Distinct pattern of TP53 mutations in human immunodeficiency virus–related head and neck squamous cell carcinoma

Human immunodeficiency virus–infected individuals (HIVIIs) have a higher incidence of head and neck squamous cell carcinoma (HNSCC), and clinical and histopathological differences have been observed in their tumors in comparison with those of HNSCC patients without a human immunodeficiency virus (HIV) infection. The reasons for these differences are not clear, and molecular differences between HIV‐related HNSCC and non–HIV‐related HNSCC may exist. This study compared the mutational patterns of HIV‐related HNSCC and non–HIV‐related HNSCC.


INTRODUCTION
Human immunodeficiency virus-infected individuals (HIVIIs) have a higher cancer incidence than the general population. Kaposi sarcoma and non-Hodgkin lymphoma are the most common neoplasms among HIVIIs and are AIDS-defining malignancies. However, since the advent of antiretroviral therapy, the incidence of other malignant tumors (non-AIDS-defining cancer) has increased, and they have become important causes of death for these patients. [1][2][3][4] Three factors are considered important for the elevated cancer incidence among HIVIIs: high smoking rates, immunosuppression, and increased susceptibility to infection by oncoviruses. Smoking is the main preventable cause of cancer but is a common habit among HIVIIs. The prevalence of tobacco smoke exposure is 2 times higher in this group versus the overall population. 5,6 Immunosuppression is the main characteristic of human immunodeficiency virus (HIV) infection, and it is considered to be an important cause of cancer development. Organ transplant patients on immunosuppressive regimens demonstrate an increased incidence of malignant tumors in comparison with the general population. Their cancer incidence is similar to the frequency observed among HIVIIs, and this suggests that immunodeficiency plays an important role in HIVII-related carcinogenesis. 1 An important common effect of tobacco abuse and immunosuppression is an increased susceptibility to infection by oncoviruses with the subsequent development of virus-related carcinogenesis. [7][8][9][10] Virus-related tumors are almost 10 times more frequent in HIVIIs versus the general US population. 11 Human herpesviruses and human papillomaviruses (HPVs) are responsible for the majority of these cases, and they lead to the development of not only sarcomas and lymphomas but also squamous cell carcinomas in the anogenital and head and neck regions. 11 Although human herpesviruses are related to AIDS-defining malignancies, HPVs are the major oncoviruses linked to non-AIDS-defining cancers. 7 All these elements are also recognized as important risk factors for head and neck squamous cell carcinoma (HNSCC) development and may explain the higher incidence of these tumors among HIVIIs. 1,2,4,[12][13][14] HNSCC is diagnosed at an earlier age and at a more advanced stage in HIVIIs, 15 and these tumors tend to be more aggressive 16 and related to worse survival rates when they express high levels of TP53 in comparison with non-HIVIIs. 17 The efficacy of current treatment approaches to HIVIIrelated HNSCC is still a matter of debate. 18 Radiotherapy treatment has been found to be less effective for the control of tumor relapse and related to worse overall survival and more treatment-related toxicity in HIVIIs. 18 Whether these differences in presentation and prognosis are related to the systemic effects of HIV-mediated immunosuppression or by particular biological characteristics of the primary tumor is still unclear.
Histopathological findings indicate that HIV-related HNSCC has unique features such as the enrichment of multinucleated giant tumor cells and the expression of HIV-related protein in some tumor cells. 19 These observations indicate unique pathological processes that are associated with these tumors. To better understand the behavior of HIV-related HNSCC and possibly develop personalized treatment approaches, it is important to determine whether this represents a distinct molecular entity.
Recently, integrated genomic studies in HNSCC have dramatically expanded our knowledge about the pathogenesis, progression, and treatment of this tumor type. For instance, these studies have revealed the genes and pathways most frequently affected in HNSCC and have identified HPV-related HNSCC as a distinct molecular entity. [20][21][22][23] In this way, we believe that a genomic analysis of HIV-related HNSCC might reveal whether this group of tumors has a distinct pattern of DNA alterations that could indicate differences in their pathogenesis and progression. To accomplish this, we compared the pattern of mutations between HIV-related HNSCC and non-HIV-related HNSCC by sequencing a panel of genes frequently mutated in head and neck cancer.

