SETD2 Haploinsufficiency for Microtubule Methylation Is an Early Driver of Genomic Instability in Renal Cell Carcinoma

Many cancer types are characterized by large-scale, characteristic alterations of chromosomal copy number. Clear cell renal cell carcinoma (ccRCC) is characterized by frequent inactivating mutations of the Von Hippel–Lindau (VHL) tumor suppressor gene (1), with accompanying loss of heterozygosity (LOH) provided by a large deletion of chromosome 3p (2). Recent large-scale multiplatform sequencing techniques identified that, in addition to VHL, other tumor suppressor genes encoding epigenetic regulators such as Polybromo 1 (PBRM1/BAF180), BAP1, and Set domain containing 2 (SETD2) are implicated in ccRCC tumorigenesis (3–5). Importantly, all of these commonly mutated genes reside in the same portion of affected chromosome 3p. Loss of 3p results in LOH for all four of these genes, which individually acquire “second-hit” mutations and loss of function of the remaining allele in subsets of ccRCC. Of these, SETD2 is of particular interest, as this tumor suppressor has been increasingly identified as a common mutation across cancer types, including lung (6), bladder (7), glioma (8), and leukemia (9).

SETD2 encodes a methyltransferase known to be the sole enzyme responsible for the trimethylation of lysine 36 on histone H3 (H3K36me3; refs. 7, 8, 10–12). Biallelic deficiency of SETD2 via deletions and inactivating mutations occur in up to 20% of primary human RCC tumors and it is associated with more advanced disease and the metastatic phenotype, typically lethal within 1–5 years (13). Biallelic loss of SETD2 has been shown to result in loss of H3K36me3 in ccRCC-derived cells and tumors (9, 14, 15). Examination of H3K36me3 status in ccRCC cells of metastatic tumor specimens suggest that SETD2 mutations may occur in over 50% of metastatic lesions (16). Furthermore, a study of ccRCC intratumoral heterogeneity identified distinct SETD2 mutations across subsections of an individual tumor, suggesting a selection bias for SETD2 mutation in the course of ccRCC development (7).

SETD2 is a multidomain-containing protein with distinct functions for each domain. The methyltransferase activity is mediated by a centrally located SET domain. Mutations in this domain are common in ccRCC (10, 14), suggesting loss of catalytic activity is a critical event in tumor development. We previously characterized a pathogenic SET domain mutation found in ccRCC, an arginine-to-cysteine mutation at residue 1625 of SETD2 (R1625C; ref. 15), which abolishes methylation activity. At its C-terminus, SETD2 also contains the Set2–Rpb1 interaction (SRI) domain (17). This domain mediates the interaction between SETD2 and the phosphorylated C-terminal domain of RNA polymerase II (RNAPII). We also identified a recurrent mutation in the SRI domain, an arginine-to-histidine mutation at residue 2510 (R2510H; ref. 15). This mutation preserves the H3 trimethylation catalytic activity of SETD2, suggesting SETD2 may have other key functions in addition to its well-characterized role as a histone methyltransferase.

We recently discovered that SETD2 also functions as a microtubule methyltransferase, in addition to the well-characterized role of SETD2 in histone methylation (18). SETD2 trimethylates α-tubulin on lysine 40 (αTubK40me3) of microtubules and loss of this marker results in genomic instability. SETD2-null cells exhibit mitotic spindle and cytokinesis defects, polyploidy, and formation of micronuclei. Importantly, pathogenic SETD2 mutations in the SET domain as well as the SRI domain were unable to methylate microtubules, and caused an increase in chromosome bridges and lagging chromosomes relative to wild-type SETD2, indicating that in addition to the catalytic domain, a functional SRI domain was also required for αTubK40me3 (18).

These mitotic alterations caused by loss of αTubK40me3 can lead to chromosomal abnormalities and genomic instability, hallmarks of tumorigenesis, and are thought to be an important source of genetic diversity and development of cell clones during tumor progression (19). In the case of the type of defects observed with SETD2 mutants deficient in microtubule methylation (lagging and bridging chromosomes), this genomic instability results in the formation of micronuclei. Micronuclei contain acentric chromosome fragments, acentric chromatid fragments, or whole chromosomes that failed to migrate during mitosis, which are enclosed by nuclear membrane (20). The presence of micronuclei is a reliable cytologic indicator of chromosome instability (21), and micronuclei are a common feature of many solid tumors and preneoplastic lesions (19, 20), but have not been studied in any detail in ccRCC to date.

