Novel functions of the ubiquitin-independent proteasome system in regulating Xenopus germline development

The primordial germ cells (PGCs), which give rise to sperm and oocytes in adult animals, are considered the ‘stem cells of the species’ (Wylie, 1999; Seydoux and Braun, 2006). Specification of PGCs occurs through different mechanisms in different species. In mammals, PGCs are induced after implantation when proximal cells in the epiblast receive BMP signals from extra-embryonic tissues. In contrast, Drosophila, Caenorhabditis elegans, zebrafish and Xenopus specify their PGCs through inheritance of germ plasm (Aguero et al., 2017b; Extavour and Akam, 2003; Saitou, 2009; Strome and Lehmann, 2007; Tam and Zhou, 1996; Lawson et al., 1999; Ying et al., 2001; Ohinata et al., 2009). Despite the differences in how PGCs are specified, ∼80% of transcripts enriched in amphibian PGCs (Butler et al., 2018) are expressed in human gonadal PGCs (Irie et al., 2015). Many aspects of PGC development are highly conserved across species. In animals such as Drosophila, C. elegans, zebrafish and Xenopus, the genetic program responsible for early PGC development is encoded by maternally expressed germ plasm RNAs and proteins. After fertilization, these factors act in a precisely regulated fashion to direct the survival, proliferation and migration of PGCs. How early germline development is orchestrated by translational and post-translational regulatory mechanisms remains unclear.

In animals that specify the germline through inheritance of the germ plasm, RNAs coding for germline determinants are transcribed during early oogenesis and asymmetrically localized in the oocyte. Translation of many germline-specific RNAs is dynamically regulated. This is particularly true for germline determinants that promote PGC development by suppressing somatic differentiation. These proteins are often expressed at very low levels, or not at all, in fully grown oocytes. Accumulation of these factors occurs after fertilization when germline development is initiated (Yang et al., 2015). Premature expression of these germline determinants in the oocyte may result in ectopic distribution of these factors, which later interferes with the development of somatic tissues in the embryo (Luo et al., 2011; Mei et al., 2013).

One such factor is Dnd1, an evolutionarily conserved RNA-binding protein essential for vertebrate germline development (Gross-Thebing et al., 2017; Weidinger et al., 2003; Youngren et al., 2005; Horvay et al., 2006; Cook et al., 2009). During early embryonic development, Dnd1 prevents somatic differentiation of PGCs (Gross-Thebing et al., 2017), at least partially through promoting the expression of Nanos1 (Kedde et al., 2007; Aguero et al., 2017a, 2018), which represses translation of somatic genes in PGCs (Nakahata et al., 2001; Sonoda and Wharton, 1999, 2001; Jaruzelska et al., 2003; Schaner et al., 2003; Deshpande et al., 1999; Sato et al., 2007; Lai et al., 2012). In Xenopus, Dnd1 protein is highly expressed during cleavage and blastula stages to promote nanos1 translation (Aguero et al., 2017a). Interestingly, a very low level of Dnd1 protein expressed in the vegetal cortex in the oocyte (Aguero et al., 2017a) maintains vegetal localization Trim36 protein by anchoring trim36 mRNA to the vegetal cortex (Mei et al., 2013). After fertilization, vegetally localized Trim36 promotes microtubule assembly during cortical rotation and subsequent dorsal development (Cuykendall and Houston, 2009). The level of Dnd1 protein must be tightly controlled in the oocyte. Knockdown of Dnd1 in the oocyte disrupts the asymmetric localization of trim36 and results in ventralized embryos (Mei et al., 2013). On the other hand, overexpression of Dnd1 triggers premature nanos1 translation in the oocyte (Aguero et al., 2017a), impairing early embryonic patterning (Mei et al., 2013). Currently, it is unclear how the expression of Dnd1 protein is precisely regulated.

The temporal and spatial expression of proteins is often orchestrated by dynamically regulated transcription, translation and protein turnover. By promoting protein turnover, the proteasome system, together with the autophagy/lysosome system, plays a major role in controlling the expression of proteins (Wong and Cuervo, 2010). The proteasome system consists of several types of macromolecular complexes capable of proteolysis inside the cell. The proteolytic activity of the proteasome resides in the 20S proteasome core particle (20S CP), a complex of 28 subunits. Protein substrates are recruited to the 20S CP by several types of proteasome activators. These include 19S regulatory particle (19S RP), proteasome activators PSME1 (PA28α), PSME2 (PA28β), PSME3 (PA28γ) and PSME4 (PA200). Among these, 19S RP, which forms the 26S proteasome together with 20S CP, recruits polyubiquitylated proteins in an ATP-dependent manner. PSME1, PSME2, PSME3 and PSME4 are responsible for recruiting non-ubiquitylated protein substrates. PSME1/PSME2 and PSME3 form hetero- or homo-heptameric complexes, respectively. PSME4 acts as a monomer to open the gate of the 20S CP (Hwang et al., 2011; Rechsteiner and Hill, 2005; Kish-Trier and Hill, 2013; Ben-Nissan and Sharon, 2014). Involvement of the ubiquitin-independent proteasome during germline development has been previously documented (Khor et al., 2006; Qian et al., 2013; Huang et al., 2016). In mice, compound mutants for PSME3 and PSME4 are completely infertile in males (Huang et al., 2016). So far, only acetylated core histones have been identified as substrates of ubiquitin-independent proteasomes during germline development (Qian et al., 2013). Additional substrates and roles of the ubiquitin-independent proteasome in germline development remain uncharacterized.

