A critical role for the COP9 signalosome subunit 3 as a gatekeeper of genome integrity in 2-cell mouse embryos

Investigations of genes required in early mammalian development are complicated by protein deposits of maternal products, which continue to operate after the gene locus has been disrupted. This leads to an underestimation of the number of genes known to be needed during the embryonic phase of cellular totipotency (up to the 4-cell stage). Here, we expose a critical role of the gene Cops3 by showing that it protects genome integrity during the 2-cell stage of mouse embryo development, in contrast to the previous functional assignment at postimplantation. This new role is mediated by a large, stable and hitherto overlooked deposit of maternal protein. Since protein abundance and stability defeat prospects of DNA- or RNA-based gene inactivation, we adopted a protein strategy of gene inactivation: antigen masking or TRIM21-mediated proteasomal degradation of COPS3. Both resulted in 2-cell embryo lethality, but the expected degradation remained outstanding, because the major fraction of the total COPS3 is secluded in a submembrane cortical rim that withstands extraction with detergent, thereby exposing a soluble vs. insoluble fraction of COPS3, which we ascribe with distinct functional properties. In mechanistic terms, transcriptomic and metabolic analyses reveal that soluble COPS3 is involved in several processes, which, however, converge on DNA endoreduplication and the accumulation of DNA strand breaks in the 2-cell nucleus, where the minor soluble fraction of COPS3 is placed. Thus, we have shown the critical role of maternal protein deposits in development, and we have also highlighted the distinction between the site of accumulation (cell cortex) Large scale data are available via Gene Expression Omnibus with the identifier GSE155205, and ProteomeXchange with the identifier PXD017212.


Introduction
Genes critical for the earliest stages of development have been known for a long time. Mouse embryos with mutations in these genes typically die during or shortly after the blastocyst stage, i.e. after the phase of cellular totipotency, which is over by the 4-cell stage as measured by the ability of isolated blastomeres to build viable blastocysts comprised of trophectoderm, primitive endoderm and epiblast 1,2 . The number of genes shown experimentally to be required prior to the blastocyst stage is relatively small. Exemplarily, of 712 candidate genes screened in a large-scale functional study using RNA interference, only 4 gene interferences resulted in cleavage stage embryo arrest and 20 interferences prevented blastocyst formation 3 . Of 347 genes whose gene ontology (GO) annotations matched the major DNA repair pathways according to the Mouse Genome Informatics Gene Ontology Project (MGI-GO) database, only 10 genes had a lethal knockout phenotype which manifests itself during the totipotent phase or by the blastocyst stage 4 . These small numbers are probably underestimated, due to an earlier accumulation of downstream product (transcript or protein), prior to experimental mutagenesis. This means that the first time point when a given gene is required for embryogenesis (the "first-time requirement") may have been overlooked in some cases, and the severity of developmental consequences is higher when this time point concerns the totipotent blastomeres.
We noted previously that the peri-zygotic transcriptional dynamics of the gene Cops3 (i.e. non-homogeneous mRNA distribution in the oocyte and difference of mRNA level between 2-cell blastomeres 5 ) is not consistent with a first-time requirement of this gene at postimplantation, as inferred from the timing of death of Cops3 knock-out embryos 6 . Furthermore, the observed in situ distribution of the Cops3 mRNA 5 is reminiscent of developmental mosaicism, which is a controversial issue in mammals 7 . Cops3 encodes a subunit of the Constitutive Photomorphogenic 9 (COP9) signalosome complex, whose subunits operate together (as a holocomplex) or as monomers 8 , as regulators of a wide range of biological processes that are ubiquitous in cells of plants and animals. Protein degradation is the process most studied, but COPS3 is also a transcriptional regulator, and a player in DNA repair and other functions [9][10][11][12][13][14] . In contrast to this variety of roles in the soma, the roles of COP9 or its subunits are still poorly characterized in the mammalian germline. Current knowledge points at germline stem cell survival 15,16 , meiosis 17 and epiblast survival in the postimplantation embryo 6 , while the preimplantation embryo seems unaffected. We were, therefore, surprised to find that when 2-cell mouse embryos were bisected and followed up in development, the Cops3 mRNA inter-blastomere difference correlated, three days later, with the number of epiblast cells found in monozygotic twin blastocysts derived from the sister blastomeres 5,18,19 .
