Research ArticleCancer

Recurrent epimutation of SDHC in gastrointestinal stromal tumors

See allHide authors and affiliations

Science Translational Medicine  24 Dec 2014:
Vol. 6, Issue 268, pp. 268ra177
DOI: 10.1126/scitranslmed.3009961


Succinate dehydrogenase (SDH) is a conserved effector of cellular metabolism and energy production, and loss of SDH function is a driver mechanism in several cancers. SDH-deficient gastrointestinal stromal tumors (dSDH GISTs) collectively manifest similar phenotypes, including hypermethylated epigenomic signatures, tendency to occur in pediatric patients, and lack of KIT/PDGFRA mutations. dSDH GISTs often harbor deleterious mutations in SDH subunit genes (SDHA, SDHB, SDHC, and SDHD, termed SDHx), but some are SDHx wild type (WT). To further elucidate mechanisms of SDH deactivation in SDHx-WT GIST, we performed targeted exome sequencing on 59 dSDH GISTs to identify 43 SDHx-mutant and 16 SDHx-WT cases. Genome-wide DNA methylation and expression profiling exposed SDHC promoter–specific CpG island hypermethylation and gene silencing in SDHx-WT dSDH GISTs [15 of 16 cases (94%)]. Six of 15 SDHC-epimutant GISTs occurred in the setting of the multitumor syndrome Carney triad. We observed neither SDHB promoter hypermethylation nor large deletions on chromosome 1q in any SDHx-WT cases. Deep genome sequencing of a 130-kbp (kilo–base pair) window around SDHC revealed no recognizable sequence anomalies in SDHC-epimutant tumors. More than 2000 benign and tumor reference tissues, including stem cells and malignancies with a hypermethylator epigenotype, exhibit solely a non-epimutant SDHC promoter. Mosaic constitutional SDHC promoter hypermethylation in blood and saliva from patients with SDHC-epimutant GIST implicates a postzygotic mechanism in the establishment and maintenance of SDHC epimutation. The discovery of SDHC epimutation provides a unifying explanation for the pathogenesis of dSDH GIST, whereby loss of SDH function stems from either SDHx mutation or SDHC epimutation.


The multimeric respiratory complex II/succinate dehydrogenase (SDH)/succinate ubiquinone oxidoreductase (SQR) is an integral component of both the mitochondrial Krebs cycle and respiratory chain [reviewed in (1)], and its four protein subunits (SDHA, SDHB, SDHC, and SDHD) are encoded by nuclear genes. Mutational inactivation of SDH is a tumor mechanism in several tumors including gastrointestinal stromal tumor (GIST), the most common mesenchymal tumor of the gastrointestinal tract. SDH deficiency is typically associated with gastric GIST in pediatric patients, leading to the term “pediatric” GIST, but it also occurs in adults as an alternative mechanism to driver mutations of signal transduction kinases such as KIT.

Several studies link disabled SDH function to tumorigenesis through pathologic activation of an otherwise physiologic hypoxia-inducible pathway of angiogenesis, glycolysis, and cell proliferation (25), as well as through nuclear epigenomic remodeling due to deranged Krebs cycle function (68). Maintenance of epigenomic integrity requires SDH-dependent catalysis of succinate to fumarate. The dioxygenase-family histone lysine demethylases (containing a jumonji C domain) and the 5-methylcytosine (5mC)–modifying enzymes in the TET family depend on the succinate/α-ketoglutarate (αKG) ratio for proper function. Pathologic elevation of the succinate/αKG ratio (secondary to SDH dysfunction) creates a “pseudo-hypoxic” state mirroring a physiologic response to hypoxia by inhibition of the oxygen-sensing prolyl hydroxylase. This triggers the hypoxia-inducible factor 1α–mediated hypoxia response, whose sequelae in dSDH GIST include highly vascularized, hypercellular tumors. Disruption of oxidative demethylation of 5mC due to inhibitory metabolites that accumulate in tumors with mutations in core metabolic pathways may be a mechanism of global DNA hypermethylation common to dSDH GIST and paraganglioma (79) and to isocitrate dehydrogenase (IDH)– and fumarate hydratase (FH)–mutant tumors of mesenchymal, epithelial, glial, myeloid, and renal (8, 1013) lineages. Along with the distinctive genome-wide CpG hypermethylation in dSDH GIST relative to KIT mutants (7), these tumors also exhibit elevated IGF1R gene expression (1416), which is of potential therapeutic interest.

Biallelic mutation of one of the four SDH component genes—SDHA, SDHB, SDHC, or SDHD (termed SDHx)—often explains SDH deactivation in dSDH GIST, because loss of function of any one of the four subunits is sufficient to disrupt the complex, evidenced by an SDHB-negative immunophenotype in histologic tumor tissue sections [reviewed in (1)]. However, a notable proportion of dSDH GISTs lack identifiable SDHx mutations; the absence of recurrent genomic copy number aberrations or loss of heterozygosity (LOH) events has thus far thwarted efforts to localize a genetic aberration that could disable the SDH complex and explain the pathologic convergence with tumors driven by SDHx mutation (17, 18).

