Attenuated TRAF3 Fosters Activation of Alternative NF-κB and Reduced Expression of Antiviral Interferon, TP53, and RB to Promote HPV-Positive Head and Neck Cancers Free

Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer, with an annual incidence of 650,000 new cases and ∼200,000 deaths worldwide (1, 2). Persistent infection with high-risk human papillomavirus subtypes HPV16 and HPV18 has been established as an important risk factor for HNSCC that develop predominantly in the oropharyngeal tonsils (3). Since 1981, there has been a notable 225% increase in HPV⁺ HNSCC, while the incidence of smoking-related HPV⁻ HNSCC has declined (4, 5). Clinically, the HPV⁺ subset exhibits better responses to therapies and survival rates than similarly advanced HPV⁻ tumors. However, HPV⁺ HNSCC are distinguished by aggressive spread and growth within regional lymph nodes, which require major surgery or toxic chemoradiotherapy regimens (2, 3). The factors that contribute to the molecular pathogenesis of these unique features of HPV⁺ HNSCC remain incomplete.

HPV16 and 18 carry early genes E6 and E7 encoding oncoproteins that target key pathways, deregulating host resistance to infection and cellular proliferation, to promote the viral life cycle. HPV E6 expression in keratinocytes can repress type-I IFN and promote proliferative genes, to enhance viral protein synthesis and proliferation of virally infected cells (6). Further studies have shown that HPV infection can induce ubiquitin carboxyl-terminal hydrolase L1 (UCHL1), which can inhibit K63 ubiquination, important in Tank binding kinase-immune response factor 3 (TBK-IRF3)-mediated type-I IFN expression (7). HPV E6 can also commandeer and activate the so-called alternative nuclear factor-κB2 (NF-κB2) transcription factors and antiapoptotic genes, which promote resistance of keratinocytes to TNF, an important mediator of antiviral immunity (8). Critically, the HPV E6 and E7 oncoproteins also strategically target for degradation the tumor suppressor proteins TP53 and RB, which control the cell cycle (9). Interestingly, however, few individuals exposed to HPV develop chronic infection and HNSCC. These observations suggest that additional genetic alteration(s) and host factors may also affect how HPV mediates suppression of IFNs, NF-κB activation, inhibition of TP53 and RB gene expression, and the malignant phenotype.

Recently, we and The Cancer Genome Atlas (TCGA) Network uncovered a subset of HPV⁺ HNSCCs that harbor deletions of the chromosome region 14q32.32, deleterious truncating mutations, and/or decreased expression, affecting the gene TNF receptor–associated factor 3 (TRAF3; ref. 10). These deletions were not due to viral integration and disruption of the TRAF3 gene. Intriguingly, TRAF3 is a unique adaptor protein and ubiquitin ligase implicated as a negative regulator of the aforementioned alternative NF-κB2/RELB pathway (11). TRAF3 promotes cIAP-mediated ubiquitination and proteasome-dependent degradation of the pivotal NF-κB inducing kinase (NIK) protein, which mediates signal activation of the alternative pathway. Lymphotoxin-β (LTβ) and other ligands, which are richly expressed in the oropharyngeal tonsils and lymph nodes where HPV⁺ HNSCC arise and spread, bind receptors to activate NIK, IKKα, processing of NF-κB2 precursor p100 to p52, and nuclear translocation of transcriptionally active NF–κB2–p52/RELB dimers. Attenuation of TRAF3 has previously been implicated in the transcription of genes affecting cell fate, proliferation, and survival of lymphoid cells and hematopoietic malignancies (11, 12). Strikingly, TRAF3 has also been shown to serve a dual function in interferon responses to other DNA viruses (11, 13). However, the functional role of genetic alterations or reduced expression of TRAF3 in modulating alternative NF-κB pathway activation, IFN expression, repression of tumor suppressors TP53 and RB, and the malignant phenotype in HPV⁺ HNSCC has not been established.