Here, we report that two inhibitory mechanisms, dnd1 translational repression and ubiquitin-independent proteasomal turnover of Dnd1, act together to prevent premature accumulation of Dnd1 protein before fertilization. These inhibitory mechanisms are relieved during the oocyte-to-embryo transition, allowing an abrupt increase in the level of Dnd1 protein. Our results further reveal that the proteasome system undergoes a dramatic relocalization during the oocyte-to-embryo transition. Consequently, proteasomes become separated from vegetally localized dnd1 and other germ plasm components in the embryo. Ectopic expression of PSME1/PSME2, PSME3 and PSME4 in the vegetal pole of the embryo impairs PGC development. Our study thus uncovers novel functions of the ubiquitin-independent proteasome in the regulation of vertebrate germline development.

We previously reported that a low level of Dnd1 protein is expressed in the oocyte (Aguero et al., 2017a), where it anchors trim36 RNA to the vegetal cortex (Mei et al., 2013). After fertilization, the expression of Dnd1 increases abruptly to promote nanos1 translation, essential for germline development (Aguero et al., 2017a). Consistent with these observations, we found that the expression level of Dnd1 protein was low in the oocyte and egg, but increased significantly after fertilization (Fig. 1A). Interestingly, after germinal vesicle breakdown (GVBD), nearly 50% of Dnd1 protein is phosphorylated (Fig. S1). To understand how the expression of Dnd1 protein is controlled, we first investigated dnd1 translational regulation. We focused on the 3′UTR of dnd1, because the 5′UTR of dnd1 is very short and unlikely to be important for translational regulation. We generated a GFP-myc-dnd3′UTR by fusing the 3′UTR of Xenopus tropicalis dnd1 to a myc-tagged GFP. GFP-myc-SV40, in which myc-GFP was fused with a SV40 polyA signal, was used as the well-translated control. RNAs (250 pg) coding for GFP-myc-dnd3′UTR and GFP-myc-SV40 were injected into fully grown oocytes. Injected oocytes were cultured for 24 h, treated with progesterone to induce oocyte maturation, and artificially activated by pricking with a glass needle. The expression of myc-GFP was monitored by western blot. As shown in Fig. 1B, the expression level of GFP-myc-dnd3′UTR was lower than that of GFP-myc-SV40 in fully grown oocytes, suggesting that dnd1 translation is repressed in the fully grown oocyte. After GVBD, we detected a dramatic increase in the expression of GFP-myc-dnd3′UTR. After egg activation, the expression level of GFP-myc-dnd3′UTR further increased by 2- to 3-fold over the 4-h period (Fig. 1B). To test directly whether the 3′UTR of dnd1 contains cis-regulatory element(s) that repress its translation in the oocyte, we inserted the complete 3′UTR of dnd1 into GFP-myc-SV40 (nucleotides 1130-1715, Fig. 1C). In the oocyte, the expression level of 1130-1715 was lower than that of GFP-myc-SV40. After GVBD, whereas the expression GFP-myc-SV40 remained unchanged, the expression of 1130-1715 increased. Using the same strategy, we tested the 3′UTR of X. laevis dnd1 and found it behaved essentially the same as the 3′UTR of X. tropicalis dnd1 (Fig. 1D). Based on these results, we conclude that the 3′UTR of dnd1 contains cis-regulatory element(s) that repress dnd1 translation in the oocyte. This repression is relieved after GVBD.

Although translational repression of dnd1 is relieved after oocyte maturation, the expression of endogenous Dnd1 protein remains at a low level in the egg. This observation suggests that protein turnover plays an important role in preventing premature accumulation of Dnd1 protein before fertilization. To assess Dnd1 protein turnover in the oocyte, we depleted dnd1 mRNA by injection of a dnd1 antisense oligo (As-oligo) (Mei et al., 2013), and monitored the levels of endogenous dnd1 mRNA and Dnd1 protein. We found that at 2 h post-As-oligo injection, Dnd1 protein was reduced by more than 50% (Fig. 2A,B), demonstrating that Dnd1 protein has a short half-life in the oocyte. In parallel to this experiment, we fused Dnd1 to GFP and compared the expression of GFP and GFP-Dnd1 in the oocyte and embryo. In the oocyte, although GFP was clearly expressed, we could barely detect GFP-Dnd1. By contrast, GFP and GFP-Dnd1 were expressed at comparable levels in the embryo (Fig. 2C). This result confirms that Dnd1 is unstable in the oocyte. To assess the half-life of Dnd1 protein directly, we blocked protein synthesis by treating oocytes and embryos with cycloheximide (CHX) and measured the level of endogenous Dnd1 protein at multiple time points. We found that the half-life of Dnd1 in the oocyte is ∼70 min. In the embryo, however, the half-life of Dnd1 protein is close to 4 h. In HEK293T cells, the half-life of overexpressed Dnd1-myc is ∼2 h (Fig. 2D). Taken together, we conclude that Dnd1 is intrinsically unstable in the oocyte, but becomes stable after fertilization.