Here, we report that SETD2′s ability to trimethylate microtubules and preserve genomic stability is dose dependent, and SETD2 haploinsufficiency or reduced dosage is sufficient to impair genomic stability and induce micronuclei formation. Using micronuclei as a readout of genomic instability in wild-type (WT) SETD2 and SETD2 disrupted human kidney proximal tubule epithelial cells (HKC), and we confirmed that SETD2 loss causes a significant increase in micronuclei. To directly demonstrate that SETD2 haploinsufficiency was sufficient to induce genomic instability (micronuclei), we induced loss of a single copy of Setd2 using Setd2^w/f mouse embryonic fibroblasts (MEF). Monoallelic loss of Setd2 was sufficient to increase lagging and bridging chromosomes, and caused a significant increase in micronuclei formation. Importantly, while promoting micronuclei accumulation, monoallelic SETD2 loss had no measurable effect on H3K36me3, but did cause loss of αTubK40me3, revealing a previously unrecognized effect of SETD2 dosage. Comparing different SETD2 mutations, we also found that the SRI domain mutation, which rescues H3K36me3, but not αTubK40me3, fails to rescue micronuclei formation and mitotic defects in SETD2-deleted cells compared with WT SETD2. Importantly, micronuclei were also a common feature of 3p-deleted ccRCC cell lines and were seen in a significant fraction of tumor cells from all patients with ccRCC. These data indicate that SETD2 is haploinsufficient for αTubK40 trimethylating activity, with loss of microtubule methylation driving early genomic instability in tumors with 3p loss, preceding the loss of histone H3K36 trimethylation, and contributing to the progression of ccRCC. Moreover, our results shed light on a new mechanism by which monoallelic inactivation of SETD2 or 3p loss can drive tumorigenesis in other cancers, and points to microtubule methylation as a key tumor suppressive activity of SETD2.

Human renal cell carcinoma ACHN, UMRC2, Caki-1, and A498 cells were acquired from the ATCC. RCC4 cells were generously provided by Dr. William Kim (University of North Carolina, Chapel Hill, NC). The SV-40–transformed human renal tubule epithelial cell line (HKC) was obtained from Dr. Lorraine Racusen, (Johns Hopkins Hospital, Baltimore, MD). All cells undergo annual short tandem repeat (STR) confirmation. Human cell lines were cultured in DMEM (Gibco/Life Technologies) supplemented with 10% FBS (Gemini Bio-Products), nonessential amino acids, l-glutamine, penicillin, and streptomycin. For Setd2^w/f and Setd2^f/f MEFs, cells were maintained in phenol red–free media (DMEM, high glucose, HEPES, no phenol red) supplemented with sodium pyruvate and GlutaMAX (Thermo Fisher Scientific), to prevent any spontaneous activation of the Cre recombinase by estrogenic compounds found in phenol red–containing media. All cultures were maintained at 37°C in 5% CO₂. The identify of all human cell lines used in this study has been authenticated using the STR profiling test offered by ATCC as a service. All cell lines have been tested for Mycoplasma contamination using the PCR-based ATCC's Universal Mycoplasma Detection test (catalog number: 30-10102K).

The detailed information for generating Setd2^w/f and Setd2^f/f MEFs was described in a previous publication (18). In brief, Setd2^w/f or Setd2^f/f MEFs were generated from 13.5 days postcoitum (d.p.c.) embryos of Setd2^w/f or Setd2^f/f mice, respectively. The cells were spontaneously immortalized via serial passaging. The immortalized cells were transfected with an ER-Cre vector expressing Cre recombinase fused with a mutated ligand-binding domain for the human estrogen receptor (ER-Cre). Stable cell lines expressing ER-Cre were generated by selection using 5 μg/mL blasticidin (Thermo Fisher Scientific).

Setd2^w/f or Setd2^f/f MEFs expressing ER-Cre were treated with 3 μmol/L 4-hydroxytamoxifen (4-HT, the active metabolite of tamoxifen; Sigma-Aldrich) or vehicle (0.01% ethanol) for 2 days (or as indicated) for efficient Setd2 knockout. The phenotype of Setd2^w/f MEFs is similar to that observed with biallelic loss (Setd2^f/f; ref. 18), for which micronuclei were counted after two days of tamoxifen treatment, while the Setd2^w/f MEFs remain viable as a culture for several passages.

To establish HKC SETD2 null cells, HKC cells were transfected as described previously (15) with a pair transcription activator–like effector nucleases (TALEN) constructs targeting exon 3 (5′-TCATGTAACATCCAGGCC-3′ and 5′-ACAGCAGTAGCATCTCCA-3′) using Amaxa cell Line Nucleofector Kit V (Lonza) and single cell sorted to isolate single clones. Individual clones were screened for loss of H3K36me3 and sequenced at the SETD2 target site to confirm the presence of inactivating mutations. The allelic sequencing confirmed the SETD2 inactivation through frameshift mutations at the TALEN target site in both SETD2 alleles.

HKC cells were infected with shRNA virus expressing either a Scramble (SCR) or shRNA against SETD2 (TRCN0000237839; Sigma). shSETD2 sequence obtained from the Broad Institute shRNA Library. Cells were passaged 24 hours postinfection in media containing puromycin (1 μg/mL). Following drug selection, cells were passaged once more at day 3 post shRNA virus infection. Subsequently, cells were fixed with either paraformaldehyde or methanol for immunostaining at 6 days postinfection.

