Acetylation of SUMO2 at lysine 11 favors the formation of non‐canonical SUMO chains

Abstract

Post‐translational modifications by ubiquitin‐related SUMO modifiers regulate cellular signaling networks and protein homeostasis. While SUMO1 is mainly conjugated to proteins as a monomer, SUMO2/3 can form polymeric chains. Poly‐SUMOylation is best understood in the SUMO‐targeted ubiquitin ligase (StUbL) pathway, where chains prime proteins for subsequent ubiquitylation by StUbLs. SUMO chains typically form in response to genotoxic or proteotoxic stress and are preferentially linked via lysine 11 of SUMO2/3. Here, we report that K11 of SUMO2/3 undergoes reversible acetylation with SIRT1 being the K11 deacetylase. In a purified in vitro system, acetylation of SUMO2/3 impairs chain formation and restricts chain length. In a cellular context, however, K11 acetyl‐mimicking SUMO2 does not affect the StUbL pathway, indicating that in cells non‐canonical chains are more prevalent. MS‐based SUMO proteomics indeed identified non‐canonical chain types under basal and stress conditions. Importantly, mimicking K11 acetylation alters chain architecture by favoring K5‐ and K35‐linked chains, while inhibiting K7 and K21 linkages. These data provide insight into SUMO chain signaling and point to a role of K11 acetylation as a modulator of SUMO2/3 chains.

Introduction

Post‐translational modification with the ubiquitin‐related modifier SUMO is a rapid and reversible way to control protein functions 1. In mammals, three SUMO paralogs (SUMO1, 2, 3) can act as protein modifiers. SUMO2 and SUMO3—denominated SUMO2/3 below—are highly similar to each other and share about 45% sequence identity with SUMO1. All SUMO forms can be conjugated to ε‐amino groups of lysine residues by a multi‐step enzymatic process, involving the heterodimeric E1 enzyme, the E2 enzyme Ubc9, and in some cases E3 SUMO ligases 2, 3. In general, the attachment of SUMO modulates the dynamics of protein networks by either inhibiting or promoting protein–protein interactions. These SUMO‐dependent interactions are typically mediated via the recruitment of binding partners that harbor specific SUMO recognition motifs termed SIMs 4. Similar to what is known of the ubiquitin system, SUMO can be conjugated as a monomer on single or multiple sites on a target protein. Moreover, like ubiquitin, SUMO is able to form polymeric chains through the attachment of one SUMO molecule to internal lysine residues of another SUMO moiety 5, 6, 7. Poly‐ubiquitylation is best known for its role in proteasomal protein degradation, but during the last decade, a striking linkage‐type specificity in both proteolytic and non‐proteolytic ubiquitin‐dependent signaling processes has been uncovered 8, 9. This versatility of ubiquitin signaling typically relies on the formation of distinct types of lysine‐linked ubiquitin chains and the recognition of these chains by specific binding modules 10. In contrast to the ubiquitin system, our understanding of SUMO chain formation and function is still very limited. It is well‐established that SUMO chains, in particular SUMO2/3 chains, are induced in response to cellular stress, such as redox, heat, or osmotic stress, pointing to a critical role of chains in coping with these situations 11. In line with this idea, expression of a lysine‐less SUMO variant in the yeast Saccharomyces cerevisiae causes sensitivity to replicative or proteotoxic stress 12. The current view in mammalian cells is that SUMO chains preferentially assemble via lysine residue 11 of SUMO2 and 3, which is located in the flexible N‐terminus of the protein 13. The preference of K11 for the formation of SUMO‐SUMO linkages has been explained by the fact that K11 resides in a so‐called SUMOylation consensus sequence, which is defined by a KxE motif 13. So far SUMO chains have been mainly studied in one particular cellular process termed the SUMO‐targeted ubiquitin ligase (StUbL) pathway. In this pathway, SUMO2/3 polymers are recognized by a distinct class of ubiquitin E3 ligases, containing tandem SIM motifs 14. The prototypical mammalian StUbL is RNF4, which is recruited to multiple‐mono‐ or poly‐SUMOylated proteins to trigger their proteolytic or non‐proteolytic ubiquitylation. The best‐studied targets of RNF4 are the promyelocytic leukemia protein, PML, and the oncogenic fusion protein, PML‐RARα. In response to the anti‐leukemogenic drug arsenic, both PML and PML‐RARα are hyper‐SUMOylated, resulting in their RNF4‐mediated ubiquitylation and subsequent proteolysis 15, 16. Poly‐SUMO chains and the StUbL pathway have also been linked to proteotoxic stress response and were shown to play an important role in the DNA damage response (DDR) pathway 17, 18, 19. In spite of these insights, our understanding of SUMO chain function, architecture, and recognition is still incomplete. One major question is whether, in analogy to the ubiquitin system, different types of SUMO chains are formed in response to distinct stimuli. Accordingly, it remains unclear as to how distinct chains are recognized and how chain architecture is regulated. In high‐throughput as well as directed mass spectrometry experiments, we and others previously identified distinct acetylated lysine residues in SUMO paralogs, including K11 in SUMO2/3 20, 21, 22. Here, we provide evidence that K11 acetylation is controlled by SIRT1 and affects SUMO chain signaling. In a purified in vitro SUMOylation system, K11 acetylation largely impairs chain formation or restricts chain length in a SIRT1‐reversible manner. In a cellular setting, however, where we found non‐canonical SUMO chains to be more prevalent, expression of an acetyl‐K11 mimicking SUMO2 variant alters SUMO chain topology. In particular, under heat stress acetyl‐K11 impairs formation of K7‐ and K21‐linked chains and favors linkages via K5 and K35.

