Embryonic Cells Redistribute SUMO1 upon Forced SUMO1 Overexpression

Embryonic stem (ES) cells exhibit unusual transcriptional, proteomic, and signal response profiles, reflecting their unusual needs for rapid differentiation and replication. The work reported here demonstrated that mouse embryonic cell lines did not tolerate the overexpression of SUMO1, the small ubiquitin-like modifier protein that is covalently attached to many substrates to alter their intracellular localization and functionality. Forced SUMO1 overexpression is toxic to ES cells, and surviving cell populations adapt by dramatically reducing the levels of free SUMO1. Such a response is not seen in differentiated cells or with SUMO2 or with nonconjugatable SUMO1 mutants or in the presence of a SUMO1 “sponge” substrate that accepts the modification. The findings suggest that excess SUMO1 modification of specific substrates is not tolerated by embryonic cells and highlight a distinctive need for these cells to control the levels of SUMO1 available for conjugation.

overexpression of SUMO1 protein capable of conjugation to substrates. Surviving cells have redistributed their SUMO1 and no longer maintain free SUMO1. In contrast, SUMO2 was readily overexpressed in both embryonic and differentiated cells. Reducing SUMO1 conjugation by eliminating the diglycine residues necessary for conjugation or by coexpression of a "SUMO sponge" or by coexpression of the desumoylase SENP1 greatly improved overexpression of free SUMO1. The results suggest that embryonic cells do not tolerate the excessive formation of the critical SUMO1-conjugated substrate(s).

RESULTS
SUMO1 cannot be overexpressed to accumulate as free SUMO1 in embryonic cells. Many studies have suggested that SUMOylation has a uniquely significant role in embryonic development (17,18) and thus might be subject to distinctive regulation in developmentally primitive cell types. To examine the consequences of increased SUMOylation in embryonic cells, we designed DNA constructs that would drive highlevel expression of SUMO1. Because embryonic cells are difficult to transfect and can silence a variety of promoters, we delivered the constructs on lentiviral vector genomes in which the EF1␣ promoter, active in embryonic cells, drove expression of Flag-tagged SUMO1 and a drug resistance protein (PuroR) designed to be translated from a single bicistronic transcript. The SUMO1 gene was positioned at the 5= end of the transcript so as to be translated by cap-dependent ribosome initiation events, while the 3= proximal puromycin resistance gene was translated separately by ribosomes initiating at an internal ribosome entry site (IRES). Constructs were generated encoding Flagtagged versions of either a wild-type (WT) full-length SUMO1 precursor, requiring processing for conjugation (Flag-SUMO1), or a truncated version lacking the six C-terminal residues, including the GG residues needed for ligation (Flag-SUMO1ΔGG). 293T cells were transfected with these vector DNAs, along with pCMVΔR8.2 DNA encoding the HIV-1 Gag and Gag-Pol proteins and pVSV-G DNA expressing the vesicular stomatitis virus G (VSV-G) envelope protein, and viral particles in the culture supernatants were collected. The virus preparations were applied to NIH 3T3 cells or F9 embryonic carcinoma cells, and transduced cells were selected with puromycin. Lysates of the pooled transduced cell cultures were prepared using harsh buffer conditions, and the levels of expression of SUMO1 were then assessed by Western blotting probed with anti-Flag antibodies. NIH 3T3 cells transduced with the wild-type SUMO1 vector accumulated both a spectrum of high-molecular-weight SUMO1 conjugates and free monomeric SUMO1 (Fig. 1A). In contrast, F9 cells transduced with the wild-type SUMO1 expressed no detectable free SUMO1 but retained all the SUMO1 in form of a few high-molecular-weight species (Fig. 1A). Many of the bands seen in NIH 3T3 cells were absent in the F9 cells. Both cell lines transduced with the SUMO1ΔGG construct, however, expressed high levels of the free monomeric mutant SUMO1.
The complete absence of free wild-type SUMO1 accumulation was seen with other embryonic cells. Transduction of the PCC4 embryonic carcinoma line or the E14 embryonic stem cell line gave results similar to those seen in F9 cells: no free wild-type SUMO1 but high levels of free SUMO1ΔGG (Fig. 1B). We note that the E14 line expressed the transduced SUMO1ΔGG at lower levels than the other cell lines, though again at much higher levels than the wild-type SUMO1. Delivering the same constructs into differentiated cell lines, both the mouse NIH 3T3 fibroblast line and human 293T cells resulted in high and comparable levels of expression of the free monomeric forms of both SUMO1 and SUMO1ΔGG (Fig. 1C).