Genomic DNA was isolated from confluent 10-cm plates using the Qiagen DNeasy Blood & Tissue Kit (Qiagen 69504) per manufacturer's standard protocols. SNP array was conducted on the Infinium Core-24 Kit BeadChip (Illumina) by UNC Mammalian Genotyping Core according to manufacturer's standard protocols. The SNP data were preprocessed with Genomestudio v2 (Illumina) and the R package, genoCNV was used to estimate the copy number states and genotype calls for all interested cell lines. The allele-specific copy number analysis of tumors (ASCAT) algorithm (version 2.5; ref. 22) was used to calculate somatic whole-genome ASCAT-profiles, as well as estimates of cell ploidy.

Cells were lysed by gentle pipetting several times and incubated on ice for 10 minutes in a cell fractionation buffer [320 mmol/L sucrose, 10 mmol/L Tris-HCl (pH 8.0), 2 mmol/L magnesium acetate, 3 mmol/L calcium chloride, 0.1 mmol/L EDTA, 0.5% NP-40, fresh 1 mmol/L dithiothreitol, protease inhibitor cocktail, 1 mmol/L PMSF, and 1 mmol/L sodium orthovanadate]. Cytoplasmic fractions were collected after centrifugation at 1,000 × g for 5 minutes and were used for Western blots using antibodies against α-TubK40me3 (18, 23) or α-tubulin (MA1-19162; Thermo Fisher Scientific). The nuclear proteins were extracted in cell lysis buffer [20 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, 1 mmol/L EDTA, 1 mmol/L EGTA, 1% Triton X-100, and protease inhibitor cocktail] by sonication for three cycles of 30 seconds on/off using Bioruptor 300 (Diagenode) and used for Western blots using antibodies against SETD2 (HPA042451; Sigma-Aldrich), H3K36me3 (61101, Active Motif), or histone H3 (ab1791; Abcam). The samples were separated on a 4%–15% SDS-PAGE gel and transferred onto a polyvinylidene difluoride (PVDF) membrane. Membranes were blocked in 5% nonfat milk in TBST (50 mmol/L Tris pH 7.5, 150 mmol/L NaCl, and 0.1% Tween 20) for 1 hour and immunoblotted using primary antibodies for overnight and econdary goat anti-mouse IgG-HRP (sc-2005; Santa Cruz Biotechnology) or goat anti-rabbit IgG-HRP (sc-2004; Santa Cruz Biotechnology) for 2 hours.

For SETD2 knockdown HKC cells, lysates were separated on a 4%–15% SDS-PAGE gel and the proteins were subsequently transferred to a PVDF membrane. Membranes were blocked in 5% nonfat milk in TBST and incubated O/N with primary antibodies in milk. Anti-β-tubulin (05-661; EMD Millipore) antibody was used as loading control.

HKC cells were cultured on chamber slides (Corning) for 16–18 hours and fixed using a 4% paraformaldehyde for 15 minutes. After three washes with cold PBS for 5 minutes each, the samples were permeabilized using a 0.25% Triton X-100 in PBS for 10 minutes and followed by 10-minute incubation with 1% hydrogen peroxide. After blocking in 5% BSA, cells were incubated at 4°C overnight with mouse anti LAP2 (1:500; Abcam) or human anti-CREST (1:6,000; Antibody Inc.) primary antibody. After four rounds of vigorous washing for 10 minutes each using PBST, cells were incubated for 1 hour at room temperature with goat anti-mouse IgG Alexa Fluor 488 (Life Technologies, 1:1,000) and goat anti-human IgG H+L 568 (Life Technologies, 1:1,000) secondary antibody. After washing, cells were counterstained using DAPI for subsequent visualization of nuclei. Coverslips were mounted into chamber slides with ProLong Gold Antifade Mountant (Life Technologies). Fluorescence signals were visualized using confocal microscopy (LSM 880, Zeiss). The number of micronuclei per 300 cells was counted and repeated for two times per coverslip. Data from three different coverslips in three independent experiments were manually quantitated.

For shRNA SETD2 knockdown of HKC cells, samples were fixed with cold methanol for 5 minutes at −20°C, followed by three 5 minutes washes in PBS. Cells were then blocked in 2.5% BSA, 1% normal goat serum (NGS) in PBS for 1 hour followed by incubation with anti-LaminB1 (ab16048; Abcam; 1:250) antibody in PBS-BSA-NGS for 1 hour at room temperature (RT). Cells were washed with PBS-0.1% Tween 20-NGS (PBST-NGS), followed by secondary antibody incubation for 30 minutes at room temperature. Cells were counterstained using DAPI (1:1,000) for 5 minutes and then mounted on slides using Prolong Gold Antifade mounting media. Images were acquired on a Zeiss 800 Upright confocal microscope using a 63×/1.4 NA under default filter settings and pinhole of 1 airy unit. For every image, approximately 10 Z-sections were acquired with an average of 0.4-μm thickness per slice. Z-stacks were processed in FIJI (Version 2.0.0) and images presented are maximum intensity projections of Z-stacks. The number of micronuclei per 150 cells per sample (shScramble and shSETD2) was counted and data from three independent experiments were manually quantitated.

