Sirtuin 2 mutations in human cancers impair its function in genome maintenance

Sirtuin 2 (SIRT2) is a sirtuin family deacetylase, which maintains genome integrity and prevents tumorigenesis. Although Sirt2 deficiency in mice leads to tumorigenesis, the functional significance of somatic SIRT2 mutations in human tumors is unclear. Using structural insight combined with bioinformatics and functional analyses, we show that naturally occurring cancer-associated SIRT2 mutations at evolutionarily conserved sites disrupt its deacetylation of DNA-damage response proteins by impairing SIRT2 catalytic activity or protein levels but not its localization or binding with substrate. We observed that these SIRT2 mutant proteins fail to restore the replication stress sensitivity, impairment in recovery from replication stress, and impairment in ATR-interacting protein (ATRIP) focus accumulation of SIRT2 deficiency. Moreover, the SIRT2 mutant proteins failed to rescue the spontaneous induction of DNA damage and micronuclei of SIRT2 deficiency in cancer cells. Our findings support a model for SIRT2's tumor-suppressive function in which somatic mutations in SIRT2 contribute to genomic instability by impairing its deacetylase activity or diminishing its protein levels in the DNA-damage response. In conclusion, our work provides a mechanistic basis for understanding the biological and clinical significance of SIRT2 mutations in genome maintenance and tumor suppression.

roid, and uterine cancer, most of which overlap with the 66 somatic SIRT2 mutations found in the Catalogue of Somatic Mutations in Cancer (COSMIC) database (17). We focused our investigation on nine SIRT2 mutations present in uterine, lung, and colorectal cancer, which were found in both data sets, predicted to be high impact mutations by cBioPortal, and occurred at appreciable frequencies within their study cohort (Fig. 1A). All mutations selected were missense mutations, occurred throughout the length of the protein, and were highly conserved across distantly related species (Fig. 1B). To validate the potential importance of these mutations, we performed in silico bioinformatics analyses for functional significance using several conservation and impact prediction platforms, including Grantham, Sorting Intolerant from Tolerant (SIFT), Polymorphism Phenotyping (PolyPhen), and ConSurf (Fig. 1C) (26 -34).
Grantham scores predict the evolutionary distance between substituted amino acids and thus the potential for missense mutations to be deleterious or neutral. This analysis was completed using the Align-GVGD program through the Huntsman Cancer Institute at the University of Utah. SIFT and ConSurf predictions are based on the degree of conservation of amino acid residues in sequence alignments derived from closely related sequences collected through Position-Specific Iterative Basic Local Alignment Search Tool (PSI-BLAST) (35)(36)(37). ConSurf determines the conservation of an amino acid and rates it on a scale of 1-9 with 9 indicating highly conserved. PolyPhen predicts the possible impact of missense mutations on the structure and function of a given human protein by comparing physical and alignment-based comparisons. Consistently, mutations P128L, P140H, R153C, A186V, and F190V Figure 1. Cancer-associated SIRT2 mutations are evolutionarily conserved and predicted to be functionally significant. A, table listing cBioPortal data for nine naturally occurring SIRT2 mutations in human cancer samples along with occurrence in patient cohort, cancer type, and study type. B, conservation of cancer-associated SIRT2 mutation amino acids among diverse species, including chimpanzee, mouse, chicken, zebrafish, fruit fly, yeast, and nematode, and their location in the protein structure. The NES is in yellow, and the deacetylase domain is highlighted in blue. C, impact prediction of cancer mutations by Grantham, SIFT, and PolyPhen, and their conservation scores as calculated by ConSurf. TCGA, The Cancer Genome Atlas; COSMIC, Catalogue of Somatic Mutations in Cancer.
were found to be highly conserved and deleterious if mutated, and notably all are located within the SIRT2 deacetylase domain. R42P was also well conserved and predicted to have a high impact on SIRT2 function with its localization within the nuclear export sequence (NES) of SIRT2 (38,39). Collectively, these data indicate that cancer-associated SIRT2 mutations are evolutionarily conserved and are predicted to be functionally significant, suggesting that they are more likely to be driver mutations that benefit cancer progression rather than random carrier mutations.