In this study, we examined and revealed a novel role for decreased TRAF3 in fostering deregulation of these pathways and promoting pathogenesis of a subset of HPV⁺ HNSCC. These findings provide new opportunities for basic and clinical research that may lead to better diagnosis, prevention, and treatment of HPV⁺ HNSCC.

TCGA Project Management has collected necessary human subjects documentation to ensure that the project complies with 45-CFR-46 (the “Common Rule”). The program has obtained documentation from every contributing clinical site to verify that Institutional Review Board approval and informed consent has been obtained to participate in TCGA. The characteristics and data repositories for 279 HNSCC specimens and 16 normal mucosa were previously reported, including 36 HPV⁺, 243 HPV⁻ HNSCC, and 16 normal samples (10). Exome-wide sequencing was performed on all samples, and normalized genomic data for 279 HNSCC and 16 normal samples displaying a squamous keratin gene signature was obtained from the TCGA Genome Data Analysis Center and downloaded on September 19, 2012. All genomic data were processed through standard TCGA analytic pipelines and accessible through TCGA data portals. Significant focal copy-number alterations were extracted from the GISTIC2.0 processing pipeline, and data aggregated on significantly altered lesions are plotted by false discovery rate (FDR) less than 5% and P values of less than 0.05. Differential gene expression analysis of tumor versus normal samples was performed using DeSeq, and all data were log₂ transformed (Bioconductor version 2.12). Molecular HPV signatures were identified using gene expression and somatic substitutions. Gene set enrichment analysis was performed using the online Cancer Genomics cBioportal database (http://www.cbioportal.org/).

UMSCC [University of Michigan (Ann Arbor, MI) series of HNSCC], UPCI (University of Pittsburgh Cancer Institute, Pittsburgh, PA), and VU (Free University Amsterdam, Amsterdam, the Netherlands) cell lines and clinical information were provided by Drs. Thomas E. Carey, Mark E. Prince, Carol R. Bradford (University of Michigan), Robert Ferris, and Susanne Gollin (University of Pittsburgh), or from previous publications (Supplementary Table S1; refs. 14–16). Cell authentication of UMSCC, UPCI, and VU lines was done at the University of Michigan by DNA genotyping of alleles for 9 loci (D3S1358, D5S818, D7S820, D8S1179, D13S317, D18S51, D21S11, FGA, vWA) and the amelogenin locus as described previously (15). A panel of 8 HPV-positive (UMSCC 47, 104, 2, 90, 152, 154, 93VU147T, and 105) and 8 HPV-negative (UMSCC 1, 9, 11A, 11B, 38, 46, 74A, and 74B) HNSCC cell lines was selected for evaluation of HPV type and TRAF3 expression. These cell lines were maintained in MEM or DMEM with 10% fetal calf serum (Life Technologies) at 37°C with 5% CO₂ for a maximum of 8 weeks. Primary human oral keratinocytes (HOK) were cultured, in accordance with the supplier's protocol (Science Cell Research). Prior to experiments, all cells were confirmed to be Mycoplasma negative (by MycoAlert kit, cat. no. M7006, Thermo Fisher). Genomic DNA was extracted using a DNeasy Blood and Tissue Kit (cat. no. 51104, Qiagen) from selected frozen cell pellets. These cell lines were further evaluated for HPV status by PCR (Human Papillomavirus Detection Set, cat. no. 6602, Takara Bio).

