Research ResourceBiochemistry

Comparative Proteomic Analysis Identifies a Role for SUMO in Protein Quality Control

See allHide authors and affiliations

Science Signaling  21 Jun 2011:
Vol. 4, Issue 178, pp. rs4
DOI: 10.1126/scisignal.2001484


The small ubiquitin-like modifiers (SUMOs) alter the functions of diverse cellular proteins by covalent posttranslational modification and thus influence many cellular functions, including gene transcription, cell cycle, and DNA repair. Although conjugation by ubiquitin and SUMO-2/3 are largely functionally and mechanistically independent from one another, both appear to increase under conditions of proteasome inhibition. To better understand the relationship between SUMO and protein degradation by the proteasome, we performed a quantitative proteomic analysis of SUMO-2 substrates after short- and long-term inhibition of the proteasome with MG132. Comparisons with changes to the SUMO-2 conjugate subproteome in response to heat stress revealed qualitative and quantitative parallels between both conditions; however, in contrast to heat stress, the MG132-triggered increase in SUMO-2 conjugation depended strictly on protein synthesis, implying that the accumulation of newly synthesized, misfolded proteins destined for degradation by the proteasome triggered the SUMO conjugation response. Furthermore, proteasomal inhibition resulted in the accumulation of conjugated forms of all SUMO paralogs in insoluble protein inclusions and in the accumulation on SUMO-2 substrates of lysine-63–linked polyubiquitin chains, which are not thought to serve as signals for proteasome-mediated degradation. Together, these findings suggest multiple, proteasome-independent roles for SUMOs in the cellular response to the accumulation of misfolded proteins.


The ubiquitin-like posttranslational protein modifiers (Ubls) are a superfamily of proteins with structural similarity to ubiquitin that form isopeptide bonds through their C-terminal carboxyl groups with the lysine side chains of target substrates. Similarities across the family extend beyond structure, with the conjugation and deconjugation of all Ubls following parallel pathways with, for most cases, distinct but structurally related enzymes (1). After ubiquitin, the small ubiquitin-like modifier (SUMO) proteins are the next most heavily studied of the Ubls. Higher eukaryotes have three SUMO paralogs, designated SUMO-1, SUMO-2, and SUMO-3, whereas yeast have only a single ortholog. The SUMO subfamily can be divided into two groups, with SUMO-2 and SUMO-3 being collectively termed SUMO-2/3 because of structural and functional differences from SUMO-1.

All SUMO proteins are conjugated to substrates by the same heterodimeric SUMO-activating enzyme (SAE1/SAE2) and SUMO-conjugating enzyme Ubc9. Paralog-specific metabolism seems to be conferred by the E3 ligases and the SUMO-specific proteases (2). The enzymes that mediate SUMO conjugation operate within a three-step pathway whereby SAE2 forms an adenosine triphosphate (ATP)–dependent thioester bond with the C terminus of SUMO in the first step, which is followed by transesterification of SUMO to Ubc9. Finally, Ubc9 catalyzes the isopeptide bond formation on the ε amino group of the lysine target commonly found within a SUMOylation consensus motif ψKxE (where ψ represents any hydrophobic amino acid residue, and x represents any residue), a step that can be accelerated by one of a number of E3 ligases (2). The SUMO proteases, also known as sentrin proteases (SENPs), are a six-member family of proteins that cleave SUMO molecules on the C-terminal side of a specific Gly-Gly sequence that is the site of conjugation to substrate proteins. Because all SUMOs are generated as nonconjugatable pro-proteins and can also form poly-SUMO chains, the SENPs are responsible for the maturation, deconjugation, and depolymerization of SUMOs (3).

SUMO is essential for normal cell function, with the Saccharomyces cerevisiae and Schizosaccharomyces pombe orthologs, Smt3 and Pmt3, being required for cell viability (4) and normal growth (5), respectively. Experiments with mouse knockouts have led to the suggestion that SUMO-2/3 can functionally compensate for the loss of SUMO-1 (6, 7); however, SUMO conjugation is required in higher eukaryotes, because deletion of Ubc9 causes early embryonic lethality in mice (8). A number of functional distinctions between SUMO-1 and SUMO-2/3 have been reported. SUMO-2/3 seem to be far more abundant than SUMO-1 in cells, and free SUMO-2/3 are incorporated into high–molecular mass conjugates in response to protein-damaging stresses, including heat shock and ethanol exposure (9). The substrate specificities of the SUMO paralogs differ substantially (10), with the most extreme examples being the nucleoporin RanGAP1 (Ran guanosine triphosphatase activating protein 1), which is almost entirely modified by SUMO-1 (9), and the promyelocytic leukemia protein (PML)–body component Sp100, which is specifically modified by SUMO-2 (10). Furthermore, unlike SUMO-1, SUMO-2/3 contain a SUMO conjugation consensus motif that enables them to accept other SUMO moieties to form polymeric chains (11), whereas the contribution of SUMO-1 to polymers in vivo seems to be restricted to the capping of SUMO-2/3 chains (12). SUMO polymers can act as discrete signals. Specifically, the really interesting new gene (RING) finger protein 4 (RNF4, also known as SNURF) is an E3 ubiquitin ligase that preferentially binds to and ubiquitinates SUMO chains composed of multiple copies of SUMO (13), which provides a direct link from SUMO to the ubiquitin modification system. Conversely, SUMOylation is also described as antagonizing ubiquitination by competing for lysine acceptors, such as in the case of the nuclear factor κB (NF-κB) inhibitor protein IκBα (14).

Proteasomal inhibition results in the accumulation of SUMO-conjugated proteins and triggers the colocalization of SUMO with ubiquitin in nuclear domains associated with PML bodies (15), which is consistent with a role for SUMO in protein turnover. Here, however, we show that the substrate-specific changes in SUMO-2 conjugation in cultured cells during treatment with the proteasome inhibitor MG132 were broadly consistent with those triggered by heat stress, which appeared to be as a secondary response to the accumulation of misfolded proteins normally destined for degradation by the ubiquitin proteasome system (UPS). We also describe time-dependent exchanges of SUMO-2 and ubiquitin among substrates during proteasome inhibition and an unexpected accumulation of Lys63-linked polyubiquitin chains in SUMO-2–conjugated proteins purified from cells after long exposure to MG132. Proteomic analysis of insoluble inclusions revealed that all SUMO paralogs accumulated in aggregates upon treatment with MG132. These data suggest that all of the SUMO proteins are involved both indirectly and directly in a complex and multifaceted response to the accumulation of unfolded and misfolded proteins within the cell.


