Distinct subgroups of HPV(+) and HPV(−) tumors harbor inactivating NSD1 mutations
In the previously published TCGA HNSCC cohorts (n = 273  or n = 421 ), NSD1 was identified as a significantly mutated gene. We found NSD1 gene alterations in 13% of HPV-negative (n = 457) and 9% of HPV-positive (n = 65) tumors (Fig. 1). The majority (61.3%) of NSD1 gene alterations in HPV(−) specimens were truncating mutations that occurred within or before the enzymatic SET domain of the NSD1 protein (Fig. 2A). The remaining alterations included homozygous gene deletions (6.7%), missense point mutations (30.7%) and a single in-frame deletion (1.3%). UMD-Predictor assigned 18 out of the 23 missense point mutations as pathogenic (Fig. 2B) . Furthermore, mutant NSD1 was found to be underexpressed in HPV(−) tumors compared to wild-type NSD1 (Fig. 3A). Likewise, of the six HPV(+) NSD1 mutations, which were comprised of truncating (33%) and missense point (66%) mutations (Fig. 1), 3 of the 4 missense point mutations were predicted to be pathogenic or probably pathogenic by UMD-Predictor (Fig. 2C). However, in contrast to HPV(−) tumors, mutant NSD1 was not underexpressed in our limited set (n = 6) of HPV(+) head and neck cancer specimens (Fig. 3B), though this finding must be examined in a larger cohort before any definitive conclusions can be drawn.
Clinical characteristics of HPV(−) and HPV(+) tumors comparing those with and without NSD1 mutations
Statistical analysis of clinical characteristics of the HPV(−) cohort demonstrated that the NSD1 mutant and wild-type groups did not differ significantly in terms of average age at diagnosis or sex, race, alcohol use, or clinical stage (Table 1). Differences in smoking history were detected, with the cohort harboring NSD1 mutations having higher proportions of current and former smokers compared to those without NSD1 mutations (p < 0.001). Although only a small number of HPV(+) tumors were available for analysis, no differences between the NSD1 mutant and wild-type groups were found in average age at diagnosis, sex, race, smoking history, alcohol use, or stage (Table 2). Thus, NSD1 mutations are not associated with any clinical characteristics, except of smoking status in HPV-negative tumors.
NSD1 mutations are enriched in laryngeal HPV(−) tumors
Classification of the TCGA HPV(−) cohort by site revealed that laryngeal cancers were overrepresented amongst those with mutant NSD1: 25% of HPV(−) tumors arose in the larynx but almost 60% of HPV(−) tumors with NSD1 mutations occurred in the larynx (Fig. 4). In contrast, oral tongue tumors were underrepresented in the NSD1 mutant cohort, comprising 27% of HPV(−) tumors but only 11% of HPV(−) tumors with NSD1 mutations. In agreement with prior reports , NSD1 mutations are enriched in laryngeal HPV(−) tumors.
NSD1 mutations correlate with improved survival in HPV(−) HNSCC but worse survival in HPV(+) HNSCC
We next performed Kaplan-Meier survival analysis of head and neck patients in the expanded TCGA cohort based on HPV and NSD1 mutation status. Interestingly, patients with HPV-negative HNSCC whose tumors contain alterations in the NSD1 gene (n = 60), showed significantly improved overall (p = 0.006) and disease-free (p = 0.007) survival compared to patients with tumors that harbor wild-type NSD1 (n = 397) (Fig. 5A). Improved survival associated with NSD1 mutations in HPV-negative HNSCC is consistent with previously published work showing that NSD1 mutations are a favorable prognostic biomarker in head and neck tumors [17, 20]. In contrast, survival analysis on HPV(+) tumors with and without NSD1 mutations surprisingly revealed that NSD1 mutations were associated with significantly worse overall (p = 0.0019) and disease/progression-free survival (p = 0.002) (Fig. 5B). These data suggest that NSD1 mutations may have different impact on patients’ survival based on HPV status, although confirmation is needed in a larger HPV-positive HNSCC cohort.