One possible explanation for the low expression of free wild-type SUMO1 in embryonic cells is that high expression is toxic and that only very few transduced cells with aberrant SUMO1 processing were surviving the drug selection. To test this, the efficiency of recovery of F9 cells after transduction was examined. Equal concentrations of virus preparations expressing wild-type SUMO1 and SUMO1ΔGG were applied to NIH 3T3 or F9 cells, the cells were plated in medium with puromycin, and the numbers of drug-resistant colonies were determined. NIH 3T3 cells yielded comparable numbers of colonies after transduction by wild-type SUMO1 or SUMO1ΔGG viruses. F9 cells, in contrast, yielded approximately 8-fold-fewer colonies after infection with the wild-type SUMO1 virus than after infection with the mutant SUMO1ΔGG virus (Fig. 1D). The resulting transduced cells were pooled and passaged for long-term culture, and the morphologies and rates of growth of the resulting drug-resistant F9 cell populations were not distinguishable. The results suggest that embryonic cells were distinctly sensitive to overexpression of SUMO1, with few clones surviving to form colonies. The few surviving clones had blocked the accumulation of unconjugated SUMO1 protein while retaining all the SUMO1 in a few high-molecular-weight conjugates. In contrast, SUMO1ΔGG was expressed in F9 cells without the equivalent toxic effects on cell survival and was fully retained in free unconjugated form.
Accumulation of free SUMO1 in embryonic cells is prevented at the posttranscriptional level. To probe the basis for the restriction in accumulation of free SUMO1, we examined the levels of SUMO1 DNA and RNA in transduced cell populations. F9 or 293T cells were transduced with the SUMO1 or SUMO1ΔGG vectors, or empty vector control, and were selected with puromycin for stable expression of the drug selection marker. The levels of SUMO1 and SUMO1ΔGG transgene DNAs were assessed by quantitative PCR (qPCR) using primers spanning exon-exon junctions, such that DNA of the endogenous SUMO1 gene would not be amplified ( Fig. 2A). Similar levels of the SUMO1 and SUMO1GG transgenes per cell were found in the F9 and 293T pooled drug-resistant cells. SUMO1 transgene DNA was not detected in control F9 cells transduced by the empty vector. We also measured the levels of the drug resistance marker DNA (puromycin resistance gene), and as expected, levels of the puromycin resistance gene were similar across all transduced cells. The levels of the SUMO1 and  puroR DNAs normalized for the differential efficiencies of amplification seen with qPCR of the original SUMO1 vector DNA and assessed by qPCR were comparable. Thus, the SUMO1 expression construct was correctly delivered to the embryonic cells and was retained in the surviving drug-resistant clones.  To test for the possibility that mutations in the SUMO1 vector were being selected during transduction, we examined the amplified PCR products from the pools of transduced cells. Genomic DNA was isolated from F9 cells transduced with the SUMO1 or SUMO1ΔGG vectors, and the SUMO1 insertion was amplified by PCR using primers that spanned the EF1a promoter and QQ region. PCR products from F9 cells transduced with SUMO1, SUMO1ΔGG, or empty vector were identical in size to the PCR products amplified from the corresponding plasmid controls (Fig. 2B). The bulk PCR products amplified from F9 cells were purified and sequenced, and no mutations differing from the wild-type sequence were detected in the SUMO1 transgenes in the pooled DNAs.
To examine the SUMO1 and SUMO1ΔGG transcript levels, we isolated RNA from F9 and NIH 3T3 cells transduced with the SUMO1 or SUMO1ΔGG vectors. RNAs were also isolated from untreated F9 and NIH 3T3 cells as negative controls. cDNA was synthesized from the RNAs, and SUMO1 or SUMO1ΔGG RNA levels were assessed by reverse transcription-quantitative PCR (qRT-PCR) using primers spanning SUMO1 exon-exon junctions. We detected similarly high levels of SUMO1 and SUMO1ΔGG transcripts in transduced F9 cells and NIH 3T3 cells and only very low levels in the untransduced control cells (Fig. 2C). Thus, the differences in the levels of utilization of the SUMO1 produced from these two constructs in embryonic cells were not due to unequal levels of transcripts.