Setd2 knockout was induced in MEFs, then fixed in 3.7% formaldehyde solution for 15 minutes, permeabilized in Triton-X 100 for 20 minutes, and incubated for an hour in blocking buffer containing 3% BSA in 1× PBS at room temperature. MEFs were then probed with human anti-CREST antibody (Antibody Inc.) and anti-rabbit CENPB (Abcam, ab25734) for an hour at room temperature. Alexafluor 568- and 488-conjugated secondary antibodies (Invitrogen) were used to detect anti-CREST and anti-CENPB, respectively, whereas DNA was stained using Hoechst 33342 (Invitrogen). Immunofluorescence images were taken using Zeiss Axiovert 200M-inverted Microscope equipped with epifluorescence optics and an Apotome structured illumination module (Carl Zeiss) or a 60 × 1.4 NA objective (Olympus) on a DeltaVision Elite Imaging System (GE Healthcare) equipped with a Cool SnapHQ2 CCD camera (Roper), and processed using ratio deconvolution in SoftWorx (GE Healthcare). Total CREST (+) micronuclei were counted per 100 micronuclei per coverslip from three independent experiments.

Cells were cultured on coverslips and fixed using ice-cold methanol for 10 minutes and followed by permeabilization using a 0.2% Triton X-100 solution in PBS buffer for 5 minutes. After three washes using PBS buffer, the cells were blocked in 2% BSA blocking buffer for 30 minutes at room temperature. Cells were incubated in blocking buffer using primary antibodies overnight at 4°C. Primary antibodies were diluted as follows; rabbit anti-H3K36me3 (1:1,000; Abcam), mouse anti-α-tubulin (1:400, clone DMIA; Sigma), and human anti-centromere (1:6,000; Antibody Inc.). After five rounds of vigorous washing for 10 minutes each using PBST, cells were incubated for 1 hour in blocking buffer with corresponding secondary antibodies [donkey anti-rabbit (clone 488; Life Technologies), goat anti-mouse IgG H&L Cy5 (Abcam), and goat anti-human IgG H+L (clone 568; Life Technologies)] at a dilution of 1:1,000. Following more than three washes, for 10 minutes each using PBST, cells were counterstained using DAPI for subsequent visualization of nuclei. Coverslips were mounted in ProLong Gold Antifade Mountant (Life Technologies) and cells visualized using confocal microscopy (LSM 880; Zeiss).

Human primary RCCs were obtained under an approved Institutional Review Board protocol with patients providing written informed consent to deidentified use of tissue. The studies were conducted in accordance with recognized ethical guidelines (e.g., Belmont Report).

Approval was granted by the Vanderbilt Biomedical Institutional Review Board. Following pathologic assessment, extra tissue was freshly provided, and the cells were dispersed by mechanical dissociation, followed by enzymatic digestion using collagenase and DNase. Single-cell suspension was passed over 70-μm strainer then subjected to red blood cell lysis. Samples were then mounted with Wescor Cytopro cytocentrifuge and fixed with 4% PFA (Electron Microscopy Sciences) in cytoskeleton buffer (10 mmol/L MES pH 6.1, 138 mmol/L KCl, 3 mmol/L MgCl₂, 2 mmol/L EGTA, 320 mmol/L sucrose) for 10 minutes. Cells were permeabilized with 0.1% TritonX-100 in PBS (PBSTx), blocked with 1% BSA in PBSTx, stained for pan-cytokeratin (AE1/AE3+5D3, Abcam ab86734, 1:100 in 1% BSA in PBSTx) and DAPI, and then mounted with Prolong Gold Antifade reagent (Thermo Fisher Scientific). For secondary antibody staining, Alexa Fluor 488 AffiniPure Donkey Anti-Mouse IgG (Jackson ImmunoResearch, 715-545-151, 1:500) was used. Images were acquired on a Zeiss LSM880 or LSM710 confocal microscopes using 63×/1.40 Plan-Apochromat oil objective and 405 and 488 nm lasers, default filter settings and pinhole of 1 airy unit or Deltavision Elite Imaging System. Images were processed in FIJI (Version 2.0.0) and images presented are maximum intensity projections of individual cells.

For H3K36me3 IHC, patient deidentified, formalin-fixed and paraffin-embedded (FFPE) ccRCC samples were sectioned and hematoxylin and eosin (H&E) stained by Vanderbilt University Medical Center Translational Pathology Shared Resource. To perform H3K36me3 IHC, FFPE sections were deparaffinized, then incubated for 45 minutes in heated citric acid buffer (10 mmol/L citric acid, 0.05% Tween-20, pH 6) for antigen retrieval. Slides were then stained with H3K36me3 antibody (Abcam, ab9050, 1:500), using the Cell and Tissue Staining Kit (Rabbit, HRP-DAB System, R&D Systems), according to the manufacturer's protocol.