In the present study, we bring to light a critical role for Cops3 during the initial and totipotent phase of mouse embryo development. When a gene product accumulates in oocytes, and even more so if the protein is abundant and stable, it becomes difficult to generate loss of function phenotypes using standard genetic knockout or RNA knockdown techniques. Therefore we harnessed two protein-based strategies to inactivate the COPS3 protein directly instead of its encoding transcript: 1) The antigen masking method, in which the protein of interest is prevented from performing its function 20 , and 2) the TRIM21-proteasomal method, in which the E3 ubiquitin-protein ligase TRIM21 (TRIpartite Motif containing-21, also known as Ro52) ubiquitinates and commits the protein of interest for degradation -hence the nickname 'Trimaway' 21 . Our COPS3 results reveal how the inactivation of a gene product on the protein level can expose a phenotype very different from that of genetic techniques and suggest that assigning genes to roles in embryogenesis may be less safe than assumed, if the protein products of these genes accumulate in oocytes. Moreover, we demonstrate that aspects of subcellular localization must be carefully considered in the design of a 'Trim-away' experiment, and in the interpretation of its results.

Results
Cops3 forms a large protein deposit, divided between a cortical and a nuclear pool, in mouse oocytes and embryos As we scrutinized the expression of Cops3 during preimplantation mouse embryo development we realized that the bulk of gene product was supplied by the oocyte to the embryo as a protein ( Figure 1A), as measured by liquid chromatography-mass spectrometry (LC-MS/MS) using the 'relative Intensity Based Absolute Quantification' (riBAQ) algorithm [22][23][24][25] . COPS3 abundance was high, in the range of the maternal-effect genes ( Figure 1A), and decreased from oocyte to blastocyst, in contrast to Cops3 mRNA ( Figure 1B). We identified two monoclonal antibodies, a rabbit (RabMab®) and a mouse one, both of which recognized a single band of the correct molecular weight in Western blots of cell lysates ( Figure 1C; the full-length blots are presented in Supplementary Figure 1). Immunofluorescence analysis revealed a cortical rim of COPS3 underneath the cell membrane, excluded from regions of cell-cell contact and from the polar body, during the first half of preimplantation development. Subsequently this cortical localization was replaced by a nuclear localization during the second half of preimplantation development ( Figure 1D). A cortical rim was also seen for another subunit of the same complex as COPS3, namely COPS5 (Supplementary Figure 2), thereby supporting the authenticity of the rim structure.
Clearly, in a case like this, trying to remove the COPS3 protein in oocytes by acting on its precursor mRNA may not result in the depletion expected. Therefore, there is scope for proposing that 1) the protein may support the biological functions associated with Cops3 even before the gene is transcriptionally upregulated, and that 2) investigations into the consequences of protein depletion should be conducted by targeting the protein COPS3 directly and not its precursor mRNA.  Cops3 mRNA and  COPS3 protein, each measured by two independent methods; data are presented as means and standard deviations  (for RNA, 100 oocytes or embryos per stage in two replicates; for protein, 200-600 oocytes or embryos per stage in  three proteomic replicates, and at least 3 oocytes or embryos per stage for immunofluorescence;  transcriptome/proteome datasets are extracted from 26 and are provided here in Supplementary Tables 1 and 2). (C). Specificity of the two monoclonal antibodies against COPS3 used in this study, RabMab® (ab79698) and mouse (DSHB-2D9), as documented by the single band in Western blot (lysate of MEFs or ES cells) and by the subcellular distribution within the same 2-cell embryos after immunofluorescence with both antibodies. The full-length blots are presented in Supplementary Figure 1. White arrow points at the non-stained polar body next to the positive rim. Note the lower background, i.e. the better signal-to-noise ratio, of the RabMab®. DNA was stained with YO-PRO-1. (D). Left: Developmental localization of COPS3 in oocyte and 2-cell stage, visualized by the RabMab® antibody, with white arrows pointing at the cortical rim and red arrow pointing at the signal exclusion from regions of cell-cell contact. Right: Developmental localization of COPS3 in preimplantation stages; note the prominence of the cortical rim up to the 4-cell stage, starting to fade thereafter, and being replaced by a nuclear signal distribution at the blastocyst stage. DNA was stained with YO-PRO-1. Abbreviations: Affymetrix stands for microarray; AU, arbitrary units; ES cells, embryonic stem cells; MEFs, mouse embryonic fibroblasts; MEG, maternal-effect genes; kDa, kiloDalton. riBAQ, relative intensity-based absolute quantification. Size bar: 50 μm.