To identify the degron that targets Dnd1 protein for rapid degradation in the oocyte, we fused various deletions of Dnd1 to GFP-myc (Fig. 3A). RNAs coding for GFP-myc and fusion proteins (200 pg each) were injected into oocytes, followed by western blot analysis (Fig. 3B). Compared with GFP, all fusion proteins carrying an evolutionarily conserved motif located between residues 107 and 127 of Dnd1 (Fig. 3C) were poorly expressed in the oocyte. In contrast, all constructs lacking this motif were expressed at levels comparable to that of GFP-myc (Fig. 3A,B). To determine whether this degron functions specifically in the oocyte, we transfected GFP_D107-127 into HEK293T cells. Whereas GFP-myc was abundantly expressed, we could detect only a trace amount of GFP_D107-127 protein (Fig. 3D). RT-qPCR revealed that transcription of GFP_D107-127 and GFP-myc was comparable (Fig. 3E). In parallel, we fused various copies of D107-127 to luciferase and expressed them in NIH3T3 cells. The luciferase assay revealed that D107-127 decreased the expression of luciferase in a dose-dependent manner (Fig. 3F). Thus, D107-127 is a potent degron capable of promoting protein degradation not only in the oocyte, but also in mammalian somatic cells. We generated a truncated Dnd1 lacking this degron (Δ_D107-119). As expected, Δ_D107-119 is much more stable than the wild-type Dnd1-myc in the oocyte and in HEK239T cells (Fig. 3G). Collectively, these results demonstrate that D107-127 targets Dnd1 for rapid degradation in the oocyte.

By blasting the Xenopus protein database using the sequence of D107-127, we found a few proteins with a similar motif. These include hnRNP R, Syncrip, A1CF, RBM47 and RBM46 (Fig. 4A). We generated myc-tagged constructs and transfected them into HEK293T cells. Whereas Dnd1 and RBM46 were poorly expressed, hnRNP R, Syncrip and A1CF were expressed at much higher levels (Fig. 4B). Because phenylalanine 110 in Dnd1 is uniquely conserved between Dnd1 and RBM46, and not in the other proteins, we generated an F110M mutant to test the significance of F110. As a control, we also mutated S117, a residue that is not uniquely conserved in Dnd1 and RBM46, by replacing the serine residue with an aspartic acid residue (S117D). When overexpressed in the oocyte, the expression of F110M was much higher than that of the wild-type Dnd1. In the embryo, F110M and the wild type Dnd1 were expressed at a comparable level (Fig. 4C). These results indicate that residue F110 is essential for D107-127 to function as a degron, and further support our conclusion that the stability of Dnd1 is differentially regulated before and after fertilization. To determine whether overexpression of F110M would affect germline development, we injected RNA encoding F110M into the vegetal pole of fertilized eggs. We found that overexpression of F110M had little effect on the number of PGCs at the tailbud stage (Fig. S2).

To understand Dnd1 turnover better, we transfected myc-Dnd1 and myc-GFP into HEK293T and GC-2spd cells, an immortalized mouse spermatocyte cell line (Hofmann et al., 1994). Transfected cells were treated with various inhibitors that selectively inhibit proteasomes, lysosomes or autophagy. In both HEK293T and GC-2spd cells, the proteasome inhibitor MG132 significantly increased the expression of myc-Dnd1 protein (Fig. 5A). Lactacystin, another proteasome inhibitor, increased the expression of Dnd1 too, albeit to a lesser extent, likely because of the difference in the potency of the two inhibitors. Both MG132 and lactacystin induced a high molecular weight smear of myc-Dnd1, indicative of polyubiquitylation of Dnd1 protein. The autophagy inhibitor 3MA increased the expression of Dnd1 in HEK293T cells, but failed to do so in GC-2spd cells, suggesting that Dnd1 may be regulated by autophagy in a context-dependent manner. Ammonium chloride, which inhibits lysosomal degradation by neutralizing the low pH of lysosome, had no detectable effect on Dnd1 in both cell lines. Because we were interested in Dnd1 turnover in oocytes, we chose to focus on the proteasome.

To determine whether D107-127 targets Dnd1 to the ubiquitin proteasome pathway, we transfected myc-Dnd1, Δ_D107-119 and GFP_D107-127 into GC-2spd cells. We found that MG132 treatment increased the expression of myc-Dnd1, Δ_D107-119 and GFP_D107-127. Surprisingly, we observed that MG132 induced a high molecular weight smear of myc-Dnd1 and Δ_D107-119, but not GFP_D107-127 (Fig. 5B). This raises the striking possibility that although Dnd1 can be degraded by the ubiquitin/proteasome system, D107-127 targets Dnd1 to the proteasome degradation pathway in a ubiquitin-independent manner. To strengthen this conclusion, we overexpressed myc-Dnd1, Δ_D107-119 and GFP_D107-127 together with HA-tagged ubiquitin (HA-Ub) and cultured cells in the presence of MG132. Myc-Dnd1, Δ_D107-119 and GFP_D107-127 were immunoprecipitated using an anti-myc antibody and probed for HA-Ubiquitin. Indeed, the high molecular weight smear of myc-Dnd1 and Δ_D107-119 was recognized by the HA antibody, demonstrating that myc-Dnd1 and Δ_D107-119 were polyubiquitylated (Fig. 5C). By contrast, GFP_D107-127 was not recognized by the HA antibody (Fig. 5C), arguing strongly that D107-127 targets Dnd1 to the ubiquitin-independent proteasome pathway.

