Two RNA binding proteins, ADAD2 and RNF17, interact to form novel meiotic germ cell granules required for male fertility

Mammalian male germ cell differentiation relies on complex RNA biogenesis events, many of which occur in RNA binding protein (RBP) rich non-membrane bound organelles termed RNA germ cell granules. Though known to be required for male germ cell differentiation, little is understood of the relationships between and functions of the numerous granule subtypes. ADAD2, a testis specific RBP, is required for normal male fertility and forms a poorly characterized granule in meiotic male germ cells. This work aimed to define the role of ADAD2 granules in male germ cell differentiation and their relationship to other granules. Biochemical analyses identified RNF17, a testis specific RBP that forms meiotic male germ cell granules, as an ADAD2-interacting protein. Phenotypic analysis of Adad2 and Rnf17 mutant mice defined a shared and rare post-meiotic chromatin defect, suggesting shared biological roles. We further demonstrated ADAD2 and RNF17 are dependent on one another for granularization and together form a previously unstudied set of germ cell granules. Based on co-localization studies with well-characterized granule RBPs including DDX4 and PIWIL1, a subset of the ADAD2-RNF17 granules are likely components of the piRNA pathway. In contrast, a second, morphologically distinct population of ADAD2-RNF17 co-localize with the translation regulator NANOS1 and form a unique cup-shaped structure with distinct protein subdomains. This cup shape appears to be driven, in part, by association with the endoplasmic reticulum. Lastly, a double Adad2-Rnf17 mutant model demonstrated loss of ADAD2-RNF17 granules themselves, as opposed to loss of either ADAD2 or RNF17, is the likely driver of the Adad2 and Rnf17 mutant phenotypes. Together, this work identified a set of novel germ cell granules required for male fertility and sheds light on the relationship between germ cell granule pools. The example described here defines a new genetic approach to germ cell granule study. AUTHOR SUMMARY To differentiate successfully, male germ cells tightly regulate their RNA pools. As such, they rely on RNA binding proteins, which often localize to cytoplasmic granules. The majority of studies have focused on a single granule type which regulates small-RNA biogenesis. Although additional granules have been identified, there is limited knowledge about their relationship to each other and exact functions. Here, we identify an interaction between two RNA binding proteins, ADAD2 and RNF17, and demonstrate mutants share a rare germ cell phenotype. Further, ADAD2 and RNF17 colocalize to the same germ cell granule, which displays two morphologically unique types. The first subset of ADAD2-RNF17 granules have similar morphologies to other characterized granules and likely play a role in the small-RNA pathway. The second granule type forms a unique shape with distinct protein subdomains. This second population appears to be closely associated with the endoplasmic reticulum. Genetic models further demonstrate the granules themselves, as opposed to the resident proteins, likely drive the mutant phenotypes. These findings not only identify a novel population of germ cell granules but reveal a new genetic approach to defining their formation and function during germ cell differentiation.


40
The male germ cell relies on complex RNA biology for successful differentiation. As a granules, the exact functional role of or relationship between them has never been defined [22]. ADAD2 granule, we immunoprecipitated ADAD2 from wildtype (n = 3) and Adad2 mutant samples but not the mutant included well characterized post-meiotic germ cell proteins along 95 with several RNA binding proteins (Fig 1B). Of the significant peptides identified in wildtype 96 only, ADAD2-derived peptides represented nearly a tenth, confirming efficacy of pulldown.

97
However, the highest number of peptides identified belong to another RNA binding protein, ring 98 finger protein 17 (RNF17). RNF17 peptides comprised over a fifth of those identified as 99 significant. RNF17, like ADAD2, is testis-specific and has been reported to form a spermatocyte 100 granule [8].