Patient Selection
The cohort of patients used in this study was obtained from the Head and Neck Cancer Specialized Program of Research Excellence Human Immunodeficiency Virus Supplement Consortium. IRB approval or exemption to share deidentified data with the study data center was obtained from sample collection sites. A detailed description of the patient recruitment, sample collection, and clinicopathologic data collection has been provided in a previous publication. 17 Briefly, formalin-fixed, paraffin-embedded (FFPE) tissue from patients diagnosed with HNSCC and an HIV infection were retrieved after a retrospective review of medical records. HNSCC patients not infected by HIV were also retrieved as an age-, subsite-, and stage-matched control group. DNA was extracted from the FFPE tissues, and only samples with sufficient DNA yields for sequencing were included in this study.

Library Preparation and DNA Sequencing
Ten nanograms of DNA was used as the input for the target DNA library preparation with the Ion AmpliSeq Library Kit. For the amplification of the targeted DNA, we used a customized pool of primers designed for the amplification of all exon regions of the following genes: AJUBA, CASP8, CCND1, CDKN2A, EGFR, FAT1, FBXW7, HLA-A, HRAS, KEAP1, NFE2L2, NOTCH1, NOTCH2, NSD1, PIK3CA, TGFBR2, TP53, and TP63. The target DNA libraries were sequenced with the Ion Personal Genome Machine sequencer platform. Variant calls were made on the Ion Reporter server with the AmpliSeq tumor-normal pair comprehensive cancer panel pipeline with customized filters. Although HLA-A is included in the sequencing assay, it is currently excluded from the analysis because of the difficulty in accurately calling mutations in this highly polymorphic gene.

TP53 Mutation Classification
TP53 gene mutations were classified according to 2 functional-impact and risk-classification systems proposed by Poeta et al 24 and Neskey et al. 25 With Poeta et al's system, the TP53 mutations were classified as disruptive and nondisruptive mutations. Disruptive mutations were those inducing a disruption of p53 protein production (nonsense, frameshift, in-frame, and splice site mutations) or any missense mutation occurring within L2 or L3 DNA binding domains (codons 163-195 and 236-251) and changing the original amino acid polarity or charge category. For Neskey et al's classification, we used the online EAp53 server, 26 which classifies TP53 mutations into high-and low-risk categories for HNSCC and determines a numeric risk score (Evolutionary Action score -EAscore). Only missense mutations are classifiable in this system.

Statistical Analysis
For the statistical analysis, we used IBM SPSS Statistics (version 22) and GraphPad Prism (version 6.07) for Windows. Associations between categorical variables were determined with Fisher's exact test. Associations between categorical and quantitative variables were determined with the Mann-Whitney test when categories had 2 values and with the Kruskal-Wallis test when categories had 3 or more values. Significant associations were considered when the P value was lower than .05. The log-rank test was used to determine differences among survival curves.