Progression through second embryonic cell cycle is severely hampered by inhibition of maternal COPS3 We adopted a dual strategy of protein inactivation, in contrast to the conventional but indirect way of interfering with the mRNA precursor or the DNA template, to ascertain the need for Cops3 during the initial phase of mouse embryo development. We operated this inactivation either before or after the stimulus that triggers development. For practical reasons, we started out with oocytes which had already received the stimulus of fertilization (zygotes), since they are more tolerant of the large microinjection volumes used in our protocol 22 compared to pre-stimulus (MII) oocytes. We microinjected zygotes with defined amounts of either of the two anti-COPS3 monoclonal antibodies ( Figure 1C), in highly purified form (see Methods), to match the amount of COPS3 present in the cell, as known from our published proteomic catalogue 22 . Fluorescence of the co-microinjected inert tracer (Oregon-Green dextran beads: OGDB) confirmed that the microinjections had been successful in all cases. Most of the microinjected zygotes arrested at the 2-cell stage (Figure 2A), which was due to the antibody, since microinjection of wt zygotes with the sole antibody buffer supported blastocyst formation, as did an antibody against GFP protein, which is not present in wt cells ( Figure 2C). These results document that COPS3 is needed earlier in development than thought previously 6 . Since this earlier phenotype precedes the transcriptional upregulation of Cops3 during embryo cleavage ( Figure 1B), it also means that the earlier phenotype hinges on the maternal protein.
Albeit impeded in its functional interactions, COPS3 had only been bound but not degraded by the antibodies. Therefore, we also employed a degradation of the COPS3 antibody complex, via TRIM21-mediated proteasomal degradation. This was achieved by overexpression of TRIM21, imposed via microinjection of mCherry-TRIM21 mRNA, thereby triggering the proteasomal commitment of the ternary complex formed between target protein, the co-injected COPS3 antibody and TRIM21 21 . Throughout the text, Trim21 mRNA and TRIM21 protein are always intended as tagged with mCherry sequence and mCHERRY peptide, respectively. Zygotes  Figure 2E; the full-length blot is presented in Supplementary Figure 1). This observation was unexpected, because the 'Trim-away' reaction should be capable of acute and rapid protein degradation 21 , as we indeed confirmed with another protein known to be depleted efficiently by 'Trim-away', SNAP23 27 (Supplementary Figure 2). To tackle these contrasting results, we took advantage of a more stringent scheme of micromanipulation, as follows. We used 2-cell embryos preloaded with mCherry-Trim21 mRNA, but we injected the RabMab® antibody in only one blastomere, while the other blastomere served as control. By comparing the sister blastomeres we confirmed that the proteasomal reaction had worked, since adding the culture medium with the potent cell-permeable proteasome inhibitor MG132 prevented the decrease of mCHERRY fluorescence in the blastomere that received the antibody (Supplementary Figure 2). More importantly, from the side-by-side comparison we saw that the proteasomal degradation of COPS3 was slow, having gone a small extent after 8 hours, and needing more than 24 hours to abolish the mCHERRY fluorescence ( Figure 2F). The treated blastomere was arrested in development, while the control blastomere progressed and formed a mini-blastocyst ( Figure 2F), consistent with the effect of inhibiting COPS3 in whole embryos (Figure 2A-C).
Together, these findings strongly support the main conclusion of this study: COPS3 is required in mouse development much earlier than thought previously 6 , that is, already after one zygotic division rather than at postimplantation. The contrasting behaviors of the sister blastomeres show that the requirement of COPS3 is cell-autonomous. However, the observed kinetics of the 'Trim-away' reaction is not up to the expectations 21 , that is, neither rapid nor acute. . mCHERRY-COPS3 fluorescence intensity was reduced but not completely abolished by the TRIM21 reaction, as visualized in eight embryos, all of which were preloaded with mCherry-Trim21 mRNA at the 1-cell stage, but only the four embryos to the right also received COPS3 antibody 6 h after the mRNA (embryos in dotted box). Note that inert OGDB was co-injected, so as to identify a posteriori which embryos received only mRNA (OGDB low) and which embryos additionally received the antibody (OGDB high). (E). Western blot analysis (200 2-cell embryos) reveals that the total amount of COPS3 is not affected by TRIM21-mediated proteasomal degradation. The full-length blot is presented in Supplementary Figure 1. (F). Side-by-side comparison of the sister blastomeres, both preloaded with mCherry-Trim21 mRNA but only one receiving also the antibody (N = 20). The kinetics of the proteasomal degradation of COPS3 is slow, as visualized by mCHERRY fluorescence after 8 hours, which is completely abolished only after 72 hours. The blastomere fails to cleave further (dotted line). **, P < 0.01. Abbreviations: OGDB, Oregon-green dextran beads; n.s., not significant. Size bar: 50 μm.