101
To confirm the ADAD2-RNF17 interaction, immunoprecipitation of either ADAD2 or RNF17 102 in an additional set of 42 dpp wildtype testes as well as in Adad2 mutant [23] and Rnf17 mutant 103 (Rnf17 M/M ) [8] testes was performed. The resulting immunoprecipitates (IPs) were probed for 104 ADAD2 and RNF17 ( Fig 1C). As expected, IP of either ADAD2 or RNF17 in wildtype testes 105 resulted in robust detection of the precipitated protein. In the case of RNF17, this includes a 106 large and small protein isoform (RNF17L and RNF17S), both of which have been detected 107 previously [8]. Further, mutation of either Adad2 or Rnf17 resulted in a complete loss of ADAD2 108 or RNF17 detection, respectively. As expected from the IP-MS analysis, IP of ADAD2 resulted more than a single focus, representing a normal chromocenter associated with PMSC.

145
However, both Adad2 and Rnf17 round spermatids had increased numbers of H3K9me3 foci 146 compared to wildtype and the increase in both mutant models was similar. This effect was 147 independent of spermatid developmental stage thus impacting the entire post-meiotic germ cell 148 population. To date, only four [27][28][29][30]

RNF17 has a distinct localization in spermatocytes that is dependent on ADAD2
154 Two RNF17 protein isoforms, large and small, have previously been reported [8]. However, appearance of early to mid-stage pachytene spermatocytes in the developing testis. Following 160 this, the abundance of ADAD2 increased dramatically at 15 dpp when the testis cellular profile is 161 highly enriched for mid-to late-pachytene spermatocytes. A very similar pattern was also 162 observed for RNF17L. In contrast, RNF17S is observed as early as 8 dpp, reaching and 163 sustaining a maximum by 10 dpp. Together, this demonstrates ADAD2 shares a very similar 164 developmental profile specifically with RNF17L and further suggests ADAD2's interaction with 165 RNF17L, even in juvenile germ cells, would not be dependent on availability as both RNF17 166 protein isoforms are present from 10 dpp onward. [8]. To better define the timing of RNF17 granule formation, we examined RNF17 localization via immunofluorescence in wildtype adult spermatocytes throughout their differentiation ( Fig  granules persist and

184
Together, these analyses demonstrate both ADAD2 granules and the large, but not small, 185 RNF17 granules are specific to mid-to late pachytene spermatocytes. Further, both ADAD2 and 186 RNF17 large granules appear at very similar developmental time points.

187
Given that RNF17 is still present in the absence of ADAD2 (S1A and S1B Figs), we sought 188 to determine whether loss of ADAD2 impacted RNF17's spermatocyte localization.

206
ADAD2 and RNF17 localize to the same granules in spermatocytes. Given this, further 207 reference will be to the large or small ADAD2-RNF17 granules.

208
The mammalian spermatocyte contains five distinct germ cell granules [3,24], as identified 209 by electron microscopy (EM). These granules are crucial for proper germ cell development 210 [5,6,31], though their exact protein composition and function are not wholly described [22]. In 211 spite of this, several proteins are observed across four of the five granules, including DDX4 [19] 212 and DDX25 [19,20]. The only granule known to be negative for DDX4 and DDX25 is referred to 213 as the "cluster of 30 nm particles", which is observed via EM from late pachytene until the end of 214 meiosis [3,24]. Previous analysis of ADAD2 granules has demonstrated ADAD2 does not associated granules. However, ADAD2's localization with DDX4 and the localization of RNF17 217 in relation to these markers is entirely undescribed.

218
To determine whether the large ADAD2-RNF17 granule represents the protein-orphan 219 "cluster of 30 nm particles" or if ADAD2 and RNF17 instead localize with one of the better 220 described granules, we examined ADAD2 and RNF17 granule co-localization with DDX4 along with RNF17's co-localization with DDX25 (S4B and S4C Figs). For both ADAD2 and RNF17, no granule, we examined the localization of a third granule associated RBP, NANOS1 [21,24]