Rare patients with dSDH GIST may develop additional tumors including paraganglioma and pulmonary chondroma, resulting in a clinical diagnosis of Carney triad, a nonfamilial multitumor syndrome whose genetic basis is unknown and for which no specific clinical tests exist (18). It was originally reported that the most frequent and greatest contiguous change in Carney triad tumors is large genomic deletion on 1q encompassing SDHC (19); however, this genomic alteration has not been observed in other studies of dSDH GIST [reviewed in (1)]. To fully dissect the molecular basis of SDH deficiency, a comprehensive molecular analysis of a large number of SDH-deficient GISTs, not limited to candidate genes and unselected for Carney triad or SDHx-mutation status, is required.

Beyond coding sequence mutations of disease target genes, epigenetic modes of gene inactivation are increasingly recognized. Some examples of these include MLH1 and MSH2 silencing via promoter hypermethylation in Lynch syndrome (20) and FMR1 promoter hypermethylation and silencing in fragile X syndrome (21). In these instances, expression and promoter DNA methylation of the causative gene are abnormal despite a lack of protein-altering mutations; moreover, these patients often exhibit epimutation in their normal somatic tissues.

In the current study, we examined the genomes, methylomes, gene expression profiles, and SDHx mutation status of a cohort of 59 dSDH GIST patients. Molecular profiling methods were adapted to routinely processed formalin-fixed, paraffin-embedded (FFPE) surgical pathology specimens to maximize patient inclusion, as well as to perform histology-guided tumor dissection for nucleic acid extraction. These analyses uncovered a recurrent gene silencing epimutation highly specific to SDHx-WT dSDH GIST, encompassing a subset of cases with manifestations of Carney triad. To distinguish between primary and secondary epimutation, we analyzed the cases for potentially causal cis-acting genomic anomalies. Finally, patient blood and saliva were examined for constitutional epimutation to gauge its developmental timing.


SDHx-mutant versus SDHx-WT subgroups of dSDH GIST

Fifty-nine SDH-disabled (dSDH) GIST cases from the National Institutes of Health (NIH) Pediatric GIST clinic were analyzed in this study (Table 1). Cases were selected on the basis of SDHB-negative immunophenotype by immunohistochemistry, a sensitive and specific marker for SDH deficiency (22). SDH deficiency was further confirmed in all tumors by genomic hypermethylation (Table 1 and fig. S1), a molecular hallmark present in all analyzed SDH-deficient GISTs that distinguishes these tumors from KIT/PDGFRA– and other kinase pathway–mutant GISTs with intact SDH function (7). Beyond immunophenotypic and molecular epigenomic confirmation of SDH deficiency, the only other inclusion criterion was sufficient available specimen for genomic DNA (gDNA) library construction for SDHx sequencing. Thus, the analyzed set of 59 GISTs represents a comprehensive cross section of SDH-deficient cases. In addition to tumor SDH deficiency, other patient annotations included age at diagnosis, sex, and whether there was a clinical diagnosis of Carney triad (Table 1).

Table 1. Characteristics of the 59 study cases.

NF, none found; Dx, diagnosis; NA, not applicable.

View this table:

Fifty-nine dSDH GIST cases were tested for SDHx coding sequence mutations by OncoVar-GIST, a custom-capture single-molecule next-generation DNA sequencing assay (NGS), performed in a Clinical Laboratory Improvement Amendments (CLIA)–certified laboratory, that sequences the complete coding sequence of SDHA, SDHB, SDHC, and SDHD from FFPE tissue after tumor dissection by a pathologist. In addition to histology-targeted dissection of tumor tissue, the fraction of malignant cells represented in tumor tissue DNA extracts was empirically calculated from the methyl deviation index (MDI), a quantitative measurement of tumor cell fraction derived from tumor-specific methylated allele frequency in tissue DNA extracts (7, 23). By these measures, all tumor extracts consisted of at least 50% malignant cells and had a median malignant cellularity of greater than 80% according to the MDI metric (Table 1 and fig. S1).

Forty tumors were positive for SDHx coding sequence mutation by the OncoVar-GIST sequencing assay, and three additional tumors had intragenic deletions by copy number and/or genotyping microarray (Table 1). A “second hit” leading to biallelic SDHx inactivation was demonstrated in 35 of 43 (81%) of SDHx-mutant tumors (Table 1 and fig. S2), consistent with the Knudson two-hit model of tumor suppressor gene inactivation (24). The eight instances of SDHx-mutant GIST without detection of a second hit were SDHA mutants (Table 1). As will be shown further below, these cases did not harbor SDHA promoter methylation. Despite the lack of LOH in these SDHA-mutant GISTs, four of them had loss of SDHA protein expression by immunohistochemistry, a biomarker of deleterious SDHA mutation in GIST (25). Beyond these exceptions to the two-hit model, all other SDHA-mutant GISTs (n = 17) harbored second hits. Thus, overall, the 59 dSDH GIST cases were split into the SDHx-mutant (n = 43) and SDHx-WT (n = 16) groups (Table 1). Regarding clinical molecular testing for SDHx mutation in GIST, our laboratory’s analyses suggest that up to one of three of SDH-deficient GISTs (16 of 59) lack sequence mutations in SDHx, a frequency that has not been well documented previously.