Total RNA of 3 primary human keratinocyte and individual HNSCC cell line was isolated by combination of TRIzol (cat no. 15-596-026, Life Technologies) and QIAGEN RNeasy Mini Kit procedure (cat. no. 74104, QIAGEN). Ribosomal RNA was depleted using the Ribo-Zero kit (cat no. MRZH11124, Epicentre). Multiplexed whole transcriptome libraries were generated by SOLiD Total RNA-Seq Kit and SOLiD RNA Barcoding Kit (cat. no. 4427046, Life Technologies), and fragmented cDNA libraries were clonally amplified by emulsion PCR. The multiplex libraries were sequenced utilizing 75 bp forward and 35 bp reverse paired-end sequencing chemistry on the ABI SOLiD system. Reads were mapped into human NCBI Build 37 reference genome (Hg19) using LifeScope v.2.5 Genomic Analysis software (https://www.appliedbiosystems.com/lifescope). The read count of the genes was normalized using DESeq (estimateSizeFactors) R package (version 3.2.0) and upper quartile normalization method. Expression analysis was performed by comparison with normal HOK cell line expression, and gene expression was quantified using RSEM (RNA-Seq by Expectation Maximization).

YFP-pTo-TRAF3 fusion expression vector was kindly provided by Dr. Christina M. Annunziata, NCI/NIH, Bethesda, MD (16). His-tagged TRAF3 wild-type was generated by GeneCopoeia. For efficient retroviral transductions, certain cell lines were engineered to express these indicated vectors. SMART pool siRNAs targeting human TRAF3 (cat. no. E-005252) and nontargeting siRNA control (M-HUMAN-XX-0005) were purchased from Dharmacon. Cell transfection was carried out using X-tremeGENE HP DNA Transfection Reagent or X-TremeGENE siRNA Transfection Reagent (cat. no. 10465500, 063662, Roche), following the manufacturer's instructions. Stably transfected cell lines were allowed to recover for 48 hours before antibiotic selection, and only confirmed surviving pools were utilized for subsequent analysis.

UPCI-SCC-90 and UMSCC47 cells were cotransfected with TRAF3 CRISPR/Cas9 KO and TRAF3 HDR plasmids (Santa Cruz Biotechnology) at equivalent ratios using lipofectamine 2000 (cat. no. 11668019, Thermo Fisher). Forty-eight hours after transfection, the cells were plated at low density with the addition of 1 μg/mL puromycin (cat. no. A1113802, Thermo Fisher) for selection of stable clones. Successful knockdown was confirmed with RT-PCR. Clonogenic assay was performed as described (17).

Data are shown as means ± SEM. Student t test or two-way ANOVA were performed using GraphPad. Statistical analysis is specifically indicated for each experiment (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001).

Additional methods of preparation of total RNA isolation, cDNA synthesis, qRT-PCR analysis, protein extracts, Western blot analysis, reporter assay, immunofluorescence staining, treatments, and cell viability are presented in the Supplementary Information.

The recent TCGA study revealed that a subset of HPV⁺ HNSCC display novel recurrent deletions, truncating mutations, and/or decreased expression of TRAF3 (10). Here, we further show that TRAF3 focal or broader deletions of chromosome 14 (Chr14q32.32) are mainly observed in HPV⁺ HNSCC (Fig. 1A). Among 36 HPV⁺ tumors in this dataset, 14% homozygous and 25% heterozygous deletions were found, whereas in 243 HPV⁻ tumors, only 1.2% homozygous and 10% heterozygous deletions were observed (Fig. 1B, Fisher exact test, P = 2.1 × 10⁻⁶). In contrast, 38% of HPV⁻ tumors exhibited TRAF3 copy-number gains compared with only 6% of HPV⁺ tumors. Together, the different distribution of these DNA copy-number variations were associated with TRAF3 mRNA expression in HPV⁺ (P = 3.52 × 10⁻⁵) and HPV⁻ HNSCC samples (P = 2.0 × 10¹⁰; Fig. 1C). Furthermore, three cases with deleterious frameshift or nonsense mutations were observed only in HPV⁺ HNSCC.