234
We further wondered whether the large versus small ADAD2-RNF17 granules are 235 molecularly distinct from one another. As a measure of this, we assessed the localization of two 236 well defined granule proteins in relation to ADAD2 and RNF17. These proteins, PIWIL1 and 237 PIWIL2, are both associated with processing of small non-coding RNAs known as piRNAs 238 (4,13,14) as well as localizing to granule structures in the pachytene spermatocytes [14,19]. To 239 date, both PIWIL1 and PIWIL2 are most closely associated with large ribonuclear complexes 240 referred to as piRNA-p-bodies [32] which can be visualized in the spermatocyte cytoplasm [33] 241 and overlap in large part with the well-defined IMC granule [18]. Additionally, RNF17 has been 242 implicated as a major regulator of piRNA biogenesis via interaction with PIWIL1 [34] suggesting 243 at least a subpopulation of RNF17 granules may also contain either PIWIL1 or PIWIL2. Co-

244
immunofluorescence was used to determine if ADAD2 and/or RNF17 localize with either PIWIL1 245 or PIWIL2 (Fig 4B and 4C) in the context of either the large or the small ADAD2-RNF17 granule. subset of granules. These PIWIL1 or PIWIL2 positive ADAD2 or RNF17 granules were, on represent the pool of small ADAD2-RNF17 granules that are observed during mid-to late 250 spermatocyte development. Given the lack of DDX4 in the small ADAD2-RNF17 granules, their 251 colocalization with PIWIL1 and PIWIL2 suggests the small ADAD2-RNF17 granules represent a 252 novel subtype of piRNA-containing granule, independent of the classically defined IMC.

253
Together, these observations demonstrate the ADAD2-RNF17 granules represent two  other granule types such as stress granules [35,36]. To better define the large ADAD2-RNF17 264 granule structure, we first examined ADAD2 localization to determine whether it displayed 265 variable localization within the granule (Fig 5A). This preliminary analysis revealed ADAD2 266 forms a distinct cup or funnel shape. The diameter of this structure can be seen as a ring, with

286
RNF17 appeared to be even less enriched in the interior of the cup than ADAD2. This 287 localization pattern held throughout the 3D structure of the granule (Fig 5C), with the RNF17 288 domain observed exterior to, but retaining the same shape of, the ADAD2 domain. Together, 289 these findings demonstrate that the ADAD2-RNF17 granule has distinct regions, and these 290 regions are defined by enrichment of either ADAD2 or RNF17.

292
The large ADAD2-RNF17 granule is associated with the endoplasmic reticulum

293
No other germ cell granule is known to have the distinct shape of the large ADAD2-294 RNF17 granule. As germ cell RNA granules are not bound by membranes of their own [3], they 295 are shaped by the interactions between their components and the cellular environment. The 296 non-spherical shape of the ADAD2-RNF17 granule suggests contact with another cellular component. We therefor sought to determine if the ADAD2-RNF17 granule associates with a 298 membrane bound organelle which may provide the surface to shape the ADAD2-RNF17 299 granule. Thus, we examined the co-localization of both ADAD2 and RNF17 with markers of the 300 nuclear membrane (Lamin A/C) [37], the mitochondrial membrane (COX IV) [38], and the 301 endoplasmic reticulum (SERCA1) [39]. Lamin was notably excluded from sites enriched for 302 either ADAD2 and RNF17 (Fig 6A) while the co-localization of ADAD2 or RNF17 with COX IV 303 showed a slightly more complex profile ( Fig 6B). COX IV was never observed near or within the 304 large ADAD2 or RNF17 granules. However, two populations of small ADAD2 and RNF17 305 granules were observed, some in close association to COX IV signal and some not. This 306 observation further supports the notion that the small ADAD2-RNF17 granule is distinct from the piRNA-associated granule.

309
In contrast to both Lamin and COX IV, SERCA1 was not excluded from large granules of 310 either ADAD2 or RNF17 and occasional regions of SERCA1 and ADAD2 or RNF17 co-311 enrichment detected ( Fig 6C) suggesting potential association of the ER with the large ADAD2-312 RNF17 granule. To better define the spatial association of the endoplasmic reticulum with the 313 ADAD2-RNF17 granule, we examined the co-localization of ADAD2 and SERCA1 using 314 confocal microscopy. These analyses clearly identified the distinct localization of the ADAD2-315 RNF17 granule along with defining the tubular structure indicative of the ER [40] (Fig 6D).   relationship between these granules. To that end, this work aimed to characterize a recently 347 identified granule that contains the RNA binding protein ADAD2, which is required for male protein also required for male fertility, RNF17 [8], as an ADAD2 interacting partner. Genetic knockout models combined with localization studies demonstrated ADAD2 and RNF17 are co-351 dependent on one another to form at least two distinct populations of ADAD2-RNF17 granules.