RESULTS
To understand the mutation frequencies and patterns in HIV1 HNSCC, we sequenced tumor samples from 20 HIV1 HNSCC patients. We also sequenced 32 HIVpatients as controls (Table 1). Among HIV1 cases, 11 (55%) were HPV-, and 9 (45%) were HPV1; among HIV-cases, 6 (18.75%) were HPV1, and 26 (81.25%) were HPV-. Sex and tissue sites did not differ among these groups, but HIV1HPV1 patients were significantly younger than HIV-HPV1 patients. All HIV1 patients were smokers or former smokers, and all except one were alcohol users. No information about tobacco and alcohol consumption was available for HIV-cases.
DNA was extracted from FFPE samples and sequenced on a custom Ion Torrent AmpliSeq panel containing 18 genes frequently altered in HNSCC. The identified mutations as well as the mutation frequencies described for HNSCC in The Cancer Genome Atlas (TCGA) 22 are listed in Table 2, and summary oncoprints are shown in Figure 1. Because HPV is known to alter the mutational landscape of HNSCC, the cohort was divided Original Article into 4 groups for most analyses based on the HIV and HPV status. The associations were assessed through comparisons of all 4 groups together and also through comparisons of HIV1 and HIV-groups with the same HPV status. Overall, 84.6% of the patient tumors were found to have at least 1 mutation in the examined genes, and the number of patients with tumors harboring mutations was significantly different among the 4 groups (P 5 .002). Among the HPV1 cases, there was no difference in the number of tumors harboring mutations when HIV1 (55.6%) and HIV-patients (66.7%) were compared. However, among the HPV-cases, the number of mutations was higher in the HIV-group (100%) versus the HIV1 group (81.8%; Fig. 2A). The number of patients harboring mutations was significantly lower overall in the virus-related groups (HIV1 or/and HPV1) versus the HPV-cases from TCGA (P < .001) as well as the HIV-HPV-group (P 5 .004). The number of genes mutated per patient and the number of mutations per patient varied among the groups (Fig. 2B,C). Both variables showed higher values among HPV-patients, regardless of the HIV status.
No mutations were detected in 4 of the studied genes that are known to have low mutation frequencies in HNSCC (AJUBA, CASP8, CCND1, and TGFBR2). Mutations in the TP53 and NOTCH1 genes were detected in all groups. When we compared the frequencies of gene mutations among the groups, only TP53 mutations were significantly different (P < .001; Table 2). HIV-HPV-patients had the highest frequency of mutations in the TP53 gene (88.5%), whereas HIV1HPV1 patients had the least number of mutations in this gene (11.1%). A strikingly high frequency of TP53 mutations was observed in HPV1 patients (33.3% of all 15 HPV1 cases) in comparison with other HNSCC HPV1 sequencing studies. We do not currently have an explanation for this high rate. Mutations in the NFE2L2 gene were found in 2 of 20 HIV1 patients (10%); this was a higher frequency of detection in comparison with the frequency for non-HIV patients (3.1%) and that reported for the TCGA cohort (6.5%). NOTCH1 mutations were observed in 3 of 9 HIV1HPV1 patients (33.3%), and this was a higher mutation frequency in comparison with the frequency of our other groups and that reported in the TCGA data set. However, statistical significance was not reached for the NFE2L2 and NOTCH1 associations. NSD1 mutations were differentially distributed among tumor sites (P 5 .025): they were observed in the larynx (2 mutated cases) and other head and neck sites (1 mutated case) but were absent in oral cavity and oropharynx cases. This is consistent with the distribution of NSD1 mutations in TCGA. The frequency of mutations in any gene did not differ significantly between HIV1 and HIV-tumors with the same HPV status.
To assess qualitative differences in TP53 mutations, we compared the types of TP53 mutations among the groups (Fig. 3). Although HPV1 cases exhibited only missense TP53 mutations, HPV-tumors also contained truncating and in-frame indel mutations. Among HIV1HPVcases, truncating mutations were the most common (66.7%), whereas these mutations represented just 26.9% of the HIV-HPV-events. Next, we used 2 TP53 functional-impact and risk-classification systems previously used for HNSCC (Table 3). With both systems, TP53 mutations related to aggressive tumors (high-risk and disruptive) were more prevalent in HIV-and HPV-cases. The mutations in virus-related cases were predominantly classified as wild-type or low-risk/nondisruptive mutations. Truncating and frameshift mutations are not classifiable in Neskey et al's system 25 and are, therefore, called other. These were more prevalent in the HIV1 group (45% of mutated cases) versus the HIV-group (30.8% of wild-type cases).