Cancer-associated mutations impair SIRT2 in vitro deacetylase activity and protein level but not localization
Each of the cancer-associated SIRT2 mutations was created by site-directed mutagenesis in a SIRT2-FLAG construct and expressed in 293T cells along with SIRT2 wild type (WT) and H187Y deacetylase-inactive mutant controls. All of the mutants were expressed at comparable levels to SIRT2 WT except for R42P, which showed lack of protein expression similar to an empty vector ( Fig. 2A), suggesting that R42P affects either protein stability or mRNA expression. SIRT2 shuttles between the nucleus and cytoplasm through active nuclear export in a CRM1-dependent manner (38). To determine whether cancer-associated mutations are important for SIRT2 localization, we examined localization of SIRT2 WT and mutants expressed in U2OS cells before and after treatment with leptomycin B (LMB), which inhibits CRM1-mediated nuclear export, by indirect immunofluorescence. All of the mutants expressed localized predominantly in the cytoplasm and demonstrated nuclear retention following treatment of cells with LMB ( Fig. 2B), suggesting that SIRT2 cancer mutants are capable of proper cytoplasmic and nuclear localization. To determine whether cancer-associated mutations are important for SIRT2 deacetylase activity, we performed in vitro measurements using the Fluor De Lys in vitro deacetylase assay by Enzo Life Sciences, which uses an acetylated p53 Lys-320 peptide as a substrate. Three of the mutants, SIRT2 P128L, P140H, and A186V, demonstrated significantly decreased deacetylase activity compared with SIRT2 WT with SIRT2 P128L and A186V impairing deacetylase activity similarly to catalytically inactive SIRT2 H187Y (Fig. 2C). All three of these mutations are located within the deacetylase domain of SIRT2 and likely affect NAD ϩ cofactor and/or substrate binding or catalytic activity.

Structural analysis of SIRT2 mutations yields insights into their functional significance
To gain further insight into how cancer-associated mutations might affect SIRT2 biological function, we examined the crystal structure of SIRT2 obtained from the Research Collaboratory for Structural Bioinformatics Protein Data Bank, specifically structure 4RMH (40). We first focused on four mutants of interest based on their predicted impact scores, deacetylase activity, and structural location: P128L, P140H, A186V, and F190V. All four were predicted to negatively impact protein function, and three of these demonstrated decreased deacetylase activity in vitro. All four were also located in the active site of SIRT2 (Fig. 3A). Pro-128 and Pro-140 make up the ends of an important ␣ helix responsible for composing part of the active site for NAD ϩ binding. Mutation of these prolines could therefore alter the orientation of the helix and result in decreased binding of NAD ϩ and thus lower deacetylase activity. Pro-140 is also important for the formation of the binding pocket for NAD ϩ along with Phe-190; these two amino acids stack on either side of NAD ϩ (Fig. 3A, blue molecule). Mutating either of these amino acids therefore would also likely result in decreased binding of NAD ϩ and lower deacetylase activity. F190V, however, does not yield a deacetylase-defective phenotype (Fig. 2C), suggesting that this mutation alone is not sufficient to inhibit cofactor binding. Ala-186 lies adjacent to the active site amino acid His-187 and thus could affect its neighboring amino acid's ability to deacetylate substrate (Fig. 3, B and C). In fact, Ala-186 lies in a tight pocket with little space for amino acid structure variation as better illustrated in the spacefilling model in Fig. 3C. Even the small addition of two methyl groups is likely enough to blow out this pocket and upset the orientation of His-187, thereby inhibiting deacetylation of substrate. Overall, these four amino acids are heavily involved in proper deacetylase activity and/or cofactor binding. The other five mutations for this study lie outside of the active site and binding pocket of SIRT2. Only four of these five amino acids (Glu-203, Arg-153, Leu-341, and Ser-73) have been verified for orientation and structure as the current crystal structures of SIRT2 lack a portion of the N terminus where Arg-42 lies (Fig.  3D). Of these amino acids, three are on the surface of the protein where they potentially interact with other proteins (Glu-203 and Arg-153) and/or perhaps undergo post-translational modification (Ser-73). Therefore, mutation of these amino acids to non-compatible substitutes could upset other aspects of SIRT2 function regulated by protein-protein interaction but that are not necessary for deacetylase activity. The fourth amino acid, Leu-341, is hydrophobic and forms a portion of an ␣ helix. Mutation of this amino acid therefore could lead to a change in structure of the helix and upset the surface structure without affecting SIRT2 deacetylase activity. Fig. 3E shows a close-up of the two types of amino acids discussed, one surface amino acid likely involved in protein-protein interaction (Arg-153) and the other a hydrophobic amino acid serving a more structural purpose (Leu-341). Fig. 3F shows a hypothetical structure and orientation of Arg-42 produced using the PyMOL Molecular Graphics System, version 1.8 (Schrödinger, LLC) that may be involved in protein-protein interaction because it lies on the protein surface and is in a more open portion of the protein. Arg-42 also lies within the NES and may be necessary for proper nuclear export (38,39).