One possible explanation for the lack of free SUMO1 accumulation in embryonic cells is protein degradation via either the proteasomal or lysosomal pathway. To test whether SUMO1 is degraded in embryonic cells by either pathway, F9 cells were transduced with the SUMO1 or SUMO1ΔGG vectors, and at 72 h postinfection the cells were treated with the proteasome inhibitor MG132 or the lysosomal degradation inhibitor chloroquine for differing times. Cell lysates were prepared and analyzed for SUMO1 expression by Western blotting. Free SUMO1 levels did not increase after 4 h of treatment with either drug (Fig. 2D). Prolonging drug treatment to 8 h, 12 h, or 24 h did not increase the low levels of SUMO1 expression (data not shown). The results suggest that proteasomal or lysosomal degradation was probably not responsible for the low levels of SUMO1 expression.
SUMO2 can be overexpressed in embryonic and differentiated cells. To determine if SUMO2 is also poorly expressed in embryonic cells, we inserted Flag-tagged SUMO2 and SUMO2ΔGG into the same pLVX lentiviral vector as was used for the SUMO1 constructs. F9, E14, and 293T cells were transduced with the SUMO2 or SUMO2ΔGG vectors and selected for drug resistance. Cell lysates were prepared and examined for SUMO2 or SUMO2ΔGG expression by Western blotting. F9 cells expressed free SUMO2 and SUMO2ΔGG proteins to roughly similar high levels, while free SUMO1 protein levels were dramatically lower than SUMO1ΔGG expression levels, as before (Fig. 3A, left blot). Embryonic E14 cells also expressed free SUMO2 and SUMO2ΔGG to similar levels, in contrast to the very low expression of free SUMO1 relative to SUMO1ΔGG (Fig. 3A, right blot). Differentiated 293T cells transduced with the SUMO2 and SUMO2ΔGG vectors expressed the two constructs to similar levels, as expected (Fig. 3B). These results indicate that poor expression of free SUMO in embryonic cells occurs specifically for SUMO1 and not for all the SUMO family members.
Reducing SUMO1 conjugation activity restores free SUMO1 expression. SUMO1ΔGG is missing the last six amino acids of SUMO1, including the diglycine residues necessary for conjugation to substrates. To test if this C-terminal tail was a target for posttranscriptional regulation, we created a construct expressing full-length Flag-tagged SUMO1 but containing alanine substitutions of the diglycine residues (SUMO1AA) that would prevent conjugation. F9 cells were transduced with the SUMO1 or SUMO1AA vector, and lysates were prepared and examined for SUMO expression by Western blotting. Free SUMO1AA was expressed well in embryonic cells to levels similar to those seen with SUMO1ΔGG and in contrast to the poor accumulation of free SUMO1 (Fig. 4A). This result suggests that the toxic effects of SUMO1 overexpression required SUMO1 conjugation to at least some substrates.
If elevated levels of particular SUMO1-conjugated substrates are toxic to F9 cells, then reducing levels of conjugation to those substrates might allow free SUMO1 to be more highly expressed. In one approach designed to reduce SUMO1 conjugation of endogenous substrates, we coexpressed SUMO1 with a highly sumoylated protein that could act as a competitive "sponge" for the overexpressed SUMO1. Tripartite motifcontaining 28 (Trim28 or KAP1), a transcriptional repressor protein, is one such highly sumoylated protein, with six lysine residues that can be conjugated by SUMO (20). Hemagglutinin (HA)-tagged Trim28 cDNA was inserted into the same pLVX lentiviral vector used for the SUMO constructs, but encoding G418 resistance, with the neoR gene in place of the puroR gene. We also designed a construct expressing Trim28 in which all six lysine residues known to act as SUMO acceptors were mutated to arginine, rendering the mutant unavailable for SUMO conjugation (Trim28 6KR ). SUMO and Trim28 constructs was introduced into cells sequentially. F9 cells were transduced and selected for the expression of the Trim28 constructs, followed by transduction and selection for the expression of the SUMO constructs, or in the reverse order. Cell lysates were prepared from these doubly drug-resistant cells, and SUMO1 expression levels were assessed by Western blotting.
Cells that were first transduced and selected for expression of the Trim28 constructs before subsequent transduction with the SUMO vectors were profoundly different with respect to the ability to express free SUMO1. Overexpression of wild-type Trim28 (Trim28 WT ), but not Trim28 6KR , allowed greatly increased subsequent detection of free SUMO1 (Fig. 4B, lanes 1 to 4). Thus, the accumulation of free SUMO1 required that the overexpressed Trim28 be capable of serving as a SUMO1 acceptor, consistent with its serving as a SUMO1 "sponge." In contrast, cells that were first transduced and selected for the expression of SUMO1 before subsequent transduction with the Trim28 vectors continued to show weak expression of free SUMO1 even in the presence of exogenous Trim28 WT or Trim28 6KR proteins (Fig. 4B, lanes 5 to 8). Thus, later expression of Trim28 could not reverse the mechanism that prevents accumulation of free SUMO1. As expected, SUMO1ΔGG was well expressed with or without the Trim28 proteins.