COPS3's cortical deposit is largely insoluble and, thereby, unsuited for degradation by Trim-away unless the cortical rim is artificially destabilized To reconcile the bold effect of COPS3 inactivation on embryo development ( Figure 2D-F), we considered that the protein substrate was refractory to degradation. This consideration is supported by two lines of evidence. Firstly, COPS3 remained present despite chemical extraction. Detergent Triton X-100 is known to extract the majority of oocytic proteins, except for those bound in fibrillar sheets (cytoplasmic lattices) or stable cortical structures [28][29][30] and is harsher than Tween 20. Consistent with the different strengths of the two detergents, the cortical rim of COPS3 withstood better the chemical extraction with 1 % Tween 20 than with 0.1 % Triton X-100 ( Figure 3A). As a second line of evidence, the cortical rim of COPS3 withstood the block of protein turnover without any apparent change of intensity or thickness ( Figure 3B) when oocytes were treated simultaneously with the two inhibitors cycloheximide (against protein synthesis) and MG132 (against protein degradation) for 72 h -a period of time spanning most of preimplantation development in mice. Thus, it is not sensible to expect that TRIM21-mediated proteasomal degradation could succeed where a harsh chemical treatment and a pharmacological block of protein turnover failed to disrupt the cortical rim.
We reasoned that if the nature of the endogenous COPS3 is what hindered the 'Trim-away' reaction, then an exogenous supply of monomeric COPS3 (i.e. not embedded in the holocomplex) should change the outcome and produce two clear effects: 1) expedite the kinetics of mCHERRY fluorescence decline; and 2) allow for complete degradation of the additional amount of COPS3.
We overexpressed an engineered version of COPS3, imposed via microinjection of mCherry-Cops3 mRNA into zygotes, thereby gaining the ability to resolve between endogenous and exogenous COPS3, thanks to the mCHERRY epitope and the different molecular sizes of the native and engineered proteins. Immunofluorescent co-staining using RabMab® and anti-CHERRY antibodies reproduced the known picture of endogenous COPS3 ( Figure 1D), but in addition, it showed that the exogenous COPS3 was mainly located in the nucleus, while only a weak signal visible in the cortical rim ( Figure 3C). The latter is a remarkable observation, given the Triton X-100 stability of the rim. As the zygotes preloaded with mCherry-Cops3 mRNA were injected with RabMab® antibody, the mCHERRY fluorescence was abolished ( Figure 3D Figure 1). Thus, we conclude that exogenous COPS3 destabilized the cortical rim, and paved the way for augmented access of the 'Trim-away' reaction to the endogenous COPS3. Further, these results suggest that the cortical rim served as a slow-release source of COPS3 for functions in the nucleus (next section). Figure 3. The cortical rim of COPS3 is refractory to spontaneous degradation and chemical extraction. (A). COPS3 is completely refractory to chemical extraction using Tween 20, and residues of COPS3 are seen also after extraction with Triton X-100. Compared to untreated oocytes, the cortical rim of COPS3 is still present, albeit thinned, in live oocytes extracted for 10 min with 1 % Tween 20 or 0.1 % Triton X-100 (N = 11, 11, 13, respectively). (B). COPS3 is refractory to spontaneous degradation, such as that taking place during oocyte aging in vitro in the presence of inhibitors of protein synthesis (CHX) and protein degradation (MG132) (MII fresh, n = 48; MII aged 72 h, n = 10; MII treated with CHX and MG132, n = 10). Representative images are shown, either raw or pseudocolored in Fiji using the LUT 'Union jack.' Data in the box plots are presented as median, 1 st and 3 rd quartiles, and range (whiskers). P values are calculated by the Wilcoxon test, setting the MII oocyte as the reference (painted red in the plot). ***, P < 0.001. (C). Immunofluorescent co-staining of endogenous and exogenous COPS3 after overexpression of mCherry-Cops3 mRNA. Images show that the exogenous COPS3 is mostly taken up in the nucleus (white arrow) but it also gains access to the cortical rim (red arrow). DNA was stained with YO-PRO-1. (D). After overexpression of COPS3, the 'Trim-away' reaction attains effectiveness as revealed by loss of mCHERRY fluorescence in live-cell imaging (dotted box) and depletion of mCHERRY-COPS3 in Western blot. The full-length blot is presented in Supplementary Figure 1 Table 2). Additionally, and as control, we profiled the influence of Trim21 mRNA over-expression on the transcriptome. We set the non-manipulated embryos as the reference, as these are the most distinct in phenotypic and molecular terms from all other experimental groups (out-group comparison). It is reassuring that the 2-cell stage transcriptome was minimally impacted by the microinjection and the overexpression of Trim21 mRNA, compared to the non-manipulated group (22 mRNAs, or 0.08 %, were differently expressed; fold change ≥ 2, p ≤ 0.01, t-test; Figure 4A).
