YBX2-dependent stabilization of oocyte mRNA through a reversible sponge-like cortical partition

Dear Editor,

Oocyte meiotic maturation and early embryogenesis occur in the absence of transcription and rely on maternal mRNA stored in oocytes.¹ In mice, the transcription is subsequently silenced from the late germinal vesicle (GV) stage until embryonic genome activation at the 2-cell stage.¹ Therefore, during early oocyte growth, the oocyte must accumulate essential mRNA transcripts to support subsequent oocyte maturation, meiosis, and early embryo development. The growth period of mouse oocytes is usually 22–24 days, and the mRNA synthesized and stored during this time is remarkably stable with a half-life of nearly 2 weeks.² In contrast, the half-life of most mRNA in mouse embryonic stem cells is less than 24 h.³ Thus, the exact mechanism for how the maternal mRNA in oocytes maintains their long-term stability remains unclear.

Here, we report that maternal mRNA incorporated with Y-Box Binding Protein 2 (YBX2) forms a reversible sponge-like cortical partition (SLCP) in mammalian oocytes, which might explain the extraordinary ability of oocytes to store and protect mRNA. YBX2 was the most abundant oocyte-specific RNA-binding protein⁴ which was highly correlated with poly(A) mRNA abundance (R = 0.95), suggesting that YBX2 plays a role in the stabilization of maternal mRNA (Supplementary information, Fig. S1a–d). Occasionally, we found that the supernatant lysis of the YBX2-eGFP expressing cells formed microparticles after storage at 4 °C for several hours. We re-centrifuged the liquid and found that the YBX2-eGFP protein all precipitated at the bottom of the tube. To understand why these microparticles were formed, we mixed and aliquoted equal volume of the lysis into Tube 1 and Tube 2. After adding RNase A to Tube 2, we found that the microparticles disappeared (Supplementary information, Fig. S2a). To make sure that these proteins were redissolved, we measured the protein concentrations of Tube 1 and Tube 2. The results showed that the protein concentration of Tube 2 was nearly double that of Tube 1 (Supplementary information, Fig. S2b). In addition, we added RNase A to Tube 1, and the microparticles were also redissolved (Supplementary information, Fig. S2a). The protein concentrations in Tube 1 and Tube 2 were not significantly different (Supplementary information, Fig. S2b). We then found an enriched disordered region in YBX2 protein sequence, which indicated that it might have phase separation properties (Supplementary information, Fig. S2c).

To determine whether YBX2 could form liquid-like puncta in oocytes, we performed immunofluorescence staining of YBX2, but no puncta signals were found. Instead, YBX2 was enriched in the cortical region and presented as sponge-like condensates (Fig. 1a). We then injected YBX2-eGFP cRNA into mouse GV oocytes and found similar condensates (Fig. 1a). Live oocyte imaging showed that adjacent condensates could assemble together after a long time (Fig. 1b; Supplementary information, Fig. S3). The fluorescence of YBX2-eGFP was efficiently recovered after a fluorescence photobleaching (FRAP) assay, which showed the liquid properties of these condensates (Fig. 1c, d). YBX2 was insoluble and retained in the cytoplasm after Triton X-100 treatment in an RNA-dependent manner (Supplementary information, Fig. S2d, e), which was consistent with the previous report,⁵ thus confirming the non-liquid properties of YBX2 condensates. The compound 1,6-hexanediol (1,6-HD) can be used as a chemical probe to distinguish liquid condensates from solid condensates.⁶ Microrchidia 3 (MORC3) is a human ATPase that forms phase-separated condensates with liquid-like properties in the cell nucleus.⁷ We found that the MORC3 presented typical liquid-like puncta in oocytes. Treatment with 5% 1,6-HD abolished almost all liquid–liquid phase separation (LLPS) of MORC3-GFP puncta in oocytes but only had a mild effect on YBX2 condensates (Fig. 1e–g; Supplementary information, Fig. S4a). The 1,6-HD treatment significantly weakened and dispersed the YBX2 signals in the cortical region of oocytes, but the YBX2 condensate signal did not disappear and a considerable part of the signal was retained in the oocytes (Fig. 1e–g; Supplementary information, Fig. S4a). In addition, SLCP of YBX2 also occurred in human oocytes, suggesting a conserved condensate structure between the species (Fig. 1h). These results suggest that in oocytes YBX2 generates an SLCP regulated by a liquid to gel-like phase transition.