Trim28 can act as an E3 ligase in the SUMO conjugation cascade (21), mediates autosumoylation (20), and contains important functions for embryonic cells (22,23). Although we found no evidence to suggest that Trim28 overexpression had any adverse effects on embryonic cells, the rescue of SUMO1 expression might have involved some specific Trim28 function. To evaluate this possibility, we repeated the

SUMO1 Overexpression Not Tolerated by Embryonic Cells
® previous experiment using a different candidate SUMO "sponge." Ran GTPaseactivating protein 1 (RanGAP1) is a regulatory trafficking protein that is commonly used in SUMO1 studies and was one of the earliest proteins to have been found to be a target of SUMO1 conjugation (5). Myc-tagged RanGAP1 cDNA was inserted into a pLVX lentiviral vector containing the G418 resistance drug selection marker. As with the Trim28 constructs, F9 cells were transduced and selected for the expression of the RanGAP1 construct before or after transduction by the SUMO1 or SUMO1ΔGG constructs. Lysates were prepared and SUMO1 expression levels were examined by Western blotting probing for the Flag tag. Expression of free SUMO1 improved markedly when RanGAP1 was overexpressed prior to SUMO1 overexpression but not when RanGAP1 was expressed after SUMO1 (Fig. 4C). Thus, the block to accumulation of free SUMO1 can be relieved by prior overexpression of multiple SUMO substrates. In another approach designed to reduce the levels of SUMO1 conjugation to substrates, we tested the overexpression of SUMO-specific peptidase 1 (SENP1). HAtagged SENP1 was cloned into a pLVX vector, and F9 cells were transduced and selected for the expression of SENP1 either before or after transduction and selection for the expression of the SUMO constructs. Lysates were prepared and analyzed for SUMO1 expression by Western blotting. SENP1 overexpression prior to expression of SUMO1 dramatically improved levels of free SUMO1 accumulation, almost reaching SUMO1ΔGG expression levels (Fig. 4D). However, SUMO1 overexpression levels did not improve when this order was reversed.
In sum, we show here that embryonic cells do not tolerate overexpression of SUMO1 and that cells surviving transduction specifically downregulate the accumulation of free SUMO1 at a posttranscriptional stage, maintaining SUMO1 in the form of selected high-molecular-weight conjugates. Reducing SUMO1 conjugation by mutation of the diglycine residues or by overexpression of a SUMO sponge or by overexpression of SENP1 dramatically increased the expression of free SUMO1. The results suggest that the toxic effects of SUMO1 overexpression are a consequence of the accumulation of SUMO1-modified proteins and not of the accumulation of free SUMO1 protein itself. Moreover, the embryonic cells surviving the forced overexpression become committed to the altered distribution of SUMO1 irreversibly, since reducing the levels of SUMO1modified proteins after the pattern of distribution had been established did not rescue the accumulation of free SUMO1.

DISCUSSION
In this report, we have provided evidence that embryonic cells do not readily tolerate forced overexpression of SUMO1 likely occurring through the effects of inappropriately high levels of SUMO1-modified substrates. Embryonic cells surviving after transduction by expression constructs have an aberrant distribution of SUMO1 between free and conjugated pools and do not accumulate free conjugation-competent SUMO1 (Fig. 1). In contrast, conjugation-defective SUMO1 mutants could be readily expressed as free protein. The redistribution of SUMO was specifically observed with SUMO1 and not SUMO2. We do not yet know the mechanism of action responsible for the altered processing of SUMO1 in cells surviving transduction, but the analysis of SUMO1 DNA and mRNA levels suggests that the changes are posttranscriptional. The relatively high frequency of recovery of surviving cells after transduction (about 10% of control) is not consistent with mutation in the host genome but rather suggests an adaptation through altered physiology or processing of SUMO1. We cannot completely rule out the possibility that there has been selection for subtle mutations in the SUMO1 expression construct that are responsible for the altered distribution, but no common mutations were detected in DNA sequences of pooled transgenes.