Inside the cell, non-ubiquitylated proteins are recruited to the proteasome by the PSME1/2 heteroheptamer, the PSME3 homoheptamer or the PSME4 monomer (Hwang et al., 2011; Rechsteiner and Hill, 2005; Kish-Trier and Hill, 2013; Ben-Nissan and Sharon, 2014). To determine whether PSME1/2, PSME3 or PSME4 are required for D107-127-mediated proteasomal turnover of Dnd1, we generated dominant negative (dn) proteasome activators. Our results reveal that overexpression of dnPSME1 (PSME1 N141Y), dnPSME2 (PSME2 N139Y), dnPSME3 (PSME3 N151Y) (Zhang et al., 1998) or dnPSME4 (PSME4^1635-1723, the BRDL region of PSME4) (Qian et al., 2013) individually did not stabilize GFP_D107-127. By contrast, GFP_D107-127 was stabilized when all four dominant negative constructs were co-overexpressed (Fig. 5D). Consistently, co-overexpression of all four dominant negative proteasome activators increased the level of endogenous Dnd1 protein in the oocyte (Fig. 5E,F). This suggests that PSME1/2, PSME3 and PSME4 function redundantly to promote ubiquitin-independent Dnd1 turnover in the oocyte.

Mutation of degron D107-127 increases the expression of Dnd1 in the oocyte, but fails to do so in the embryo (Figs 3G and 4C). This indicates that the ubiquitin-independent proteasomal turnover of Dnd1 is inhibited after fertilization. In principle, this may be caused by inhibition of the ubiquitin-independent proteasome pathway after fertilization. Alternatively, a regulatory mechanism may be activated after fertilization to protect Dnd1 from ubiquitin-independent proteasomal degradation in the embryo. To distinguish between these possibilities, we assessed the expression of several Dnd1 GFP fusion constructs in the embryo. All of these constructs contain degron D107-127, which is located in RRM1, but vary in their N- or C-terminal sequences (Fig. 5G). Our results revealed that GFP_D96-247, GFP_D96-C, GFP_D1-127 and GFP_D96-127 were poorly expressed in the embryo. In contrast, the expression of GFP_D1-247, GFP_D72-151 and GFP-Dnd1 was similar to that of myc-GFP (Fig. 5H). It appears that all Dnd1 deletion constructs lacking an intact RRM1 are unstable in the embryo, whereas constructs containing an intact RRM1 are stable (Fig. 5G). These results demonstrate that the ubiquitin-independent proteasome pathway is active in the embryo, and argue for the existence of a regulatory mechanism that acts on the RRM1 to prevent Dnd1 from ubiquitin-independent proteasomal degradation after fertilization.

Inspired by the above findings, we examined the expression of psme1, psme2, psme3 and psme4. RT-PCR results revealed that total levels of psme1, psme2, psme3 and psme4 RNAs remain unaltered during the oocyte-to-embryo transition (Fig. 6A). In stark contrast, the subcellular localization of these RNAs changes dramatically during the oocyte-to-embryo transition. In the oocyte, psme1, psme2, psme3 and psme4 RNAs form a smooth gradient along the animal-vegetal axis, with the highest concentration of RNAs being detected in the animal pole (Fig. S3). In the vegetal hemisphere, the trailing end of psme1, psme2, psme3 and psme4 overlaps with pgat (Hudson and Woodland, 1998) and dnd1 (Horvay et al., 2006), germ plasm RNAs that are transported to the vegetal pole during oogenesis through the early and late localization pathways, respectively (Kloc et al., 2001; Kloc and Etkin, 2005). In the mature egg, psme1, psme2, psme3 and psme4 RNAs become restricted to the animal hemisphere, with a boundary at the equator that separates them from vegetally localized pgat and dnd1. This pattern persists after fertilization. At the 2-cell stage, RNAs coding for proteasome activators remain restricted to the animal hemisphere (Fig. 6B).

We extended our analysis by examining the expression of psmc6 and psma2, which encode components of the 19S regulatory particle (RP) and 20 S core particle (CP), respectively. In situ hybridization results revealed that the expression pattern of psmc6 and psma2 was essentially the same as that of psme1, psme2, psme3 and psme4, forming a gradient along the animal-vegetal axis in the oocyte, but becoming uniformly distributed in the animal hemisphere after maturation (Fig. 6B). We also examined the expression of eIF4A1 and eIF4E, two housekeeping genes that encode key components of the translational machinery. eIF4A1 and eIF4E can be detected in the entire oocyte, but are more abundant in the animal hemisphere. This pattern remains unchanged during the oocyte-to-embryo transition (Fig. 6B).

We further examined the subcellular distribution of 20S CP during the oocyte-to-embryo transition by immunofluorescence. Similar to the dynamic expression pattern of RNAs coding for proteasome components, we found that 20S CP formed an animal-to-vegetal gradient in the oocyte. After oocyte maturation, the staining signal became stronger in the animal hemisphere. In the 2-cell-stage embryo, the animal/vegetal difference is more obvious, most likely because of new synthesis of proteasome components in the animal hemisphere (Fig. 6C). It appears that the proteasome system undergoes a dramatic vegetal-to-animal relocalization during the oocyte-to-embryo transition. To further confirm the graded distribution of the proteasome system in early embryos, cytoplasm was collected from animal and vegetal hemispheres of 1-cell-stage embryos using a micropipette-based nanofluidic device (Saha-Shah et al., 2015). The harvested cytoplasm was analyzed directly by shotgun proteomics (A.S., P.S.K. and B.G. unpublished). We detected several proteasome components, including PSMA2, PSMA3, PSMA4, PSMA5, PSMA6, PSMB1, PSMB3, PSMB4, PSME1 and PSME3 (Table S1). Consistent with immunofluorescence of 20S CP, we found all PSMAs, PSMBs and PSME1 were enriched in the animal hemisphere. Compared with the above proteasome components, PSME3 was detected at low levels by mass spectrometry (Fig. 6D) and hence quantified with lesser confidence. We believe that the lower apparent enrichment of PSME3 in the animal hemisphere relative to other components of 20S CP is due to the difficulty in detecting this low abundance protein. Collectively, our immunofluorescence and proteomic data support the conclusion that proteasomes are enriched in the animal hemisphere in early Xenopus embryos.