To determine if there are HPV⁺ HNSCC cell lines with decreased TRAF3 expression suitable for study, we investigated TRAF3 expression in a panel of eight HPV⁺ and HPV⁻ HNSCC cell lines (Supplementary Table S1), using RNA sequencing data. Overall, we observed significantly increased TRAF3 mRNA expression in HPV⁻, compared with normal HOK and HPV⁺ cell lines (P < 0.05, Fig. 1D), consistent with the few HPV⁺ tumors and cell lines with low expression, as well as high number of HPV⁻ tumors and cell lines with gain of expression of TRAF3. We further compared TRAF3 protein levels in the 8 HPV⁺ HNSCC cell lines with normal human oral keratinocytes (HOK). UMSCC47 and UMSCC104 cell lines exhibited lower TRAF3 protein levels as assessed by Western blot (Fig. 1E), and these lines were selected for further study of the effects of decreased TRAF3 expression and modulation.

As TRAF3 was previously identified as a negative regulator of the alternative NF-κB pathway in immune and hematopoietic malignant cells (11, 12), we asked whether altered TRAF3 is associated with NF-κB pathway alterations in HNSCC tissues. Using the HNSCC TCGA datasets, we identified an apparent co-occurrence between TRAF3 genomic and expression alterations with the mRNA expression of alternative NF-κB pathway components (Fig. 2A left, top), but not with those in classical NF-κB pathway (Fig. 2A left, bottom), in HPV⁺ HNSCC tumors. Conversely, no such co-occurrence was observed in HPV⁻ HNSCC tumors (Fig. 2A, right). Supporting this, we observed a significantly higher average percentage of alterations in the NF-κB alternative pathway in HPV⁺ tumors, compared with HPV⁻ tumors, whereas no significant difference was observed in the classic NF-κB pathway (Fig. 2B). Finally, the co-occurrence and mutual exclusivity of paired alterations were calculated, and the relationships with statistical significance are indicated as color intensities (Fig. 2C; Supplementary Table S2). We observed more frequent co-occurrence of alterations in the key components of the alternative than the classic NF-κB pathway in HPV⁺ HNSCC tumors (Fig. 2C, left). In contrast, fewer tumor samples with strong and intermediate co-occurrences were observed in HPV⁻ HNSCC (Fig. 2C, right). These analyses indicate that deficient TRAF3 is associated with increased expression of alternative NF-κB pathway components in HPV⁺ HNSCC tumor specimens.

To test the hypothesis that decreased TRAF3 is permissive for activation of the alternative NF-κB pathway in HPV⁺ HNSCC, we examined the expression of alternative pathway components and the effects of transiently expressing His-tagged TRAF3 (His-TRAF3) or control (His-Control) vectors in the HPV⁺ UMSCC47 and UMSCC104 cells that displayed relatively lower endogenous TRAF3 expression. TRAF3 mRNA was significantly increased in both cell lines after TRAF3 transfection (Fig. 3A, top). Next, we evaluated the mRNA expression levels of key alternative pathway subunits RELB and NF-κB2, which are modulated at transcriptional and posttranslational levels, as well as classic (RELA/p65) NF-κB subunit, which is constitutively expressed and modulated posttranslationally (18). Consistent with this, transient TRAF3 overexpression in both HPV⁺ cell lines resulted in a significant reduction of RELB and NF-κB2 mRNA expression, without affecting RELA expression (Fig. 3A). We then evaluated the effect of TRAF3 expression on protein levels of key alternative NF-κB pathway components in the same cell lines and an additional HPV⁺ HNSCC cell line, 93VU174T, following TRAF3 transfection (Fig. 3B; Supplementary Fig. S1A). TRAF3 expression significantly reduced protein levels of components involved in the alternative NF-κB pathway, including its co-factor and ubiquitinase cIAP1 together with its substrate NIK, and processed NF-κB2/p52 and RELB subunits (Fig. 3B and C; Supplementary Fig. S1A). We observed comparatively smaller decreases in TRAF3-cIAP co-factor TRAF2 or alternate pathway kinase IKKα that mediates NF-κB2-p52/RELB (Supplementary Fig. S1B). Taken together, these results indicate that increasing expression of TRAF3 in HPV⁺ HNSCC cells with lower TRAF3 can suppress mRNA and/or protein levels of components of the alternative NF-κB pathway.