Fig. 1. YBX2 formed a reversible SLCP in oocytes stabilizing oocyte mRNA.

a YBX2 was enriched in the cortical region and formed an SLCP. YBX2-eGFP also formed similar irregular structures. And YBX2 SLCP transformed into liquid-like puncta in oocytes after RNase A injection. YBX2: n = 50. YBX2-eGFP: n = 30. YBX2 + RNase A injection: n = 40. b Representative images of gel-like condensates assembly. n = 5. c Intensity analysis of the FRAP data for YBX2-eGFP condensates. d Representative images of FRAP experiments with YBX2-eGFP condensates. n = 3. e Intensity analysis of the YBX2 signals in the cortical region of the oocytes with (n = 20) or without (n = 28) 1,6-HD treatment. f Width of the YBX2 signals in the cortical region of the oocytes with (n = 20) or without (n = 28)1,6-HD treatment. g 1,6-HD treatment decreased the YBX2 signals in the cortical region of the oocytes, and a considerable part of the YBX2 condensate signal was retained in the oocytes. h YBX2 SLCP in human oocytes. n = 3. i The spatiotemporal changes of the gel-like condensates transforming into liquid-like puncta. n = 15. j In vitro RNA decay analysis of YBX2 (WT-eGFP, ΔNTD-eGFP, ΔCSD-eGFP, ΔCTD-eGFP, Mutant 1-eGFP, Mutant 2-eGFP, and Mutant 1+ Mutant 2-eGFP) after GLPS formation. n = 3. k Pattern diagram of the conversion process for YBX2 GLPS and YBX2 LLPS in GV oocytes.

To explore the dynamic of YBX2 condensates in oocytes, we performed immunostaining of the oocytes and embryos at postnatal day 12, GV, MI, MII, 2 pronuclear (2PN), and 2-cell stage. The results showed that the gel-like condensates had already formed in the growing oocyte at postnatal day 12, and a few liquid-like puncta were also detectable (Supplementary information, Fig. S5). During the GV, MI, and MII, the liquid-like puncta disappeared, and the gel-like condensates were enriched in the cortical region. The YBX2 condensates disassembled and decreased sharply at 2PN as the maternal mRNA largely decayed (Supplementary information, Fig. S5). At the 2-cell stage, almost no YBX2 signal was detected (Supplementary information, Fig. S5). Considering the observation that YBX2 can form condensates when binding with RNA in vitro, we asked whether the cortical condensates observed by YBX2 immunofluorescence were RNA-dependent. We injected RNase A into the GV oocytes to degrade maternal RNA and surprisingly found that these gel-like condensates dissociated and assembled into larger condensates, and then these condensates deformed into spherical puncta (Fig. 1a, i). These liquid-like puncta were also enriched in the cortical region (Fig. 1a). These data strongly suggested that the formation of gel-like condensates was RNA-dependent, especially because these condensates could reversibly transform into liquid-like puncta after RNA degradation. We found that the 1,6-HD treatment abolished all the liquid-like puncta induced by RNase A injection (Supplementary information, Fig. S4b). This suggested that the droplets were sensitive to 1,6-HD. The different responses of the YBX2 condensates and the liquid-like puncta to 1,6-HD imply a liquid to gel-like phase transition.

We then reconstituted the liquid to gel-like phase transition and gel-like phase separation (GLPS) in vitro. YBX2 underwent an LLPS on the glass slide (Supplementary information, Fig. S6a, b). To identify the role of mRNA in the formation of the LLPS, we added extra mRNA and found that the droplets aggregated and transformed into GLPS (Supplementary information, Fig. S6b, d, e). Meanwhile, YBX2 was colocalized with Cy5-mRNA in these gel-like structures (Supplementary information, Fig. S6c). Interestingly, this process was highly reversible, and after RNase A treatment the gel-like condensates transformed into LLPS again (Supplementary information, Fig. S6b). We also observed that the YBX2 LLPS was sensitive to salt ion concentration (Supplementary information, Fig. S7a, b). The GLPS decreased with a serial dilution in mRNA concentration and YBX2 concentration (Supplementary information, Fig. S8a, b).