Results and Discussion

Acetylation of SUMO2/3 at K11 limits the formation of SUMO chains

In previous work, we had focused on the characterization of acetyl‐SUMO sites that cluster around the SIM‐binding region of SUMO family members 23. In line with data from high‐throughput acetylome studies, our directed mass spectrometry (MS)‐based analysis of SUMO2 also revealed an acetylation of lysine residue 11 20, 21, 22. K11 resides in the flexible N‐terminal domain of SUMO2 and is conserved in SUMO3 (Fig 1A and B). Since K11 is described as the preferred site for poly‐SUMO2/3 chain conjugation, we hypothesized that an acetylation on SUMO2/3 at K11 could affect chain formation. The formation of free poly‐SUMO2/3 chains can be readily detected in an in vitro modification system, where recombinantly expressed SUMO is incubated with SUMO E1 and E2 enzymes in the presence of ATP 13, 24. To follow the idea that K11‐acetyl SUMO2/3 can interfere with this process, we set up an in vitro SUMOylation reaction using either unmodified SUMO2 or SUMO2^K11Ac as the respective modifier. SUMO2^K11Ac was produced in E. coli that harbor an orthogonal N^ε‐acetyl‐lysine‐tRNA synthetase/tRNA_CUA pair. We have previously used this system for the incorporation of specific acetyl‐lysine residues in SUMO paralogs and accordingly used it for generation of K11‐acetyl‐SUMO2 23. Proper acetylation of SUMO2 at K11 was validated by MS (Fig EV1A). Additionally, immunoblotting with a highly specific, custom‐made K11 acetyl‐SUMO2 antibody confirms the presence of acetyl‐K11 in recombinant SUMO2 (Fig EV1B–D). We next investigated SUMO2 chain formation with either unmodified SUMO2 or K11 acetylated SUMO2 by in vitro SUMOylation assays. A SUMO2^K11Q variant was included in the experiments, because a K‐to‐Q exchange mimics N^ε‐acetyl‐lysine residues 25. Wild‐type SUMO2 is entirely converted to SUMO conjugates and forms polymeric chains of different length in the SUMOylation assay (Fig 1C). Separation of the reaction products by SDS–PAGE and immunoblotting with an α‐SUMO2/3 antibody detects almost no more free SUMO2 (~17 kDa), but reveals SUMO2 polymers corresponding to 2xSUMO2 (~35 kDa), 3xSUMO2 (~48 kDa), and higher molecular weight conjugates. Since the reaction contains an excess of E1 over Ubc9, some of the higher molecular weight conjugates correspond to mono, multiple‐mono and poly‐SUMOylated AOS1‐UBA2 (dimeric E1 SUMO activating enzyme) as has been described previously 26. When compared to unmodified SUMO2, acetyl‐K11 SUMO2 as well as the SUMO2^K11Q variant showed a strongly reduced ability to form, poly‐SUMO chains. This is reflected by the presence of remaining free SUMO2 (~17 kDa) and decreased intensity of bands migrating at ~35 kDa (2xSUMO), ~48 kDa (3xSUMO), and higher molecular weight SUMO2 chains (Fig 1C). We next tested whether addition of the E3 SUMO ligase ZNF451, which triggers chain formation, can also act on SUMO2^K11Ac 27. To monitor the effect of ZNF451, assays were performed in the presence of limited amounts of Ubc9. Consistent with published data, an N‐terminal fragment of ZNF451 induces the formation of wild‐type SUMO2 polymers (2xSUMO2, 3xSUMO2) 27. However, ZNF451 does not stimulate polymerization of SUMO2^K11Ac (Fig 1D). Altogether, these data support the idea of an inhibitory function of SUMO2^K11Ac and SUMO2^K11Q on chain formation. Next, we wanted to address how acetylation of SUMO affects chains, when both modified and unmodified SUMO2 are competing with each other. Therefore, we performed an in vitro modification assay in the presence of unmodified SUMO2 and increasing concentrations of SUMO2^K11Ac. Ratios ranged from 3:1 (unmodified/K11Ac) to 1:2 (unmodified/K11Ac). With increasing concentrations of SUMO2^K11Ac, we observed a decrease in chain length as monitored by α‐SUMO2/3 immunoblotting (Fig 1E). Notably, immunoblotting with the α‐SUMO2K11Ac antibody shows increasing signals for free SUMO2^K11Ac, indicated by a band migrating at ~17 kDa (1xSUMO2) and increasing signals for mono‐SUMOylated E1 at ~100 kDa. These data imply a potential function of SUMO2^K11Ac in terminating SUMO chains, thereby restricting their length. A similar role as chain restrictor or terminator has been proposed for SUMO1, which lacks the equivalent of K11 in SUMO2/3 5. In a complementary approach, we investigated whether K11 acetylation of SUMO2 can also modulate chain formation on a target protein. As a model, we selected the well‐established SUMO target p53, which harbors a single SUMO conjugation site at K386 28, 29, 30. To investigate SUMO modification of p53, the protein was generated by in vitro transcription/translation, using a rabbit reticulocyte system, and in vitro SUMOylation was performed as described above. Notably, Western blot analysis with an α‐p53 antibody showed SUMO modification of p53 even without the addition of ATP due to the presence of ATP in the rabbit reticulocyte lysate. Regardless of that, upon addition of ATP, mono‐ as well as poly‐modified p53 can be detected with unmodified wild‐type SUMO2, as indicated by signals at ~75 kDa (p53‐SUMO2), ~80 kDa (p53‐2xSUMO2), and ~100 kDa (p53‐3xSUMO) (Fig 1F). Importantly, when using SUMO2^K11Ac the bands at ~80 kDa (p53‐2xSUMO) and ~100 kDa (p53‐3xSUMO) were strongly reduced, whereas p53 mono‐SUMOylation (~75 kDa) was not affected. This demonstrates a specific reduction in poly‐SUMO chains on p53, but no change in mono‐modification. The spurious detection of SUMO polymers in the K11Ac‐SUMO2 containing reactions is likely due to the presence of wild‐type SUMO2 in the reticulocyte lysate. In summary, the data demonstrate that acetylation of K11 strongly reduces the formation of free poly‐SUMO chains and diminishes poly‐SUMOylation of substrates in a reconstituted in vitro SUMOylation system. In addition, the acetylation of SUMO2K11 also limits chain length, when unmodified SUMO2 and SUMO2^K11Ac compete for conjugation. These results are in line with the current view that K11‐linked SUMO chains are the most prevalent species formed in vitro.

Figure 1. Acetylation of SUMO2/3 at K11 limits the formation of SUMO chains.

1. Structure of SUMO2. Lysine residues within the flexible N‐terminus (K5, K7, K11) are marked in magenta, lysine residue 21 is colored in teal and lysine residues involved in SUMO/SIM interactions (K33, K35, K42, K45) are colored in green (PDB: 2N1W). The red circle indicates K11. Ac. indicates the acetylation at K11.
2. Comparison of the amino acid sequences for SUMO1/2 and 3 and the yeast variant Smt3. Lysine residues are marked in blue. SUMO consensus motifs are marked in red. SUMO2/3 and Smt3 display a conserved consensus motif, whereas SUMO1 shows an inverted consensus motif.
3. Substrate‐independent in vitro SUMOylation assay with SUMO2^WT, SUMO2^K11Ac, or SUMO2^K11Q was performed and analyzed by immunoblotting with the indicated antibodies.
4. Substrate‐independent in vitro SUMOylation assay using either SUMO2^WT or SUMO2^K11Ac in the presence of increasing amounts of ZNF451‐N. Reaction products were analyzed by immunoblotting with the indicated antibodies. Note that when compared to (C), assays were done at reduced concentrations of E1 and E2 (see Materials and Methods).
5. In vitro SUMOylation assay using a fixed amount of SUMO2^WT (220 ng) with increasing concentrations of SUMO2^K11Ac. Ratios ranged from 3:1 (unmodified/K11Ac) to 1:2 (unmodified/K11Ac)
6. In vitro SUMOylation assay, using SUMO2^WT or SUMO2^K11Ac, with p53 as a substrate. p53 was generated by in vitro translation/transcription in a rabbit reticulocyte lysate system. SUMO chain formation on p53 was analyzed by immunoblotting with α‐p53 antibodies.