The total RNAs were isolated from Setd2^w/w, Setd2^f/w, and Setd2^f/f MEFs treated with 4-HT for 0, 3, and 6 days using RNeasy Mini Kit (Qiagen), and their complementary DNAs were synthesized using Superscript III reverse transcriptase (Invitrogen). The quantities of Setd2 mRNAs were assessed by real-time quantitative PCR using specific TaqMan probes (Thermo Fisher Scientific, Mm01250234_m1) amplifying exon 6, which is the region to be deleted by Cre recombination, and a TaqMan Fast Universal master mix on QuantStudio Flex Real-Time PCR system (Thermo Fisher Scientific). The quantities of Setd2 mRNAs were normalized to those of glyceraldehyde 3-phosphate dehydrogenase, which was used as an endogenous control. The following set of conditions were used for each real-time reaction: 95°C for 20 minutes followed by 40 cycles of 1 second at 95°C and 20 seconds at 60°C. The real-time PCR reactions were performed in triplicate twice and quantified using the −ΔΔ cycle threshold (C_t) method.

Densitometry analyses using ImageStudio Ver2.0 were performed on immunoblots of H3K36me3 and αTubK40me3 normalized to H3 and αTubulin, respectively. Significance was inferred with Tukey multiple comparison test. To determine significance of micronuclei count or percentage and of mitotic defects from triplicates, we performed a two-way ANOVA test, which shows significant interactions between samples. We performed Student t test using a paired, two-tailed distribution to determine the significance of qRT-PCR and micronuclei count in SETD2 knockdown HKC datasets. A reported P value of less than 0.05 is considered significant. All analyses were performed using Prism 7 (GraphPad) and Excel (Microsoft) software.

Previous studies have linked SETD2 inactivation to genomic instability in RCC utilizing tumor-derived analysis (24). We first asked whether loss of this tumor suppressor alone was sufficient to promote genomic instability in nontumor kidney epithelial cells. To answer this question, and to examine the domain(s) of SETD2 required for maintenance of genomic stability in normal kidney epithelial cells, we utilized a cell line generated from immortalized normal human kidney proximal epithelial cells (HKC), which retain a normal chromosome 3 (Supplementary Fig. S1A). SETD2 was inactivated in the parental HKC line (SETD2^−/−), and a series of rescue lines generated by reexpressing SETD2. This panel included lines rescued with a fully functional wild-type N-terminal truncated SETD2 (tSETD2), a tSETD2 catalytically dead SET domain-mutant (R1625C), and a tSETD2 SRI domain-mutant (R2510H). Genomic instability was a prominent feature of SETD2^−/− HKC cells expressing empty vector (GFP), catalytically dead SET domain mutant tSETD2 R1625C, or the SRI domain mutant tSETD2 R2510H (Fig. 1A). SETD2-deficient cells exhibited a significant increase in both lagging chromosomes and chromosome bridge defects (Fig. 1A–C). Thus, as previously seen in RCC-derived 786-0 cells (18), in normal kidney epithelial cells both the SET and SRI domains were required to maintain genomic stability: an increased incidence of mitotic defects occurred when SETD2 activity is absent or when it carries pathogenic mutations in either of these two functional domains.

This genotype–phenotype correlation was extended using micronuclei as a readout of genomic instability. Micronuclei were evident as punctate, DAPI-positive structures that were also immunoreactive with CREST (anti-centromere) antibody. Only wild-type SETD2 (tSETD2) could rescue genomic instability in SETD2^−/− HKC cells and reduce micronuclei count to normal levels (Fig. 2A). Neither empty vector (GFP), catalytically dead SET domain R1625C, nor SRI domain R2510H SETD2 mutants rescued the increase in micronuclei observed in SETD2^−/− HKC. Furthermore, SETD2^−/− HKC cells showed both CREST (+) and CREST (−) micronuclei, indicative of entire or fragmented chromosomes (Fig. 2B and C). This was also the case in SETD2^−/− HKC cells reconstituted with catalytically dead SET domain (R1625C) and SRI domain (R2510H) tSETD2 mutants (Fig. 2B and C).

Micronuclei were also a common feature of a panel of ccRCC-derived cell lines, relative to ACHN cells, a known wild-type VHL line cytogenetically similar to papillary type renal cell carcinomas (25). Single nucleotide polymorphism (SNP) arrays using Log2R and B allele frequency score confirmed that all of the ccRCC-derived cells displayed chromosome 3p deletion. However, ACHN and proximal tubule kidney epithelial HKC cells (26) retained two normal copies of chromosome 3, and had a lower micronuclei count relative to the other ccRCC-derived cells (Supplementary Fig. S1A–S1D).

One question that arises is whether the cells that have experienced genomic instability as a result of SETD2 depletion remain in the population and perpetuate these genomic changes as potential clonal features furthering the process of tumorigenesis. As an indicator of this process, we tested the copy number variation (CNV) of both SETD2 proficient (+/+) and SETD2 deficient (−/−) HKC cells. ASCAT profiles of both HKC cell lines harboring empty GFP-containing vector revealed an increase of genomic instability upon SETD2 loss characterized by gains and losses of parts of, or entire, chromosomes, supporting the notion that changes in SETD2 activity contribute to the genomic instability of tumor cells (Supplementary Fig. S2A and S2B).