To determine which domain of YBX2 functions in mediating LLPS or GLPS, we constructed variants missing the N-terminal domain (ΔNTD), the cold shock domain (ΔCSD), and the C-terminal domain (ΔCTD) (Supplementary information, Fig. S9a). In vitro LLPS assays showed the turbidity and the ratio of the pellet to supernatant were sharply decreased with the ΔCTD variant, indicating that it failed to undergo the LLPS and GLPS (Supplementary information, Fig. S9b–e). The ΔCSD also had decreased pellet-to-supernatant ratio but had increased turbidity, indicating the highly increased ability to undergo LLPS and decreased ability to undergo GLPS (Supplementary information, Fig. S9b–e). The overexpressed mutants in live oocytes showed that the ΔCTD presented a diffused signal throughout the whole oocyte, indicating that it failed to form condensates (Supplementary information, Fig. S9f). In addition, the ΔCSD-eGFP was shifted from the cytoplasm to the nucleus and underwent punctate signals after long-term expression (Supplementary information, Fig. S9f). The RNA pull-down assay using oligo(dT) showed that ΔCTD abolished the RNA-binding ability of YBX2, while ΔCSD had only a slight effect on its RNA-binding ability (Supplementary information, Fig. S10a, b). Previous work⁵ and our current work (Supplementary information, Figs. S9, S10) showed that the CTD of YBXs was required both for LLPS and RNA binding. However, using RNA binding-deficient mutants in CTD could not distinguish the effects of LLPS and RNA-binding ability. Thus, we aimed to design the mutations in CSD which decreased the GLPS and maintained the RNA-binding ability. The mutant 1 (R105A & F110A) and mutant 2 (D115A & D140A) in CSD were associated with the cytoplasm retention ability in an RNA-dependent manner for unknown reasons.⁵ In the current study, we proved that the cytoplasm retention ability of YBX2 might be due to the liquid to gel-like phase transition (Supplementary information, Figs. S2, S9, S10). We found that the mutant 1, mutant 2, and mutant 1 + mutant 2 variants all had delayed and decreased formation of the GLPS with little effect on the RNA-binding ability of YBX2 (Supplementary information, Fig. S10c, d and Table S1). Both mutants failed to form the sponge-like condensates and showed primarily nuclear localization with only a weak signal in the cytoplasm (Supplementary information, Fig. S11a). These data showed that the SLCP might be determined by mutant 1 and mutant 2 in CSD without affecting the RNA-binding ability of YBX2.

YBX2 is predicted to be a nuclear localization protein with a canonical nuclear localization signal,⁸ but how YBX2 maintains its cytoplasmic localization in the oocyte is unknown. One possibility is that YBX2 and mRNA undergo a stable GLPS to maintain their cytoplasmic partition. To prove this, we injected RNase A into the oocyte cytoplasm and observed that wild-type (WT) YBX2 was partly transferred to the nucleus at 30 min after RNase A treatment (Supplementary information, Fig. S11b). It has been indicated that the material properties and solid phase separation can be manipulated and abolished by tethering the FUS LC domain with the target protein.⁶ With this method, we showed that YBX2-eGFP was in the cytoplasm and no signal was in the nuclear region, while part of the YBX2-FUS-eGFP was translocated to the nuclear region (Supplementary information, Fig. S11c, d). These results indicated that the sponge-like cytoplasmic localization of YBX2 depended on its ability to undergo GLPS with mRNA.

To evaluate the effect of GLPS on the ability of YBX2 to protect mRNA, we performed an in vitro mRNA decay assay. The ability of YBX2 to protect mRNA significantly increased after forming the GLPS, while the loss of GLPS formation in varying degrees of the mutant proteins decreased this RNA-protecting ability (Fig. 1j). This indicated that the GLPS significantly enhanced the RNA stability of YBX2. To confirm the essential role of SLCP in oocyte maturation and female fertility, we generated the mutant 1 (R105A & F110A) knock-in mice. However, both the male and female F0 mice were sterile. The bilateral ovaries of the F0 female mice were significantly smaller than the WT controls (Supplementary information, Fig. S12), and no oocytes were found after dissection of the ovaries. In previous work, YBX2 knockout and transgenic RNAi-mediated knockdown both caused significant decay of maternal mRNA in developing oocytes leading to problems with oocyte maturation and early embryogenesis,^9,10 and thus we concluded that the GLPS increased the ability of YBX2 to protect mRNA, which might be essential for oocyte growth and maturation.