The above results raise the striking possibility that the vegetal-to-animal relocalization of the proteasome system may be a prerequisite for the germline development in the embryo. We thus determined whether overexpression of wild-type PSME1, PSME2, PSME3 and PSME4, which prevent premature expression of Dnd1 in the oocyte, would interfere with PGC development. We injected psme1+psme2 (1 ng each), psme3 (1 ng), psme4 (1 ng), or a combination of all four mRNAs (330 pg each) into the vegetal pole of 1-cell-stage embryos. These embryos were harvested at the late tailbud stage and analyzed by in situ hybridization for pgat, a widely used marker for PGCs (Hudson and Woodland, 1998). We found that the number of pgat-positive PGCs were significantly reduced in PSME1+PSME2-, PSME3- and PSME4-overexpressing embryos (Fig. 7A,B). Some of PSME4-overexpressing embryos were cultured until stage 48. Importantly, we did not detect any morphological or histological abnormalities in PSME4-overexpressing embryos (Fig. S4). It appears that overexpression of ubiquitin-independent proteasome activators in the vegetal pole specifically affects germline development. We conclude that the vegetal-to-animal translocation of psme1, psme2, psme3 and psme4 during the oocyte-to-embryo transition is crucially important for PGC development after fertilization.

Because Dnd1 is protected from the ubiquitin-independent proteasome after fertilization (Fig. 5G,H), we speculated that the ubiquitin-independent proteasome is capable of degrading other PGC regulators in the embryo. We tested three germ plasm components that are abundantly expressed in PGCs (Butler et al., 2018): Trim36 (Cuykendall and Houston, 2009), Dazl (Houston and King, 2000; Houston et al., 1998) and Nanos1 (Zhou and King, 1996; Lai et al., 2012, 2011; Luo et al., 2011). At the 1-cell stage, RNAs encoding myc-Trim36 (100 pg), myc-Dazl (100 pg) and myc-Nanos1 (100 pg) were injected into the vegetal pole alone, or together with wild-type psme4 (1 ng). At stage 9, embryos were harvested for western blot analysis. As shown in Fig. 7C, overexpression of PSME4 reduced the expression of Trim36, without affecting Dazl and Nanos1. This suggests that the ubiquitin-independent proteasome interferes with PGC development by promoting degradation of a subset of germ plasm components.

During early embryogenesis, Dnd1 regulates PGC development (Weidinger et al., 2003; Horvay et al., 2006; Gross-Thebing et al., 2017) by protecting germline-specific RNAs and promoting the expression of PGC regulators (Kedde et al., 2007; Koebernick et al., 2010; Aguero et al., 2017a). In Xenopus, a low level of Dnd1 protein is expressed in the oocyte (Aguero et al., 2017a) and plays a pivotal role in anchoring trim36 to the vegetal cortex (Mei et al., 2013). After fertilization, vegetally localized Trim36 regulates polymerization of microtubules during cortical rotation, allowing transportation of dorsal determinants to the future dorsal side of the embryo (Cuykendall and Houston, 2009). Knockdown of Dnd1 in the oocyte disrupts microtubule assembly during cortical rotation and results in ventralized embryos. Interestingly, overexpression of Dnd1 in the oocyte impairs early embryonic development (Mei et al., 2013), presumably by triggering premature nanos1 translation in oocytes (Aguero et al., 2017a), which interferes with the development of soma (Luo et al., 2011). The goal of the current study was to understand how Dnd1 protein is so precisely regulated at the translational and post-translational levels.

The results presented here reveal two mechanisms that act together to prevent premature accumulation of Dnd1 protein before fertilization: translational repression, mediated by the inhibitory element(s) in the 3′UTR of dnd1 RNA, and ubiquitin-independent proteasome degradation of Dnd1 protein. Translational repression of dnd1 is relieved after GVBD. Dnd1 protein is protected from the ubiquitin-independent proteasome pathway in the embryo, possibly as a result of post-translational modification of the RRM1 or the expression of an RRM1-binding protein after fertilization, which reduces the proteasomal accessibility of Dnd1. As a consequence of inactivation of these inhibitory mechanisms during the oocyte-to-embryo transition, Dnd1 protein accumulates in the embryo. Interestingly, we found that Dnd1 can be regulated by the ubiquitin-dependent proteasome and autophagy in vitro. It is possible that these protein turnover mechanisms each contribute to overall regulation of Dnd1 protein during early germline development.