In cells lacking SETD2, both microtubule and histone methylation are abrogated (18). To determine the impact of reduced SETD2 dosage (monoallelic loss) on microtubule versus histone methylation, we used tamoxifen-inducible ER-Cre wild-type Setd2 (w/w), heterozygous (w/f), and homozygous (f/f) Setd2-mutant MEFs. These cells lack the genetic diversity found in cancer cell lines, thus allowing for Setd2 alterations to be examined in isolation from other genetic aberrations, and for drawing causal associations between mono- and biallelic Setd2 loss and genomic stability.

We first compared the effects of mono- and biallelic Setd2 loss on histone (H3K36me3) and microtubule (αTubK40me3) methylation. Strikingly, although Setd2^w/f MEFs exhibited reduction of αTubK40me3, observed in multiple clones whose relative Setd2 mRNA levels were shown to be partially reduced compared with those of Setd2^w/w MEFs (Supplementary Fig. S3A and S3B), monoallelic loss of Setd2 caused no change in H3K36me3 levels (Fig. 3A). Conversely, Setd2^f/f MEFs showed a complete loss of both microtubule and histone methylation, consistent with complete loss of SETD2 methyltransferase activity (Fig. 3A). Thus, monoallelic SETD2 was haploinsufficient for microtubule, but not chromatin, methylation.

To explore the effect of SETD2 monoallelic loss on genomic stability, we quantitated acute chromosomal defects in Setd2^w/w and Setd2^w/f MEFs following excision of Setd2. Relative to Setd2^w/w MEFs, Setd2^w/f MEFs exhibited evidence for chromosome lagging and bridging, the same mitotic defects seen in SETD2-deficient HKC cells, and a significant increase in micronuclei (Fig. 3B–D). These data demonstrate that when Setd2 gene dosage is reduced by monoallelic loss of Setd2, this deficit is sufficient to decrease αTubK40me3, and cause both mitotic chromosome defects and a statistically significant increase in micronuclei. The mitotic phenotype of these cells is similar to that observed with biallelic loss (18).

As noted above, micronuclei can contain whole chromosomes or acentric chromosome fragments, both of which can occur as a result of mitotic defects (lagging and bridging chromosomes). Micronuclei of Setd2^f/f and Setd2^w/f MEFs stained with Hoechst, CENPB, and CREST also exhibited both CREST (+) whole chromosome and CREST (−) acentric chromosome staining (Fig. 4A). Both Setd2^f/fand Setd2^w/f MEFs showed increased total micronuclei number, but homozygous Setd2 MEFs (Setd2^f/f) showed a more dramatic increase in Hoechst (+) CENPB/CREST (−) micronuclei, containing acentric chromosome fragments, compared with Setd2^w/f MEFs (Fig. 4B; Supplementary Fig. S4A). Remarkably, MEFs with monoallelic loss of Setd2 (Setd2^w/f) showed the most dramatic increase in Hoechst/CENPB/CREST (+) micronuclei, suggesting these micronuclei arose predominantly from segregation defects involving whole chromosomes (Fig. 4B; Supplementary Fig. S4A and S4B). Importantly, given that MEFs with monoallelic loss of Setd2 (Setd2^w/f) were proficient for H3K36me3 histone methylation, these data suggest that genomic instability associated with haploinsufficiency is independent of the histone methyltransferase activity of SETD2. In agreement with the latter results, partial SETD2 knockdown (∼30% decrease) using shRNA on SETD2-proficient HKC cells further demonstrates that reduced SETD2 levels significantly increase micronuclei count (Supplementary Fig. S5A–S5D). Furthermore, as noticed with Setd2^w/f MEFs, the same partial SETD2 knockdown was sufficient to promote mitotic defects in HKC cells (Supplementary Fig. S5E).

To link the causal associations seen in vitro between monoallelic loss of SETD2 and genomic instability, we utilized tumor cells derived from patients with ccRCC to determine whether the nearly ubiquitous and early loss of chromosome 3p seen in these tumors resulted in loss of genomic stability. Using micronuclei as an indicator for loss of genomic stability, tumor-derived, and adjacent normal kidney cells from five patients with ccRCC were stained with DAPI (DNA) to quantitate micronuclei, and pan-cytokeratin as a marker for normal epithelial and ccRCC cells (Fig. 5). Remarkably, tumor cells from all five ccRCC patient-derived samples exhibited a significantly increased frequency of micronuclei (Table 1). The frequency of micronuclei in tumor cells ranged from 1.1% to 30%. Micronuclei were detected in only one matched normal sample, but in that individual, the frequency of micronuclei was 2-fold higher in tumor versus matched normal cells. The histology of these tumors was confirmed to be clear cell, and nuclei of these tumors stained positive for histone methylation (Supplementary Fig. S6A), confirming that these tumors retained H3K36 trimethylating activity. In addition, sequencing available tissue from the ccRCC tumor samples identified none with deleterious SETD2 mutations (Supplementary Fig. S6B).