SDHC promoter hypermethylation in SDHx-WT GIST

The DNA methylomes of 59 individual dSDH GISTs were profiled by Illumina 450K Infinium methylation beadarray [Gene Expression Omnibus (GEO) platform GPL13534]. In a t-test two-group comparison of SDHx-WT versus SDHx-mutant tumors, the methylation levels of 11 CpG targets (of 485,577 total) demonstrated significant hypermethylation (q < 0.05) in SDHx-WT GIST (Fig. 1A). All 11 targets mapped to a 656–base pair (bp) genomic interval encompassing the two SDHC promoter CpG islands (CGIs) on chromosome 1q23.3 (Fig. 1B). Outside the SDHC promoter, there was no significant (q < 0.05) CpG target hypermethylation elsewhere in the SDHx-WT GIST genome (Fig. 1A), including SDHA, SDHB, and SDHD (fig. S3). Although all hypermethylated targets mapped to the SDHC promoter, scattered CpG targets registered as significantly hypomethylated (q < 0.05), yet were close to the border of statistical significance (fig. S4); future studies on additional SDHx-WT tumors may further clarify the biological relevance, if any, of these hypomethylated targets to pathogenesis.

Fig. 1. SDHC epimutation in SDHx-WT dSDH GIST.

(A) Janus plot showing upward facing CpG target hypermethylation (–log10q) and downward facing gene probe set hypoexpression (log10q) in SDHx-WT versus SDHx-mutant GIST, calculated from t-test comparison of genome-wide methylation and expression profiles, plotted on genome coordinates. q-value significance thresholds (q = 0.05, gray dotted lines) are indicated. (B) UCSC (University of California, Santa Cruz) genome browser display shows genomic position (hg19 coordinates), CGIs (green bars), SDHC promoter (red bar), and SDHC 5′UTR/exon 1 in relation to hypermethylated and hypoexpressed targets. The lower browser track shows the position of each 450K Infinium methylation CpG target in the region. (C) Heatmaps show methylation β values (legend at bottom right) of the 11 significant hypermethylated Infinium 450K CpG targets spanning the SDHC promoter (q < 0.05, red bar over upper heatmap) and upstream and downstream flanking regions. SDHCme-positive (upper heatmap) or SDHCme-negative (lower heatmap) GIST samples are ordered according to SDHx mutation. (D) 450K Infinium methylation data from 854 normal and tumor reference tissues from the GEO database (project IDs shown on the left of the heatmap). ETMR/PNET, embryonal tumor with multilayer rosettes/peripheral neuroectodermal tumor; PGL/Pheo, paraganglioma/pheochromocytoma; GBM, glioblastoma multiforme.

We next categorized the zygosity (homozygous versus heterozygous) of SDHC promoter methylation in SDHCme tumors for correspondence with SDHx mutation status. Because the malignant cell fraction in tumors may be expected to influence the measurement of SDHC promoter–methylated alleles in whole tumor extracts (the nonnormalized SDHCme allele frequency), tumor SDHC promoter methylation was normalized to malignant cell fraction in tumor lysates to derive the tumor SDHCme zygosity index (SDHC promoter–methylated allele frequency normalized to tumor malignant cell purity). Here, an index of ~1 would indicate homozygous/biallelic SDHCme, and ~0.5 would correspond to heterozygous/monoallelic methylation at the SDHC promoter. As anticipated, hierarchical clustering of dSDH GISTs according to SDHCme index segregated the tumors into three SDHC promoter methylation (SDHCme) groups: homozygous/fully methylated (n = 12), heterozygous/hemimethylated (n = 7), and negative (n = 40) (Fig. 1C and fig. S5). All 12 homozygous SDHCme GISTs were from female SDHx-WT patients and included five of eight Carney triad cases (Fig. 1C and fig. S5). Hemimethylation encompassed four GISTs with SDHC heterozygous mutation (Fig. 1C and fig. S5); moreover, by genotyping and/or array comparative genomic hybridization (aCGH), these four tumors had no evidence of SDHC LOH, consistent with a heterozygous mutant tumor with the second hit by promoter hypermethylation (Table 1 and fig. S2); the phenomenon of second hit SDHC promoter methylation/silencing of the wild-type (WT) allele in heterozygous SDHC-mutant GIST has been previously reported (17). Three of seven hemimethylated SDHCme GISTs in our study were SDHx WT and included one Carney triad patient.

Promoter methylation of pathogenicity genes in fragile X and Lynch syndromes often stems from genomic anomalies proximal to FMR1 and MSH2, respectively (20, 21). Thus, we performed deep sequencing of 12 SDHCme GISTs for mutations within a 130-kbp window encompassing SDHC and including its promoter and flanking genomic sequence. This effort did not reveal any recognizable recurrent genomic aberrations in the tumors. For completeness, additional GIST pathogenicity genes, including KIT, PDGFRA, BRAF, and NF1 [four genes more typically associated with non-dSDH GIST (7)], were sequenced and found to lack germline or somatic mutation in these SDHx-WT dSDH GISTs. Thus, on the basis of the absence of identified genomic aberration, SDHC promoter methylation in GIST is consistent with primary epimutation, although future discovery of a molecular mechanism may favor reclassification as secondary epimutation. No SDHA (n = 22), SDHB (n = 12), or SDHD (n = 1) mutants had SDHCme (fig. S3). Among the SDHCme-negative cases were two SDHC mutants; although these tumors lacked SDHC promoter hemimethylation, both had second hits to SDHC due to chromosome 1q23 LOH encompassing SDHC and resulting in biallelic mutation in the tumor (Table 1 and fig. S2C).