The finding that Dnd1 is degraded by the ubiquitin-independent proteasome pathway is intriguing. Although functions of the 26S proteasome, which degrades polyubiquitylated substrates, have been documented by a large body of literature (for reviews, see Muller and Schwartz, 1995; DeRenzo and Seydoux, 2004; Bowerman and Kurz, 2006; Karabinova et al., 2011; Tsukamoto and Tatsumi, 2018), little is known about the ubiquitin-independent proteasome during development. Only a handful of substrates for the ubiquitin-independent proteasome have been identified (Hoyt and Coffino, 2004). Some recent studies uncovered roles of the ubiquitin-independent proteasome during spermatogenesis (Huang et al., 2016; Qian et al., 2013) and identified acetylated core histones as substrates for the ubiquitin-independent proteasome (Qian et al., 2013). However, roles of the ubiquitin-independent proteasome during germline development are largely unknown. Our results demonstrate that, in Xenopus, Dnd1 and Trim36 are novel substrates for the ubiquitin-independent proteasome. In the case of Dnd1, degron D107-127, a small motif located in the RRM1, targets Dnd1 to the ubiquitin-independent proteasome pathway. Mutation of this degron leads to stabilization of Dnd1 protein in the oocyte. The sequence of D107-127 is similar to a motif found in TRP2 (amino acid residues 181-193), which is processed by the PSME1/PSME2-containing proteasome (Murata et al., 2001). Consistent with our observation, Li et al. recently identified several mutations that promote proteasomal degradation of Dnd1. These mutations are located in the RRM1 of Dnd1 and alter the structure around degron D107-127 (Li et al., 2018). We show that protein degradation mediated by degron D107-127 could be blocked in Xenopus oocytes when dnPSME1, dnPSME2, dnPSME3 and dnPSME4 were co-expressed. This demonstrates that PSME1/PSME2-, PSME3- and PSME4-containing proteasomes function redundantly to prevent accumulation of Dnd1 protein in the oocyte, highlighting a novel function of the ubiquitin-independent proteasome in preventing premature initiation of germline development.

Our work reveals that psme1, psme2, psme3 and psme4, and other proteasome RNAs (psmc6 and psma2) form an animal-to-vegetal gradient in the oocyte. The distribution of these RNAs changes dramatically during the oocyte-to-embryo transition. By the 2-cell stage, these RNAs are restricted to the animal hemisphere and become separated from vegetally localized germ plasm. This observation is consistent with results from a recently published RNAseq analysis (Sindelka et al., 2018). We assessed the subcellular localization of 20S CP by immunofluorescence and observed similar dynamic changes, albeit with slower kinetics. Although 20S CP is only slightly enriched in the animal hemisphere in the oocyte and egg, we detected a much larger animal/vegetal difference by the 2-cell stage, likely due to new synthesis of proteasome in the animal hemisphere. We collected cytoplasm from animal and vegetal hemispheres of fertilized eggs and performed proteomic analysis. Quantitative proteomic measurement confirms that proteasome components are animally enriched at the 1-cell stage. Our results further indicate that overexpression of ubiquitin-independent proteasome activators in the vegetal pole of embryos reduced the expression of Trim36 during blastula stages and decreased the number of PGCs at the tailbud stage. These observations support our model that relocalization of ubiquitin-independent proteasomes during the oocyte-to-embryo transition reduces proteasomal turnover rate in the vegetal pole, creating a permissive environment for germline development after fertilization (Fig. 7D).

We have noticed that the graded distribution of the proteasome was not detected in a recently published proteomic study (Sindelka et al., 2018). This discrepancy is likely caused by the difference in sampling methods to harvest cytoplasmic proteins. In our study, we used two orthogonal approaches to assess the subcellular localization of 20S CP. First, we demonstrated animal hemisphere enrichment by immunofluorescence. Second, we collected cytoplasm from animal and vegetal hemispheres of otherwise unperturbed fertilized eggs using a micropipette and cytoplasm was then analyzed directly by mass spectrometry. In contrast, Sindelka et al., extracted proteins from relatively large pieces of eggs using a bottom-up proteomic sample preparation approach coupling NP40 extraction with filtering (Sindelka et al., 2018). Although this method effectively reduces contamination from yolk proteins, other proteins, including proteasomal proteins, may also have been removed during sample preparation.

It is worth mentioning that the germline-specific RNAs dnd1 and pgat remain vegetally localized during the oocyte-to-embryo transition. The distribution of eIF4A1 and eIF4E, which encode proteins essential for translation, is not changed at all. Thus, not all maternal RNAs undergo vegetal-to-animal translocation during the oocyte-to-embryo transition. Nevertheless, it seems unlikely that this reorganization event is specific to proteasome RNAs. The vegetal-to-animal translocation of the proteasome is reminiscent of the massive rearrangement of yolk platelets and deep cytoplasmic components during the oocyte-to-embryo transition (Danilchik and Denegre, 1991; Imoh, 1995). In 1991, Danilchik and Denegre hypothesized that this dramatic intracellular rearrangement during the oocyte-to-embryo transition might represent an important mechanism to regulate the functions of some maternal factors (Danilchik and Denegre, 1991). In agreement with this view, we found that separation of the ubiquitin-independent proteasome from vegetally localized germ plasm is important for germline development. It will be of interest to study further how this intracellular rearrangement occurs and prepares the oocyte for embryonic development.