To explore whether increase in micronuclei count is observed in other types of kidney cancer characterized by retaining chromosome 3p, we DAPI-stained tumor-derived and adjacent normal kidney cells from two papillary RCCs (pRCC) and two patients with oncocytoma and quantitated micronuclei. In these patient-derived samples, we found lower micronuclei counts compared with those in ccRCC-derived tumor cells (Table 1).

Thus, ccRCC cell lines and ccRCC patient-derived tumor cells exhibit both the genomic instability (micronuclei) associated with monoallelic SETD2 deletion, as a consequence of chromosome 3p loss, as well as the dependence on the SRI domain for nonhistone methylating SETD2 function. Together, these data demonstrate that micronuclei formation occurs with monoallelic loss of SETD2 in vivo, in line with our in vitro observations, pointing to genomic instability and micronuclei as early and common features caused by the 3p loss in ccRCC.

The vast majority (in excess of 95%) of ccRCCs has SETD2 haploinsufficiency as a consequence of chromosome 3p loss, and this chromosomal event is widely accepted to be an early event in ccRCC transformation (27). On the basis of our previous and current data, we postulate that SETD2 haploinsufficiency plays a crucial role as a driver during early ccRCC disease development. In addition, an independent report, using high-resolution profiling of H3K36me3 in metastatic ccRCC tumor samples, suggests that decreased SETD2 mRNA expression, after monoallelic loss, is not sufficient to alter SETD2′s histone H3 lysine methyltransferase activity (16). Our findings consistently demonstrate SETD2 acting as a dual-function methyltransferase, here identifying a scenario in which staged loss of function separates these activities. Heterozygosity for SETD2 provides an opportunity for genomic instability to promote early disease development, potentially creating conditions permissive to the harsher setting of biallelic loss, and complete loss of trimethylation functionality. On the basis of these and other data, we propose a stepwise model for loss of SETD2 functionality as a driver in ccRCC. First, with the loss of 3p and haploinsufficiency for SETD2 as an early event contributing to genomic instability, and upon biallelic inactivation, the additional loss of H3K36 trimethylating activity (Fig. 6).

SETD2 is mutated or lost in a growing list of tumors (6–9), and notably mutations in SETD2 are enriched in chemotherapy-refractory pediatric acute lymphocytic leukemia (9, 28). The frequency with which these mutations occur in one allele without loss of heterozygosity has not been examined in detail. Furthermore, 3p is a commonly deleted genomic region, raising the possibility that haploinsufficiency for SETD2 function may be a more common event in tumorigenesis than previously considered. It is important to consider, however, that in ccRCC tumors (that routinely lose 3p), haploinsufficiency for a number of important proteins may be playing a role in permitting tumorigenesis. Indeed, preliminary studies have implicated BAP1, another 3p-localized gene, as a potential mediator in the haploinsufficient state (29). Future studies will need to consider the effect of multigene haploinsufficiency in promoting tumorigenesis.

Beyond loss of genomic stability, biallelic SETD2 inactivation may contribute to tumorigenesis in several ways, in particularly in mediating DNA repair (30). SETD2′s expanding function as a methylating enzyme has revealed that, in addition to the H3K36me3 mark on histones, SETD2 trimethylates lysine 40 on α-tubulin of mitotic microtubules (18) and the transcription factor STAT1 (31). In the case of SETD2′s microtubule methylating function, failure to trimethylate αTubK40, in the setting of acute Setd2 loss, is accompanied by loss of the αTubK40me3 mark from mitotic microtubules, and catastrophic mitotic defects (18). These acute effects we previously observed following SETD2 loss were unlikely to be accounted for by loss of histone methylation, as conversion of histone methylation is considered to be a slow process over several cell divisions, and RNA-seq revealed no consistent alterations in gene expression associated with acute SETD2 loss (18). Here, we report that monoallelic SETD2 loss is sufficient to cause genomic instability (lagging and bridging chromosomes and micronuclei), which is linked to SRI domain defects that uncouple SETD2′s histone methylating from its newly discovered tubulin methylating activity. Other targets of SETD2 methylation are emerging, notably STAT1 (31), for which the impact of SETD2 haploinsufficiency is unknown. Furthermore, genomic instability in human ccRCC tumors that were haploinsufficient for SETD2, but that retained H3K36 trimethylation, further supports the linkage between microtubule methylation by SETD2 and genomic stability. These data also point to micronuclei formation as a cytologic tool to reveal novel aspects of SETD2 function as a multifunctional tumor suppressor critical to genome maintenance.

Micronuclei were identified and described in erythrocytes first by W. Howell (1890) and later on by J. Jolly (1905; ref. 32) as small, extranuclear chromatin bodies surrounded by a nuclear envelope, which originate from acentric chromosome fragments or entire chromosomes lagging behind the main chromatin mass during anaphase (33, 34). Today, it is accepted that micronuclei represent a faithful cellular indicator of chromosome instability (or genomic instability), as high micronuclei frequency is associated with chromosome instability observed in many solid tumors and preneoplastic lesions (35, 36). Importantly, micronuclei have also been documented in primary cancers (33, 35), although ccRCC has not been well characterized in this regard. We show that micronuclei are a common feature not only among ccRCC-derived cell lines, but more importantly among ccRCC patient–derived tumors. Considering the typical low mitotic rate of primary RCC tumors, the accumulations of micronuclei at the rates we observed implicate chromosome instability as a major driver of ccRCC tumorigenesis.