All SUMO paralogs respond to proteasome inhibition with a general increase in the extent of their conjugation

In cultured cells, SUMO-2 and SUMO-3, unlike endogenous SUMO-1, respond to inhibition of the proteasome by exhibiting increased conjugation (9, 16). Furthermore, links between the SUMO-2/3 and ubiquitin pathways have been uncovered that suggest that some SUMO-conjugated proteins may be targeted for degradation by the UPS (17). To better understand the dynamics of this response, we analyzed the changes in SUMO-1, SUMO-2/3, and ubiquitin conjugation in cultured human cells over a 7-hour period of proteasomal inhibition. HeLa cells were exposed to MG132 (20 μM) for different time periods, and crude cell extracts were analyzed by Western blotting. As expected, conjugates containing ubiquitin (Fig. 1A) and SUMO-2/3 (Fig. 1B) accumulated in a time-dependent manner, although the ubiquitin response was more rapid than that of SUMO-2/3 (compare the 1-hour time points between Fig. 1, A and B).

Fig. 1

Analysis of total SUMO and ubiquitin conjugation kinetics after inhibition of the proteasome with MG132. (A to D) HeLa cells were treated with MG132 (20 μM, +MG132) or DMSO alone (−MG132) for the indicated time periods, and whole-cell lysates were prepared. Protein (20 μg) from each sample was resolved by SDS-PAGE (on 4 to 12% polyacrylamide gels) before undergoing Western blotting analysis for the detection of (A) ubiquitin, (B) SUMO-2/3, (C) SUMO-1, and (D) α-tubulin. The lower panel in (C) shows a longer exposure of the blot above to more clearly show unconjugated SUMO-1. (E) Densitometric analysis of the relative abundances of free unconjugated modifiers over the time courses of MG132 treatment depicted in (A) to (C). Data are representative of multiple similar experiments.

In response to MG132, the global changes in the extent of conjugation, as judged by Western blotting, were less extensive with SUMO-1 than with SUMO-2/3 (Fig. 1C, upper panel). However, despite the lower proportion of SUMO-1 in the unconjugated state (Fig. 1C, upper panel) compared to that of SUMO-2/3, the disappearance of the unconjugated protein was qualitatively and kinetically conserved among all SUMO paralogs (Fig. 1, A to E). This suggested that the mechanism responsible for enhanced SUMOylation upon MG132 treatment was at least partially paralog-independent and that the major determinant of the extent of SUMOylation was the quantity of unconjugated SUMO molecules in the cell. This suggestion is supported by experiments involving cells transfected to overexpress exogenous SUMO-1, which showed a more SUMO-2/3–like accumulation into conjugates after MG132 treatment than did endogenous SUMO-1 (15, 18). Nevertheless, considering the changes in the abundance of conjugates, the greatest responses of the endogenous SUMO paralogs in these cells were exhibited by SUMO-2/3.

SUMO-2 conjugation shows dynamic changes during short- and long-term proteasomal inhibition

The accumulation of ubiquitin conjugates upon proteasomal inhibition can be explained by the reduced turnover of polyubiquitinated conjugates by the proteasome; however, because SUMO is not thought to directly target proteins for degradation, the simplest explanation for accumulation of SUMO-2/3 conjugates is that these proteins are normally turned over by the UPS, through either a SUMO-dependent or a SUMO-independent mechanism (17). This being the case, it was unclear why this accumulation was temporally distinct from that of ubiquitinated proteins in general (Fig. 1). To gain a deeper insight into the relationship between SUMO-2 and proteasome inhibition, we performed a time-resolved quantitative proteomic study to monitor changes in the conjugation status of SUMO-2 substrates during proteasomal inhibition (Fig. 2). HeLa cells stably expressing tandem affinity protein (TAP)–tagged SUMO-2 (19) were grown in medium to enable stable isotope labeling by amino acids in cell culture (SILAC) (20). In this experiment, the mock-treated cells were grown in isotopically normal conditions (K0R0), and K4R6- and K8R10-labeled cells were treated with MG132 (20 μM) for 2 or 7 hours, respectively (Fig. 2). These are more generally regarded as “light” (L), “medium” (M), and “heavy” (H) isotopic conditions, respectively. These time points were selected to follow the full extent of SUMO-2 conjugation (7 hours), as well as ~50% completed (2 hours) (Fig. 1E). Western blotting analysis with antibodies against TAP (anti-TAP antibodies) showed that TAP–SUMO-2 responded to proteasomal inhibition in a manner similar to that of endogenous SUMO-2 (fig. S1).

Fig. 2

A quantitative proteomics experiment to monitor changes in SUMO-2 conjugation after 2 and 7 hours of MG132 treatment. HeLa cells stably expressing TAP–SUMO-2 were cultured in SILAC medium containing unlabeled lysine and arginine amino acids (K0R0), +4-dalton lysine and +6-dalton arginine (K4R6), or +8-dalton lysine and +10-dalton arginine (K8R10). K0R0 cells were the “untreated” (DMSO) control, and the K4R6 and K8R10 cells were treated with MG132 (20 μM) in DMSO for 2 or 7 hours, respectively. Fifty dishes of cells for each treatment were lysed under denaturing conditions, and extracts were mixed in a 1:1:1 ratio by protein mass. Of this “crude” extract, 75 μg was resolved over three lanes of an SDS-PAGE gel (left gel shows one example), and the remaining 550 mg was used for TAP affinity purification (see Materials and Methods for details). The entire sample of purified proteins was then resolved in a single lane by SDS-PAGE (right gel). Gel lanes were excised into nine sections, and each slice was subjected to tryptic digestion and peptide extraction before undergoing MS analysis and quantitation with MaxQuant software. For each protein, three ratios were generated for the relative abundance of that protein in each experimental condition: Medium/Light (2 hours MG132/untreated), Heavy/Light (7 hours MG132/untreated), and Heavy/Medium (7 hours MG132/2 hours MG132). The entire data set for two ratios can be represented on a single scatter plot [tsMAP (19)], in which each point represents a protein and its coordinates reflect the relative change in abundance between the compared conditions. tsMAPs showing log2 ratios for the entire data sets for both “Crude” and “Purified” preparations are shown. Data are representative of a single SILAC experiment with a minimum of two SILAC ratio counts per protein. Ratio counts for each protein can be found in Supplementary File 1.