Cancer-associated mutations impair SIRT2 deacetylation of DDR substrates in vitro and in cells but not interaction with substrate
To determine whether the impairment in deacetylase activity of SIRT2 P128L, P140H, and A186V observed with an acetylated p53 Lys-320 peptide used in the Fluor De Lys assay is substrate-specific, we tested for their ability to deacetylate additional SIRT2 substrates involved in the DDR in vitro. Consistent with findings using an acetylated p53 Lys-320 peptide, SIRT2-FLAG P128L showed the greatest impairment in in vitro deacetylase activity compared with SIRT2-FLAG WT when SIRT2 cancer mutations are functional using acetylated ATRIP Lys-32 and CDK9 Lys-48 as substrates (Fig. 4, A and B). SIRT2-FLAG P140H and A186V failed to show the same degree of impairment in deacetylation of ATRIP Lys-32 and CDK9 Lys-48 (Fig. 4, A and B) compared with p53 Lys-320 (Fig. 2C), suggesting that these mutants are less impactful on SIRT2 deacetylase activity and/or impair deacetylation of specific substrates. To determine whether cancer-associated mutations impair the deacetylation of DDR substrates in cells, we expressed SIRT2-FLAG WT and mutants in 293T cells and determined their ability to deacetylate CDK9-HA and endogenous ␣-tubulin. Again, SIRT2-FLAG P128L but not P140H demonstrated significant impairment in deacetylation of CDK9 Lys-48 and ␣-tubulin Lys-40 in cells (Fig. 4, C and E).
Interestingly, SIRT2 A186V also demonstrated decreased deacetylase activity against these substrates in cells (Fig. 4, C and E), implying that A186V impairs SIRT2 deacetylase activity Figure 2. Cancer-associated mutations impair SIRT2 deacetylase activity and protein level but not localization. A, Western blot analysis demonstrating protein levels of SIRT2-FLAG WT and mutants expressed in U2OS cells. B, U2OS cells were transfected with SIRT2-FLAG WT or mutants. 72 h after transfection, cells were incubated with or without 5 nM LMB for 4 h to inhibit SIRT2 nuclear export, fixed, and processed for indirect immunofluorescence using anti-FLAG antibodies (green) and DAPI staining. Representative images are shown. C, in vitro measurements of SIRT2 deacetylase activity using the Fleur de Lys deacetylase assay are shown. SIRT2-FLAG WT and mutants were purified from 293T cells and eluted from FLAG M2 beads. Protein concentration was measured by Coomassie staining with a BSA standard control. 120 ng of SIRT2-FLAG WT and mutants was incubated in a reaction with an acetylated p53 Lys-320 peptide as a substrate. The mean of the relative fluorescence measured from three replicas is shown. Error bars represent S.D. Deacetylase activities of mutants were compared with that of SIRT2-FLAG WT. *, p Ͻ 0.05; **, p Ͻ 0.01. IB, immunoblotting.
for these substrates to a greater degree in cells potentially due to factors present in cells but not in vitro conditions. To determine whether the impairment in deacetylation of CDK9 Lys-48 in cells by SIRT2 P128L and A186V may be due to differences in binding with substrate, we analyzed their interaction by co-IP. No differences in co-IP of CDK9-HA with SIRT2-FLAG WT and mutants was observed despite impairment in deacetylation of co-immunoprecipitated CDK9-HA by SIRT2-FLAG P128L and A186V (Fig. 4D), providing strong evidence that the impairment in deacetylation of CDK9 Lys-48 by SIRT2-FLAG P128L and A186V mutants is not due to an impairment in binding with substrate. As the majority of cancer-associated SIRT2 mutations are in a heterozygous state in patients, to determine whether they exert dominant negative effects on SIRT2 WT deacetylase activity, we performed an in vitro deacetylase assay in which acetylated CDK9-GFP was incubated with SIRT2-FLAG WT alone or with equal amounts of both SIRT2-FLAG WT and one of four SIRT2-FLAG mutants. We observed no significant difference in deacetylation of CDK9 Lys-48 by SIRT2-FLAG WT and mutants compared with SIRT2-FLAG WT alone in this in vitro deacetylase assay (Fig. 4F), suggesting that SIRT2 mutants do not exert a dominant negative effect on SIRT2 WT activity in this context.