In principle, elevated levels of either free SUMO1 or SUMO1-modified proteins could be problematic with respect to high expression in embryonic cells. We found that SUMO1 mutants that could not conjugate to substrates (SUMO1ΔGG and SUMO1AA) were readily expressed in embryonic cells to high levels ( Fig. 1A and B and 4A), suggesting that the accumulation of free SUMO1 protein per se was not toxic but rather was tolerated so long as the free SUMO1 protein could not be conjugated to substrates. To test whether SUMO1 conjugated substrates initiated the altered utilization, we blocked or reduced SUMO1 conjugation to endogenous substrates using various approaches. One approach was to coexpress SUMO1 with a highly sumoylated protein to act as a "SUMO sponge" for excess SUMO1 protein. We chose RanGAP1 and Trim28 as SUMO sponges because Trim28 is known to contain multiple sites for sumoylation (20), and RanGAP1 was one of the first known prominent substrates of sumoylation (24). In the background of Trim28 and RanGAP1 overexpression, expression of free SUMO1 in embryonic cells was greatly improved (Fig. 4B and data not shown). Another approach was to overexpress SUMO1 along with SENP1, the enzyme responsible for deconjugation of SUMO1 from its substrates (16). We found that overexpression of SENP1 greatly improved exogenous SUMO1 expression as well. These results indicate that SUMO1 overexpression in embryonic cells is prevented as a response to the accumulation of SUMO1-modified proteins and not as a response to the overexpression of SUMO1 protein per se (see Fig. 5 for model).
An intriguing aspect of these experiments is that free SUMO1 accumulation improved only when the SUMO sponge or SENP1 was overexpressed prior to transduction with the SUMO1 vector. Reversing this order did not result in improved accumulation of free SUMO1; once the pattern of utilization was established, it was not relieved by subsequent lowering of SUMO1 conjugation. SUMO1 modification of substrates was previously reported to have had long-term effects even when SUMO1 was no longer conjugated to the substrate (24), and perhaps the SUMO1 modification of certain critical embryonic substrates creates long-term effects that are not immediately reversible. Though none were detected, any mutations of the SUMO1 transgene selected for during transduction would also not be reversed.
The results described here suggest that the accumulation of SUMO1-modified proteins was toxic or inhibited replication or caused cell death and that only a few cells survived transduction by expressing low levels of free SUMO1. One possible explanation for the observations would be the sequestration of the SUMO1 in an intracellular location that precludes its extraction, but we consider this unlikely given the harsh conditions used for lysis. We favor the possibility that an increase in the levels of specific SUMO1-modified proteins triggers a mechanism for lowering free SUMO1 expression in the surviving embryonic cells. This mechanism could in principle act at any of several stages of expression: at the retention of the transgene, at transcription, at translation, or at a posttranslational step. Examination of the SUMO1 transgene and transcript levels in embryonic cells transduced with the SUMO1 vector showed that both DNAs and RNAs were present at levels equivalent to those in embryonic cells transduced with the SUMO1ΔGG vector ( Fig. 2A and C). Thus, the likely mechanisms are posttranscriptional and could have involved protein stability. We found no evidence for SUMO1 protein degradation by the proteosomal or lysosomal pathways (Fig. 2D). The most likely mechanism is an altered course of SUMO1 processing, with all the detectable SUMO1 being distributed to high-molecular-weight conjugates that are tolerated, and with no accumulation of free SUMO1. This could be achieved by redirecting the SUMO1 to acceptable substrates or by increasing the levels of these substrates or by decreasing the levels of critical substrates that become toxic to cell viability upon excessive SUMO1 conjugation.
A striking aspect of the SUMO redistribution is the specificity for SUMO1 and not another SUMO family member. SUMO2 was overexpressed in both embryonic and differentiated cells as efficiently as SUMO1ΔGG and accumulated to high levels ( Fig. 2A). It is possible that SUMO1 is redistributed because it is more actively conjugated to critical substrates whereas SUMO2 is less efficiently conjugated or because addition of SUMO1 to critical substrates is inherently more toxic. The results are consistent with previous studies showing that SUMO1 knockout is tolerated (13) but that the increase in steady-state levels of SUMO1 resulting from SENP1 knockout is embryonic lethal (16). The fact that the cells adapt to SUMO1 overexpression by reducing SUMO1 availability for conjugation suggests that an alternative mechanism of escape-reducing the levels of the critical substrate itself-is not a viable option. The identity of the crucial substrate (or substrates) in embryonic cells is not known, but several proteins known to be modified by addition of SUMO1 are candidates. These include regulators of stem cell differentiation such as Oct4 and Sox2 (25), which control levels of expression of the Nanog protein, and it is plausible that inappropriate levels of their modification could be toxic. Further studies exploring the distribution of SUMO1 to its substrates in embryonic cells will be important for understanding the complex nature of embryonic cells.