In sum, 58 of 59 (98%) dSDH GISTs demonstrated SDHx mutation, SDHC promoter hypermethylation, or SDHC mutation plus SDHCme as a second hit, whereas one tumor remained SDHx-WT and SDHCme-negative (Table 1). As in all SDHx-WT cases in our study, this case does not harbor SDHA, SDHB, or SDHD promoter hypermethylation (fig. S3) and does not have tumor LOH that would point to overlying genetic mutation.

Regarding Carney triad, six of eight tumors with this clinical annotation demonstrate SDHCme (five homozygous and one hemimethylated; fig. S5), but the tumors from two patients designated Carney triad (due to co-occurrence of dSDH GIST and pulmonary chondroma) were found to harbor SDHA mutation and negative SDHCme (Table 1, Fig. 1C, and figs. S3 and S5). We interpret the data as most consistent with alternative SDHCme and SDHx-mutant molecular subtypes of Carney triad as defined by anatomic criteria. Although some investigators may reclassify patients with GIST and pulmonary chondroma as not having Carney triad owing to the finding of SDHA mutation, others may favor keeping the Carney triad diagnosis on the basis of anatomical-pathology criteria. Why Carney triad shows a greater correlation with SDHC epimutation than SDHx mutation (P = 0.0025) and whether the Carney triad definition should change from purely morphologic criteria to incorporate SDHC epimutation status are interesting topics for future study. Because currently there is no diagnostic marker specific for Carney triad, longitudinal observation of dSDH GIST patients may demonstrate utility of SDHCme as a biomarker of increased risk for paraganglioma and/or pulmonary chondroma.

SDHx-WT GIST patients, including those with and without Carney triad manifestations, were significantly younger than mutant counterparts (average age, 17 years versus 29 years; P = 0.003; Table 1). SDHx-WT patients with Carney triad manifestations were not significantly older at the time of initial GIST diagnosis than those without manifestations (average age, 19 years versus 16 years; P = 0.57; Table 1).

None of the SDHC-epimutant GISTs in our study demonstrated hypermethylated SDHA, SDHB, or SDHD promoters by 450K methylation microarrays (Fig. 1A and fig. S3), which target multiple CpG sites in each respective promoter. In addition, none of the dSDH GISTs we studied, including SDHx-mutant and SDHx-WT, had large contiguous deletions on chromosome 1q (fig. S2H). Stand-alone SDHC epimutation in SDHx-WT GIST in our study contrasts with a recently proposed molecular triad of simultaneous chromosome 1q deletion and SDHB and SDHC promoter methylation that were based on a high–cycle number (60-cycle) bisulfite polymerase chain reaction (PCR) assay applied to a series of three Carney triad GISTs (26). Therefore, it seems uncharacteristic that tumors could simultaneously harbor both SDHB and SDHC promoter hypermethylation. One case in our study (dSDH38) had a relatively small 8.4-kbp germline intragenic deletion in SDHC, identical to that previously reported in a paraganglioma (27) (known as the “Pittsburgh mutation”), with tumor LOH in the remaining allele; however, outside of relatively rare instances of germline SDHC intragenic deletions in patients with paraganglioma and/or GIST that then complete the mutation with LOH, there are no events in our cases that would classify as chromosome band size deletions on 1q in dSDH GIST (fig. S2H).

To evaluate the more general specificity of SDHCme to SDHx-WT GIST, we analyzed 450K Infinium methylation profiles from a reference panel of more than 2300 normal and cancer tissues. We evaluated multiple project data sets from the GEO and The Cancer Genome Atlas (TCGA) databases, including non-GIST pediatric tumors; mesenchymal, neuroectodermal, and epithelial tumors; and specimens with SDHx, IDH1/2, or TET mutations and associated with a hypermethylator epigenotype. These were evaluated for SDHC promoter methylation using all 11 microarray SDHC promoter CGI targets. Beyond the SDHx-WT and SDHC-mutant GISTs in our study, we identified no cancers in GEO with a hypermethylated SDHC promoter (Fig. 1D). One cancer specimen out of multiple TCGA data sets—an instance of papillary renal carcinoma (1 of 226 total papillary kidney cancers with methylation array data)—was found to have hemimethylated SDHCme (specimen ID: TCGA-F9-A8NY-01A-11D-A369-05; fig. S6). Although SDH-deficient renal tumors have been recently described, they do not typically have papillary morphology, making this particular case even more of an anomaly. Beyond such rare cases, the absence of SDHC promoter hypermethylation in reference tissues suggests that, in both normal and tumor tissues, the SDHC gene promoter is strongly protected against methylation, and heightens its pathobiological relevance in dSDH GIST. Furthermore, the tumor-specific convergence in dSDH GIST of mutually exclusive SDHCme and SDHx mutation meets the criteria for driver epimutation as recently discussed by Sproul and Meehan (28). One feature of embryonic stem cells and induced pluripotent stem cells relative to somatic differentiated tissues and sperm is that the SDHC distal upstream promoter CGI becomes partially methylated (fig. S7); however, complete SDHC promoter hypermethylation spanning the CGI most proximal to the transcription start site (TSS) remains specific to SDHC-epimutant GIST (Fig. 1 and fig. S7).