Cancer-associated SIRT2 mutations fail to rescue replication stress response (RSR) defects of SIRT2 deficiency
To determine whether cancer-associated mutations are important for the RSR activities of SIRT2, we analyzed recovery from replication arrest in HCT-116 SIRT2-depleted cells complemented with SIRT2-FLAG WT and mutants (Fig. 5A). Ten  hours following release from HU-induced replication arrest, SIRT2-depleted cells expressing SIRT2-FLAG P128L, P140H, and A186V showed a similar impairment in recovery from replication arrest to SIRT2-depleted cells expressing deacetylaseinactive H187Y compared with WT (Fig. 5, B and C). SIRT2depleted cells expressing SIRT2-FLAG P128L, P140H, and A186V also showed comparable HU hypersensitivity to H187Y compared with WT (Fig. 5D), suggesting that P128L, P140H, and A186V impair SIRT2's activity in responding to replication stress. Moreover, SIRT2-FLAG P128L and A186V but not P140H failed to alleviate the impairment in HU-induced GFP-ATRIP foci of SIRT2 deficiency to a similar level as SIRT2-FLAG WT (Fig. 5, E and F), suggesting that these mutants impair the recruitment of ATRIP to stalled replication forks.

Cancer-associated SIRT2 mutations fail to rescue genomic instability of SIRT2 deficiency
To determine whether cancer-associated mutations contribute to impairment in genome stability, we generated U2OS SIRT2 knock-out (KO) cells utilizing CRISPR/Cas9 and performed complementation experiments with expression of SIRT2-FLAG WT and mutants (Fig. 6A). U2OS SIRT2 KO cells demonstrated a significant increase in spontaneous ␥H2AX foci, a marker for DNA double-strand breaks, which can form from the collapse of stalled replication forks, compared with U2OS SIRT2 WT cells, and this was reduced by expression of exogenous SIRT2-FLAG WT (Fig. 6, B and C). Expression of SIRT2-FLAG P128L, P140H, and A186V all showed impairment in rescue of the spontaneous induction of spontaneous ␥H2AX foci observed with SIRT2 deficiency, suggesting that P128L, P140H, and A186V impair SIRT2's ability to maintain genome integrity. As a more direct measure for genomic instability, we analyzed induction of spontaneous micronuclei, a marker for chromosomal breaks, in U2OS SIRT2 KO cells complemented with SIRT2-FLAG WT and mutants. Expression of SIRT2-FLAG P128L, P140H, and A186V also failed to reduce the spontaneous micronuclei of SIRT2 deficiency to a similar extent as expression of SIRT2-FLAG WT (Fig. 6, D and E). Collectively, these data suggest that P128L, P140H, and A186V impair SIRT2's activity in the DDR, which leads to genomic instability.