Fifty 150-mm dishes of cells grown in each SILAC condition were prepared, and the denaturing extracts were mixed in ~1:1:1 ratio by protein mass. Of this mixture, 75 μg was fractionated by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) to give a “crude” sample (Fig. 2, left gel), whereas the remaining material was used for TAP purification to generate a “purified” sample of the SUMO-2 subproteome (21), also fractionated by SDS-PAGE (Fig. 2, right gel). Both gels were excised into nine slices at about the same apparent molecular mass (MWApp) boundaries. Proteins were subjected to in-gel tryptic digestion (22) and analyzed multiple times by nanoscale liquid chromatography coupled to high-resolution, hybrid mass spectrometers: linear trap quadrupole (LTQ) Orbitrap or Fourier transform ion cyclotron resonance (FT-ICR) devices. Relative changes in protein amounts were quantified by MaxQuant analysis of SILAC triplets (23). From the crude sample, 2467 proteins were identified by at least one unique peptide and quantified with a minimum of two SILAC ratio counts, and we detected and measured 1355 proteins in the purified sample by the same criteria; of these, 1001 proteins were quantified in both samples. The data are represented as triple SILAC maps (tsMAPs) (19) (Fig. 2, bottom charts), in which raw ratios are converted to log2 values to provide coordinates for each protein represented as a point on the scatter chart. Proteins identified in the crude sample were mainly clustered in the same region of the chart, whereas proteins from the TAP–SUMO-2 proteome were spread over a far larger area. This indicates that most proteins from the crude sample did not substantially change in total abundance over the 2 or 7 hours of MG132 treatment, whereas the abundances, and hence SUMO-2 conjugation state, of many proteins from the purified sample varied greatly between time points.

MG132-responsive SUMO-2 substrates are separable from other components through a crude data-dependent filtering method

Because of inevitable sample mixing errors and incomplete isotope incorporation, it is necessary to normalize raw SILAC ratios derived from quantitative proteomics experiments. MaxQuant software normalizes SILAC ratios and identifies statistically significant outliers by an internal calibration method that relies on the frequency distribution of the SILAC ratios of all of the identified proteins being unimodal (23). As a consequence of the unusually high stringency of the TAP–SUMO-2 purification, such distributions are typically not unimodal (19), and so alternative methods are required to normalize ratios and to identify significant outliers (that is, those proteins whose change in abundance is likely to be real). To facilitate this here, we developed a crude data-dependent method of normalization and filtering. This method relies on the assumption that the frequency distributions of the SILAC ratios of proteins whose abundances were unaffected by treatment with MG132 were the same in both the crude and the purified sample preparations from the same experiment. Frequency histograms of the crude data for each SILAC ratio were unimodally, normally distributed (fig. S2A), which should reflect the equivalent distributions of SILAC ratios of proteins from the purified sample that did not significantly change in abundance. The data from the crude fraction were fitted to normal distributions to define M/L, H/L, and H/M ratio boundaries that encompassed >95% of the proteins (fig. S2, A and B, and see the Supplementary Materials for details). These were then applied to the purified data set to filter proteins from the TAP-purified fractions whose abundance significantly changed after 2 or 7 hours of MG132 treatment (that is, putative SUMO-2 substrates whose conjugation status changed at either time) (fig. S2C). This approach provided a list of 564 proteins from the 1355 quantified in the TAP–SUMO-2 sample, which we propose to be SUMO-2 conjugates that responded to MG132. The calculated mean values of the crude normal distributions enabled normalization of the purified ratios (fig. S2D).

Bioinformatic comparisons between the putative substrates and the remaining proteins showed that the MG132-responsive putative substrates group was enriched for proteins that contained SUMO conjugation consensus motifs (table S1), proteins already known to be SUMO conjugates (fig. S3), and proteins that resolved in the SDS-PAGE gel to a molecular mass higher than that expected by sequence alone (fig. S4). Together, these analyses confirm that this crude data-dependent filtering method was successful in identifying the SUMO substrates from the mixture of proteins identified from the purified data set. This method should be universally applicable to triple-label SILAC experiments for which “internal” normalization and filtering methods cannot be used.

The SUMO-2 conjugation response to MG132 is dynamic and substrate-dependent

We prepared a color-coded tsMAP of the normalized data with marginal histograms in which putative SUMO-2 conjugates are shown in blue and filtering rejects in red (Fig. 3). About half of the putative substrates and 90% of the rejects were also identified in the crude sample (fig. S5A). Changes in protein abundances in the purified sample were a consequence of altered SUMOylation extent rather than altered total protein amounts, because there was no statistically significant correlation between normalized crude and purified ratios for both the 2-hour data (fig. S5B) and the 7-hour data (fig. S5C). Exceptions to this included ubiquitin, SUMO-2, SUMO-3, TAP–SUMO-2, and NEDD8, whose SILAC ratios in both preparations were close to equivalent, which implies that the change seen in the purified sample was largely a consequence of increased total protein amounts. SUMO-1 did not follow this trend, with an increase in abundance only in the TAP-purified preparation, suggesting that SUMO-1 was conjugated to SUMO-2 or SUMO-3 substrates after MG132 treatment.

Fig. 3

Global quantitative analysis of the SUMO-2 conjugate proteome after 2 and 7 hours of MG132 treatment. A tsMAP (19) is shown with marginal frequency histograms of 564 proteins, which were accepted by crude, data-dependent filtering (fig. S2) giving 564 putative SUMO-2 substrates (blue), and 742 proteins rejected as nonsubstrate “contaminants” or substrates that do not change in modification state during the analysis (red). Percentages of putative substrates in each quarter of the chart are shown in blue. Data are representative of a single experiment with a median of 10 and minimum of 2 SILAC ratio counts per putative substrate protein. Ratio counts for each protein can be found in Supplementary File 1.