SDHC silencing in SDHCme tumors

The connection between DNA methylation and gene expression was evaluated by profiling 20 FFPE dSDH GISTs with Affymetrix U133P2 arrays, containing 54,675 probe sets. Whole-genome expression analysis of 7 SDHx-WT and 13 SDHx mutants identified 4 (of 54,675 total) probe sets with significantly decreased expression (q < 0.05) in SDHx-WT dSDH GIST (Fig. 1A). These probe sets all map to the SDHC locus, have the SDHC promoter as the nearest transcription start site, and have an average 4.3-fold reduced expression in SDHx-WT tumors (Figs. 1A and 2). In addition to these SDHC probe sets, two other SDHC probe sets showed hypoexpression with q values between 0.05 and 0.25. No gene expression differences other than SDHC met the significance threshold in the SDHx-WT versus SDHx-mutant comparison (Fig. 1A and fig. S4). SDHC expression silencing was independently validated using an alternative RNA amplification and labeling protocol before array hybridization (fig. S8) and by quantitative reverse transcription PCR (RT-PCR) (fig. S8). Thus, three independent gene expression profiling methods validated SDHC silencing in SDHx-WT GIST. Additional analyses of sex chromosome–linked gene expression further demonstrated the biological legitimacy of the combination of samples and expression array analyses (fig. S8).

The Janus plot demonstrates reciprocal epimutation and expression silencing of SDHC within the SDHx-WT tumor group (Fig. 1A), and hierarchical clustering according to SDHC expression perfectly segregated SDHCme-positive (n = 8) and SDHCme-negative (n = 12) samples (Fig. 2A), clearly demonstrating the coupling of the epimutation with gene silencing within individual specimens. Similarly, this SDHC expression–based hierarchical clustering nearly perfectly segregated specimens according to SDHx-WT versus SDHx-mutant status, with the exception of a heterozygous SDHC mutant with promoter methylation as the second hit (Fig. 2A), whose reduced SDHC expression is consistent with gene inactivation via tandem monoallelic silencing and mutation.

Fig. 2. SDHC silencing in SDHx-WT dSDH GIST.

(A) Heatmap shows the expression (legend on the right) of all probe sets identified by Affymetrix U133P2 microarrays as being significantly different between SDHx-mutant GIST (n = 13) and SDHx-WT GIST (n = 7) (q < 0.05). The four probe sets map to SDHC. CTr, Carney triad. (B) Boxplots of SDHC probe set expression demonstrate an average 4.3-fold silencing in SDHx-WT GIST.

Expression microarray data were available for three cases with the clinical annotation of Carney triad (Fig. 2A): an SDHx-WT case demonstrated homozygous SDHCme and expression silencing, and two SDHA-mutant cases were SDHCme-negative with conserved SDHC expression. Thus, as we found for SDHC promoter methylation, we identified molecular heterogeneity in the transcription of SDHC within patients classified as Carney triad. Our results suggest that SDHC epimutation—evidenced by promoter methylation and gene expression silencing—is common among Carney triad GISTs. However, it is important to note that some patients with a morphology-based Carney triad diagnosis may have SDHx mutation, negative SDHCme, and maintenance of SDHC transcription, and that most GIST cases with SDHCme and silencing are not associated with Carney triad at the time of diagnosis. Although case studies have reported SDHC promoter methylation (coupled to other changes not evident in our series, such as SDHB promoter methylation and large contiguous deletions on chromosome 1q) in selected Carney triad patients (26), analysis of 59 dSDH GISTs unselected for Carney triad status has revealed SDHCme as a frequent recurrent anomaly that is neither perfectly sensitive nor specific for Carney triad. Because Carney triad is a post hoc diagnosis based on the emergence of a spectrum of tumors over the lifetime of a patient, longitudinal follow-up is necessary to establish the prognostic specificity of GIST SDHC epimutation status for subsequent development of Carney triad sequelae.

Mosaic constitutional SDHC epimutation in patients with SDHC-epimutant GIST

Epimutation may be encountered soma-wide in patients with disorders like fragile X and Lynch syndromes. We tested blood and saliva from SDHx-WT GIST patients for constitutional SDHC epimutation, the manifestations of which could provide insight into the developmental timing/establishment of the epimutant DNA methylation mark. Whole blood from SDHx-WT GIST patients showed a modest but significant 2.7% mean elevation in SDHC promoter methylation versus the SDHx-mutant group (P = 0.003, Fig. 3). SDHC promoter methylation was further evident in saliva (fig. S9), which like blood showed a significant (P = 0.018) mean 2.4% elevation of SDHC methylation. The blood and saliva of several SDHC-epimutant GIST patients had up to 10% elevation in SDHC promoter methylation (fig. S9).

Fig. 3. SDHC promoter methylation in blood.

Boxplot of SDHC promoter methylation in blood of SDHx-WT (n = 11) and SDHx-mutant (n = 14) GIST patients shows a 2.7% mean elevation in the SDHx-WT group (P = 0.003). SDHC promoter methylation in blood from individual patients is shown in fig. S9.