Figure EV1. Generation of recombinant SUMO2 ^K11Ac and characterization of a SUMO2^K11Ac‐specific antibody.

1. Mass spectrum for acetylated SUMO2K11. In‐gel tryptic digest of recombinant SUMO2^K11Ac, followed by MS analysis (Orbitrap Elite). Two‐dimensional representation of the relative abundance (right y‐axis) and intensity (left y‐axis) as well as the mass‐to‐charge ratio (m/z, x‐axis) of the depicted SUMO2K11 acetylated peptide. Peptide fragment ions are indicated by y‐ (red) and b‐ (blue). Acetylation site (green).
2. Workflow of antibody production. The antibody was generated by the company Immunoglobe. The peptides were synthesized, and the animal was immunized three times. Subsequently, the serum was subjected to affinity purification via a two‐column system.
3. Antibody test with purified SUMO2^K11Ac and SUMO2^WT and Western blot analysis with α‐SUMO2K11Ac and α‐SUMO2/3.
4. For antibody characterization, SUMO2^K11Ac was separated by SDS–PAGE (5‐200 ng as indicated) and immunoblotted with α‐SUMO2K11Ac.
5. Antibody test with purified SUMO2^K33Ac, SUMO2^K35Ac, and SUMO2^WT and Western blot analysis with α‐SUMO2K33Ac, SUMO2K35Ac, and α‐SUMO2/3.

SIRT1 catalyzes the deacetylation of SUMO2^K11Ac

Our initial findings on a potential role of K11 acetylation in the control of SUMO chain formation raised the question how acetylation of SUMO2/3K11 is controlled. Interestingly, global quantitative acetylomic profiling points to an involvement of the deacetylase SIRT1 in controlling the acetylation status of SUMO2K11 31, 32. We, therefore, set out to experimentally validate this idea and first tested for potential deacetylase activity of SIRT1 on K11‐acetyl SUMO2 in an in vitro deacetylation assay. To this end, recombinantly expressed SIRT1 was incubated with either unmodified SUMO2 or SUMO2^K11Ac for 2 h at 37°C with or without addition of the SIRT1 cofactor NAD⁺. Subsequently, samples were separated by SDS–PAGE and the acetylation status of SUMO2 was monitored by immunoblotting with the α‐K11‐acetyl SUMO2 antibody. In control samples without SIRT1 as well as samples lacking the cofactor NAD⁺, acetyl‐K11 SUMO2 was detected (Fig 2A). Importantly, however, a complete deacetylation of SUMO2^K11 was observed in the presence of SIRT1 and NAD⁺. At the same time, the total levels of SUMO2, as monitored by immunoblotting with α‐SUMO2/3 antibodies, remained unchanged. We next investigated whether SIRT1 exerts specificity toward SUMO2^K11Ac, or targets other known acetylation sites in SUMO2. We, therefore, performed in vitro deacetylation assays on SUMO2^K33Ac and SUMO2^K35Ac, which were expressed in E. coli, using the method described above. Acetylation levels of SUMO2^K33 and ‐K35 were analyzed by immunoblotting with specific antibodies for the respective acetylation sites (Fig EV1E). Importantly, SIRT1 catalyzes deacetylation of SUMO2^K11, but does not affect acetylation levels of SUMO2^K33 and SUMO2^K35 (Fig 2B). To further test whether SUMO2^K11Ac is also targeted by other SIRT family members, we performed in vitro deacetylation assays with active SIRT2 (Fig EV2). Immunoblotting analysis against SUMO2^K11Ac demonstrates that SIRT2 cannot catalyze deacetylation of SUMO2^K11 (Fig 2C). Next, we investigated whether SIRT1‐mediated deacetylation of SUMO2^K11 restores SUMO2 chain formation. To address this point, we performed in vitro deacetylation followed by in vitro SUMOylation. Unmodified SUMO2 or K11‐acetyl SUMO2 was pre‐incubated with SIRT1 in the presence or absence of NAD⁺ and subsequently E1 and Ubc9 were added to trigger chain formation. The reaction was stopped by addition of SDS sample buffer, and samples were separated by SDS–PAGE and immunoblotted with α‐SUMO2/3. While chain formation of unmodified SUMO2 was not influenced by SIRT1, a net increase in chain conjugation after SIRT1‐mediated deacetylation of SUMO2^K11Ac was observed (Fig 2D). Similar to the reaction with unmodified SUMO2, acetyl‐K11 SUMO2 was converted into 2x, 3x, and 4xSUMO2, detected by bands at ~35 kDa, ~48 kDa, and ~63 kDa, indicating that poly‐SUMO chain conjugation is restored after deacetylation of SUMO2^K11Ac by SIRT1. In an analogous experiment, we were able to restore poly‐SUMOylation on p53 when SUMO2^K11Ac was incubated with SIRT1 (Fig 2E). In summary, these data demonstrate that SIRT1 specifically catalyzes deacetylation of SUMO2^K11Ac. Having established the regulatory effect of SIRT1 on deacetylation of SUMO2^K11 in vitro, we aimed to get deeper insights into this process in a cellular context. To address this question, HeLa cells were left untreated or treated with nicotinamide (NAM) to inhibit SIRT1. After 5 h, cells were lysed under denaturing conditions and acetylated proteins were enriched with α‐acetyl‐lysine antibody coupled to agarose beads. Immunoprecipitates were analyzed by immunoblotting with α‐SUMO2/3 or α‐SUMO2K11Ac antibodies. As an additional specificity control, we used extracts of cells depleted from SUMO2/3 by siRNA. Remarkably, we were able to detect specific α‐SUMO2/3 reactive bands in α‐acetyl‐lysine pulldowns from NAM‐treated cells, which were not present in control cells and also absent upon siRNA‐mediated depletion of SUMO2/3 (Fig 2F). These bands migrate between 180 and 245 kDa, indicating that they represent cellular proteins that are modified by acetylated SUMO forms. Most importantly, a subset of these bands was also recognized by the α‐SUMO2K11Ac antibody, indicating that upon inhibition of SIRT1 SUMO2/3K11Ac is found in SUMO conjugates. Altogether, our data provide evidence for a direct involvement of SIRT1 in SUMO2K11 deacetylation in vivo and point to a regulatory role of SUMO2K11 acetylation in cells. Even though the amount of acetyl‐K11 SUMO2 modified targets appeared to be rather low, the modification may specifically regulate a defined subset of cellular proteins or may restrict the formation or length of K11‐linked SUMO chains in a timely or spatially regulated manner.

Figure 2. SIRT1 catalyzes the deacetylation of SUMO2^K11Ac .