Analysis of the tsMAP for purified material (Fig. 3) shows that, consistent with the Western blotting data (Fig. 1B), the extent of the general response to MG132 was related to the duration of incubation. Specifically, an incubation time of 2 hours caused a relatively small increase (2-fold) in the abundances of a subset of proteins (Fig. 3, horizontal histogram), whereas 7 hours of incubation promoted a far greater increase (8- to 16-fold) in abundance (Fig. 3, vertical histogram). As previously reported (16), long exposures to MG132 caused an increase in conjugation to many SUMO-2/3 substrates, concomitant with the deconjugation of others (Fig. 3, vertical histogram); however, when we considered these data in the context of the shorter treatment, a more complex picture emerged, where for a large proportion of substrates, the change in SUMOylation state was not linear with respect to time. Inspection of the tsMAP (Fig. 3) shows that most of the putative SUMO-2 substrates (76%) are found in the top half, indicating the general increase in conjugation after 7 hours of MG132 exposure. However, for the remaining 24% that were deconjugated from SUMO-2 after extended proteasome inhibition (Fig. 3, lower half of tsMAP), most previously experienced a period of increased conjugation after only 2 hours of inhibition (22.7% of all substrates versus 1.2% deconjugated at both time points). Also, most of the 31.4% of identified putative SUMO-2 substrates that exhibited deconjugation after 2 hours of MG132 incubation (Fig. 3, left half of tsMAP) experienced increased SUMOylation 5 hours later (30.2% versus 1.2%). Evidence supporting this unexpected nonlinearity of SUMOylation change over time can be found in the Western blotting analysis of crude cell lysates from this experiment, in which a TAP-specific antibody-reactive species was increased in abundance after 2 hours, then reduced in abundance five hours later (fig. S1, asterisk). This shows that there was not only a general increase in SUMO-2 conjugates after proteasome inhibition but also a temporally resolved flux of SUMO-2/3 among substrates. Furthermore, consistent with the published literature (24), comparisons between MWApp in crude and in purified preparations suggest that for most proteins, only a small proportion of the total pool was modified by SUMO and responded to proteasome inhibition (fig. S6).

The observed changes to the SUMO-2 subproteome in response to MG132 treatment are a consequence of proteasome inhibition

Although widely used as a proteasome inhibitor, MG132 also inhibits the activities of calpains and some lysosomal cysteine proteases (25, 26). To confirm that the observed effects of MG132 on the SUMO-2 conjugate subproteome were a consequence of its proteasome-inhibiting activity, we performed a quantitative proteomic experiment to compare the effects of MG132 with those of lactacystin, another proteasome inhibitor (25). In these experiments, we compared the effects of 8 hours of exposure to lactacystin (40 μM, K4R6) or MG132 (4 μM, K8R10) with those of dimethyl sulfoxide (DMSO, K0R0) on the SUMO-2 conjugate subproteome in experiments with the previously characterized 6His–SUMO-2 HeLa cell line (27) (fig. S7). We identified and quantified 507 proteins from nickel affinity chromatography purifications from these cells (Supplementary File 6). Similar to the TAP–SUMO-2 experiment (Fig. 2), we saw that proteins showed a wide range of changes in abundance after treatment, and that there was a strong correlation in the responses to lactacystin and MG132 (fig. S7). This confirmed that the changes in the SUMO-2 conjugate subproteome stimulated by MG132 were as a result of its inhibition of the proteasome.

MG132 triggers a similar SUMO-2 conjugation response to that of heat stress

We previously used the same TAP–SUMO-2 quantitative proteomics protocol described here to study the effect of heat shock on SUMO-2 conjugation (19). Similar to extended exposure to MG132, heat shock induces a general increase in SUMO-2 conjugation, in spite of which a minority of proteins undergo deconjugation (19). Gene ontology (GO) analysis showed that putative substrates identified in each study (fig. S8) were similarly represented in different functional groups. Specifically, proteins with nucleic acid–binding activity, such as transcription factors, chromatin remodelers, mRNA-processing proteins, and DNA-metabolizing proteins, were highly overrepresented in the putative SUMO-2 substrates lists, whereas underrepresented groups largely contained proteins that function at or close to the plasma membrane. In one heat shock TAP–SUMO-2 experiment, 587 putative SUMO-2 substrates were identified, of which 399 were also identified here (Fig. 4A). Comparison of the normalized ratios of all 399 proteins showed that they correlated closely, with a Pearson coefficient of 0.74, although the extent of modification seems to have been greater for heat shock than for MG132 treatment (Fig. 4B). These data show not only that heat shock and long-term MG132 exposure regulated the SUMOylation state of proteins with similar functions but also that individual proteins were regulated in a quantitatively similar way.

Fig. 4

Changes in SUMO-2 conjugation as a consequence of proteasome inhibition correlate with those that occur after heat shock. (A) Protein identification overlap between this study (yellow) and a study of the change in TAP–SUMO-2 conjugation upon heat stress (HS, pink) (19). (B) Scatter plot of all of the 399 proteins common to the two studies showing log2(HS/untreated) (abscissa) and log2(7 hours MG132/untreated) (ordinate). Equations of linear regression, R2, and Pearson correlations are shown. Note that 572 proteins were designated as “putative substrates” for the MG132 experiment from this analysis rather than the previously reported 564 proteins (Fig. 3). This is due to the inclusion of the heat shock data, which increased the total number of protein identifications and also increased the number of proteins from the MG132 experiment that met the 1% FDR setting during MaxQuant processing.