Previous studies have shown that dSDH GISTs are collectively characterized by genome-wide hypermethylation and low cytogenetic complexity (1, 7); the absence of both SDHx coding mutations and LOH in many cases left the SDH-deficient tumor phenotype unexplained.

In the current study, we demonstrate that SDHx-WT GISTs, including several from patients with Carney triad, commonly arise from SDHC epimutation, evidenced by highly focal SDHC promoter CGI hypermethylation and transcriptional silencing. SDHC epimutation was the only distinction between SDHx-WT and SDHx-mutant GIST by genome-wide DNA methylation and expression profiling and was mutually exclusive of SDHx mutation, with the notable exception of SDHC heterozygous mutant tumors with silencing of the second allele by SDHC promoter methylation. SDHC epimutation does not appear to be a polymorphism or passenger epiallele, because a reference panel of more than 2300 benign and tumor tissues, including malignancies with a hypermethylator epigenotype, show conservation of an unmethylated SDHC promoter, indicating strong protection from methylation and silencing, perhaps as expected for a gene encoding an essential mitochondrial protein. Regarding a recently proposed molecular triad of hypermethylation of both SDHB and SDHC promoters coupled with 1q deletion as a mechanism of SDH deficiency in Carney triad tumors (19, 28), we found no instances of simultaneous SDHC and SDHx comethylation and no cases with large chromosome 1q deletion. Instead, SDHC promoter methylation and/or gene expression silencing alone is pathognomonic for an SDH-deficient tumor. These results establish SDHC epimutation as the molecular pathologic basis for disabling the SDH complex in most of SDHx-WT GIST patients, encompassing those with Carney triad.

Blood and saliva from patients with SDHC-epimutant GIST manifest detectable but sub-hemimethylated levels of SDHC epimutation as well. Although distinct from those Lynch syndrome patients with soma-wide MLH1 hemimethylation (20), low to modest levels of statistically significant SDHC methylation in patients with SDHC-epimutant GIST are reminiscent of mosaic low-level MLH1 methylation (2 to 5%) in the blood of other Lynch patients (20). Overall, we interpret SDHC hypermethylation in the blood and saliva of SDHx-WT/SDHC-epimutant GIST patients to be biologically relevant and consistent with a mosaic constitutional epimutation that is clonally expanded in tumor cells; moreover, this somatic mosaicism is consistent with postzygotic SDHC promoter methylation reprogramming. Future studies are required to explore and further validate the manifestations of SDHC epimutation mosaicism in the soma of these patients. Such mosaicism in the soma is consistent with a postzygotic onset of epimutation, rather than germline inheritance, which would be expected to result in soma-wide SDHC hemimethylation, unless partially erased in somatic lineages. In either case, the resulting mosaicism points to a process of SDHC reprogramming during early development, which is then maintained so as to be detectable in blood and saliva later in life. Methylation data from additional constitutional anatomic sites from our patients are needed to further evaluate the extent of somatic mosaicism. More generally, the finding of elevated SDHC promoter methylation in the soma of SDHx-WT GIST patients highlights an emerging need in the epimutation field, namely, a large number of methylomes from saliva and other fluids to establish reference values for gene-specific methylation.

The finding of SDHC epimutation is important for several reasons. First, in the clinical genetic evaluation of GIST patients, our data indicate that SDHC epimutation frequency is comparable to SDHx coding sequence mutation and should be considered in dSDH GIST cases, particularly those that prove SDHx-WT. It is reasonable to hypothesize that individuals who present with an SDHC-epimutant GIST may be at greater risk for additional tumors, including paraganglioma and pulmonary chondroma. The identification of SDHC epimutation raises the possibility that demethylating agents such as decitabine could restore SDHC expression and SDH function in SDHx-WT GISTs, which currently lack targeted chemotherapy. Finally, the discovery of SDHC epimutation now provides a unifying explanation for the pathogenesis of almost all cases of dSDH/methyl-divergent GIST, where loss of SDH function arises through either mutation or epimutation.

For now, detection of SDHC epimutation in surgical pathology specimens will require DNA methylation and/or gene expression analysis because presently there are no published methods for SDHC protein immunohistochemistry on fixed tissue, and we have attempted but so far were not able to optimize such an assay. At the same time, whole genome expression and DNA methylation profiling proved technically feasible and successful as in this study, which is encouraging for finding pathogenicity genes in rare tumor types or archival collections where only FFPE material may be available.

The rarity of trans-generational heritability of Carney triad (29), coupled with homozygous tumor SDHC epimutation without LOH in SDHx-WT GIST patients, raises the possibility of primary/de novo SDHC epimutation. A clue to the mechanism of SDHC silencing may be that homozygous SDHCme GISTs in our study arose only in females, implicating a potential role for sex chromosome or hormone biology in the mechanism of SDHC epimutation. The ability to positively identify SDHC-epimutant cases without awaiting the various Carney triad manifestations will enable future studies to test mechanistic hypotheses for this phenomenon.


Study design

Fifty-nine cases from the NIH GIST clinic were identified and included in this study on the basis of molecular and pathologic demonstration of SDH deficiency. The overall goal of the study was to elucidate the molecular mechanism of SDH deficiency in GIST, particularly in SDHx-WT cases. Molecular profiling assays were adapted to archival FFPE surgical pathology specimens, and included NGS, Illumina 450K methylation microarray profiling, and Affymetrix gene expression microarray profiling.

