The deacylase SIRT5 supports melanoma viability by influencing chromatin dynamics

Abstract

Cutaneous melanoma remains the most lethal skin cancer, and ranks third among all malignancies in terms of years of life lost. Despite the advent of immune checkpoint and targeted therapies, only roughly half of patients with advanced melanoma achieve a durable remission. Sirtuin 5 (SIRT5) is a member of the sirtuin family of protein deacylases that regulates metabolism and other biological processes. Germline Sirt5 deficiency is associated with mild phenotypes in mice. Here we showed that SIRT5 was required for proliferation and survival across all cutaneous melanoma genotypes tested, as well as uveal melanoma, a genetically distinct melanoma subtype that arises in the eye and is incurable once metastatic. Likewise, SIRT5 was required for efficient tumor formation by melanoma xenografts and in an autochthonous mouse Braf Pten-driven melanoma model. Via metabolite and transcriptomic analyses, we found that SIRT5 was required to maintain histone acetylation and methylation levels in melanoma cells, thereby promoting proper gene expression. SIRT5-dependent genes notably included MITF, a key lineage-specific survival oncogene in melanoma, and the c-MYC proto-oncogene. SIRT5 may represent a druggable genotype-independent addiction in melanoma.

Keywords: Apoptosis; Cell Biology; Melanoma; Metabolism; Molecular biology.

Figure 1. Increased SIRT5 copy number in human melanoma.

(A) Gain of extra SIRT5 copies in melanoma. BRAF, NRAS, PTEN, MITF, NF1 and other sirtuins are shown for comparison (n = 287; data from TCGA, Provisional, analyzed on cBioPortal). ND, not determined. Percentage of samples with any genomic alteration (Any) or amplification or gain (Amp/Gain) is indicated. Graphed are any alterations queried for the indicated gene. Copy number gain indicates a low-level gain of a single additional copy, and amplification refers to high-level amplification (multiple extra copies). Results from the query (GENE: MUT AMP HOMDEL GAIN HETLOSS) in cBioPortal were analyzed and plotted. (B) Kaplan-Meier analysis of overall survival in melanoma patients with or without copy number gain or amplification of SIRT5. Overall survival was analyzed using the query, “SIRT5: AMP GAIN.” (C) SIRT5 (6p23) and centromere 6p (Cen6p) amplification (amp) or coamplification (Co-amp) in melanoma, as assayed by FISH staining (n = 32). (D) Sirtuin gene copy number (CN) in human melanoma samples, as assayed by high density SNP array (n = 139). (E) SIRT5 mRNA expression levels in melanoma correlate with Clark’s level (P = 0.0044, linear regression; P = 0.037, 1-way ANOVA). (F) SIRT5 protein levels are increased in melanoma relative to benign melanocytic lesions (P = 0.0333, χ²; n = 14 nevi, n = 87 melanoma). See also Supplemental Figure 1 and Supplemental Table 1.

Figure 2. SIRT5 is required for melanoma cell growth and survival.

(A) The BRAF or NRAS mutant melanoma cell lines indicated were infected with a nontargeting shRNA (control) or 1 of 2 SIRT5 shRNAs (KD1 or KD2). Equivalent cell numbers were then plated 48 hours after transduction into 96-well plates in the presence of puromycin. Cell mass was determined at the indicated time points via WST-1 assay, with absorbance measured at 450 nm. Average results (n = 6/time point) are graphed. Error bars represent standard deviation. Representative of 5 of 5 SIRT5 shRNAs tested (see also Figure 3B). (B) SIRT5 KD results in significantly (P < 0.0001, 1-way ANOVA) impaired colony formation by A2058 and SK-MEL-2 cells 12 days after transduction. Cell mass was assayed using crystal violet staining, with absorbance measured at 590 nm. Average of n = 12 technical replicates is plotted. Error bars represent standard deviation. Representative (n = 4) crystal violet–stained wells are shown. Bottom, representative immunoblot analysis demonstrating SIRT5 KD. (C) Top, viability of A2058 cells transfected with the indicated CRISPR guide RNA (Control or G1–G4). Cell mass was assayed using crystal violet staining, with absorbance measured at 590 nm. Average of n = 9 technical replicates is plotted. Error bars represent standard deviation. Significance calculated using 1-way ANOVA. Bottom, representative immunoblot analysis confirming CRISPR-mediated SIRT5 loss (Control: empty vector).

Figure 3. SIRT5 depletion rapidly induces apoptosis in melanoma cells.