When TRIM21 was combined with the COPS3 RabMab® antibody, it induced a vast perturbation of the transcriptome, even though the COPS3 depletion was partial: We found an altered (≥ 2-fold, p ≤ 0.01) expression of 1280 mRNAs, of which 1185 mRNAs were unique for the depletion, 91 mRNAs were shared with the antibody-alone group, and 4 were shared with the Trim21 mRNA only group ( Figure 4B). Compared to the 1280 mRNAs of the RabMab® antibody with TRIM21, the antibody alone perturbed 236 mRNAs, of which 145 mRNAs were unique for this group. It, thus, appears that the 2-cell phenotype is due, to a minor extent, to antigen-based masking (antibody alone) and, to a major extent, to antigen removal (TRIM21-mediated proteasomal degradation). In total, aggregating the results and discarding the duplicates, 1421 mRNAs (n = 145 + 91 + 1185) were perturbed by anti-COPS3 with or without TRIM21. These significantly perturbed mRNA did not include Cops3 itself, thereby indicating that acting on COPS3 did not induce a compensatory transcriptional response at this locus.
We subjected the set of 1421 mRNAs to GO analysis, using the software ReViGO to aggregate and visualize the GO enrichment results 31 . ReViGO returned a large number of hits related to a large number of biological processes and sub-compartments of the cell, for example ranging from 'nuclear part' (GO:0044428) to 'cytosolic part' (GO:0044445), and from 'DNA repair' (GO:0006281) to 'translation' (GO:0006412). The most prominent hits were those related to RNA metabolism ( Figure 4B). Having found this process is consistent with the window of time of this study, when degradation of (certain) maternal mRNAs is required shortly after fertilization in order to enable de novo transcription during the embryonic genome activation (EGA) 32 at the 2-cell stage. In a nutshell, this is the window of time of the maternal-to-embryonic transition. We, therefore, compared and contrasted the 1421 mRNAs with a list of mRNAs that are known to be degraded (oocyte-specific, n = 149) or upregulated (major wave of EGA, n = 3480) in 2-cell mouse embryos (Datasets S1 and S6 in 33  Therefore, we examined the capacity of COPS3-depleted 2-cell embryos for total RNA, DNA and protein synthesis, as determined by Click-IT chemistry after the incorporation of the respective precursors in cultured embryos. In contrast to unchanged protein and RNA synthesis ( Figure 5A, B), DNA synthesis was significantly higher (2X) in COPS3-depleted 2-cell embryos relative to control embryos ( Figure 5C), in conjunction with a significant increase in the number of DNA double-strand breaks inferred from the number of histone γH2A.X nuclear foci 14 ( Figure 5D). Given the factor 2X, we hypothesized that the increased DNA content was due to DNA endoreduplication, and that the increased DNA damage might be a manifestation of replication stress, possibly in conjunction with reduced levels of genes that superintend to DNA repair. We went back to our transcriptome data to look for relevant changes in related gene activities to corroborate or confute this hypothesis. In line with the latter, the levels of Geminin mRNA -the factor required to prevent endoreduplication -are significantly reduced in the COPS3-depleted 2-cell embryos relative to control embryos (-88 %, p = 0.02; Figure 5E); whereas the levels of CyclinE2 mRNA -the factor required to permit endoreduplication 36 Figure 5E). This observation can be extended to the transcripts of two additional, well-known DNA repair activities (i.e. Rad21, -60 %, p = 0.02; Brcc3, -43 %, p = 0.04; ttest; Figure 5E). Some of the mRNAs above were also confirmed as proteins. Specifically, the proteins of Geminin and Brcc3 were detected in unmanipulated and in Trim21 mRNA-injected embryos but not in COPS3-depleted embryos (see LC-MS/MS data in Supplementary Table 1), consistent with the significant reduction of the two mRNAs (-88 %, -43 %). We could not verify the stable levels of the mRNAs of 'housekeeping' genes 34 in COPS3-depleted 2-cell embryos, consistent with the report that these genes are not constantly expressed but, in fact, regulated 33 .