To further characterize how TRAF3 modulates the cellular distribution of alternative RELB/NF-κB2 subunits in HPV⁺ HNSCC, we examined cytoplasmic and nuclear expression of these protein subunits by Western blot. Aberrant nuclear localization of RELB, NF-κB2/p52, as well as p100 was detected in control nuclear as well as cytoplasmic fractions (Fig. 4A), consistent with a prior report demonstrating enhanced nuclear expression of both p100/52 in HPV⁺ SCC and E6-expressing keratinocytes (19). Transfection of TRAF3 increased distribution of NF-κB2 p100/52 proteins in the cytoplasmic fraction while decreasing nuclear p100/p52 and RELB protein levels in independent experiments (Fig. 4A; Supplementary Fig. S2A and S2B). We further tested functional NF-κB activity by utilizing a NF-κB luciferase reporter gene assay in the UMSCC47 cell line. Consistent with the reduced nuclear RELB/p52 protein levels, we observed a significant reduction in functional NF-κB reporter activity in TRAF3 cotransfected UMSCC47 cells (Fig. 4B). Conversely, TRAF3 knockdown using three different siRNAs, significantly reduced TRAF3 mRNA expression in UMSCC47 cells (Fig. 4C). Consistent with reduced TRAF3 mRNA expression in Fig. 4C, we observed reduced TRAF3 and increased RELB protein by Western blot (Supplementary Fig. S2C), as well as enhanced nuclear RELB by immunofluorescence staining (Fig. 4D). TRAF3 knockdown was accompanied by increased NF-κB reporter activity in UMSCC47 cells (Fig. 4E). These results demonstrate that TRAF3 can inversely modulate alternative NF-κB pathway RELB and NF-κB2/p52 expression, cellular localization, and functional activity in HPV⁺ HNSCC cells.

HPV⁺ HNSCCs display rapid growth and spread within the tonsils and regional lymph nodes, where ligands activating the alternative NF-κB pathway are expressed (4, 11). To investigate the upstream receptor-mediated activation of NF-κB in HPV⁺ lines, we first established if HPV⁺ and HPV⁻ HNSCC cell lines express mRNAs for the alternative pathway receptor lymphotoxin-beta receptor (LTβR), and two classic NF-κB pathway TNF receptors (TNFR1A and TNFR1B). We detected elevated mRNA expression of LTβR in HPV⁺ compared with HPV⁻ HNSCC cell lines (Fig. 5A, left, HOK cells as dotted line), while no significant difference was observed for either TNFRs (Fig. 5A, right). Next, we compared the functional effects of recombinant LTβ or TNFα protein on NF-κB reporter activity in HPV⁺ UMSCC47 cells and HPV⁻ UMSCC1 cells, establishing that LTβ as well as TNFα could induce NF-κB reporter activity in both (Fig. 5B; Supplementary Fig. S3A). We found that LTβ treatment attenuated TRAF3 and NF-κB2/p100 precursor protein, while enhancing RELB and processed NF-κB2/p52 protein levels (Fig. 5C; Supplementary Fig. S3B). Conversely, after knockdown of TRAF3, basal and LTβ-induced levels of RELB and NF-κB2/p52 proteins were substantially enhanced (Fig. 5C; Supplementary Fig. S3B), supporting the role of TRAF3 as a negative regulator of the alternative NF-κB pathway in HPV⁺ HNSCC cells.