(A) Immunoblot analysis demonstrating induction of caspase 3 cleavage 72 and 96 hours after transduction with shRNAs SIRT5-KD1 -KD2 in A2058 and SK-MEL-2 cell lines. (B) Viability of MP-41, A2058, or YUMM5.2 cells infected with control (C) or 1 of 5 SIRT5 shRNAs (KD1–KD5) against human SIRT5 (top and middle panels) or mouse Sirt5 (bottom panel). Average results (n = 6/time point) are graphed. Error bars represent standard deviation. Right panels: immunoblot analysis demonstrating loss of SIRT5 and induction of caspase 3 cleavage following SIRT5 KD. (C) Flow cytometric analysis of A2058 cells stained with Annexin V and propidium iodide (PI), as indicated, showing an increased fraction of Annexin V–positive cells 96 hours after SIRT5 KD. (D) Average of n = 3 technical replicates is plotted. Error bars represent standard deviation. Significance calculated using unpaired 2-tailed Student’s t test. Increased Annexin V⁺ staining is observed in both the PI-positive and PI-negative populations.

Figure 4. SIRT5 loss-of-function inhibits melanoma tumor growth in vivo.

(A) SIRT5 depletion in A2058 cells results in attenuated xenograft tumor growth. Quantification of tumor size was initiated on day 13 after initial injection of cells (left panel). Tumor size was recorded with Vernier calipers on the days indicated. Each point represents the measurements on n = 5 mice for each condition (C, KD1, or KD2). Pairwise representation of endpoint tumor size in each mouse within each group is plotted (right panel). Average tumor mass measurements at day 28 are plotted (P < 0.05, paired 2-tailed t test for each group). Error bars represent standard deviation. (B) Mice were sacrificed, and tumors were dissected at 28 days after initial injection. Scale bar below tumors: 2 cm. (C) SIRT5 deficiency attenuates tumor formation in an autochthonous melanoma model. Sirt5-deficient mice were bred into the Braf^CA Pten^fl/fl Tyr:CreER background (55). Melanomas were induced in littermate male Sirt5-WT or Sirt5-KO mice as shown by topical application of 4HT at ages 4 to 9 weeks; tumors were weighed following euthanasia. Averages of 5 sets of male mice are plotted (P < 0.05, paired 2-tailed t test). Mean ± standard deviation are shown. (D) SIRT5 immunoblot of a representative tumor from a Sirt5-WT or -KO male or female mouse (left). Representative tumor from a Sirt5-WT or KO male mouse, as indicated, after 4HT induction (right). Scale bar: 1 cm.

Figure 5. Bioenergetics are maintained upon SIRT5 loss in melanoma cells.

A2058 and A375 cells maintain glycolytic function (A), glucose-dependent mitochondrial respiration (B), and ATP production (C) upon SIRT5 depletion compared with control cells. Mitochondrial respiration, glycolytic stress tests, and ATP production rates were measured at 72 hours after transduction with shRNAs against SIRT5 using a Seahorse XFe96 Analyzer. All rates are normalized to total protein content per sample (n = 6 for A and C, n = 5 for B). OCR, oxygen consumption rate; ECAR, extracellular acidification rate. (D) Mitochondrial membrane potential is stable in A2058 cells after SIRT5 loss (C, control cells, n = 6). Cells were incubated with JC-1, a dye that exhibits membrane potential–dependent accumulation in mitochondria, indicated by a fluorescence emission shift from green to red. Mitochondrial depolarization is indicated by a decrease in the red/green (aggregate/monomer) fluorescence intensity ratio. FCCP, a mitochondrial uncoupler, depolarizes mitochondrial membrane potential and is used as a positive control. Error bars represent standard deviation. Significance calculated using 1-way ANOVA.

Figure 6. Transcriptomic analysis reveals MITF dependency on SIRT5 expression.

Genes (A) upregulated or (B) downregulated upon SIRT5 KD. Only genes significantly (P < 0.05) altered in both KDs in each cell line, as indicated, were scored. (C) Expression levels of DEGs (qadj < 0.05) in response to SIRT5 KD were correlated with SIRT5 gene expression using Spearman’s rank correlation coefficient in 443 sequenced human skin cutaneous melanoma (SKCM) samples, identifying DEGs with significant clinical correlation with SIRT5 expression (q < 0.01). Labeled genes represent oncogenes or extremely correlated genes most significantly altered by SIRT5 KD (q < 0.0001, log₂ fold change > 2). (D) Expression of SIRT5, MITF, and the MITF target, PPARGC1A, are positively correlated in melanoma clinical samples (P < 0.0001, data from TCGA, analyzed on cBioPortal; see Figure 1A).