1. In vitro deacetylation of recombinant SUMO2^WT or SUMO2^K11Ac (2.65 μM each) with recombinant SIRT1 in the presence or absence of NAD⁺.
2. In vitro deacetylation of recombinant SUMO2^WT, SUMO2^K11Ac, SUMO2^K33Ac, and SUMO2^K35Ac. Immunoblotting was performed with antibodies as indicated.
3. In vitro deacetylation of SUMO2^K11Ac with increasing concentrations of SIRT1 or SIRT2 (0.74, 1.5, 3 μM), followed by immunoblotting with α‐SUMO2K11Ac.
4. Combination of in vitro deacetylation and substrate‐independent in vitro SUMOylation with SUMO2^WT and SUMO2^K11Ac in the presence or absence of SIRT1 (1.5 μM).
5. In vitro deacetylation of SUMO2^WT or SUMO2^K11Ac with or without addition of SIRT1, followed by substrate‐dependent in vitro SUMOylation of p53.
6. Enrichment of acetylated proteins after SIRT1 inhibition in HeLa cells. Western blot analysis of SUMO2^K11Ac after treatment of cells with the SIRT1 inhibitor NAM (10 mM) for 5 h and subsequent enrichment for acetylated proteins via acetyl‐lysine immunoprecipitation. As a control, cells were either treated with control siRNA or siSUMO2/3. The samples were analyzed by immunoblotting with α‐SUMO2/3 and α‐SUMO2K11Ac antibodies. The α‐SUMO2/3^Ac and α‐SUMO2^K11Ac reactive species migrating at ˜200 kDa are marked with an asterisk.

Figure EV2. Activity assay to monitor SIRT1 or SIRT2 activity.

The activity of SIRT1 and SIRT2 was measured with the HDAC fluorometric Assay Kit (Enzo). Activity was measured at a concentration of 100 nM. Where indicated, the SIRT inhibitor NAM was added. One representative dataset of three independent experiments is shown.

Degradation of PML by the StUbL pathway is not affected by K11 acetylation

After having established that acetylated or acetyl‐mimicking SUMO2 impairs SUMO chain formation, we aimed to investigate the functional consequences of this process in a cellular context. We focused on the StUbL pathway and choose PML as the best characterized model target of this pathway. PML is SUMO modified on at least three lysine residues (K65, K160, K490) 33, 34. As outlined in the introduction, upon treatment of cells with arsenic poly‐ and/or multiple‐mono‐SUMOylation of PML is induced 15, 16. This triggers the recruitment of the E3 ubiquitin ligase RNF4, thereby catalyzing the ubiquitylation and proteasomal degradation of PML (Fig EV3A). To study K11 acetylation in this context, we first wanted to make sure that K11‐linked SUMO2/3‐polymers are found on PML after arsenic treatment. To this end, we used a HeLa cell line stably expressing HA‐tagged PML and depleted endogenous SUMO2/3 by siRNA. The re‐expressed SUMO2 construct harbors an N‐terminal His‐tag and an exchange of threonine at position 90 to lysine, which makes SUMO2^T90K accessible to GlyGly proteomics 35, 36, a well‐established method in the ubiquitin field 37. Cleavage of proteins modified by SUMO2^T90K with endoproteinase LysC generates peptides that harbor a GlyGly remnant at the SUMO‐modified lysine residue. Peptides carrying these signatures can be specifically enriched by monoclonal antibodies (α‐GlyGly IP) that are directed against this motif. Analysis of these peptides by MS allows the identification of SUMOylation sites in a given protein and also permits the identification of SUMO‐SUMO chain linkages. Cells were either left untreated or treated with arsenic for 2 h to induce formation of SUMO chains on PML, but not its degradation. HA‐PML was immunopurified under denaturing conditions by an α‐HA‐affinity matrix, the immunoprecipitated material was digested with LysC, α‐GlyGly IP was performed, and enriched peptides were analyzed by MS (Fig 3A). The analysis revealed a peptide with a GlyGly remnant at K490 of PML, thus confirming SUMOylation of PML at this residue (Fig EV3B). Importantly, we also detected a SUMO2/3 peptide harboring a GlyGly remnant at K11, indicative for K11‐linked SUMO chains at PML. This peptide was about fivefold more abundant in arsenic‐treated than in control cells supporting the idea that arsenic induces K11 chains on PML (Fig 3B). To now unveil the impact of K11 acetylation on arsenic‐induced degradation of endogenous PML, we established a knockdown‐complementation system, in which cells were depleted for endogenous SUMO2/3 by siRNA and SUMO2^WT or SUMO2^K11Q were re‐expressed under the control of an inducible promoter. We therefore generated doxycycline (DOX)‐inducible cells expressing siRNAi‐resistant variants of either His‐SUMO2^WT or His‐SUMO2^K11Q. Cells expressing the lysine‐less, chain‐deficient SUMO variant SUMO2^K0 were used as an additional control. The cell lines were used to monitor arsenic‐driven degradation of endogenous PML. Since the anti‐PML immunoblot shows several bands, we had previously confirmed by siRNA experiments that a 100‐kDa species corresponds to endogenous PML 24. In accordance with this, we validated that this 100‐kDa species is degraded upon arsenic treatment in a SUMO2/3‐dependent process (Fig EV3C). As expected, the degradation of PML after arsenic treatment was abolished following siRNA‐induced SUMO1/2/3 knockdown, but could be rescued by DOX stimulated re‐expression of His‐SUMO2^WT (Fig EV2D). By contrast, re‐expression of SUMO2^K0 could not completely restore this process confirming our previous finding that PML degradation cannot be mediated by multiple‐mono‐SUMOylation alone 24. Surprisingly, however, re‐expression of SUMO2^K11Q restored degradation of PML, indicating that the arsenic‐triggered signaling cascade leading to RNF4‐mediated PML degradation does not strictly depend on K11‐linked SUMO2 chains (Fig 3C). These data indicate that in the absence of K11 non‐canonical SUMO chains can trigger the StUbL pathway. In line with this idea, we could demonstrate in a GST‐pulldown experiment that a RNF4‐derived poly‐SIM module can capture non‐canonical SUMO chains from heat‐stressed cells expressing a SUMO2^K11R mutant (Fig EV4A). Moreover, the residual SUMO chains that are still present upon knockdown of ZNF451 and most likely correspond to non‐canonical chains are retained on the RNF4‐derived poly‐SIM. Notably, in this cellular setting lack of ZNF451 reduces the amount of both canonical and non‐canonical chains (Fig EV4B). Since our in vitro data did not reveal a direct involvement of ZNF451 in formation of non‐canonical chains, this indicates that in cells ZNF451 may cooperate with another E3 ligase. To further investigate the contribution of alternative chains in arsenic‐induced PML degradation, we focused on K5 and K7 and generated a cell line expressing a SUMO2^K5,7,11R variant. Importantly, expression of this variant in SUMO1‐/2‐/3‐depleted cells impairs arsenic‐induced degradation of PML (Fig EV5A and B), indicating that the lack of these three residues cannot be compensated by other chain linkages. Altogether, these data demonstrate that in response to arsenic K11‐linked chains do form on PML, but are not essential for its degradation. This explains why K11‐acetylation of SUMO2 does not affect this particular pathway.

