extraction and RT-qPCR were dechorionated, flash frozen in liquid nitrogen, and stored at -80˚C.

Single molecule fluorescence in situ hybridization (smFISH)
smFISH was performed according to Abbaszadeh and Gavis [51].smFISH probe sets consisting of 20 nt oligonucleotides were designed with Stellaris Probe Designer and synthesized by Biosearch Technologies.Probes complementary to egfp (32 oligos) were conjugated to Atto 647N dye (Sigma-Aldrich) and purified by HPLC as previously described [52].smFISH probes for sfgfp (31 oligos) were purchased already conjugated to Quasar 670 fluorophore from Biosearch Technologies.Probes were labeled with far-red fluorophores to avoid detecting fluorescence from Osk-sfGFP and Nos-EGFP proteins.For quantification of total localized fluorescence intensity, 1 μL of probes per 100 μL of hybridization buffer was used; for colocalization analysis and particle quantification, 3 μL of probes per 100 μL of hybridization buffer was used.Ovaries or embryos samples were mounted under #1.5 glass coverslips (VWR) in Vectashield Mountant (Vector Laboratories) for quantification of total localized fluorescence intensity or in Prolong Diamond Antifade Mountant (Thermo Fisher Scientific) for particle quantification.

Nile Red staining
Ovaries were fixed as described above but not treated with methanol.Ovaries were washed 2×5 min.with PBST and 2×10 min.with BBT (PBST, 0.1% globulin-free BSA), then incubated in 2 mg/ml Nile Red in BBT for 20 min.Ovaries were rinsed once with PBST, then mounted in Prolong Diamond Antifade Mountant.

Microscopy and image quantification
Confocal imaging for experiments shown in Figs 2, 3, 6 and 7 was performed using a Leica SP5 laser scanning microscope with a 63× 1.4 NA oil immersion objective and GaAsP "HyD" detectors.For the smFISH experiment in Fig 5, confocal imaging was performed using a Nikon A1 microscope with a 40x 1.3NA oil immersion objective and GaAsP detectors.All imaging parameters were kept identical within each experiment.For experiments in Figs 1, 4, S1, and S2, confocal and 2D-STED imaging was performed on a Nikon A1R-STED with a 100x oil immersion objective.The sfgfp Quasar 670 probes were detected with the STAR RED setting.Nile Red was detected using the STAR ORANGE setting.Deconvolution of STED images was performed using the Nikon NIS-Elements built-in deconvolution tool.
For quantification of total fluorescence intensity, z-series with a 2 μm step size were used to capture the entire germ plasm-localized signal.Image processing and analysis were done in Fiji [53].Z-projections were made with the "sum slices" function and the threshold adjusted so the entire localized signal was included.The total fluorescence intensity of the localized signal (integrated density function in Fiji) was then measured.For the time series analysis, an ROI was drawn to encompass the germ plasm and fluorescence intensity was measured for the ROI, then the ROI was moved to an anterior region of the oocyte and background fluorescence intensity was measured and subtracted from the germ plasm ROI measurement.
For quantification of localized mRNA particle number and size (number of mRNAs per particle), HyD detectors were used in photon counting mode and z-series covering half of the thickness of an oocyte or embryo were captured with a 340 nm step size.Particles were identified and quantified as previously described [40] using a threshold of 0.5 for sfgfp probes and 0.75 for egfp probes.Only particles containing �4 mRNAs were included in the quantification [18].
Data are displayed as mean ± standard deviation.For 2-sample comparisons, the Student's t-test was used; for multiple comparisons, a one-way ANOVA was used with either Tukey or Dunnett's post-hoc test as indicated in the figure legends.Statistical analysis was performed using GraphPad Prism software.
Fig) and at the posterior of the oocyte from stage 8 to stage 13 by single molecule fluorescence in situ hybridization (smFISH) and STED microscopy (Figs S1