Discussion
Using structural insight combined with bioinformatics and functional analyses, we found that naturally occurring cancerassociated SIRT2 mutations at evolutionarily conserved sites disrupt its deacetylation of DDR proteins by impairing its cat-alytic activity or protein level but not localization or binding with substrate, demonstrating that cancer-associated SIRT2 mutations are functionally significant. We further found that SIRT2 mutant proteins fail to restore the replication stress sensitivity, impairment in recovery from replication stress, and spontaneous induction of DNA damage of SIRT2 deficiency in cancer cells, providing evidence that cancer-associated SIRT2 mutations impair genomic integrity of cancer cells. Thus, somatic SIRT2 mutations in human tumors contribute to genomic instability by impairing its deacetylase activity or protein level in the DDR, providing a model for the loss of the tumor suppressor function of SIRT2 in human tumors.
Of the original nine SIRT2 mutations selected, only four demonstrated functional significance by impacting either deacetylase activity or protein level. These four mutations are located within the deacetylase active site, NES, or NAD ϩ cofactor-binding site. R42P resulted in decreased protein levels and likely affects either protein stability or expression. P128L, P140H, and A186V resulted in decreased SIRT2 deacetylase activity but had no significant effect on SIRT2 protein levels or localization.
P128L showed the greatest deficit in SIRT2 deacetylase activity against all substrates tested, including ATRIP Lys-32, CDK9 Lys-48, ␣-tubulin Lys-40, and p53 Lys-320. In addition, P128L was unable to rescue the replication stress sensitivity, impairment in recovery from replication arrest, impairment in HUinduced ATRIP focus accumulation, and spontaneous induction of ␥H2AX foci and micronuclei of SIRT2 deficiency. Structural analyses indicate that this lack of activity is due to decreased binding of the SIRT2 cofactor NAD ϩ . The P128L proline is essential to the proper orientation of the ␣ helix involved in forming the binding pocket for NAD ϩ and for the formation of the stacking interaction of Pro-140 with NAD ϩ . Consequently, its mutation would result in improper orientation/folding and therefore proper interaction with the essential cofactor for deacetylation reactions.
P140H appears to be partially substrate-specific as it was unable to deacetylate p53 Lys-320 in vitro to the same extent as SIRT2 WT but was able to deacetylate other known substrates CDK9 Lys-48, ATRIP Lys-32, and ␣-tubulin Lys-40 both in vitro and in cells. Consistently, P140H rescued the impairment in HU-induced ATRIP focus accumulation of SIRT2 deficiency to a comparable level as SIRT2 WT. However, P140H was unable to rescue the genome maintenance defects of SIRT2 deficiency, suggesting that its effect on the DDR activities of SIRT2 may be mediated through p53 or an untested substrate.

SIRT2 cancer mutations are functional
Interestingly, even in the absence of exogenous damage, SIRT2depleted cells complemented with SIRT2 P140H showed a significant increase in the percentage of cells in S and G 2 /M phases even compared with SIRT2 depletion alone, suggesting a gain of function in impairing cell cycle progression. Structural analyses indicate that this mutation impairs the ability of SIRT2 to bind NAD ϩ as it affects the orientation of the same ␣ helix as P128L and is involved in a stacking interaction between Phe-190 and NAD ϩ in the cofactor-binding pocket. It is likely that this mutation is not as impactful as P128L in disrupting the binding of NAD ϩ to impair deacetylase activity for the majority of known SIRT2 substrates but may impart a gain of function in impairing cell cycle progression, which leads to increased genomic instability.
A186V also exhibited decreased deacetylase activity for p53 in vitro but only displayed decreased deacetylase activity on CDK9 Lys-48 and ␣-tubulin Lys-40 in cells but not on CDK9 Lys-48 or ATRIP Lys-32 in vitro. A186V was unable to rescue the deficiency in ATRIP focus formation in response to replication stress after SIRT2 knockdown to the same extent as SIRT2 WT and P140H likely due to its reduced deacetylase activity of RSR substrates in cells. Additionally, there was an increase in the percentage of cells in S and G 2 /M phases as seen with the P140H mutation as well as impairment in SIRT2 DDR phenotypes. Structural analyses reveal that Ala-186 lies adjacent to H187Y, which is an essential catalytic amino acid for deacetylation reactions. Specifically, Ala-186 lies in a narrow pocket that is size-restricting. The addition of methyl groups found in the valine amino acid structure would interfere with the structure of the deacetylation pocket and orientation of His-187, leading to decreased deacetylation of substrate. It is possible that this mutation may be more impactful on certain substrates in cells due to the presence of an unknown regulatory partner, for example CDK9 complexes with several regulatory cyclins in cells.
The other five mutations, S73R, R153C, F190V, E203G, and L341M, did not significantly affect SIRT2 protein levels, localization, or deacetylase activity. S73R and L341M lie outside of well defined functional SIRT2 protein domains, and although it is possible that S73R may affect protein function as a site of post-translational modification by phosphorylation, it was predicted to be a benign mutation by PolyPhen and ConSurf based on conservation and structural analyses. Leu-341 is a hydrophobic amino acid likely involved in maintenance of a C-terminal ␣ helical structure and, although highly conserved, was predicted to be a tolerated mutation by SIFT and Grantham prediction. Its higher score on the ConSurf scale is likely due to its functionality in an ␣ helix structure that seems to lack an essential role in SIRT2 deacetylase activity. R153C, F190V, and E203G lie in the SIRT2 deacetylase domain, but only F190V lies in the active site (specifically in the NAD ϩ -binding pocket). Despite its seemingly essential placement in the stacking interaction with NAD ϩ and Pro-140, the mutation of Phe-190 alone was not sufficient to induce a deacetylase-deficient phenotype. R153C and E203G lie on the surface of the protein structure as opposed to inside the active site and are likely involved in protein-protein interactions that are not critical for SIRT2's ability to bind and deacetylate substrate.
Interestingly, we found no evidence that cancer-associated SIRT2 mutants exert a dominant negative effect on the deacetylase activity of SIRT2 WT using an in vitro deacetylase assay. However, it is possible that this could differ in a heterozygous state in human cancers in vivo.
In summary, our findings support a model for the tumorsuppressive function of SIRT2 in which somatic mutations in SIRT2 contribute to genomic instability by impairing its deacetylase activity or protein level in the DDR and provide a mechanistic basis for understanding the biological and clinical significance of SIRT2 mutations in genome maintenance and tumor suppression. In addition, as SIRT2 depletion in cancer cells results in hypersensitivity to replication stress caused by many types of cancer therapies (18 -20), our findings provide insight into the rationale-driven design of targeted therapies against SIRT2 activity and potential resistance mechanisms in tumors that may be utilized for individualized cancer therapy.