NSD1 mutations are associated with increased sensitivity to cisplatin
To begin exploring the factors underlying the survival differences between patients harboring tumors with wild-type and mutant NSD1, we used Cancerrxgene.org, a database in which > 1000 genetically characterized human cancer cell lines have been screened against a wide range of anti-cancer drugs . Although the database did not include enough head and neck cancer cell lines for separate cohort analysis, the pan-cancer cohort showed significantly increased sensitivity (p = 0.0198) to cisplatin for cell lines with NSD1 mutations (n = 18) compared to those with wild-type NSD1 (n = 798) (Fig. 6A). Cisplatin, a DNA crosslinking agent, is the standard chemotherapy drug used as a radiation sensitizer for locally advanced HNSCC . Increased sensitivity to cisplatin may in part explain why NSD1 mutations confer a survival advantage compared to wild-type NSD1 in HPV(−) tumors. In contrast to improved sensitivity to cisplatin, NSD1 mutant cell lines had the same sensitivity as NSD1 wild-type cells to several other types of DNA damaging agents: bleomycin (Fig. 6B), a radiomimetic drug that produces DNA single and double strand breaks ; etoposide (Fig. 6C), a topoisomerase II inhibitor that generates DNA double strand breaks (DSBs) [25, 26]; and paclitaxel (Fig. 6D), a DNA damage-inducing antimicrotubular drug.
Inactivation of ERCC5 has been shown to result in increased sensitivity to cisplatin in a variety of cancers [27,28,29]. The TCGA cohort was queried to determine if NSD1 mutations correlated with ERCC5 expression. Expression of ERCC5 was significantly lower in HPV(−) tumors with NSD1 mutations, compared to HPV(−) tumors with wild-type NSD1 (q = 0.0314) (Fig. 7A). However, for the limited number of HPV(+) tumors (n = 6 for NSD1 mutants), there was no difference in ERCC5 expression between those tumors with wild-type and mutated NSD1 (q = 0.975) (Fig. 7B).
NSD1 knockdown increases cell sensitivity to DNA-crosslinking agents, but not to other DNA-damaging agents
With indications that NSD1 mutant and wild-type cells respond differently to cisplatin (Fig. 6A), we depleted NSD1 to determine the effect of NSD1 on survival of HNSCC cells or immortalized keratinocytes following treatment with platinum or other types of DNA-damaging agents. We first depleted NSD1 with specific shRNAs in FaDu, SCC35 (both are human head and neck cancer cell lines) cells, and HaCat (human immortalized, non-tumorigenic epithelial) cells, and selected knockdown clones using different shRNAs for each line that had > 50% depletion of NSD1 expression compared to parental cells or cells expressing control shRNA (Fig. 8).
Cell viability assays of NSD1 wild-type (control) and NSD1-depleted (NSD1 sh) FaDu cells showed that NSD1 knockdown rendered cells more sensitive to increasing concentrations of carboplatin and cisplatin (Fig. 9A top and bottom). FaDu, HaCat, or SCC35 cells expressing another NSD1 shRNA (NSD1 sh a) were also more sensitive to cisplatin than cells expressing wild-type NSD1 (Fig. 9B, C, and D). Clonogenic survival of FaDu and HaCat cells demonstrated that increasing concentrations of carboplatin also inhibited survival of NSD1-depleted cells (Fig. 9E). Thus, NSD1 knockdown caused increased sensitivity of epithelial cells to platinum-based chemotherapy drugs in both short-term cell viability and long-term cell survival assays.
As opposed to platinated drugs, cell viability assays showed that depletion of NSD1 did not sensitize cells to etoposide, zeocin (a chemical analog of radiomimetic drug bleomycin), paclitaxel, or mirin, an MRE11 inhibitor that prevents activation of a DNA damage sensor molecule Ataxia Telangiectasia Mutated (ATM) and DNA damage signaling (Fig. 10 and B). Conversely, NSD1 knockdown cells were relatively resistant to treatment with increasing concentrations of zeocin, paclitaxel, and mirin. Etoposide had a concentration-dependent effect on FaDu cells, with NSD1 knockdowns being more resistant to etoposide at low concentrations (5 μM and 10 μM), but more sensitive to etoposide at higher concentrations (20 μM and 25 μM).
Expression of ERCC5 was decreased in HPV(−) tumors with NSD1 mutations, compared to HPV(−) tumors with wild-type NSD1 (Fig. 7A). ERCC5 mRNA levels were also lower in HaCat and FaDu, but not in SCC35 cells expressing NSD1 shRNAs, as compared to control shRNA expressing cells (Fig. 11).