Figure EV3. As₂O₃ triggers poly‐SUMO‐dependent degradation of PML.

1. Model of the StUbL‐dependent proteasomal degradation of PML in response to As₂O₃ treatment. Upon As₂O₃ stimulation, PML is poly‐SUMOylated and subsequently recognized by the StUbL RNF4. RNF4 binds to the SUMO chain via its tandem SIM region, ubiquitylating PML and the SUMO chain and thereby targeting PML for proteolysis.
2. The PML SUMOylated peptide, identified in the MS analysis, is shown in Fig 3B. The SUMOylation site is marked with SUMO2 (green), and peptide fragment ions are indicated by y‐ (red) and b‐ (blue).
3. HeLa cells, treated with the indicated siRNAs, were stimulated with As₂O₃ for 6 h and lysed in SDS sample buffer. Proteins were separated by SDS–PAGE, and protein levels were monitored by immunoblotting with the indicated antibodies. Immunoblotting with α‐tubulin or α‐PCNA served as a loading control.
4. The degradation of the PML reactive band (˜120 kDa) was determined by immunoblotting. HeLa cells were either mock transfected or transfected with siRNA directed against SUMO1/2/3. The expression of RGS‐His‐tagged SUMO2^WT or SUMO2^K0 was induced with DOX for 12 h. Prior to lysis, cells were treated with 1 μM As₂O₃ for 6 h. Western blot analysis for wild‐type SUMO2 and SUMO2^K0 was done with α‐RGS‐His antibodies. Immunoblotting with α‐vinculin or α‐PCNA served as a loading control.

Figure 3. Degradation of PML by the StUbL pathway is not affected by K11 acetylation.

1. HeLa cells constantly expressing HA‐tagged PML were depleted for endogenous SUMO1/2/3 by siRNA. Subsequently, they were transfected with RGS‐His‐tagged SUMO2^T90K for 72 h. Prior to lysis, cells were treated with As₂O₃ for 2 h. HA‐tagged PML was immunoprecipitated via a denaturing HA‐IP, which was followed by LysC digestion. Peptides harboring GlyGly remnants, corresponding to SUMOylation sites, were enriched by GlyGly IP, and the samples were analyzed by LC‐MS/MS.
2. Western blot analysis with α‐RGS‐His and α‐HA was performed to monitor poly‐SUMOylation of HA‐PML. The remaining samples were measured by LC‐MS/MS, and final data analysis was done with MaxQuant. A K11 SUMOylated SUMO2 peptide was identified and is depicted with an Andromeda score of 49.3. Intensities for this SUMO2 peptide for the control (gray) and arsenic (purple)‐treated samples are plotted as a bar graph. The diagram shows a representative dataset derived from two replicates. The corresponding peptide is displayed below the graph. The SUMOylation site is marked with SUMO2 (green), and peptide fragment ions are indicated by y‐ (red) and b‐ (blue).
3. The degradation of the PML reactive band (˜120 kDa) was determined by immunoblotting. HeLa cells were either mock transfected or transfected with siRNA directed against SUMO1/2/3. The expression of RGS‐His‐tagged SUMO2^WT or SUMO2^K11Q was induced by DOX for 12 h. Prior to lysis, cells were treated with 1 μM As₂O₃ for 6 h. Western blot analysis for wild‐type SUMO2 and SUMO2^K11Q was done with α‐RGS‐His antibodies. Immunoblotting with α‐tubulin served as a loading control. Immunoblots for knockdown controls (SUMO1 and SUMO2/3) can be found in Fig EV5B.

Figure EV4. Enrichment of SUMO2 chains on poly‐SIM modules.

1. HeLa cells were depleted for SUMO2/3 by siRNA for 24 h. Subsequently, they were transfected with SUMO2^WT, SUMO2^K11R, or SUMO2^K0 for additional 48 h. Prior to the pulldown, cells were incubated at 43°C for 1 h. Poly‐SUMOylated conjugates were captured with GST‐control or GST‐SUBE out of the cell lysate. SUBEs are composed of tandem repeats of RNF4‐derived SIM2 and SIM3 motifs 46. Samples were immunoblotted with α‐RGS‐His (SUMO2 variants) and α‐vinculin as loading control.
2. HeLa cells were depleted for SUMO2/3 by siRNA for 24 h. Simultaneously, siRNA‐mediated knockdown for siZNF451 or siContr. was performed. Subsequently, cells were transfected with SUMO2^WT, SUMO2^K11R for additional 48 h. Prior to the pulldown, cells were incubated at 43°C for 1 h. Poly‐SUMOylated conjugates were captured with GST‐control or GST‐SUBE out of the cell lysate. The ability of poly‐SUMO2 chain formation with SUMO2^WT and SUMO2^K11R, in the absence and presence of ZNF451, was analyzed by immunoblotting with α‐RGS‐His (SUMO2 variants), α‐ZNF451, α‐SUMO2/3, and α‐vinculin as loading control.

Figure EV5. Contribution of K5‐ and K7‐linked SUMO2 chains to As₂O_3‐‐induced degradation of PML.

1. The degradation of the PML reactive band (˜120 kDa) was determined by immunoblotting as described before. Western blot analysis for wild‐type SUMO2 and SUMO2^K0 and SUMO2^K5,7,11R with α‐RGS‐His. Immunoblotting with α‐tubulin or α‐PCNA served as a loading control.
2. Corresponding knockdown controls for the experiment in Fig 3C. Western blot analysis was performed with α‐SUMO1 and α‐SUMO2/3 antibodies to verify siRNA knockdown.
3. Corresponding second experiment to Fig 4C. Bar graph, comparing the identified SUMO‐SUMO2 linkages of control and heat shock samples for SUMO2^WT (gray) and SUMO2^K11Q (purple). The relative abundance (y‐axis) of the indicated SUMO‐SUMO2 linkage sites (x‐axis) is shown. The measured intensity for K11 in the SUMO2^WT under control conditions was set to 100%, and the relative abundance for the individual residues was calculated in relation to this value.