Original Article
We also assessed the types of nucleotide changes in each studied group for all genes as well as the TP53 gene individually (Fig. 4). C>T mutations were the most prevalent nucleotide changes observed in all groups and for the majority of the genes. Only the genes NFE2L2, NSD1, and PIK3CA showed more C>G, C>A, and T>C nucleotide changes, respectively. No significant difference in the distribution of nucleotide changes was observed among the 4 groups (P 5 .93; Fig. 4A) or between HIV1 and HIV-groups with the same HPV status (P 5 .24 for HPV1 tumors and P 5 .98 for HPVtumors). No difference in the pattern of nucleotide change distribution was observed when patients were grouped according to their HIV status (P 5 .95; Fig. 4B).
As for nucleotide changes in the TP53 gene, C>T changes were the most commonly observed, but no statistical difference was observed among the 4 groups (P 5 .80; Fig. 4C) or between HIV1 and HIV-samples with the same HPV status (P 5 .34 for HPV1 tumors and P 5 .92 for HPV-tumors). Although the frequency of C>T changes was higher among tumors with an HIV infection (66.7%) versus HIV-tumors (43.3%), the difference between them was not significant (P 5 .95; Fig. 4D).
Considering the high frequency of C>T mutations in these samples, we investigated the 5 0 flanking nucleotide for each mutation (Fig. 4E). HIV1HPV-samples exhibited a higher frequency of TpC>T mutations  (26.3%) than tumors of HIV-HPV-patients (10.2%) when all mutated genes were taken into account. As for HPV1 tumors, TpC>T changes were observed only among HIV1 patients (42.9% of all mutations) and were absent among HIV-ones (P 5 .02). When tumors were grouped according their HIV status, regardless of the HPV infection status, the frequency of TpC>T was significantly higher among HIV1 patients (34.8%) versus HIV-patients (10.9%; P 5 .02; Fig. 4F). The same trend was observed when only the TP53 gene was considered: only HIV1 tumors exhibited TpC>T nucleotide changes (P 5 .02; Fig. 4H).
Differences in 5-year survival rates were determined for patients with respect to their HIV/HPV group, TP53 status (wild-type vs mutated), and TP53 mutation type (insertion/deletion, transition, or transversion). No significant differences between survival curves were observed in any comparison (P 5 .45, P 5 .30, and P 5 .46, respectively). The presence of mutations in other genes (alone or in combination) was also not associated with survival in these patients. Because of the absence of TNM information for HIV-patients, the influence of the tumor clinical stage on these results was not able to be determined.