As we observed above that TRAF3 inhibited alternative NF-κB activation and cIAP1, which have been previously linked to proliferation, and TNFα and chemotherapy resistance (20), we evaluated if TRAF3 overexpression could inhibit tumor cell proliferation, or enhance the inhibitory effects of TNFα, LTβ, or chemotherapy agent cisplatin used for HNSCC. Transient TRAF3 expression resulted in a modest but significant decrease in cell proliferation and showed greater inhibitory effects in combination with TNFα, cisplatin, or both at day 3 (Supplementary Fig. S4A) and day 5 (Fig. 5D). In cell lines selected for TRAF3 and control vector expression, TRAF3 enhanced sensitivity of cells to the combination of TNFα and CDDP, but greater resistance and growth in response to TNFα alone (Supplementary Fig. S4B). While sensitization to TNFα and cisplatin was associated with TRAF3-inhibitory effects on alternative NF-κB components and cIAP1 above, we observed minimal modulation of BCL family members that contribute to survival (Supplementary Fig. S4C). To further examine the functional role of alternative NF-κB pathway, we knocked down IKKα and RELB to evaluate cell proliferation in HPV⁺ UMSCC47 cells. We observed a significant reduction in the cell proliferation after knockdown of either gene alone, with RELB knockdown having a greater effect than IKKα knockdown on cell proliferation (Fig. 5E). As HPV⁺ HNSCCs show increased propensity for malignant spread, we next performed a cell migration assay. We observed a significant delay in cell migration over a 24-hour time course in UMSCC47 cells expressing TRAF3 (Fig. 5F), before antiproliferative effects ∼10% of control are observed in XTT assay by day 3 (Supplementary Fig. S4A). Together, these data demonstrate that TRAF3 inhibits LTβ enhanced alternative pathway NF-κB2/p52 and RELB protein levels, increased sensitivity to TNFα and chemotheraputic agent as well as inhibits migration of HPV⁺ HNSCC cells.

As TRAF3 appeared to reduce proliferation by 5 days following transient transfection, we enriched for UMSCC47 cells expressing TRAF3 or vector plasmid control by G418 selection, and examined the colony-forming capacity and proliferation of HPV⁺ UMSCC47 cells. We observed a significant reduction in colony formation of UMSCC47 cells that expressed increased TRAF3, compared with control (Fig. 6A and B). UMSCC47 cells selected for expression also demonstrated significantly reduced cell proliferation in TRAF3 versus vector control cells (Fig. 6C). Interestingly, we observed that TRAF3 expression declined with passage despite selection, consistent with the antiproliferative activity of a tumor suppressor (Supplementary Fig. S5). As the cell line UPCI-SCC-90 expressed higher endogenous levels of TRAF3 (Fig. 1E), we next examined effects of CRISPR knockout of TRAF3 on colony formation and cell proliferation. CRISPR depleted TRAF3 markedly enhanced colony formation and proliferation of UPCI-SCC-90 (Fig. 6D–F). Together, our data demonstrate that TRAF3 expression suppresses the proliferation of HPV⁺ HNSCC cells.