The SUMO-2 response to MG132 is blocked by the inhibition of protein synthesis

A major role for ubiquitin-mediated protein degradation in the cell is the specific degradation of misfolded proteins (28); almost 30% of newly synthesized proteins are thought to be misfolded upon initial synthesis at the ribosome (29). Because heat shock triggers protein unfolding, it is possible that the accumulation of misfolded proteins is the common feature between heat shock and MG132 treatment to which SUMO-2 conjugation responds. However, it is also conceivable that proteasome inhibition and heat shock both result in the stabilization of a protein responsible for SUMO conjugation, thus triggering the accumulation of conjugates. If the accumulation of newly synthesized misfolded proteins is the trigger for the SUMO-2/3 response to MG132, then inhibition of new protein synthesis should block the response to MG132 but not that to heat shock. To test this hypothesis, we pretreated HeLa cells with cycloheximide and puromycin to block ribosome function and monitored the accumulation of SUMO-2 and ubiquitin conjugates by Western blotting analysis of crude extracts from cells exposed to heat shock or MG132 (Fig. 5). Inhibition of protein synthesis completely abrogated the SUMO-2/3 response to MG132 but not that to heat shock, confirming that the SUMO-2/3 response to MG132 was indeed dependent on protein synthesis. Even in the absence of new protein synthesis, ubiquitin still responded to MG132 by accumulation into high–molecular mass conjugates, albeit to a lesser extent than occurred under normal conditions (Fig. 5). This was likely because of the stabilization of proteins other than those that were newly synthesized. Consistent with misfolded proteins being a trigger for SUMO-2 conjugation, the nonprotein amino acid l-canavanine, which competes with l-arginine for incorporation into nascent polypeptides, thus creating misfolded canavanyl proteins (30), also triggered both SUMO-2/3 and ubiquitin conjugation (fig. S9).

Fig. 5

Inhibition of protein synthesis blocks the SUMO-2 response to MG132 but not heat shock. HeLa cells were treated with either DMSO or cycloheximide (CHX, 30 μg/ml) and puromycin (30 μg/ml) for 4 hours before treatment with heat shock (at 43°C) or MG132 (20 μM). Cells were harvested at the indicated time points after treatment, and 3 μg of each crude lysate was analyzed by Western blotting with antibodies against SUMO-2/3 (upper panels) or ubiquitin (lower panels).

Ubiquitination associated with SUMO-2 conjugates is temporally and qualitatively distinct from the ubiquitination of total proteins during proteasome inhibition

Like others (16), we found that SUMO-SUMO branched conjugates increased in abundance during proteasome inhibition, with the increase being largely proportional to the duration of inhibition (Fig. 6A, right panel). There appeared to be substantial differences in the abundances of ubiquitin and ubiquitin polymers between unpurified extracts and samples from TAP–SUMO-2 purifications. In crude samples, the amounts of ubiquitin were modestly increased at 2 and 7 hours (Fig. 6A, left panel), as were branched conjugates of ubiquitin through Lys48 (Fig. 6B, upper panel), whereas the abundances of Lys63-linked ubiquitin-ubiquitin branches were only modestly affected by proteasome inhibition (Fig. 6B, upper panel). These findings are consistent with previous work (31) that showed that Lys63-linked ubiquitin chains are largely unaffected by proteasome inhibition; however, this pattern was not mirrored in the TAP–SUMO-2 purified preparations from the same cells. Whereas extended exposure to MG132 caused an increase in the amount of ubiquitin associated with TAP–SUMO-2, shorter exposure resulted in the reduction of ubiquitin amounts (Fig. 6A, right panel). This pattern was also found for ubiquitin-ubiquitin conjugates (Fig. 6B, lower panel) and ubiquitin–SUMO-2/3 branched forms (Fig. 6C), the abundances of which were reduced after 2 hours of treatment with MG132 and increased after longer exposure. Although we observed the expected increase in Lys11- and Lys48-linked ubiquitin polymers derived from the TAP–SUMO-2 preparation (Fig. 6B, lower panel), we also observed a 10-fold increase in Lys63-linked ubiquitin polymers (Fig. 6B, lower panel). This suggested that, whereas MG132 triggered a time-dependent increase in overall ubiquitination, the ubiquitin associated with SUMO-2 and possibly its conjugates underwent an early period of deconjugation before a later increase in conjugation and that the Lys63-linked ubiquitin-ubiquitin chain was the most substantially increased in abundance, despite it not being thought to trigger proteasomal degradation (31, 32).

Fig. 6

The ubiquitination of SUMO-2 substrates is temporally and qualitatively distinct from the global ubiquitination that occurs after proteasome inhibition. (A) Changes in the abundances of total proteins and SUMO-SUMO branched peptides in the crude (left) and TAP-purified (right) samples after MG132 treatment. (B) Changes to GG-linked adducts on ubiquitin in crude (upper) and TAP–SUMO-2–purified (lower) extracts after MG132 treatment. (C) Changes to GG-linked adducts on SUMO-2(/3) from TAP–SUMO-2–purified extracts after exposure to MG132. The bold region of the amino acid sequence SUMO-2 is identical to that of SUMO-3. SILAC ratio counts for each protein or peptide are shown in parentheses. SUMO-SUMO branched conjugates, ubiquitin-ubiquitin linkages through Lys6 and Lys11, and ubiquitin-SUMO linkages through Lys11, Lys20, Lys32, Lys41, and Lys44 were not detected in crude extracts. Slice-by-slice details of these proteins and peptides can be found in figs. S11 and S12. Abbreviations for the amino acids are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; and Y, Tyr.

SUMO conjugates accumulate in insoluble inclusions upon proteasome inhibition

Lys63-linked polyubiquitin chains play a role in the metabolism of misfolded proteins (33). If SUMO-2/3 conjugation responds to the accumulation of misfolded proteins, and conjugates accumulate on SUMO-2 substrates, it is possible that a subpopulation of SUMO molecules is directly targeting misfolded or aggregating proteins for conjugation. To determine whether SUMO accumulated on aggregating proteins and to put any such accumulation into context with other proteins, we conducted a quantitative proteomic experiment to monitor the change in abundance of detergent-insoluble proteins over a 7-hour time course of MG132 exposure (Fig. 7A). Proteins insoluble to a buffer containing 1.2% deoxycholate and 1.2% Triton X-100 (see Supplementary Materials for details) represented ~2 to 5% by mass of the total cellular protein and were detected and quantified by liquid chromatography–mass spectrometry (LC-MS) coupled with Andromeda-based MaxQuant (MaxQuant version; (34). We detected and measured 538 proteins in all three mixes (Supplementary File 5), which enabled us to calculate the relative abundance changes for the entire 7-hour reaction. Data are presented as a tsMAP for the 2- and 7-hour time points (Fig. 7B) and show a progressive increase in insolubility for most proteins. The validity of this approach was confirmed by the measured enrichment of molecular chaperones and chaperone regulators, such as Bcl-2–associated athanogene 2 (BAG2), and a range of heat shock proteins (Fig. 7B, insets). The proteins with the greatest increase in abundance in insoluble preparations after 7 hours of MG132 treatment were SUMO-2/3 and ubiquitin (Fig. 7B); surprisingly, SUMO-1 also accumulated in these preparations. All these ubiquitin-like modifiers and ubiquitin were characterized by a sigmoidal change in accumulation over time, whereas other proteins largely accumulated in a linear manner (Fig. 7B, insets), suggesting an uncommon means of accumulation, possibly indicative of polymerization. Western blotting analysis of insoluble extracts confirmed the MS data analysis and showed an apparently broad spectrum of covalent conjugates for all of these modifiers accumulating over time (Fig. 7C). This confirms that SUMOs accumulate in detergent-insoluble protein inclusions during proteasome inhibition in a manner analogous to that of ubiquitin, which is an important regulator of insoluble protein metabolism.