Cell cycle recovery
HCT116 cells were transfected with ATR, NS, or SIRT2 siRNA and 24 h later with SIRT2-FLAG WT (Addgene 13813) (21) or mutants, treated with or without 3 mM HU 72 h postknockdown for 20 h (arrested) (41), and released into 0.5 g/ml nocodazole (Fisher Scientific) for 10 h. Cells were harvested and fixed in ice-cold 70% ethanol, and DNA was stained with 25 g/ml propidium iodide (Sigma-Aldrich) with 100 g/ml RNase A (Qiagen). DNA content was measured by flow cytom- etry using a BD FACSCanto II flow cytometer and then analyzed by the FlowJo software gating analysis tool (Tree Star).

Immunofluorescence
For SIRT2 localization experiments, U2OS cells were transfected with SIRT2-FLAG WT or mutants, treated 72 h posttransfection with 5 nM leptomycin B (Sigma) for 4 h, fixed in 4% paraformaldehyde for 10 min, and permeabilized in 0.5% Triton X-100 for 10 min. Cells were blocked in 5% BSA and immunostained with anti-FLAG (Cell Signaling Technology, 2368P) and Alexa Fluor 555 anti-rabbit secondary antibody (Invitrogen) followed by DAPI stain (Southern Biotech). For GFP-ATRIP experiments, a stably transfected GFP-ATRIP U2OS cell line was used. Cells were transfected with nontargeting siRNA or siSIRT2-5 followed by transfection with pcDNA3.1 empty vector (EV) or SIRT2-FLAG WT or mutants. SIRT2-FLAG constructs contained wobble mutations to protect exogenous expression from knockdown. Cells were treated with 3 mM HU for 24 h prior to fixation, fixed, and blocked as described above. Cells were immunostained with anti-FLAG (Cell Signaling Technology, 2368P) followed by Alexa Fluor 555 secondary antibody (red) and DAPI Stain. GFP-ATRIP did not require antibody (green). SIRT2-FLAG-positive cells were used for quantitation of these experiments with the exception of EVtransfected conditions. For spontaneous ␥H2AX focus experiments, we utilized SIRT2 WT or KO U2OS cells transfected with or without SIRT2-FLAG WT or mutants (42). Cells were fixed and blocked as described above, immunostained with anti-FLAG (Cell Signaling Technology, 2368P) and anti-␥H2AX (Millipore, 05-636) followed by Alexa Fluor 488/555 secondary antibodies and DAPI stain. SIRT2-FLAG-positive cells were used for quantitation of these experiments with the exception of non-SIRT2-transfected conditions. The percentage of cells positive for GFP-ATRIP or ␥H2AX foci was counted from three replicas of 100 cells each. Micronuclei experiments were fixed and processed as described above. Only SIRT2-FLAG-positive cells were used in quantitation analysis. The percentage of cells positive for micronuclei was counted from three replicas with 100 cells each. All images were captured on a Zeiss Observer Z1 microscope equipped with AxioVision Rel 4.8 software.