Figure 7. Expression of MITF and MITF target genes is dependent upon SIRT5.

(A) Immunoblot demonstrating loss of MITF expression 96 hours after transduction with shRNAs SIRT5-KD1 or -KD2 compared with a nontargeting control in 5 cutaneous and 1 uveal melanoma cell line, as indicated. (B) Relative FPKMs in A2058 and SK-MEL-2 cells demonstrate a loss of MITF (bar graphs, upper panels) and several MITF target gene transcripts upon SIRT5 KD (heatmaps, lower panels). Scale bars adjacent to heat maps indicate linear fold change (control set to 1). Error bars represent standard deviation. Significance calculated using 1-way ANOVA. C, control.

Figure 8. SIRT5 promotes histone acetylation in melanoma.

(A) Heatmap of z scores calculated from metabolic reaction fluxes predicted by genome-scale modeling to be differentially active (P < 0.01) after SIRT5 KD. (B) Total histone acetylation is reduced 96 hours after transduction with shRNAs SIRT5-KD1 or -KD2 compared with a nontargeting control in melanoma cell lines. Lanes were run on the same gel but are noncontiguous. (C) Immunoblot demonstrating loss of H3K9ac and H4K16ac 96 hours after transduction with shRNAs SIRT5-KD1 or -KD2 compared with a nontargeting control in A2058 cells. (D) H3K9ac is reduced within the promoter regions of MITF and c-Myc in SIRT5-depleted A2058 cells via CUT&RUN followed by qRT-PCR. Signal (Ct values) relative to input DNA were normalized to control samples for each primer set. Graphed are averages of n = 9 replicates. Error bars represent standard deviation. Significance calculated using 1-way ANOVA. Acetylation (E) and MITF expression (F) are restored in A2058 cells lacking SIRT5 after 4 weeks of continual culture in puromycin. (G) Total cellular acetyl-CoA levels are increased in A2058, A375 and SK-MEL-2 cells 96 hours after SIRT5 depletion. Acetyl-CoA abundance was quantified by liquid chromatography–high resolution mass spectrometry and normalized to cell number. Plotted are average (n = 5) acetyl-CoA levels as pmol acetyl-CoA/10⁵ cells. Error bars represent standard deviation. Significance calculated using 1-way ANOVA. C, control.

Figure 9. SIRT5 promotes histone methylation and reduced cellular ROS levels in melanoma.

(A) LC-MS/MS-based metabolite profiling followed by MetaboAnalyst pathway analysis demonstrate alterations in glycine and serine and methionine biosynthesis pathways in melanoma cells upon SIRT5 depletion. (B) Perturbations in 1C metabolite levels in response to SIRT5 loss in the cell lines shown. Each column represents the mean of 3 independently prepared biological replicates. Metabolite levels in SIRT5-depleted (KD1 and KD2, as indicated) samples are normalized to control. SAM, S-adenosyl-methionine; SAH, S-adenosylhomocysteine; GSH, reduced glutathione; GSSG, glutathione disulfide. (C) H3K4me3 and H3K9me3 immunoblot in melanoma cells 96 hours after transduction with shRNAs SIRT5-KD1 or -KD2 compared with a nontargeting control. (D) H3K4me3 and H3K9me3 levels are restored in A2058 cells lacking SIRT5 after 4 weeks of continual culture in puromycin. (E) Flow cytometric analysis of DCFDA-stained A2058 cells 96 hours after transduction with shRNAs SIRT5-KD1 or -KD2 reveals increased ROS compared with a nontargeting control, P < 0.005. Left panel, average mean fluorescence intensity of DCFDA positive populations in n = 3 samples. Error bars represent standard deviation. Significance calculated using 1-way ANOVA. Right panel, representative (n = 6) flow cytometric analysis of A2058 cells stained with DCFDA. (F) SIRT5 interacts with MTHFD1L in A2058 cells. Increasing amounts of anti-SIRT5 antibody increases SIRT5-MTHFD1L coprecipitation compared with normal rabbit IgG control. Basal expression of SIRT5 and MTHFD1L in whole-cell extract (1% of initial amount used for immunoprecipitation) is shown for comparison. (G) Proposed model of promotion of MITF and c-MYC expression via SIRT5-dependent chromatin modifications in human melanoma. Me, methylation; Ac, acetylation. C, control.