Fig. 7

SUMO conjugates accumulate in insoluble inclusions upon proteasome inhibition. (A) Insoluble protein preparations were made from three batches of HeLa cells grown in SILAC conditions and treated with MG132 (20 μM) as indicated (left). Proteins were identified and quantified from proteins fractionated by SDS-PAGE (right). Data are representative of a single experiment with a median of 10 and minimum of 2 SILAC ratio counts per protein. Ratio counts for each protein can be found in Supplementary File 5. (B) Changes in the abundances of 538 proteins after 2 or 7 hours of treatment with MG132 are shown as a scatter plot. Insets show data for the entire time course for selected proteins. Insets show the time in hours on the abscissa and the fold change along the ordinate. (C) Western blotting analysis of ubiquitin, SUMO-2/3, and SUMO-1 in insoluble protein preparations from HeLa cells treated with MG132 (20 μM) for the indicated times. Western blotting data are consistent with similar non-SILAC experiments.


With the discovery of SUMO-targeted ubiquitin ligases (STUbLs), which provide a direct link between SUMO and the UPS (17), a previously uncharacterized role for SUMO in the control of protein stability has been established. Consistent with this, proteasome inhibition resulted in the accumulation of SUMO conjugates (9, 15, 18), although the kinetics of the accumulation of SUMO conjugates were slower than that of ubiquitin conjugates (Fig. 1) (15, 16). To better understand the relationship between SUMO and ubiquitin, we used a quantitative proteomics approach to monitor the changes in SUMO-2 modification in cultured cells after 2 and 7 hours of treatment with MG132 (Figs. 2 and 3). Although these results are consistent with those of a previous smaller study (fig. S13) (16), we showed that changes to the SUMO subproteome upon proteasome inhibition were not explicable by the stabilization of SUMO conjugates destined for UPS degradation alone. This was facilitated by identifying a far greater number of putative conjugates than in the previous work (16), studying the changes in SUMO-2/3 conjugation during short- and long-term exposures to MG132, and comparing findings across proteomics studies.

Substrate-specific changes in the SUMO-2 proteome in response to MG132 correlated well with the response to heat shock (Fig. 4), suggesting a common trigger and a common outcome. Because MG132 inhibits the degradation of newly synthesized, misfolded proteins (29), and MG132-triggered SUMOylation is dependent on protein synthesis (Fig. 5), the simplest explanation is that in both cases SUMO is responding to an accumulation of unfolded or misfolded proteins. Bearing in mind the functional relevance of the SUMO-2 substrates to the cellular response to heat shock (fig. S8) (19), we find it reasonable to conclude that for the most part, SUMO-2/3 is acting in a signal transduction role as a general response to the accumulation of unfolded or misfolded proteins. In so doing, it is likely that SUMO is contributing to the alteration of the cellular environment to reduce the cytotoxicity of aggregated and dysfunctional proteins.

It appears paradoxical that the activity of a largely nuclear modifier such as SUMO (9) can depend on protein synthesis in the cytoplasm; however, we found that free, unconjugated SUMO was solely cytoplasmic in HeLa cells, whereas covalently conjugated SUMO was almost exclusively nuclear in localization (fig. S15). This suggests that the regulation of free SUMO partitioning in the cytoplasm may be a step on the pathway from the misfolded protein signal to the SUMOylation response and provides a potential mechanism for the transduction of the signal across the nuclear membrane.

The extent of the SUMOylation response to MG132 was substrate-specific. Although for about half of the substrates a progressive increase in conjugation to SUMO-2 was seen during MG132 exposure, many proteins underwent cycles of deconjugation followed by reconjugation, or vice versa, when comparing short-term to long-term periods of proteasome inhibition (Fig. 3). Similarly, ubiquitin seemed to cycle on and off some proteins during proteasomal inhibition. Considering SUMO and possibly its substrates, these proteins underwent an early phase of net deubiquitination followed by a later increase in conjugation (Fig. 6), even when apparently all of the ubiquitin was incorporated into conjugates (Fig. 1). Both of these observations can be explained by cellular mechanisms that prioritize the modifiers when different demands are placed on the finite pools of the proteins. For example, at early time points, there is likely to be a high demand for ubiquitin to modify misfolded proteins (29), whereas later, the priority may be shifted toward other systems and SUMO conjugates (Fig. 6 and fig. S12). Thus, it appears that ubiquitin and SUMOs are prioritized among different roles during a proteotoxic response.