Impact of K11 acetylation on SUMO chain architecture under heat stress

To more generally investigate the impact of K11 acetylation in stress‐induced SUMO chain formation, we used an unbiased MS‐based approach. We concentrated on the question how the expression of the acetyl‐mimicking mutant SUMO2^K11Q affects the induction of SUMO2 chains in response to heat stress. Therefore, we monitored chain linkages in control HeLa cells or cells exposed to 43°C for 1 h. To compare wild‐type SUMO2 with the acetyl‐mimicking SUMO2^K11Q variant, we again turned to a knockdown‐complementation protocol, in which cells were depleted for endogenous SUMO2/3 by siRNA and SUMO2^WT or SUMO2^K11Q were re‐expressed by transient transfection (Fig 4A). Both SUMO variants harbor an N‐terminal His‐tag and an exchange of threonine at position 90 to lysine to perform SUMO‐GlyGly proteomics allowing the identification of SUMO‐SUMO chain linkages. After heat shock, cells were lysed and an aliquot was taken to control for equal transfection levels of wild‐type SUMO2 and SUMO2^K11Q by immunoblot analysis. The immunoblot revealed an increase in SUMO2 conjugation after heat shock, but no major difference between cells expressing SUMO2 or SUMO2^K11Q (Fig 4B). The remaining sample was digested with LysC and enriched for SUMOylated peptides by α‐GlyGly‐remnant IP. Peptides were measured by LC‐MS/MS, and data were analyzed using MaxQuant software. The experiment was done in two independent biological replicates, and data are presented in Figs 4C and EV4. In cells expressing wild‐type SUMO2, SUMO2‐SUMO2 linkages at K5, K7, K11, K21, K33, K35, and K42 were detected. Under control conditions, K11‐linked SUMO peptides were the most abundant species, but K35‐linked chains were also found in significant amounts. By contrast, other linkage types were only detected at low intensity. Upon heat stress, both K11 and K35 linkages were about threefold to fivefold induced. Interestingly, however, the strongest induction in response to heat stress was revealed for K5‐linked peptides (Fig 4C). Altogether, these data support the prevalence of K11 as the major site for SUMO linkages, but also indicate the presence of alternative linkage types under normal growth conditions and upon heat stress. Overall, these findings are largely in line with previous high‐throughput MS screenings 5, 35, 36, 38, 39. As expected, in cells expressing SUMO2^K11Q a drop of K11‐linked chains is observed. Intriguingly, however, K7 and K21 linkages were also no longer detectable, indicating that K11Q also impairs attachment of SUMO to these sites. Under control conditions, the intensity of K35‐linkages was fairly comparable between SUMO2^K11Q and wild‐type SUMO2. Under heat stress, however, both K5‐ and K35‐linkages were highly enriched in SUMO2^K11Q‐expressing cells (Figs 4C and EV5C). Comparing the heat stress response of wild‐type SUMO2 and the SUMO2^K11Q variant revealed a significantly enhanced level of K5‐ and K35‐linked SUMO in the acetyl‐mimicking SUMO2 variant in both replicates. Importantly, even under heat stress SUMO2^K11Q does not form K7‐ and K21‐linked chains (Figs 4C and EV5C). In conclusion, these data show that in the absence of K11, K35 serves as major attachment sites for SUMO‐SUMO linkages under normal growth conditions, while both K5 and K35 are used in stress response signaling. The complete absence of K7 and K21 linkages in SUMO2^K11Q‐expressing cells points to a regulatory role of K11 acetylation beyond the mere blockage of K11 residues. Acetylation of K11 may directly prevent recognition of these sites by the SUMO machinery. Alternatively, enhanced acetylation of K5 or K35 may impair chain formation via K7 or K21.

Figure 4. Impact of K11 acetylation on SUMO chain architecture under heat stress.

1. HeLa cells were depleted for endogenous SUMO2/3 by siRNA and transfected with SUMO2^T90K (WT or K11Q). Prior to lysis, cells were heat shocked at 43°C for 1 h. The lysate was digested with LysC, followed by GlyGly enrichment. Subsequently, samples were measured by LC‐MS/MS and analyzed with MaxQuant software.
2. Immunoblotting of samples with α‐RGS‐His (SUMO variants) and α‐PCNA as loading control. Sample preparation was performed as described in the text.
3. Bar graph, comparing the identified SUMO‐SUMO2 linkages of control and heat shock samples for SUMO2^WT (gray) and SUMO2^K11Q (purple). The relative abundance (y‐axis) of the indicated SUMO‐SUMO2 linkage sites (x‐axis) is shown. The measured intensity for K11 in the SUMO2^WT under control conditions was set to 100%, and the relative abundance for the individual residues was calculated in relation to this value. One of two independent replicates is shown. Data from the second replicate are shown in Fig EV5C.
4. Model depicting the possible impact of an acetylation of SUMO2 at K11 on formation of SUMO signaling. SUMO2/3^K11Ac can restrict chain length by acting as a K11 chain terminator (A), SUMO2/3^K11Ac can function as a binding platform for acetyl‐lysine‐binding proteins (B), SUMO2/3^K11Ac can trigger the formation of alternatively linked SUMO2/3 chains (C).

Altogether, our data are consistent with a potential role of K11 acetylation of SUMO2/3 in modulating the architecture and/or length of SUMO2 chains (Fig 4D). Because K11 of SUMO2/3 is not only the preferred site for SUMO‐SUMO linkages, but also serves as the site for the formation of mixed SUMO2‐SUMO3 chains as well as SUMO‐ubiquitin chains, K11 acetylation could also interfere with the formation and architecture of these heterotypic chains 40. We provide evidence that SIRT1 acts as a highly specific deacetylase of SUMO2/3 at K11. In line with this finding, systemwide acetylomic studies identified K11 acetylation as the only upregulated acetylation site on SUMO after genetic or chemical inhibition of SIRT1 31, 32. It is tempting to speculate that the activation of SIRT1 under stress conditions may contribute to the stress‐induced formation of K11 SUMO chains or allow the formation of heterotypic SUMO‐ubiquitin chains. Interestingly, SIRT1 is itself a SUMO target and SUMOylation of SIRT1 has been linked to its activation 41. Moreover, high‐throughput MS data indicate that SUMOylation of SIRT1 is induced under heat stress raising the intriguing possibility of a positive feed‐forward loop in the heat stress response 36, 38. Considering the relatively low abundance of K11 acetylation, it likely regulates SUMO chain topology or length in a spatially or timely controlled manner. The observation that SIRT1 is recruited to PML nuclear bodies and regulates SUMOylation of PML supports the idea of spatial regulation of SUMOylation by SIRT1 42. In addition to the discovery of K11 acetylation, our data generally expand the concept of non‐canonical SUMO chains in SUMO signaling. A major question will be to determine the degree of redundancy or specificity in chain linkages. An elegant, systematic study in the yeast S. cerevisiae revealed critical roles for polymeric SUMO chains, but redundancy in specific chain linkages 43. Here, we provide evidence that degradation of PML by the RNF4‐StUbL pathway depends on K5‐, K7‐ and K11‐linked lysine chains. Chains connected via these adjacent N‐terminal residues likely share a similar topology that can be recognized by the poly‐SIM of RNF4. By contrast, other chain types, such as K33‐, K35‐, or K42‐linked chains will likely exhibit a different architecture that is read by distinct binding modules that remain to be discovered. Notably, the acetyl‐lysine residues on SUMO paralogs may not only alter chain architecture, but may also provide an alternative docking site for acetyl‐lysine‐binding proteins of the bromodomain family. The combinatorial use of these PTMs thus provides an enormous versatility allowing a balanced response to cellular stress.