Together, these data provide the molecular rationale for the phenotypic consequences of immunological COPS3 inactivation. This inactivation elicits a wide range of changes in the biological processes, including housekeeping, many of which would suffice on their own to affect the viability of 2-cell embryos. In particular, we observe that the nuclear processes of DNA replication and repair of DNA are affected (endoreduplication + histone γH2A.X nuclear foci) and this takes place in parallel with the observation of nuclear relocalization of the cortical deposit from the 2-cell stage onward.  protein synthesis to the same extent as non-manipulated controls, while the protein synthesis inhibitor cycloheximide reduced the incorporation (N = 12, 14, 11, respectively). (B). COPS3-inactivated zygotes incorporated the EU precursor for RNA synthesis to the same extent as non-manipulated controls, while the RNA polymerase II inhibitor α-amanitin reduced the incorporation (N = 5, 6, 4, respectively). (C). COPS3-inactivated zygotes incorporated the EdU precursor for DNA synthesis to a higher extent than non-manipulated controls, while the DNA polymerase inhibitor aphidicolin reduced the incorporation (N = 5, 7, 11, respectively). (D). COPS3-inactivated zygotes presented more histone γH2A.X foci per nucleus than non-manipulated controls, while the topoisomerase inhibitor etoposide increased the number of foci even more (N = 11, 11, 8, respectively). (E). Transcript levels of key genes involved in the prevention of DNA endoreduplication and actuation of DNA double-strand break repair as well as housekeeping genes (from transcriptome data, dataset GSE155205; Supplementary Table 2). Data are presented as means and standard deviations. P values are calculated by t-test. *, P < 0.05. **, P < 0.01. ***, P < 0.001. Abbreviations: wt, wildtype; OPP, O-propargyl-puromycin; EU, 5-ethynyl uridine; EdU, 5-ethynyl-2'-deoxyuridine; ns, not significant. Size bar: 50 μm.

Discussion
So far, COPS3 has been understood mainly as a factor necessary for epiblast survival at the time of mouse embryo implantation 6 -a role shared with other mutated members of the COP9 signalosome complex i.e. COPS2, COPS5, COPS6 and COPS8 [37][38][39][40] . The results of our immunological COPS3 inactivation reveal that prior to supporting epiblast survival in the postimplantation embryo, COPS3 accumulates in the oocyte and is required for protecting DNA integrity during the 2-cell stage. Cops3 is thereby recruited to the group of only 10 genes (reviewed in 4 ) which have a lethal mutant phenotype caused by defective DNA repair during the very early phase of development that harbors cellular totipotency. The reassigned function of COPS3 is mediated by an abundant, stable and hitherto overlooked protein deposit, which is present in oocytes in the form of a cortical rim -a cytological theme found also in maternal genes such as those of the subcortical maternal complex (SCMC) 41,42 . The rim is mostly insoluble and, thereby, unsuited for degradation by the 'Trim-away' method in its conventional form, but it coexists with a less abundant fraction in the nucleus, which is the actual point of use of COPS3 in early mouse embryos.
Embryonic arrest at the 2-cell stage after co-injection of anti-COPS3 antibody with or without Trim21 mRNA is a phenotype which is reminiscent of so-called maternal-effect genes, such as Mater (Nlrp5), Zar1, Hsf1 and Brg1 (Smarca4) [43][44][45][46] . However, Cops3 is known to function in germinal-vesicle mouse oocytes to support meiosis 17 , whereas mutations of maternal-effect genes should affect the embryogenesis without affecting the gametogenesis (neither oogenesis nor spermatogenesis) 47 . Looking for the main cause of the 2-cell arrest following immunological COPS3-inactivation proved challenging because it turned out that COPS3 is involved in numerous molecular functions of many different cellular processes. Recalling the case of Brg1 was helpful to look for a cause in the case of Cops3. When Brg1 was deleted in mouse oocytes 46 , the arrested 2cell embryos were affected in about 30 % of the α-amanitin-sensitive genes, which are bona fide markers of EGA. For this reason, we interrogated the transcriptome of COPS3-inactivated embryos. It is reassuring that the number of mRNAs which were perturbed by the microinjection of the sole Trim21 mRNA was very small, given the known interaction partners of TRIM21, for example, the gap-junction subunit CONNEXIN 43 48 , the tumor suppressor p53 49 or the histone deacetylase HDAC6 50 . By contrast, the transcriptome was vastly perturbed -including 'housekeeping' mRNAs -after COPS3 inactivation, and such vastness requires an explanation. The gene Cops3 has paralogs whose proteins take part in the proteasome lid and the translation initiation complex, for example, Psmd3 51,52 . These complexes determine the fate of hundreds of proteins and, therefore, malfunctioning of each can cause the failure of many cellular processes, including basic biosynthetic ones -among these, DNA metabolism.
Among the many things that can go wrong predictably because of COPS3-inactivation, given the gene ontologies associated with this gene, the embryos had a doubled DNA content with an increased number of histone γH2A.X foci, while RNA and protein synthesis remained unchanged. The double DNA content per se may not be blamed for the 2-cell arrest of COPS3inactivated embryos, since tetraploid embryos form blastocysts 53 , whereas genotoxic stress (DNA damage) is well-known to cause 2-cell arrest 54 . Similar effects on DNA replication and repair in 2cell mouse embryos have previously been reported after genetic ablation of DDB1-and CUL4associated factor 2 (Dcaf2) and Cullin 1 (Cul1) [55][56][57][58] , however, the operating mode of COPS3 may or may not be the same. Whereas the inactivation of CUL1 resulted in an accumulation of pan-CYCLIN E 55 , the inactivation of COPS3 resulted in an accumulation of Cyclin E2 mRNA but not of Cyclin E1 mRNA; and whereas COPS3 features a cortical deposit in oocytes, this structure could not be demonstrated in Dcaf2-deficient oocytes (due to the lack of a suitable antibody) and the microinjection of mCherry-tagged Dcaf2 mRNA in oocytes did not result in the assembly of a cortical rim 56 . Refining the mechanistic picture of the role of COPS3 in DNA metabolism, our study also indicated possible mediators of the nuclear DNA phenotype: Downregulation of genes such as Geminin and Brcc3, which superintend to DNA endoreduplication and DNA repair, respectively.