Unexpectedly, although polyubiquitin chains that formed through Lys6, Lys11, and Lys48 (the canonical UPS degradation signal) increased in abundance after 7 hours of MG132 treatment in the SUMO-2–purified sample (1.3-, 2.4-, and 3-fold, respectively; Fig. 6B), Lys63-linked polymers were the most increased in abundance (10-fold). This occurred despite the modest increase (1.6-fold) in the abundance of Lys63-linked chains in whole-cell extracts over the same time period (Fig. 6B) and was surprising, given previous work that suggested that Lys63-linked polyubiquitin chains are not signals for proteasomal degradation (31, 32). This implies that at least a subpopulation of SUMO-2 substrates bucks the global trend and accumulates forms of polyubiquitin chains not thought to target proteins for degradation by the proteasome (32). Given the well-established role of SUMO in the regulation of protein activity, it is possible that the Lys63-linked polyubiquitin chains associated with SUMO-2 substrates are derived from proteins that function within signaling or DNA repair pathways (3537). Additionally, consistent with recent findings (38), it is possible that for some of these proteins, SUMO-2/3 is involved in indirectly targeting to the proteasome, although involvement in the regulation of degradation pathways other than the UPS cannot be excluded: Growing evidence indicates that when the production of misfolded proteins exceeds the capacity of the chaperone and UPS systems, misfolded proteins are targeted for degradation by the aggresome-autophagy pathway (39, 40). Furthermore, Lys63-linked polyubiquitin chains synthesized by the Parkinson’s disease–associated ubiquitin E3 ligase parkin are required for targeting misfolded proteins to the aggresome (41). These possibilities are not mutually exclusive, but it is noteworthy that Lys63-linked ubiquitin polymers did not increase in abundance in SUMO-2 purifications after heat shock (fig. S8) (19) when proteasome function was not inhibited.

In support of this theory, SUMO-2/3 was the protein that most significantly increased in abundance in insoluble inclusions in cells treated with MG132 (Fig. 7). SUMO-1 was also increased in abundance in these fractions, implicating all SUMO paralogs in this phenomenon; however, whether SUMOs play a causal or consequential role remains to be established. Together, these findings are interesting, because SUMOs, ubiquitin, and Lys63-linked ubiquitin are diagnostic markers for nuclear inclusions that are characteristic of many neurodegenerative disorders (4247), suggesting that links between the SUMO and the ubiquitin systems may be important factors in the pathological or symptomatic accumulation of insoluble proteins. In summary, our work highlights the need for careful interpretation of proteasome inhibition studies involving SUMOs because of the roles of the modifiers not only in SUMO-targeted degradation through the UPS but also in a range of direct and indirect responses to the formation of protein aggregates. A deeper understanding of the regulation and roles of SUMO under these conditions is likely to be of broad biological importance.

Materials and Methods

Cell culture and protein labeling

Quantitative proteomics experiments were performed with the SILAC technique broadly as described previously (48). Cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) with l-lysine and l-arginine replaced with stable isotope (SILAC) forms (Cambridge Isotope Laboratories) depending on the treatment (see below). SILAC DMEM was supplemented with 10% dialyzed fetal calf serum (FCS). The experiment described in Fig. 2 used three SILAC conditions: (i) mock, DMSO-treated TAP–SUMO-2 HeLa cells were grown in SILAC DMEM containing isotopically “normal” amino acids (light conditions, K0R0); (ii) TAP–SUMO-2 HeLa cells exposed to MG132 (20 μM) for 2 hours were grown in SILAC DMEM containing 4,4,5,5-D4 lysine and 13C6 arginine (medium conditions, K4R6); and (iii) TAP–SUMO-2 HeLa cells exposed to MG132 (20 μM) for 7 hours were grown in SILAC DMEM containing 13C615N2 lysine and 13C615N4 arginine (heavy conditions, K8R10). Cells were grown in 150-mm-diameter dishes, and 50 dishes were used for each SILAC condition. DMSO or MG132 in DMSO was added by first diluting in culture medium to 20 times the final concentration before vortex mixing and adding to dishes. For Western blotting analysis of cells treated with cycloheximide and puromycin, MG132, or l-canavanine, HeLa cells were treated as described in the appropriate figure legends before being washed twice in phosphate-buffered saline (PBS), lysed in Laemmli’s sample buffer, resolved by SDS-PAGE, transferred to blots, and analyzed with the appropriate specific antibodies.

Sample preparation for MS analysis

To analyze the response to MG132 of the nonpurified subproteome, as well as of the TAP–SUMO-2 conjugate subproteome, two preparations were made from the same cells. When cells were at ~90% confluency, 50 dishes for each SILAC condition were washed twice with PBS and cells were lysed in 1 ml per dish of 50 mM tris-HCl (pH 8.0), 2% SDS, 10 mM iodoacetamide, and 1 mM EDTA, with complete protease inhibitor cocktail (Roche). Lysates were sonicated, protein concentrations were determined by BCA (bicinchoninic acid) assay (Pierce), and equivalent amounts of protein from each lysate were mixed. The pooled lysate was centrifuged at 180,000g at 20°C for 1 hour. The cloudy upper phase was removed, clarified through a 0.22-μm filter, and added to the remaining supernatant to give a final sample volume of ~150 ml, containing 550 mg of protein. For the crude sample, 75 μg of this mixture was diluted 1:1 with 2 × Laemmli’s sample buffer, and the total sample was run over three lanes of a NuPAGE bis-tris 10% polyacrylamide gel run in Mops buffer (Invitrogen). The remaining sample was purified as described previously (21), and the resultant sample (20 to 30 μg of protein) was fractionated by the same gel system. All protein-containing lanes were sliced into nine sections. Protein bands were subjected to in-gel digestion with trypsin (Promega) essentially as described previously (22). The resulting peptide mixtures were desalted and concentrated with self-made reversed-phase C18 stop-and-go extraction (STAGE) tips (49). Dried peptide samples were resuspended in 32 μl of 1% formic acid or 0.5% acetic acid.

Quantitative MS

MS analysis of the nonpurified (crude) sample was performed by nanoscale liquid chromatography–tandem mass spectrometry (LC-MS/MS) with a linear ion trap–FT-ICR hybrid mass spectrometer (LTQ FT-ICR Ultra, Thermo Fisher Scientific), and the TAP-purified (purified) sample was analyzed by LC-MS/MS with a linear ion trap–orbitrap hybrid mass spectrometer (LTQ Orbitrap, Thermo Fisher Scientific). Both devices were equipped with a nanoelectrospray ion source (Proxeon Biosystems) coupled to a Proxeon Easy nLC HPLC (high-performance liquid chromatography) system (Proxeon Biosystems) and operated in the data-dependent mode, switching automatically between MS and MS/MS acquisition. Peptides were injected into an in-house made 75-μm reversed-phase C18 column with a flow of 500 nl/min and eluted with a linear gradient of 98% solvent A (0.5% acetic acid in H2O) to 50% solvent B (80% acetonitrile and 0.5% acetic acid in H2O) and a flow of 250 nl/min. For LTQ FT analysis, the five most intense ions from the FT survey scan [mass/charge ratio (m/z) 300 to 1650; target value 5,000,000; R = 100,000 at m/z 400] were fragmented through collision-induced dissociation (CID) and acquired in the LTQ (target value of 20,000). Samples were analyzed three times to give 27 raw data files. For Orbitrap analyses, full-scan spectra (m/z 300 to 1650) were acquired in the orbitrap with resolution R = 60,000 at m/z 400 (after accumulation to a target value of 1,000,000). The five most intense ions were fragmented by CID and recorded in the linear ion trap (target value of 5000) based on a lower-resolution preview scan (R = 15,000). The orbitrap analyzer was operated with the “lock mass” option to improve the mass accuracy. Samples were analyzed four times with different gradient times to provide 36 raw data files. All of the raw data files, including those from supplementary experiments, are available upon request.