Materials and Methods

Cell culture, plasmids, siRNA, transfection, and drug treatments

HeLa cells (ATCC) were grown under standard conditions in DMEM high glucose, 10% FCS, and 100 U/ml penicillin and 100 U/ml streptomycin. Cells stably expressing HeLa HA‐PML III were maintained with G418 (150 μg/ml, Roth). Where indicated, cells were treated with 1 μM As₂O₃ (Roche). To inhibit SIRT1, cells were treated with 10 μM NAM (Sigma) for 5 h prior to cell lysis. HeLa Flp‐In/T‐Rex cells expressing RGS‐His‐tagged variants of SUMO2 under the control of doxycycline‐inducible promoter were generated as described 23. Expression of the respective proteins was induced by addition of doxycycline (Sigma) to a final concentration of 1 μg/μl 24 h prior to cell lysis. siRNA‐mediated knockdown experiments were done as described 23.

For transient expression of the RGS‐His‐epitope‐tagged SUMO2 variants (SUMO2^T90K, SUMO2^WT or SUMO2^K11Q), the respective cDNA sequences were inserted into the pCI vector (Invitrogen). To generate SUMO2^T90K and SUMO2^K11Q, site‐directed mutagenesis was carried out using the QuikChange Mutagenesis Kit (Stratagene) according to the manufacturer's instructions. For bacterial expression of GST‐fusion proteins (SIRT1, SIRT2), cDNA sequences were inserted into pGEX (GE Healthcare) vectors. Expression of RGS‐His‐tagged SUMO2^WT and the acetyl‐variant in E. coli was done as described 23. Plasmids encoding p53, Ubc9, SAE1/SAE2, SUMO2^K33Ac, and SUMO2^K35Ac were described previously 23, 28, 44. The plasmid encoding ZNF451 was described previously 27. All plasmids were sequenced (Eurofins Genomics) to confirm the intended sequence.

Antibodies

The following antibodies were used for immunoblotting: PML (H238, Santa Cruz, sc‐5621), SIRT1 (D793, Cell Signaling, 2493), acetyl‐lysine (Cell Signaling, 877‐616), PCNA (D3H8P, Cell Signaling, 13110), SIRT2 (Sigma, S8447), HDAC1 (Sigma, H3284). Flag (M2, Sigma, F1804), SUMO2/3 (1E7, MBL, M114‐3), RGS‐His (Qiagen, 34610), HA (16B12, Covance, MMS‐101R), SUMO1 (GMP1, Clone 21C7, Invitrogen, 33‐2400), tubulin (B7, Santa Cruz, sc‐5286), Ubc9 (Biocompare, Clone 50, 610748), ZNF451 (Sigma, SAB 2108741), vinculin (hVIN1, Sigma, V9131). Anti‐SUMO2K11Ac antibody was generated by Immunoglobe (Germany) using a SUMO2‐derived peptide (‐EKPKEGVK(ac)TENC‐) for immunization of rabbits. The acetylated peptide was used for three successive rounds of immunization and boosting. Thereafter, serum was collected and affinity purified.

Expression of recombinant proteins, in vitro SUMOylation assays and pulldown of poly‐SUMOylated conjugates

Expression of non‐acetylated and acetylated recombinant proteins in E. coli as well as in vitro SUMOylation assays were performed as described previously 23, 24, 44. For the expression of ZNF451‐N, the LB medium was supplemented with 500 μl zinc chloride. The purified recombinant protein was stored in the following buffer 20 mM HEPES, pH 7.3, 110 mM potassium acetate, 150 mM NaCl, 2 mM magnesium acetate, 0.5 mM TCEP, and 10% glycerol. Substrate‐independent in vitro SUMOylation assays in the absence of a ligase were performed in the presence of 265 nM Ubc9, 3.6 μM E1, and either SUMO2^WT (2.65 μM), SUMO2^K11Ac (2.65 μM), or SUMO2^K11Q (2.65 μM). Experiments with ZNF451 were done with as described using 100 nM Ubc9 and 2 μM SUMO2^WT or SUMO2^K11Ac 20 μl. ZNF451‐N was used at concentrations of 15, 60, 90 nM). Single‐turnover assay was performed as described 27. The charging was performed as described by Wiechmann et al 24. Ubc9 was charged with SUMO2K11R. Charging was stopped by the addition of apyrase (2 U per 20 μl, Sigma). Discharge was performed in a 2.5‐fold‐diluted reaction in 20 μl in the absence or presence of the E3 and the indicated substrate (SUMO2^wt or SUMO2^K11Ac). For pulldown of poly‐SUMOylated conjugates, cells were depleted from SUMO2/3 by siRNA and transfected with plasmids encoding the respective SUMO2 variants. After cells lysis, poly‐SUMOylated proteins were captured by SUMO affinity traps (SUBE) as described 24.

In vitro deacetylation assay

Deacetylation of the recombinant SUMO2^WT or acetylated SUMO2 variants by SIRT1, SIRT2, or HDAC1 was performed in deacetylation buffer (50 mM Tris/HCl [pH 9.0], 4 mM MgCl₂, 1 mM DTT (Roth), 50 μM NAD⁺ [freshly added, only for SIRT1 and SIRT2]). Therefore, 2.65 μM SUMO2 was mixed with 1.5 μM of the indicated deacetylase, filled up with buffer to a final volume of 30 μl, and incubated for 2 h at 37°C. The reaction was stopped by adding 5 μl 6× SDS sample buffer. The samples were heated up at 95°C for 5 min. Activities of SIRT1 and SIRT2 were measured by using the HDAC Fluorometric Assay Kit (Enzo) according to the manufacturers' instructions.

Enrichment of acetylated proteins by acetyl‐lysine IP

1.5 × 10⁵/ml HeLa cells were seeded on 4 × 10 cm dishes/sample. Cells were either mock transfected or depleted for endogenous SUMO2/3 by siRNA‐mediated knockdown. 48 h after siRNA transfection, cells were treated with 10 mM NAM for 5 h. Subsequently, cells were washed twice with ice‐cold PBS and lysed in 400 μl with 1% SDS‐IP buffer (1% SDS, 25% glycerol, 0.25 M Tris/HCl [pH 8.0]). The lysate was collected and rotated for 20 min at 4°C. Afterward, the lysate was sonicated 3× for 1 s with an amplitude of 70%. The lysate was subsequently diluted with PBST to a final SDS concentration of 0.1% and centrifuged at 20,817 g for 20 min at 4°C. One‐tenth of the sample was taken as an input sample for Western blot analysis. Anti‐acetyl‐lysine‐coupled beads (10 mg/ml) (ImmuneChem) and empty protein A/G agarose beads (Santa Cruz) were once washed with PBS. As control, 20 μl of empty protein A/G agarose beads was used. To enrich for acetylated proteins, 5 μl acetyl‐lysine beads was mixed with 15 μl empty protein A/G agarose beads and added to the sample. After acetyl‐lysine‐IP overnight, the samples were washed thrice in diluted IP buffer. Subsequently, beads were eluted with 2.5× SDS sample buffer, heated up to 95°C for 5 min, and analyzed by immunoblotting.