Fig 1 .
Fig 1. Founder granules comprise agglomerates of osk RNPs.The top row shows single confocal sections of the posterior region of oocytes at stages 10 to 13, with osk or nos detected by smFISH.The yellow boxes indicate ROIs imaged using STED microscopy as shown in the panels in the bottom row.Panels in the middle row show confocal images of each ROI prior to STED for comparison.Images are rendered using the Red Hot lookup table in Fiji and scale bars are indicated.https://doi.org/10.1371/journal.pgen.1010877.g001

Fig 4 .
Fig 4. SOLE mutants affect accumulation of osk RNPs in founder granules.The top row shows single confocal sections of the posterior region of stage 13 oocytes expressing osk-sfgfp or osk-sfgfp with the indicated SOLE mutations.The transgenic mRNAs were detected by smFISH using probes for sfgfp.The yellow boxes indicate ROIs imaged using STED microscopy as shown in the panels in the bottom row.Panels in the middle row show confocal images of each ROI prior to STED for comparison.Images are rendered using the Red Hot lookup table in Fiji and scale bars are indicated.https://doi.org/10.1371/journal.pgen.1010877.g004

Fig 5 .
Fig 5.The osk 3 0 UTR is required for late-phase localization.(A) Structure of the osk-sfgfp-K10_3 0 UTR transgene.The K10 3'UTR is indicated by the thinner black bar.(B) Confocal z projections of transgenic stage 10 (B) and stage 13 (C) oocytes, with osk-sfgfp-K10_3 0 UTR mRNA detected by smFISH.Anterior is toward the left.Some osk-sfgfp-K10_3 0 UTR mRNA accumulates at the anterior of the oocyte up to stage 10, likely due to the residual localization elements in the K10 3'UTR, but the majority is distributed throughout the ooplasm.�10 oocytes were imaged for each stage, with similar results.Scale bars are indicated.https://doi.org/10.1371/journal.pgen.1010877.g005

Fig 6 .
Fig 6.The osk 3 0 UTR is sufficient for founder granule targeting.(A) Structure of transgenes.Gray boxes: osk exonic sequences; blue boxes: nos exonic sequences; white boxes: introns; red bars: sequences creating the SOLE; green boxes: sfgfp or egfp; thinner boxes indicate 3 0 UTRs.(B-I) Confocal z-series projections of early embryos (� nuclear cycle 4).Anterior is toward the left.Transgenic mRNAs are detected by smFISH with probes for sfgfp or egfp (green).CycB smFISH (magenta) marks germ granules (C, E, G, I) and anti-Stau immunofluorescence (magenta) marks founder granules (B, D, F, H).Insets show enlargements of ROIs indicated by yellow boxes.Note that CycB is found in about 50% of all germ granules [18].(J) Quantification of colocalization between transgenic mRNAs and either germ granules or founder granules; n = 4-5 embryos each.Individual data points and mean ± standard deviation are shown; p value for each pair <0.0001 as determined by 1-way ANOVA with Tukey's post-hoc test and by Student's t-test.Scale bars are indicated.Source data for the graphs in Fig 6J are provided in S1 Data.https://doi.org/10.1371/journal.pgen.1010877.g006

Fig 7 .
Fig 7. Additional features are required to maintain osk after stage 12. (A-B) Confocal z-series projections of early embryos showing the entirety of the germ plasm (nuclear cycle 3 to 5).(C-K) Confocal z-series projections of oocytes at stage 10 (C, F, I), stage 12 (D, G, J), and stage 13 (E, H, K) capturing the entirety of the germ plasm.Anterior is toward the left.Transgenic mRNAs were detected by smFISH with probes for sfgfp or egfp.(L) Quantification of fluorescence intensity of the germ plasm from stages 10 to 13; n = 9-22 oocytes each.Individual data points and mean ± standard deviation are shown; **p < 0.01 as determined by Kruskal-Wallis one-way analysis of variance, with Dunn's post-hoc test.Scale bars are indicated.Source data for the graphs in Fig 7L are provided in S1 Data.https://doi.org/10.1371/journal.pgen.1010877.g007