DISCUSSION
In this study, we observed an overall decreased frequency of mutations in HIV1 HNSCCs, and HIV-related HNSCCs had a distinct pattern of genetic mutations in the TP53 gene characterized by truncating mutations and TpC>T nucleotide changes.
HIVIIs are a unique group of patients in the context of HNSCC. They have greater exposure to tobacco and alcohol, have impaired immune function, and consequently are more prone to infection by high-risk HPV. [4][5][6][7]14 In this way, these patients have multiple key risk factors for the development of cancer and particularly HNSCC. Not surprisingly, HIVIIs have higher rates of HNSCC than the general population. It has been reported that HIVIIs with HNSCC present at a more advanced tumor stage and have reduced survival rates. 15,16,18,27 However, it is not clear whether these  findings are caused by intrinsic features of the tumor cells or are related to host factors related to the HIV infection. 22,23,28 We aimed to investigate this by comparing the patterns of mutations in HIV-related and non-HIVrelated tumors in genes frequently mutated in HNSCC.
Because an HPV infection causes a distinct subtype of HNSCC, 22,23,28 we considered the HPV status as a key variable as well.
Our observations regarding genetic alterations in HNSCC from HIVIIs are unique in the literature. The only similar analysis that we could find was performed by Souza et al, 29 who compared the patterns of TP53 mutations in DNA obtained from cervical swabs (normal and altered cytologies) from HIV-infected and uninfected patients (all HPV1). They found that the frequencies of TP53 mutations were similar in the 2 groups (approximately 19%).
Souza et al 29 also found that the genomic position of TP53 mutations changed with the HIV status. HIV1 patients had more mutations located in exon 7, whereas HIV-patients had more mutations in exon 6. Our small sample size made it difficult to compare the genomic locations of TP53 mutations, but we did examine the types of mutations. HIV1HPV-patients presented with a higher frequency of truncating mutations, and the missense mutations were more frequently classified as low-risk and nondisruptive. Many TP53 mutations have been studied for their gain-of-function properties, which can provide novel characteristics to the tumors. Although missense mutations in the p53 DNA binding domain can promote a loss of DNA binding activity, this event may also change the protein structure and lead to new potential protein interactions, which in turn may result in p53 gain of function. 30 The pattern of TP53 mutations in HIV1 cases suggests that there are fewer gain-of-function TP53 mutations, and it is intriguing to speculate that this could be related to the immune status of the individuals or HIV itself.
Another interesting qualitative observation is the different patterns of nucleotide changes in all mutated genes but especially in the TP53 gene among HIV1 and HIV-tumors. The presence of C>A changes in the TP53 gene is considered a hallmark of tobacco-related DNA mutations in lung cancer and also in some HNSCCs. 21,22,31,32 In fact, only 1 HPV1 case in our cohort exhibited C>A transversions. Among the HIV1HPV-cases, the percentage of C>A changes was similar to that among the HIV-HPV-cases (12.5% vs 9.6%), and this might be expected because all the HIV1 patients were smokers. C>T transitions have been demonstrated as being more predominant in virally transformed tumors, including HNSCC tumors. 22,33 However, we demonstrated that a specific subtype of C>T changes, the TpC>T mutations, were enriched in the HIV-infected patients, and they conferred a virus-related mutation fingerprint for these cases. Interestingly, the increased number of TpC>T changes among HIV1 tumors was independent of the presence of an HPV coinfection.
Cytosine deamination is a defense mechanism against retrovirus infections exerted by the APOBEC family of enzymes. Human APOBEC3G is induced by an HIV infection to impair virus infectivity by promoting mutations in its DNA. [34][35][36] However, the mutagenic potential of APOBEC3G is not restricted to the viral genome, and cytosine deamination may also occur in human DNA. The APOBEC-related pattern of mutations, characterized by TpC>T transitions, can be observed in numerous human cancers, including HNSCC, and they are commonly described in the TP53 gene. [37][38][39][40][41] Interestingly, oncogenic pathogens such as HPV and Helicobacter pylori have been proved to cause APOBEC3G-mediated mutations in oral and gastric epithelium. 37,41 The enrichment of TpC>T changes in HIV-related tumors might suggest that HIV infection contributes to the TP53 mutation pattern observed in this cohort.
Some studies have demonstrated that HIV-encoded proteins may interact with the p53 protein and modulate its function in different ways. [42][43][44] One of these HIVencoded products, the Nef protein, is believed to interact with p53 and inhibit its function and thus promote HIV infection and replication. Greenway et al 45 observed that Nef inhibits p53-dependent apoptosis, and McLemore et al 19 found the expression of this protein in the cytoplasm of 7 HNSCC samples. These results might indicate that an HIV infection would have a direct role in HNSCC pathogenesis independent of an HPV infection. 46 HIV DNA, RNA, and viral particles have been detected in the oral mucosa of HIVIIs, 47 and it has been demonstrated that epithelial cells from the oral mucosa lining are susceptible to HIV infection. 48 These findings might suggest a direct oncogenic effect of the virus on oral epithelial cells. More extensive sequencing would be necessary to confirm this.
Despite the genetic differences observed among these patients, no significant difference in survival was observed in our analysis with respect to the HIV and HPV status, TP53 mutation status, and TP53 mutation type. We believe that our small sample size might limit definitive conclusions on the impact of such variables on the survival of HNSCC patients. The impact of such variables should be further investigated in larger cohorts.
Considering these results, we conclude that HIVrelated HNSCC likely represents a distinct genomic entity. The alterations observed in this study must be validated in larger cohorts, but they suggest that these tumors in HIVIIs are biologically distinct. More studies are needed to understand the unique etiology, pathogenesis, and biology of these tumors and to determine whether there are unique therapeutic modalities that would benefit these HNSCC patients.

FUNDING SUPPORT
This study was funded by the Head and Neck Cancer Specialized Program of Research Excellence Human Immunodeficiency Virus Supplement Consortium (the National Cancer Institute) and the American Recovery and Reinvestment Act through the following grants: P50 CA097248 to the University of Michigan; University of Michigan Cancer Center Core Grant P30 CA46592; 5P50 CA097007 to The University of Texas MD Anderson Cancer Center; P50 CA097190 to the University of Pittsburgh; P50 DE019032 and 3P50 DE019032-14S2 to Johns Hopkins University; and P50 CA128613 to Emory University. Sarah I. Pai is supported by grant 1R01 DE025340.