352
Protein composition studies of these ADAD2-RNF17 granules further showed molecularly 353 distinct subpopulations not related to the best characterized granule, the IMC (Fig 8). Lastly,

360
Most genetic studies of male germ cell granules to date have utilized single gene 361 knockout models [11,31,[41][42][43]. Although powerful for defining the function of the targeted 362 protein, these approaches fail to inform on the function of the granules themselves as they also 363 impact any non-granule functions of the chosen target [22]. Supporting this notion, single gene 364 knockouts of granule-associated proteins often lead to phenotypes prior to formation of the 365 granule [13,31]. Single loss of ADAD2 and RNF17 avoids these complications as both result in 366 phenotypes after granule formation. Further, their reliance on one another for granule formation 367 means reciprocal studies in the two models provide an opportunity to study the granule-specific  size and apparent density makes it likely to be detected by EM [3]. Further, lack of co-388 localization with the granule protein DDX4 strongly suggests it is the protein-orphan granule 389 "cluster of 30 nm particles", which is the only granule known to be negative for DDX4 [24]. As 390 such, this report represents the first identification of proteins associated with this exclusively 391 EM-defined structure. Further, localization of NANOS1 to the structure provides the first hint 392 towards a potential function. The Drosophila melanogaster homolog of NANOS1 is known to 393 facilitate localized translation regulation in the embryo [47] while in Xenopus, Nanos1 facilitates 394 translational repression in the germline [48]. Together, the emerging model is that NANOS1 397 That NANOS1 appears to be shared between the large ADAD2-RNF17 granule and the 398 other EM-defined granules is also the first biological connection between the orphan granule 399 and the broader granule population. For the first time, this suggests that all the EM-defined granules may share one or more components and calls into question the commonly held belief that germ cell granules represent distinct pools within the germ cell cytoplasm. This notion is supported by co-localization studies of the small ADAD2-RNF17 granule. As all small ADAD2-403 RNF17 granules contain the piRNA processing proteins PIWIL1 and PIWIL2, they are likely 404 associated with either piRNA processing or piRNA action. To date, the IMC is the only known 405 site for piRNA processing in meiotic male germ cells [14]. However, a subset of small ADAD2-406 RNF17 granules is separate from mitochondria, demonstrating them to be independent of the hypothesis. First, as discussed above and recently reviewed [22], there are multiple proteins 413 observed across many granules. Second, multiple granules have already been shown, by both 414 EM and immunodetection, to actively share components [3,19,24,49]. Although this hypothesis 415 will require continued rigorous testing, it represents an exciting future avenue of germ cell 416 granule research.

417
The last findings of importance to the broader germ cell granule field herein are those 418 focused on the structure of the ADAD2-RNF17 granule. First, localization of ADAD2 within the 419 large granule identified distinct regions of high and low density, in particular along the rim of the 420 structure and relative to the interior. This is reminiscent of what is observed in stress granules, 421 especially for the stress granule protein G3BP1 [50][51][52]. Stress granules are well-described,

422
RNA-rich, non-membrane bound structures that form under specific cellular stresses [53]. As 423 such, they represent an excellent model system to better understand granule formation. Recent 424 work has demonstrated that nucleation of G3BP1 drives stress granule formation, which is 425 ultimately the result of liquid-liquid phase separation [54]. It is unknown whether this is a primary separation other observations suggest additional levels of regulation. In contrast to stress 431 granules, which are spherical or oblong in nature and are not known to interact with any specific 432 membrane-bound organelles [56], the ADAD2-RNF17 granule displays multiple facets that 433 appear flat. This suggests some sort of physical or mechanical constraint on granule formation, 434 perhaps similar to the constraints mitochondrial tethering puts on IMC components [44,60].