Specimens and annotations

FFPE tumor tissues from 59 distinct NIH GIST clinic cases were tested. All cases received histopathologic diagnosis including immunohistochemistry for SDHA and SDHB protein expression where feasible [National Cancer Institute (NCI) Laboratory of Pathology]; only SDH-disabled tumor cases as evidenced by SDH-negative immunophenotype and/or hypermethylation epigenomic signature [methyl-divergent signature (7)] were included in the study. Malignant cell fraction in tumor tissue DNA extracts was empirically measured using the MDI as follows: the top 100 hypermethylated targets in dSDH GIST were identified by t-test comparison to KIT-mutant/SDHB+ immunophenotype tumors from GEO data set GSE343877. In dSDH GIST, these 100 CpG targets approach methylation β values of 1.0, versus 0.0 in KIT-mutant/SDHB+ GIST, indicative of biallelic/homozygous de novo methylation in the former (fig. S1). The MDI was calculated as the average β value of these 100 targets and termed MDI100, the empirically measured malignant cellularity of tumor tissue DNA extracts (Table 1).

SDHx mutation analysis

Cases received CLIA-certified laboratory testing for coding sequence mutations in SDHA, SDHB, SDHC, SDHD, KIT, PDGFRA, BRAF, and NF1 genes [OncoVar-GIST assay, NCI Clinical Molecular Profiling Core (CMPC)]. The mean read depth for targeted gene coding sequences was >100×. Sequence reads were aligned with Burrows-Wheeler Aligner (BWA), and variants were called by mpileup and by visual inspection of alignments in Integrated Genome Viewer (IGV). Cases were assayed for genomic copy number aberrations and copy-neutral LOH with the Illumina FFPE CytoSNP assay after DNA treatment with FFPE restoration solution (Illumina) (Table 1 and fig. S2). Tumor genotyping microarray data were visualized with Nexus software (Biodiscovery Inc.). Cases were thereby annotated as SDHx-mutant versus SDHx-WT for subsequent statistical group comparisons. Additional case annotations included patient age, sex, and diagnosis of Carney triad (Table 1). The diagnosis of Carney triad was provided by treating physicians, and required clinical evidence of pulmonary chondroma in addition to GIST.

DNA methylation profiling

DNA extraction and genome-wide DNA methylation profiling by Illumina 450K Infinium assay were performed as previously described (7). Briefly, microdissected FFPE tissue (paraffin block needle core or glass slide razor scrape) was lysed in a cocktail containing mineral oil (for deparaffinization), proteinase K, and ATL lysis buffer (Qiagen), and resultant lysates were filter-purified by Qiagen DNA enrichment columns. Purified DNAs were treated with FFPE DNA restoration solution (Illumina) and then analyzed with the standard protocol Illumina 450K methylation beadarray assay.

Gene expression analysis

Total RNA was extracted from 0.6- to 1.0-mm FFPE tissue cores (paraffin block needle core or glass slide razor scrape) with the RNeasy FFPE Kit (Qiagen) according to the manufacturer’s instructions. RNA quantity and purity were determined with a NanoDrop spectrophotometer (ND-1000, Thermo Scientific). RNA integrity was evaluated with the Bioanalyzer RNA 6000 Nano Kit (Agilent). One hundred nanograms of total RNA sample was converted to complementary DNA (cDNA) and SPIA-amplified with the Ovation FFPE WTA System (NuGEN) according to the manufacturer’s instructions. SPIA-amplified cDNA was purified with the Agencourt RNAClean XP Kit (Beckman Coulter Genomics) according to the Ovation FFPE WTA System supplementary protocol. Purified cDNA was quantitated with a NanoDrop spectrophotometer. After quantitation, 4 to 5 μg of purified cDNA sample were subjected to fragmentation and biotin labeling with the Encore Biotin Module (NuGEN) according to the manufacturer’s instructions. An aliquot of unfragmented and fragmented cDNA was evaluated for product size with the RNA 6000 Nano Assay. A sample hybridization cocktail, consisting of the biotin-labeled fragmented sample, hybridization controls, and hybridization buffer, was prepared according to Encore Biotin Module instructions for standard array format. The sample hybridization cocktail was applied to a GeneChip Human Genome U133 Plus 2.0 array (Affymetrix) and hybridized for 18 hours at 45°C with rotation in an Affymetrix GeneChip Hybridization Oven. The following day, the array “wash and stain” procedure was performed with the automated Affymetrix GeneChip Fluidics Station 450 and fluidics script FS450-0004. The arrays were scanned with the Affymetrix GeneChip Scanner 3000 7G.