Quantitative data analysis

For the experiment described in Fig. 2 and the data shown in Fig. 4, MS raw data files were processed with MaxQuant (version (23) and with the Mascot search engine (Matrix Science, version 2.1.04). Further details of the algorithms and principles applied in the MaxQuant package are described by Cox and Mann (23). Enzyme specificity was set to trypsin/P. Cysteine carbamidomethylation was selected as a fixed modification and methionine oxidation and protein N-acetylation, and Gly-Gly adducts of lysine were searched for as variable modifications. The data were searched against a target/decoy human international protein index (IPI) database (version 3.24) (50). Initial maximal allowed mass deviation was set to 7 parts per million (ppm) for peptide masses and 0.5 dalton for MS/MS peaks. The maximum peptide length was set to six amino acid residues, and a maximum of three missed cleavages and three labeled amino acid residues was allowed. MS/MS spectra determined to be SILAC-labeled in the MaxQuant pre-search were searched with the fixed modifications Arg6 and Lys4 or Arg10 and Lys8; for spectra with a SILAC state not determinable before the database search. Arg4, Arg10, Lys6, and Lys8 were taken as variable modifications. A 1% false discovery rate (FDR) was required at both the protein level and the peptide level. In addition to the FDR threshold, proteins were considered identified if they had at least one unique peptide. Errors in quantitation were minimized by the requirement of at least two quantified SILAC triplets for each protein and by reporting the median, which is a robust estimate of protein ratio. Ratio counts are reported by MaxQuant and are available for scrutiny in the Supplementary Files. The SUMO-2 branched peptides were quantified by modification of the MASCOT database with linear fusions of the SUMO C terminus with the SUMO-2 target peptide, allowing for a missed cleavage at Lys11, as described in principle previously (51). To create a single spreadsheet containing data that provide SILAC ratios for the total change in protein abundance and for changes in each slice, two MaxQuant analyses were made, which were then subsequently merged into a single file. Both analyses used the same results of the Quant program and MASCOT. The first run defined all of the raw file entries as either crude or purified in the “Experiment” column of the “experimental design template” file. The results of this analysis provided overall SILAC ratios for each protein in both preparations. For the second analysis in the “Experiment” column of the “experimental design template” file, each raw file was designated as information pertaining to both the source of the sample (that is, crude or purified) and the slice of the gel from which it originated (for example, “crude 5” or “pure 8”). This analysis provided SILAC ratios for every slice of the gel for each crude or purified sample. These two data sets were merged into a single spreadsheet with protein identifiers to synchronize entries (see Supplementary File 1).

Supplementary Materials

Materials and Methods

Fig. S1. TAP–SUMO-2 conjugation response to MG132 treatment under SILAC conditions.

Fig. S2. Filtering TAP–SUMO-2 protein identifications and normalizing with SILAC ratio data from crude cell lysates.

Fig. S3. Known SUMO substrates are enriched in the list of putative SUMO-2 substrates in comparison with those proteins rejected from the analysis.

Fig. S4. “Putative substrates” run in SDS-PAGE gels at molecular weights consistent with posttranslational modification, whereas “internal rejects” do not.

Fig. S5. Changes in protein ratios during MG132 treatment are due to altered SUMOylation and not changes in overall protein concentrations.

Fig. S6. Only a small proportion of the total cellular pool of most substrates is modified by SUMO-2 and responds to MG132 with altered SUMOylation.

Fig. S7. Lactacystin and MG132 trigger similar changes in SUMO-2 conjugation in HeLa cells.

Fig. S8. Comparisons of the functions of SUMO-2 substrates identified here and in a previous study.

Fig. S9. Canavanine induces SUMO-2/3 conjugation and turnover.

Fig. S10. Comparison between the effects of heat shock and MG132 on proteins involved in SUMO and ubiquitin metabolism purified by TAP–SUMO-2 from HeLa cells.

Fig. S11. Slice-by-slice analysis of data shown in Fig. 6A.

Fig. S12. Slice-by-slice analysis of data shown in Fig. 6, B and C.

Fig. S13. Comparison of the quantitative data between proteins common to this study and that of Schimmel et al.

Fig. S14. A working model for the response of SUMO and ubiquitin to proteasome inhibition.

Fig. S15. The SUMO conjugation signal in response to MG132 transfers from the cytoplasm to the nucleus.

Table S1. Frequency analysis of SUMO conjugation consensus sequences from proteins identified in this study compared with those of previous studies.


Files 1 to 6.

References and Notes

  1. Acknowledgments: We thank C. Cole (University of Dundee, UK) for providing the tools for SUMO consensus site analysis and R. Gourlay and D. Campbell (University of Dundee, UK) for MS data acquisition for analysis shown in fig. S7. Funding: This work was funded by Cancer Research UK. I.M. is a Sir Henry Wellcome Postdoctoral Fellow. Author contributions: M.H.T. designed and conducted the experiments and analyses, processed data, interpreted results, and wrote the manuscript. I.M. acquired MS data and advised over quantitative proteomic analysis. M.M. provided advice and support for MS data acquisition. R.T.H. conceived the project and interpreted the results. I.M., M.M., and R.T.H. edited the manuscript. Competing interests: The authors declare that they have no competing financial interests.
View Abstract

Stay Connected to Science Signaling

Navigate This Article