435
Supporting this, the ADAD2-RNF17 granule appears to be in close contact with the ER 436 membrane, which may be providing the necessary mechanical force to generate the flat 437 surfaces observed in the ADAD2-RNF17 granule. Whether ADAD2 and/or RNF17 are directly or 438 indirectly tethered to the ER remains a question for future work.
Green -RNF17, red -SYCP3, blue -DAPI. 400x magnification. Co-immunofluorescence of ADAD2 and RNF17 in adult wildtype testes across selected pachytene spermatocyte developmental stages demonstrating colocalization. Roman numerals -testis tubule cross-section stage (VIII containing mid-stage pachytene spermatocytes, IX through XI containing late-stage pachytene spermatocytes). Asterisks -small granules and arrowheads -large granules. Red -RNF17, green -ADAD2, and blue -DAPI. Coimmunofluorescence of B. PIWIL2 and C. PIWIL1 with ADAD2 or RNF17 in adult wildtype testes demonstrating co-localization in a subset of ADAD2 or RNF17 granules. Red -PIWIL2 or PIWIL1, green -ADAD2 or RNF17, and blue -DAPI. Asterisks -small ADAD2 or RNF17 granules. Arrowheads -large ADAD2 or RNF17 granules. 630x magnification for all images.  Co-immunofluorescence in adult wildtype testes of the nuclear membrane marker Lamin A/C or B. the mitochondrial marker COX IV and ADAD2 or RNF17 demonstrating neither ADAD2 nor RNF17 large granules colocalize with either the nuclear membrane or mitochondria but a subset of small granules colocalize with the mitochondria. Red -Lamin A/C or COX IV, green -ADAD2 or RNF17, and blue -DAPI. Asterisks -small ADAD2 or RNF17 grnaules and arrowheadslarge ADAD2 or RNF17 granules. 630x magnification for above images. C. Co-immunofluorescence in adult wildtype testes of the endoplasmic reticulum marker SERCA1 and ADAD2 or RNF17 showing clustering of the SERCA1 signal at large ADAD2 or RNF17 granules. Red -SERCA1, green -ADAD2 or RNF17, and blue -DAPI. Asterisks -small ADAD2 or RNF17 grnaules and arrowheads -large ADAD2 or RNF17 granules.   Supp. Figure 7. Western blot loading controls. SYPRO-Ruby stained membranes for blots shown in A. S1 Fig, B. S2 Fig, and C. S5 Fig showing equal loading across lanes.
Stage indicated by Roman numerals. I-II/III: SYCP3 stains small pachytene spermatocytes (*) with threads and multiple large blobs of signal and strongly stains the chromocenter of round spermatids. IV: SYCP3 stains small pachytene spermatocytes similar to II/III and the chromocenter of round spermatids, but weakly. V: SYCP3 stains larger pachytene spermatocytes with distinct threads and a few large blobs of signal. VI: SYCP3 weakly stains the B-spermatogonia (*) and large pachytene spermatocytes with distinct threads. VII: SYCP3 stains preleptotene spermatocyte nucleoplasm (*) and large pachytene spermatocytes with distinct threads. VIII: SYCP3 strongly stains preleptotene spermatocyte nucleoplasm along with intensely stained patches (*) and larger pachytene spermatocytes with distinct threads that are further apart than in VII. IX: SYCP3 strongly and unevenly stains leptotene spermatocyte nucleoplasm (*) more strongly than large pachytene spermatocytes which display distinct threads. X: SYCP3 stains leptotene spermatocytes (*) which display weak threads and large pachytene spermatocytes with slightly more diffuse threads. XI: SYCP3 stains zygotene spermatocytes (*) with mostly intact threads and large diplotene spermatocytes with slightly diffuse threads. XII: SYCP3 stains zygotene spermatocytes (*) with intact threads and weakly stains M2 spermatocytes with diffuse signal.