Generation of CRISPR/Cas9 SIRT2 knock-out cells
U2OS/HCT116cells were transfected with plasmid containing guide RNAs targeting an early SIRT2 exon or a nontargeting control and Cas9-GFP construct (Sigma). 72 h after transfection, cells were harvested and sorted by FACS into 96-well plates with one cell per well based on a high level of GFP expression. Single cells that survived and gave rise to cell lines were tested for SIRT2 knock-out by Western blot analysis.

In vitro or cellular deacetylation experiments
For in vitro deacetylation analysis of CDK9 and ATRIP, 293T cells were transiently transfected with CDK9-GFP or HA-ATRIP and histone acetyltransferase (p300/CBP/pCAF) or SIRT2-FLAG WT or mutants and treated with 0.5 M TSA and 20 mM nicotinamide for 12 h. Cells were lysed with IP buffer (20 mM Hepes, pH 7.4, 180 mM KCl, 0.2 mM EGTA, 1.5 mM MgCl 2 , 20% glycerol, 1.0% Nonidet P-40) supplemented with 1 M TSA and fresh protease inhibitors. Acetyl-CD Lys-9 or acetyl-ATRIP were immunoprecipitated using anti-GFP (Abcam, Ab6556) and protein A-agarose beads (Invitrogen) or anti-HApreconjugated beads (Sigma), washed with IP buffer containing 1 M TSA to remove nicotinamide, and washed an additional two times with deacetylation buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 1 mM MgCl 2 ). SIRT2-FLAG constructs were immunoprecipitated using anti-FLAG M2-agarose beads (Sigma), washed with IP buffer and TBS (50 mM Tris, pH 7.5, 150 mM NaCl), and eluted with 0.15 mg/ml 3ϫ FLAG peptide (Sigma) for 30 min at 4°C. Protein concentration was determined by Coomassie staining. Substrate on beads (HA-ATRIP or CDK9-GFP) was incubated with or without 1 g of SIRT2-FLAG WT or mutants in 24 l of deacetylation reaction buffer (1 M TSA, 50 mM Tris⅐HCl, pH 7.5, 150 mM NaCl, 1 mM MgCl 2 , 1 mM NAD ϩ ) at 30°C for 3 h. For the dominant negative assay, CDK9-GFP was incubated with 1 g of SIRT2 WT or with 0.5 g of SIRT2-WT and 0.5 g of one of the four mutants tested. The reaction was stopped by the addition of 5ϫ SDS loading buffer, and samples were analyzed for acetylation by Western blotting with anti-Ac-ATRIP Lys-32 (19 -21) and Ac-CDK9 Lys-48 (18). For cellular deacetylation analysis, 293T cells were transiently cotransfected with CDK9-HA, histone acetyltransferases (p300/CBP/pCAF), and SIRT2-FLAG WT or mutants. Cells were cultured with 0.5 M TSA for 12 h prior to being lysed with IP buffer containing 1 M TSA, and protein lysates were immunoprecipitated using HA-preconjugated beads. The immunocaptured protein were analyzed for deacetylation by immunoblotting with anti-AcCDK9 K48, anti-HA, ␣-tubulin, and Ac-tubulin Lys-40. For the Flour de Lys in vitro experiments, SIRT2-FLAG WT and mutants were purified from 293T cells and eluted from FLAG M2 beads as described above. 120 ng of SIRT2-FLAG was used following the manufacturer's instructions (Enzo Life Sciences). All deacetylation experiments, both in vitro and in cells, were performed at least four times.

Cell viability assay
HCT116 cells were transfected with 40 nM siRNA using RNAiMAX reagent and split followed by transfection with SIRT2-FLAG WT or mutants, treatment with or without 1.6 mM HU 72 h postknockdown for 24 h, and recovery for 24 h. Viability was measured using alamarBlue (Thermo Fisher) on a Spectra Plus plate reader. A ratio of treated/untreated viability was calculated and normalized to that of nontargeting siRNA (43).

Sequence alignment
Protein BLAST (blastp) was used through the NCBI website to compare sequence homology among closely and distantly related organisms with the human SIRT2 protein sequence obtained from UniProt and to align protein sequences.