Enrichment of SUMO substrates by GlyGly immunoprecipitation

For the identification of SUMOylation sites, we used a modified version of a published protocol 35, 36. 1.5 × 10⁵/ml HeLa cells were seeded on 10 × 10 cm dishes/sample. Cells were depleted for endogenous SUMO2/3 by siRNA‐mediated knockdown. 24 h after siRNA transfection, cells were transfected with SUMO2^WT,T90K or SUMO2^K11Q,T90K. Seventy‐two hours after siRNA transfection, cells were heat shocked at 43°C for 1 h. Cells were lysed with 500 μl 4% SDS buffer (4% SDS, 0.1 M Tris/HCl [pH 7.6]) supplemented with 25 mM NEM (Sigma) The lysate was sonicated 3× for 1 s with an amplitude of 70%, and samples were incubated at 90°C for 10 min. For in‐solution digest, 15 mg starting material was used. Proteins were precipitated by the addition of four volumes of ice‐cold acetone (100%), followed by an incubation at −20°C for 2 h. Afterward, precipitates were centrifuged at 17,949 g and 4°C for 10 min, the supernatant was discarded and the pellet was washed with 1 ml ice‐cold acetone (90%), followed by another centrifugation step. The pellet was dissolved in 3 ml 8 M urea [6 M urea + 2 M thiourea in 0.1 M HEPES (pH 7.4)]. Reduction and alkylation were carried out as described previously 45. Proteins were digested with the endopeptidase LysC (Wako) (1:100, enzyme to protein ratio) overnight at RT. Digestion was stopped by the addition of 1% trifluoroacetic acid (TFA) for 10 min at RT. Peptides were desalted using a 1 g C18 Sep‐Pak SPE cartridge (Waters). The eluate was dried overnight by lyophilization at −20°C. Lyophilized peptides were resuspended in 6 ml of ice‐cold IAP buffer (50 mM MOPS [pH 7.4], 10 mM Na₂HPO₄, 50 mM NaCl). Samples were incubated with 40 μl slurry of anti‐GlyGly‐coupled beads (Cell Signaling) for 2 h at 4°C. Unbound material was removed by washing the beads thrice with 1 ml of ice‐cold PBS and thrice with 1 ml ice‐cold Mili‐Q water. Bound K‐ε‐GlyGly peptides were eluted twice with 20 μl 0.15% TFA, incubated for 10 min at RT, and purified as described previously 45.

Double immunoprecipitation (HA‐GlyGly‐IP)

For the analysis of SUMOylation sites on PML, a double IP (HA‐GlyGly‐IP) was performed. 1.5 × 10⁵/ml HeLa cells, stably expressing HA‐tagged PML, were seeded on 10 × 10 cm dishes/sample. Cells were depleted for endogenous SUMO2/3 by siRNA‐mediated knockdown. 24 h after siRNA transfection, cells were transfected with SUMO2^T90K. Seventy‐two hours after siRNA transfection, cells were treated with 1 μM arsenic for 2 h. For immunoprecipitation cells were lysed under denaturing conditions. Therefore, cells were washed twice with PBS and lysed with 1% SDS‐IP buffer [1% SDS, 25% glycerol, 0.25 M Tris/HCl (pH 8.0)]. The lysate was collected and rotated for 20 min at 4°C. Afterward, the lysate was sonicated 3× for 1 s with an amplitude of 70%. The lysate was subsequently diluted with PBST to a final SDS concentration of 0.1% and centrifuged at 20,817 g for 20 min at 4°C. Anti‐HA‐coupled beads (3.5 mg/ml) (Roche) were once washed with PBS, and the amount was adjusted to the protein concentration (5 μg beads per 500 μl protein lysate). After HA‐IP overnight, the samples were washed thrice in diluted IP buffer, 1/10 slurry was taken as an input sample for Western blot analysis. Subsequently, beads were eluted in 8 M urea, 2× for 15 min. Reduction, alkylation, and LysC digestion were carried out as described previously 45. The next day, the digest was stopped by the addition of 1% TFA for 10 min at RT. For desalting, the peptide samples were loaded onto a 50 mg C18 Sep‐Pak SPE cartridge. The eluate was dried overnight by lyophilization at −20°C. Lyophilized peptides were resuspended in 50 μl of ice‐cold IAP buffer and enriched for GlyGly‐modified peptides, using 5 μl slurry of anti‐GlyGly‐coupled beads. Bound K‐GlyGly peptides were eluted twice with 20 μl 0.15% TFA, incubated for 10 min at RT and purified using C18‐based STAGE tips as described 45.

Liquid chromatography and mass spectrometry analysis

Analysis by LC‐MS/MS was performed as previously described 45. Peptides were eluted by increasing the relative amount of buffer B from 10 to 38% in a linear gradient within 35 min at a column temperature of 40°C. This was followed by an increase to 95% buffer B within 10 min. Gradients were completed by a re‐equilibration to 5% buffer B. Q Exactive HF settings: MS spectra were acquired using 1 × 10⁶ as an AGC target, a maximal injection time of 20 ms, and a 60,000 resolution at 200 m/z. The mass spectrometer operated in a data‐dependent Top5 mode with subsequent acquisition of HCD (higher‐energy collisional dissociation) fragmentation MS/MS spectra of the top 5 most intense peaks. Resolution for MS/MS spectra was set to 30,000 at 200 m/z, AGC target to 5 × 10⁶, maximal injection time to 120 ms, and the isolation window to 1.6 Th.

Data analysis with MaxQuant

Data were processed as described 45. For protein assignment, electrospray ionization‐tandem mass spectrometry fragmentation spectra were correlated with the UniProt human database (v. 2016), including a list of common contaminants and the sequences for the used SUMO2 variants. Searches were performed with LysC specifications and default settings for mass tolerance for MS and MS/MS spectra. Carbamidomethyl at cysteine residues was set as a fixed modification and methionine oxidation, acetylation at the N‐terminus as well as GlyGly‐modified lysine residues were set as variable modifications. The match between run feature was used with a time window of 0.7 min. SUMOylated site identification and quantitative information were obtained from the MaxQuant GlyGly(K) sites table.

Author contributions

AG performed most experiments, prepared Figures, and drafted parts of the manuscript. KW contributed experiments to the revised version. SH performed mass spectrometry measurements and helped in analyzing MS data. KK provided experimental help for the revised version. MSR provided materials and protocols for SUBE experiments. SM designed the study, supervised the work, and wrote the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Review Process File

EMBO Reports (2018) 19: e46117