As an alternative to and validation of Nugen-derived gene expression profiles, RNAs extracted from FFPE tissues were amplified with the SensationPlus FFPE Amplification and 3′ IVT Labeling Kit (Affymetrix) and hybridized to GeneChip Human Genome U133 Plus 2.0 Array (Affymetrix) expression arrays (fig. S7). Briefly, according to the manufacturer’s protocol, 200 ng of total RNA sample was subjected to one round of cDNA synthesis followed by sense RNA amplification, a second round of cDNA synthesis, fragmentation, and biotin labeling. Subsequent array hybridization and processing steps were the same as described above for Nugen-amplified RNAs. SDHC expression silencing was also validated by quantitative RT-PCR (fig. S8) as follows. cDNA was generated from FFPE-extracted total RNA with the MessageBOOSTER Whole Transcriptome cDNA Synthesis Kit for quantitative PCR (Epicentre, cat. no. MBWT80510) according to the manufacturer’s instructions. SDHC primers, tccagaccggaacccaagat (forward) and cgaccaacgtgtctcagcaa (reverse), were used to amplify a 50-bp spliced transcript; the ACTB PrimePCR SYBR Green assay (Bio-Rad) was amplified as an internal control for expression level normalization. SsoAdvanced Universal SYBR Green Supermix (Bio-Rad, cat. no. 172-5270) was used to amplify SDHC and ACTB cDNA targets according to the manufacturer’s protocol with the CFX96 instrument (Bio-Rad), operated by CFX Manager software. Cycling conditions included a polymerase activation step of 98°C for 2 min and 40 cycles of 98°C for 2 s and annealing/extension at 60°C for 5 s with melt curve analysis from 65° to 95°C in 0.5°C increments.

Deep sequencing of SDHC and flanking regions

A 130-kb window centered on the SDHC gene, including its promoter and flanking genes (hg19 chr1:161246130-161408301), was targeted for deep sequencing in SDHC-epimutant cases using tiling custom-capture baits applied to GIST gDNA libraries constructed according to standard protocols (Illumina Tru-Seq). The custom capture baits consisted of biotinylated probes derived by Klenow amplification of SDHC-containing BAC (bacterial artificial chromosome) clone RP11-122G18. Posthybridization libraries were sequenced on the MiSeq to an average >500× coverage.

Microarray processing, normalization, and statistical analyses

Infinium 450K methylation array image files (.idat) were imported to GenomeStudio software (Illumina Inc.), using the methylation module; Cy3/Cy5 color channel normalization and background subtraction were performed according to the manufacturer’s instructions; sample methylation β values (which approximate the percent methylation at genomic CpG loci) were computed with the GenomeStudio methylation analysis function, and β values were exported from GenomeStudio for subsequent statistical analyses and visualization with Qlucore Omics Explorer v.3 software (QOE). For each patient’s blood or saliva samples, target methylation β values were calculated from 450K methylation array data using the group methylation profile in GenomeStudio; this function computes the average target methylation β value for any sample technical replicates. Tumor SDHCme zygosity and blood and saliva SDHCme levels were based on the average β value across SDHC promoter CpG targets mapping to the TSS-proximal CGI (cg00576014, cg01931502, cg11221265, cg12036621, and cg17496230). Normalized gene expression profiles were generated from Affymetrix .CEL files by importing .CEL files into QOE and selecting the RMA (robust multi-array average) normalization algorithm. The QOE QC report tool was used to verify the quality of individual expression arrays and exclude outliers. Samples were annotated into SDHx-mutant versus SDHx-WT groups as described above (SDHx mutation analysis), and the between-group significance of each 450K methylation CpG target variable or U133P2 probe set variable [q value, a false discovery rate–corrected P value (30)] was computed in QOE, which uses a t test for two-group comparisons. CpG target and expression probe set variables were annotated for hg19 genomic map position and closest gene, using microarray manufacturer annotations and probe sequence information. –Log10(q) methylation (upward facing hypermethylation) and log10(q) gene expression (downward facing hypoexpression) were plotted as a function of genomic map position using R (Fig. 1A). The validity of gene expression microarray statistical approaches was verified by analysis of sex-linked gene expression, which showed data consistent with specimen male/female sex annotation (fig. S7).


Fig. S1. dSDH GIST methyl deviation index.

Fig. S2. Genotypes of SDHx-mutant and SDHx-WT dSDH GIST.

Fig. S3. SDHA, SDHB, SDHC, and SDHD promoter methylation in SDHx-WT versus SDHx-mutant GIST.

Fig. S4. Noncorrespondence of hypomethylation and hyper-expression in SDHx-WT GIST.

Fig. S5. Homozygous, hemimethylated, and negative SDHCme GIST groups.

Fig. S6. Lone case of SDHC promoter methylation identified in GEO and TCGA databases.

Fig. S7. Analysis of SDHC promoter methylation in placenta, sperm, embryonic stem cells, and induced pluripotent stem cells.

Fig. S8. Validation of gene expression profiling.

Fig. S9. Analysis of constitutional SDHCme.


  1. Funding: We thank the Life Raft Group and the Intramural Research Program of NIH, NCI, Center for Cancer Research, for funding this study. Author contributions: Designed and performed experiments: J.K.K., M.M., R.L.W., J.J.W., N.N., P.R., Z.Y., W.I.S.J., C.C.L., M.P., J.W., H.S., C.S., Z.W., J.L., S.Y.K., S.A.B., L.J.H., and P.S.M.; analyzed data: J.K.K., M.M., Y.W., Y.J.Z., W.I.S.J., M.S.K., S.Y.K., S.A.B., L.J.H., and P.S.M.; wrote the paper: J.K.K., M.M., W.I.S.J., M.S.K., L.J.H., and P.S.M. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The original gene expression data shown in this publication have been deposited in the NCBI (National Center for Biotechnology Information) GEO and are accessible through GEO Series accession no. GSE56670.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article