A previous study reported that YBX2 CSD preferentially bound to m⁵C-modified RNA oligos, and LLPS of YBX2 occurred in YBX2 over-expressing HeLa cells.¹¹ To explore whether the SLCP of YBX2 was affected by m⁵C RNA modification, we performed a colocalization analysis of YBX2 and NSUN2 (a critical RNA methyltransferase for adding m⁵C to mRNA). We showed that YBX2 was colocalized with NSUN2 (Supplementary information, Fig. S13a), and NSUN2 knockdown decreased the protein level of YBX2, but the SLCP still existed (Supplementary information, Fig. S13b–d). To further explain whether m⁵C RNA modification mediates the formation of YBX2 SLCP, we generated the m⁵C-binding site mutant (W101F) according to Wang et al.¹¹ and injected the mutant cRNA in oocytes to evaluate the effect of m⁵C RNA modification on YBX2 SLCP. The results showed that the mutant (W101F) had no effect on SLCP (Supplementary information, Fig. S13e).

In this study, we proposed an SLCP model in which YBX2 and mRNA form a hydrogel-like condensate in the subcortical region. This hydrogel-like condensate has sponge-like properties that accumulate and store the mRNA in a highly reversible manner (Fig. 1k). Previous studies also indicated that the phase transition of RNA-binding proteins and mRNA played an essential role in RNA storage in Drosophila oocytes.⁶ A similar sponge-like structure has been reported in Drosophila oocytes regulated by a liquid to solid phase transition.⁶ The SLCP has not been described previously in mouse and human oocytes. This SLCP enhanced RNA protection and maintained the cytoplasmic partition. Furthermore, YBX2, which is specifically expressed in oocytes, is the most abundant RNA-binding protein (approximately 2% of total oocyte proteins), and it degrades at the 2-cell stage, corresponding to the degradation cycle of maternal mRNA.⁴ In addition, YBX2 knockout and knockdown have a global decrease in mRNA in developing oocytes, leading to oocyte maturation and early embryogenesis problems.^9,10 Together these findings suggest that the unique YBX2 SLCP is key to explaining the extraordinary ability of oocytes to store and protect mRNA.

While this manuscript was in preparation, Cheng et al. reported that a mitochondria-associated membraneless compartment (MARDO) regulated by ZAR1 stores maternal mRNA in the oocytes of various mammalian species.¹² This study found that several RNA-binding proteins co-localized with mitochondria to form the MARDO and maintain RNA stability,¹² while our work demonstrated that YBX2 and maternal mRNA underwent a gel-liquid phase transition both in living oocytes and in vitro, which led to the accumulation and storage of mRNA in a highly reversible manner. The dynamic of YBX2 condensates was different from that of the ZAR1-related MARDO, which disassembled at the MI-MII transition (Supplementary information, Fig. S5). This indicated a different role between SLCP and MARDO in regulating maternal mRNA storage. To further clarify the relationship and difference between the YBX2 condensates described here and ZAR1 condensates as MARDO,¹² we performed a series of additional assays. The results indicated that YBX2 was colocalized with ZAR1 and mitochondria at the cortical region in GV oocytes. After RNase A injection, the ZAR1 signal was not incorporated into the YBX2 liquid-like puncta (Supplementary information, Fig. S14a). Moreover, the mitochondria were also not colocalized with YBX2 puncta (Supplementary information, Fig. S14b). In contrast, ZAR1 was highly colocalized with the accumulated mitochondria after RNase A treatment (Supplementary information, Fig. S14c). Furthermore, the dynamic of YBX2 condensates was different from the ZAR1-related MARDO (SLCP disassembled at 2PN vs the ZAR1-related MARDO disassembled at the MI-MII transition¹²) (Supplementary information, Fig. S5). Besides, the ZAR1 knockout mice had a major effect on the embryo development and had little effect on the oogenesis and oocyte maturation,¹² while the YBX2 knockout obvious affected the oogenesis and oocyte maturation.¹⁰ This also suggested a different role of YBX2 and ZAR1 in female infertility. Combing all this evidence, we supposed that SLCP might be a complementary or independent mechanism compared with MARDO during the assembly and dynamics of RNA and RNA-binding proteins.

In summary, we have demonstrated that maternal mRNA is incorporated with YBX2 in the form of a reversible SLCP, and this indicates the essential role of the SLCP in mRNA storage in mouse and human oocytes. These findings suggest a mechanism for maintaining the stability of maternal mRNA in mammalian oocytes.

Author contributions

Z.Z., Q.S., and L.W. conceived and designed the study. Z.Z., R.L., and H.Z. performed the experiments and bred the mice. J.F. contributed to the collection of human GV oocytes. R.L. performed the oocyte microinjection. Y.L. and H.Z. constructed the plasmids. T.W. provided assistance with the confocal imaging. B.C. and J.M. provided assistance with the experiments and data analysis. Z.Z., Q.S., and L.W. wrote the manuscript.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41422-023-00824-0.