In showing that SETD2 haploinsufficiency and partially reduced levels of SETD2 are sufficient to initiate a wide range of mitotic defects, resulting in lagging chromosomes and micronuclei formation, our data further underscore a much more complex role for SETD2 in genome maintenance than previously appreciated. Although some of the more dramatic examples may represent mitotic failures, the persistence of micronuclei implies that some aberrant mitoses are permitted, possibly as a result of a failed checkpoint. In cells lacking SETD2, we were able to show evidence of an unstable genome, although this model does not fully address the haploinsufficient state. The bypassed checkpoint allowing these cells to persist remains an object of future investigation. In this regard, the shared genomic instability phenotype seen with SETD2 haploinsufficiency and SRI-domain defects that abrogate microtubule but not histone methylation argues that maintenance of genomic stability by SETD2 is linked to its function as a microtubule methyltransferase. These data also suggest αTubK40 trimethylation is more sensitive to SETD2 stoichiometry fluctuations than H3K36 trimethylation. The effect of SETD2 haploinsufficiency on any other targets is not known. These data strongly support the notion of αTubK40me3 playing a specialized role in maintaining genomic stability during mitosis. Dissecting the tubulin dynamics that govern this process, and the specialized role of SETD2 and the SRI domain in regulating αTubK40 methylation will be an important future goal.

It is now well accepted that the main drivers of cancer are not limited to highly penetrant dominant and recessive mutations, but also include heterozygous losses giving rise to haploinsufficiency states (37). There is an ever-increasing number of tumor suppressor genes known to display haploinsufficiency, such as PTEN and p53 (among others), with some pathways showing a higher tendency to dosage sensitivity than others (37–39). Another example of haploinsufficiency in cancer progression is PERK, an endoplasmic reticulum transmembrane protein kinase, displaying pro-tumorigenic activity in melanomas harboring activated Braf^V600E mutation. However, deletion of one allele of PERK is permissive for Braf^V600E-dependent transformation, whereas excision of both alleles did not cooperate with Braf^V600E in melanoma tumorigenicity (40). Interestingly, PERK mutations decreasing its activity are seen at a 7% frequency (40). Therefore, our compelling results showing SETD2 dose-dependent methyltransferase activity are in line with the work previously done in other cancer types, and demonstrate a crucial, and previously unappreciated, role of SETD2 haploinsufficiency caused by 3p loss in progression of ccRCC.

No potential conflicts of interest were disclosed.

Conception and design: Y.-C. Chiang, R. Dere, C.L. Walker, W.K. Rathmell

Development of methodology: Y.-C. Chiang, I.-Y. Park, E.A. Terzo, D.N. Tripathi, C.C. Fahey, M. Karki, R.T. Powell, C.L. Walker, W.K. Rathmell

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y.-C. Chiang, I.-Y. Park, D.N. Tripathi, F.M. Mason, C.B. Shuster, B.-H. Sohn, R.T. Powell, B.D. Strahl, C.L. Walker, W.K. Rathmell

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y.-C. Chiang, F.M. Mason, R.T. Powell, Y.S. Tsai, A.A. de Cubas, A. Khan, I.J. Davis, B.D. Strahl, J.S. Parker, R. Dere, W.K. Rathmell

Writing, review, and/or revision of the manuscript: Y.-C. Chiang, I.-Y. Park, E.A. Terzo, D.N. Tripathi, F.M. Mason, C.C. Fahey, M. Karki, P. Chowdhury, R. Ohi, A.A. de Cubas, A. Khan, I.J. Davis, J.S. Parker, R. Dere, C.L. Walker, W.K. Rathmell

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): I.-Y. Park, C.C. Fahey, P. Chowdhury, A.A. de Cubas, C.L. Walker, W.K. Rathmell

Study supervision: C.B. Shuster, C.L. Walker, W.K. Rathmell

We used the Zeiss LSM880 Airyscan and LSM 710 microscopes (NIH S10; 1S10OD021630-01) at the Vanderbilt University Cell Imaging Shared Resource (CISR) facility, the Integrated Microscopy Core at Baylor College of Medicine (NIH DK56338, and CA125123), CPRIT (RP15078), the Dan L. Duncan Comprehensive Cancer Center, and the John S. Dunn Gulf Coast Consortium for Chemical Genomics. We used the Flow Cytometry Core Facility at UNC supported in part by the Cancer Center Core Support Grant (P30 CA16086) to the UNC Lineberger Comprehensive Cancer Center. The authors acknowledge support from the NIH (R01GM086610 to R. Ohi; R01CA203012 to W.K. Rathmell and C.L. Walker; R01CA198482 to W.K. Rathmell, I.J. Davis, and B. D. Strahl; T32CA119925 to E.A. Terzo; T32CA009156 to Y.C. Chiang). R. Ohi acknowledges support from the Leukemia and Lymphoma Society.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.