TRAF3 contributes to expression of type I IFN in response to other DNA virus infections in other tissues (13), but the expression of type-I IFN and relation to lower TRAF3 observed in HPV⁺HNSCC has not been examined. We analyzed RNA-seq data from our HNSCC cell lines and TCGA HNSCC tumors to compare the mRNA expression profiles for type-I interferon IFNA gene in HPV⁺ and HPV⁻ cell lines, as well as tumors. The HPV⁺ HNSCC cell lines displayed lower IFNA1 expression than normal HOK or HPV⁻ HNSCC cell lines (Fig. 7A, left). IFNA1 mRNA in HPV⁺ HNSCC tissues were also expressed at significantly lower levels when compared with HPV⁻ HNSCC tissues from TCGA data (Fig. 7A, right). To test the hypothesis that decreased TRAF3 may contribute to reduction in IFNA and be reversed by TRAF3, we compared IFNA1 expression in the two HPV⁺ HNSCC cell lines with decreased TRAF3 expression, UMSCC47 and UMSCC104, following transient transfection with control and TRAF3 vectors. TRAF3 induced a small but significant increase in the expression of IFNA1 in both cell lines (Fig. 7B). IFNA1 protein was also found to be modestly significantly increased in cell culture supernatants isolated from HPV⁺ UMSCC47 cells stably transfected with TRAF3 compared with control vector (Fig. 7C). UMSCC47 stably expressing increased TRAF3 and IFNA1 as above exhibited reduced proliferation and displayed enhanced sensitivity to exogenous IFNA1 or 2a (Fig. 7D). We explored if TRAF3 modulated mRNA expression of interferon response factors (IRF), which are transcription factors that mediate the transcription of type-I IFN (7, 11). However, we did not see consistent changes in IRF3, 5, or 7 mRNA expression in the HPV⁺ line selected for expression of the TRAF3 gene (Supplementary Fig. S6A). Further, while TRAF3-regulated IRF3 nuclear protein was also decreased in HPV⁺ UMSCC47 cells, the lack of nuclear IRF3 did not appear to be due to lower TRAF3 protein level alone, as expressing TRAF3 vector induced a minimal increase in expression of nuclear IRF3 in replicate experiments (Supplementary Fig. S6B). Moreover, even in HPV⁺ UMSCC47 with lower TRAF3 expression and low nuclear IRF3, nuclear IRF3 and type-I IFN remained inducible with a non-HPV RNA virus mimic and TLR3 ligand poly-I:C, unless TRAF3 was completely knocked out by CRISPR (Supplementary Fig. S7A–S7D). Conversely, UMSCC47 CRISPR/CAS9 TRAF3-9 KO rescued with TRAF3 vector showed no increase in baseline IRF3 or IFN, but did so following stimulation with poly-I:C (20 μg/mL; Supplementary Fig. S8A–S8D). Together, these findings are consistent with either a lack of strong activation of TRAF3-IRF3 signaling by integrated HPV in UMSCC47, or that IRF3 activation and type-I IFN expression may be further disrupted downstream of TRAF3 expression, as found in HPV E6/7-expressing keratinocytes in prior studies (6, 7).

As repression of tumor suppressor proteins TP53 and RB by HPV contributes to the proliferation of HPV⁺ cancers (21, 22), we examined if TRAF3 may modulate expression of TP53 and RB. Compared with the HPV⁻ mtTP53 line UMSCC46, UMSCC47 expressed low basal levels of TP53 and RB protein expression (Fig. 7E). Unexpectedly, TP53 and RB expression was reproducibly enhanced in UMSCC47 cells selected for expression of TRAF3 or after transient transfection (Fig. 7E; Supplementary Fig. S9A). UMSCC104 cells with the lowest TRAF3 expression also showed enhanced TP53 and RB expression after transient transfection with TRAF3 (Supplementary Fig. S9B). In UMSCC47 TRAF3-expressing cells, the increase in TP53 was not attributable to increased mRNA, while RB mRNA was increased ∼2-fold as measured by qRT-PCR (Supplementary Fig. S9C). TP53 siRNA knockdown attenuated the antiproliferative effects of TRAF3 (Supplementary Fig. S7D). The increase in TP53 was associated with enhanced expression of TP53-regulated cyclin-dependent kinase CDKN1A(p21), and proliferation was enhanced by siRNA knockdown of CDKN1A(p21) in UMSCC47 expressing TRAF3 (Fig. 7F and Supplementary Fig. S9E), supporting a contribution of this cell-cycle regulator to the antiproliferative effects of TRAF3 and TP53.

As HPV E6 oncoprotein mediates suppression of TP53 (22), we explored whether TRAF3 may affect HPV E6 levels protein in UMSCC47 cells selected to express TRAF3 or control vector. HPV E6 protein localized predominantly in the nucleus in the control cell line and was partially decreased in the TRAF3-expressing UMSCC47 cell line (Fig. 7G). We observed minimal change in transcription of E6 mRNA (Supplementary Fig. S10A). However, we observed only a minimal increase in E6 protein after transfection with an E6 expression vector that increased mRNA ∼400-fold (Fig. 7H; Supplementary Fig. S10B), indicating that the very low level of E6 detected is not primarily determined by transcriptional regulation (Fig. 7H). Conversely, we also observed only a minimal decrease in E6 protein at 72 hours after transient transfection with TRAF3 (Fig. 7H) and did not observe higher or lower MW bands, or enhancement by proteasome inhibitors, that would support a ubiquitination or proteasome-dependent mechanism (Supplementary Fig. S10C). Together, these observations suggest that the increase in TP53 with TRAF3 within 24 hours after transfection is likely dependent on mechanisms other than modulation of E6 protein expression alone.