Together, these novel findings abide by the notion that maintenance of genome integrity is a precondition for totipotency 59 , documented in numerous cases, for example, the perturbation of totipotency marker genes (e.g. Zscan4d) leading to genomic instability and karyotype abnormalities 60 .
In spite of the 2-cell lethality of COPS3 inactivation, this phenotype originated from a partial depletion: Only a minute fraction of the total COPS3 was depleted, while most of it was still there. As we are going to discuss, this was not an experimental, but a biological issue, linked to the stability and accessibility of the cortical rim of COPS3. However, this is a potential limitation of our study that prompts at least two questions: 1) Why was the depletion partial and 2) how could a partial depletion be so detrimental? To answer the first question, we recall that oocytes' structures persisting during preimplantation embryos and even after extraction with detergents have been known for a long time, such as cytoplasmic lattices and SCMC 29,30,61 , which are subcortically located in the ooplasm similar to COPS3. The cortical rim of COPS3 withstood the extraction with Tween 20 and Triton X-100. This gave us the idea to introduce an additional amount of COPS3 into the cell, via microinjection of Cops3 mRNA (overexpression), in order to create a free pool which should be more accessible for the 'Trim-away' reaction. Not only was the exogenous amount of COPS3 depleted efficiently, but this happened also to the endogenous COPS3. It thus appears that the exogenous COPS3 had destabilized the endogenous COPS3, allowing TRIM21 and antibody to gain access to both. In addition, we appreciated a distinct localization of exogenous COPS3 also in the nucleus. This observation leads us to envisage an apportionment of total COPS3 in two pools: One pool is sequestered (insoluble) in the cortical rim and is, thus, not accessible for enzymatic proteolysis, the other pool is soluble in the cytosol, from where it can also access the nucleus and operate in it. To answer the second question, namely how a partial depletion could be so detrimental, we propose that the rim is not the point of use of COPS3 in the early embryo, and that COPS3 in the rim cannot be easily transported to the 2-cell nucleus -where its function is first needed, as revealed by our DNA assays. The nuclear amount is very small and this would be why our Western blot analysis of the total native amount (no overexpression) was unable to detect change after TRIM21-mediated proteasomal degradation of COPS3. This consideration elicits a speculation that the phenotype is not controlled by the total resource of COPS3 present but by the part of it needed at the point of use (limiting factor)reminiscent of the 'law of the minimum' formulated by Justus von Liebig in the 19 th century 62 .
From drawing this parallel, we can better understand the idea that an authentic phenotype can be exposed by acutely reducing -but, in fact, not completely removing -gene activity 63 , since only a fraction of the total is biologically relevant and the biological system had no time to mount adaptive responses (not to mention the phenotypes that may otherwise remain hidden due to loss-of-function embryonic lethality 64 ). Together, these two answers predict that a complete depletion by TRIM21-mediated proteasomal degradation may not work, and may not be required to work (depending on one's aim), for every protein. Indeed, it may be noted that out of thirteen studies which applied the TRIM21-mediated proteasomal degradation on oocytic or embryonic proteins in mammals 21,22,27,[65][66][67][68][69][70][71][72][73][74] , four studies reported complete depletion (ITPR1 67 ; SNAP23 27 ; G6PD, PKM, GFPT1 65 ; EG5 also known as KIF11 21 ), four studies reported almost complete depletion (BTG4 68 ; TACC3 69 ; RACGAP1 also known as CYK4 and PLK1 70 ; SMC3 72 ) and five studies reported incomplete depletion (RCC1 66 ; CENPF 71 ; TEAD4 22 ; aPKC also known as PRKC 73 ; HUWE1 74 ). Thus, complete removal of the protein of interest seems to be more the exception than the rule (with current technology).