It is important to note that we do not understand the biological function of animally enriched proteasomes. During the oocyte-to-embryo transition, a series of important events happen in the animal hemisphere, including the completion of meiosis II and fusion of pronuclei. During cleavage, animal blastomeres divide slightly faster than vegetal blastomeres. It is possible that relocalization of the proteasome during the oocyte-to-embryo transition results in a higher protein turnover rate in the animal hemisphere, which facilitates some of these events. Interestingly, restriction of the proteasome to a specific cellular compartment occurs during mouse oocyte-to-embryo transition as well. Before oocyte maturation, proteasomes are distributed in the germinal vesicle and cytoplasm of mouse oocytes. After fertilization, however, proteasomes become exclusively nuclear during cleavage. Proteasomes can be detected in the cytoplasm only when the blastocyst is formed (Evsikov et al., 2004; Solter et al., 2004). We speculate that relocalization of the proteasomes during the oocyte-to-embryo transition may be an evolutionarily conserved phenomenon. Future studies are needed to understand fully its biological function.

In this work, we use X. laevis oocytes, eggs and embryos for our investigation. All Xenopus procedures were approved by the University of Illinois at Urbana-Champaign Institutional Animal Care and Use Committee (IACUC), under animal protocol #17199, and performed in accordance with the recommendations of the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. Embryos were obtained as described (Sive et al., 2000). When performing microinjection, embryos were placed in 0.5× Marc's Modified Ringer's (MMR) with 3% Ficoll. In all experiments, we injected 10 nl of solution into embryos using a Narishige IM300 microinjector. After injection, embryos were cultured in 0.2× MMR and harvested at the desired stages. Stage VI oocytes were collected from an ovary by manual defolliculation (Houston, 2018) and cultured in oocyte culture medium (Heasman et al., 1994). The antisense oligo against dnd1 has been previously published (Mei et al., 2013; Aguero et al., 2017a). To induce oocyte maturation, oocytes were treated with 2 nM progesterone overnight. Oocyte maturation was judged by the appearance of a white spot on the animal hemisphere of the oocyte. To induce artificial egg activation, eggs were pricked with a glass needle in the animal hemisphere. Eggs with obvious surface contraction were chosen for further analysis. We used Gastromaster (Nepagene, Japan) to dissect oocytes into animal and vegetal halves for RT-PCR analysis. All RNAs used for microinjection were synthesized from linearized plasmid templates using the mMESSAGE mMACHINE Kit (Ambion).

HEK293T and GC-2spd (Hofmann et al., 1994) cells were cultured in DMEM supplemented with 10% fetal bovine serum, penicillin and streptomycin. Cells were cultured at 37°C in a humidified, 5% CO₂ atmosphere. Transfection was performed using Lipofectamine 2000 as previously described (Jin et al., 2010).

GFP-myc-SV40 was described (Jin et al., 2009). To generate GFP-myc-x.l.dnd3′UTR, the 3′UTR of X. laevis dnd1 (Aguero et al., 2017a) was PCR amplified and cloned between the XhoI and NotI sites of GFP-myc-SV40. All other Dnd1 expression constructs used in this study were derived from X. tropicalis dnd1 (Mei et al., 2013) using standard PCR cloning methods. F110M and S117D were generated by site-directed mutagenesis from myc-Dnd1. Myc-RBM46 (IMAGE:7689323), myc-A1CF (IMAGE:3398290) and myc-hnRNP R (IMAGE:6316097) were generated by standard PCR cloning methods. pCS2-myc-Syncrip was cloned by RT-PCR using cDNA from stage 33 embryos. The ORFs of psme1, psme2 and psme3 were PCR amplified from X. laevis oocyte cDNA and cloned into pCS2 to generate pCS2-PSME1, pCS2-PSME2 and pCS2-PSME3. Dominant-negative proteasome activators dnPSME1 (PSME1 N141Y), dnPSME2 (PSME2 N139Y) and dnPSME3 (PSME3 N151Y) (Zhang et al., 1998) were generated by site-directed mutagenesis. pCMV-SPORT6-PSME4 (IMAGE:6637082) was used for overexpression of PSME4. To generate dnPSME4, the BRDL region of PSME4 (residues 1635-1723) was PCR amplified and cloned into pCS2. All cloning was carried out according to standard protocols. Briefly, plasmids or PCR products were digested with appropriate restriction enzymes (New England Biolabs). After digestion, DNAs were separated on agarose gel. We recovered the desired DNAs from the agarose gel slice using the GeneJETTM Gel Extraction Kit (K0692, Thermo Fisher Scientific). The ligation reaction was set up using a 2:1 insert/vector ratio. After overnight ligation, transformation was carried out using custom-made DH5 alpha competent cells. Colonies were screened and the correct constructs were sequenced.

Antibodies used in this study were: anti-myc (1:1000; 9E10, #5546, Sigma-Aldrich), anti-FLAG (1:1000; M2, #F1804, Sigma-Aldrich), anti-β-tubulin (1:2000; #T5293, Sigma-Aldrich), mouse anti-Hsc70 (1:10,000; Sc7298, Lot# C1716, Santa Cruz Biotechnology), rabbit anti-proteasome 20S core subunits (1:100; BML-PW8155-0100, Lot# 08081644, Enzo Life Sciences), rabbit anti-proteasome 20S α+β (1:100; ab22673, Lot# GR3218416, Abcam) and rabbit anti-Dnd1 (1:500), which has previously been reported (Mei et al., 2016).