In conclusion, totipotency is generally regarded as a special property which hinges on extraordinary and elusive 'master' genes. We have recruited COPS3 -hitherto known for serving rather ordinary, basic, not to say 'housekeeping' duties -to the family of maternal (but not maternal-effect) factors that are necessary to protect genome integrity during the totipotent phase, i.e. the 2-cell stage of mouse embryo development. The case of Cops3 is probably not limited to this gene. Similar to Cops3, also Sox2 and Max had null (-/-) blastocysts with the same immunostaining intensity as +/-or +/+ counterparts after intercross of heterozygotic parents 6,75,76 , suggestive of relevant oocytic deposits persisting until the stage of embryo implantation.
Similarly, the proteins of the SCMC genes were found to be still present at the blastocyst stage, as revealed by Western blotting, although the alleles had been inactivated in the zygote and the transcripts degraded at the 2-cell stage 42 . More recently, immunofluorescence revealed that the pyruvate dehydrogenase complex translocated from the cytoplasm to the nucleus of the 2-cell stage mouse embryos, and this translocation was necessary for EGA 77 . The common trait in these various cases was that new insight came from direct inspection of the protein products of those genes, thereby, questioning the safety of transcriptomic screening to identify the molecular signature of totipotency, since this signature could be partially made of proteins, which escape transcriptomic screening. This advises us that gene role assignments in embryogenesis may be less safe than assumed when the proteins of those genes are accumulated in oocytes, and suggests that further identification of key maternal proteins and their functions via single-cell proteomics and 'Trim-away', respectively, could greatly facilitate studies of totipotency. However, we need to be aware of the prerequisites posed by this still young method, beyond the availability of excellent antibodies: A quantitative knowledge of the oocyte proteome and of its subcellular distribution. Proteome and transcriptome analysis of oocytes, zygotes and preimplantation embryos. To assess the developmental profile of COPS3 we made use of already existing datasets that we had published previously 22,26 . To assess the immediate impact of TRIM21-mediated COPS3 depletion on the protein composition of the 2-cell stage we generated a new proteomic dataset using the same pipeline as described in 22  Block of oocyte's protein turnover or chemical extraction with detergents. MII oocytes were subjected to either of these two treatments, as follows. To block protein turnover, oocytes were cultured for 3 days in α-MEM supplemented with inhibitor of protein synthesis (cycloheximide 50 µg/mL 82 ) and inhibitor of proteasomal activity (MG132, peptidaldehyde Z-Leu-Leu-Leu-CHO, 10 µM 83 ). The proteasome inhibitor was also applied alone on 2-cell embryos to demonstrate that the TRIM21 reaction depends on proteasome function. For chemical extraction, MII oocytes were treated with buffer containing 0.1M KCl, 20 mM MgCl2, 3 mM EGTA, 20 mM HEPES (pH 6.8), 1 x Complete Protease Inhibitor Cocktail (Roche) and 0.1% Triton X-100 or 1% Tween 20 for 10 min 61 . After each treatment, embryos were fixed with 3.7% formaldehyde for 15 min, and processed for immunofluorescence and imaging as described (see "Immunofluorescence analysis of COPS3, COPS5, mCHERRY and γH2A.X expression"). The native images were normalized using identical settings for contrast and brightness for reasons of comparability before being used for analysis.

Methods
Images were pseudocolored for demonstration (not for analysis!) purposes [84][85][86]   were washed four times and concentrated at 4 °C using Amicon Ultra-0.5 100 KDa centrifugal filter devices (Merck Millipore, cat. no. UFC100) to remove salts and preservatives (e.g. sodium azide) and stabilizers (e.g. albumin), and to replace the buffer with water. In case of blastomere injection, zygotes were preloaded with a mixture of mRNA and dextran beads without antibody, followed by the antibody on the next day at approximately 39 h post-hCG. Microinjection was conducted on the stage of a Nikon TE2000U microscope fitted with a piezo drill (PrimeTech), using a bluntend glass needle (inner diameter 6-7 microns, outer diameter 8-9 microns) filled with 2-3 microliters mercury at the tip. Volumes were pressure-injected into the zygote or blastomere using a Gilmont GS-1200 micrometer syringe operated manually. During the microinjection, cells were kept in a 200-300 microliters drop of Hepes-buffered CZB medium 87 on a glass-bottomed (Nomarski optics) dish at a room temperature of 28 °C. After microinjection, zygotes or embryos were allowed to recover in the drop for 5-10 min, before returning them to KSOM(aa) medium.
Immunoblotting analysis of COPS3, mCHERRY and SNAP23. Oocytes or embryos were centrifuged in protein-free HCZB medium at 700 rpm for 10 min to form a tiny pellet. The supernatant was carefully aspirated using a mouth-operated micropipette, and replaced by RIPA buffer containing  For additional datasets cited in this study (PXD012613, DNA Databank of Japan Sequence Read Archive DRA005956 and DRA006335) we refer to the original publication 26 .