Protocols for western blots were described previously (Jin et al., 2009). Briefly, oocytes or embryos were homogenized in NP40 lysis buffer (50 mM Tris pH 7.5, 125 mM NaCl, 1 mM EDTA, 1% NP40) supplemented with protease inhibitor cocktails (Sigma, P8340) (40 µl lysis buffer per oocyte or embryo). Lysates were cleared by centrifugation at 500 g for 5 min, then at 20,000 g for 5 min. Cleared lysates were mixed with 2× SDS sample buffer, boiled for 3 min, and loaded on SDS-PAGE. Proteins were transferred to PVDF membrane for western blot according to the standard protocol. To detect endogenous Dnd1 by western blot, we enriched Dnd1 protein from 25 oocytes or embryos by immunoprecipitation before western blotting.

To detect proteasome 20S core particles by immunofluorescence, oocytes, eggs and embryos (2-cell stage) were fixed with MEMFA [0.1 M MOPS, 2 mM EGTA, 1 mM MgSO₄, 3.7% formaldehyde (pH 7.4)], embedded in paraffin wax, and sectioned according to a standard protocol (Sive et al., 2000). For immunostaining, sections were blocked with blocking buffer (0.2% bovine serum albumin, 0.1% Triton X-100 in 1× PBS) with 10% goat serum for 1 h at room temperature, and then stained with anti-proteasome 20S core subunits antibody (1:100) overnight. Sections were washed with PBS + 0.1% Tween-20 three times and stained with goat anti-rabbit Alexa-594 fluorescent antibody (A11012, Thermo Fisher Scientific; 1:500) for an hour. Samples were washed again with PBS + 0.1% Tween-20 three times before mounting for imaging. We performed staining using two different antibodies (BML-PW8155-0100, Lot# 08081644, Enzo Life Sciences; ab22673, Lot# GR3218416, Abcam) and obtained essentially the same staining results.

For in situ hybridization, embryos and oocytes were fixed in MEMFA at room temperature for 1 h, washed with PBS twice, and dehydrated in methanol. To assess the expression of proteasome components, we cut stage VI oocytes, ovulated eggs and 2-cell-stage embryos into halves using a razor blade. Hemi-sectioned oocytes, eggs and 2-cell-stage embryos were placed in the same vial for in hybridization as described (Rorick et al., 2007). Plasmids used for labeling pgat (Hudson and Woodland, 1998) and dnd1 (Aguero et al., 2017a) in situ probes have been previously published. pCS2-PSME1, pCS2-PSME2, pCS2-PSME3, pCMV-SPORT6-PSME4, and EST clones TGas131c05, TEgg045c15, TGas116f12 and TTpA010e07 were used for making in situ probes for psme1, psme2, psme3, psme4, psmc6, psma2, eIF4a1 and eIF4e, respectively.

Real-time PCR was performed using the SYBR Green PCR Master Mix (A25742, Applied Biosystems) according to a standard protocol (Aguero et al., 2017a). PCR primers were: odc (Yang et al., 2003), psme1 (5′-AGCTGAACTGGACATACCAATTCC-3′; 5′-CTCTTTTGATTCCTGGATTTCACTC-3′), psme2 (5′-ATGTGGAGGACCTAGCAACC CTTC-3′; 5′-ATCCTTCTCTCAGGCTTCGTATCTC-3′), psme3 (5′-CTGATTGAGAAGTGCAACACGG-3′; 5′-TGCTCTAGTGATGTAGTATCTGG-3′) and psme4 (5′-GTGAGAAACAAAGCTCAGCAGGC-3′; 5′-CTGCACAATACACTCCCAGTCATGG-3′).

Micropipettes were fabricated in-house and utilized for localized sample (cytoplasm) collection from X. laevis embryos as per the protocol published in literature (Saha-Shah et al., 2015). Briefly, micropipettes were utilized to puncture the embryonic membrane and collect nanoliter volumes of cytoplasm via pressure actuation from the animal and vegetal hemispheres of individual embryos 1 h post-fertilization. The samples from animal and vegetal hemispheres were prepared separately for bottom-up mass spectrometry following standard protocols. The standard protocol includes treatment of the sample with dithiothreitol to reduce the proteins followed by alkylation of newly reduced thiols using iodoacetamide and finally digestion of the proteins to peptides using trypsin. The peptides were de-salted using c-18 stage tips prior to LC-MS/MS. An UltimateTM 3000 RSLCnano System nano liquid chromatography system was coupled to a Q-Exactive HF-X Hybrid Quadrupole-Orbitrap Mass spectrometer (Thermo Scientific) for proteomic data acquisition. Proteomic data analysis was performed using Spectronaut 10 (Biognosys) software. The mass spectrometry data was searched using Xenbase protein FASTA file (downloaded on September 2017). Proteomics data from five replicates or embryos were used in this study. Details of this method will be published separately (A.S., M. Esmaeili, S. Sidoli, H.H., J.Y., P.S.K. and B.G.).

Peptide abundances from LC-MS/MS data were log₂ transformed, normalized by the average of the peptide abundance within each sample and finally raised to the power of 2 to reverse the initial log₂ transformation. Protein abundance was obtained by averaging all detected peptides corresponding to a protein. Fig. 6D plots the average of protein abundance across all conditions and replicates in the y-axis against log₂ of protein abundance ratio between animal and vegetal hemispheres. The mass spectrometry proteomics data have been deposited in the ProteomeXchange Consortium via the PRIDE Archive under accession number PXD013446.

We thank Dr Romana Nowak for providing GC-2spd cells.