TRAF3 copy loss, mutation, and/or decreased expression was significantly enriched in a subset of HPV⁺ HNSCC tumors (10), suggesting that TRAF3 could be a novel suppressor of HPV viral and tumor pathogenesis. Previously, TRAF3 deficiency was implicated in impaired Toll-like receptor 3 responses and increased susceptibility to herpes simplex virus (HSV)–induced encephalitis (23), hematologic malignancies, and Epstein–Barr virus (EBV)–related nasopharyngeal cancers, but not HPV cancers. We observed that oropharyngeal HNSCC linked to HPV exhibit aberrant nuclear activation of the alternative NF-κB pathway (24), but the significance and function and role of deficient TRAF3 in alternative NF-κB activation have not been investigated previously in HPV⁺ HNSCC cell lines and tissues. Here, we demonstrate for the first time that decreased TRAF3 expression is linked to aberrant expression and activation of alternative NF-κB2/RELB transcription factors, contributes to reduced expression of antiviral type-I IFN and tumor suppressor TP53 and RB proteins, which are deregulated in HPV⁺ HNSCC (Supplementary Fig. S11). Supporting this role, ectopic TRAF3 expression in HPV⁺ HNSCC with decreased endogenous TRAF3 inversely modulated these key targets and suppressed several hallmarks of cancer, including cell colony formation, proliferation, and migration, while sensitizing HPV⁺ HNSCC to immune IFNs, TNFα, and chemotherapy agent cisplatin. Complementing our findings, one of our laboratories currently obtained evidence that deletions or mutations in TRAF3 in HPV⁺ HNSCC tumors are associated with increased expression of an NF-κB transcriptional program, and episomal HPV infection (25). Interestingly, TRAF3-deficient HPV⁺ HNSCC also appear to be linked to a better prognosis (25), than HPV⁺ or HPV⁻ tumors with PIK3CA, ΔNp63, FADD, cIAP1/2, and other genomic alterations (i.e., TP53; ref. 10), implicated in modulation of NF-κB or other pathways important in pathogenicity (26–29). The findings in the present study support the addition of TRAF3 as a tumor suppressor in HPV⁺ HNSCC.

The recent discovery of deep deletions on chromosome 14q32.32 containing the TRAF3 coding region and 3 deleterious loss-of-function mutations by TCGA project was previously unrecognized in HPV⁺ HNSCC (10). Here, we further defined an increased frequency of hetero- and homozygous deletions of TRAF3 in HPV⁺ HNSCC, in contrast to HPV⁻ tumors, which often display copy gain. Furthermore, we observed that TRAF3 genetic alterations and expression significantly correlated with expression of alternative NF-κB pathway components in HNSCC tissues. TRAF3 gene copy loss, inactivating mutations, and decreased expression of TRAF3 have been previously observed only in non-HPV cancers. These include multiple myeloma (MM; ref. 16), B-cell lymphomas (30), and EBV–infected human nasopharyngeal carcinomas (NPC; ref. 31). Annunziata and colleagues found that 4.4% of 451 patients with MM had TRAF3 genetic defects, such as silencing, homozygous deletion, or somatic mutation, and 17 MM cases exhibited TRAF3 mRNA expression 13-fold below the median level of the cohort, predominantly in association with these genetic alterations (16). Chung and colleagues detected deletion or missense mutations of TRAF3, TRAF2, and A20 in 3 of 33 (9.1%) of primary NPC tumor specimens (31). Together, these observations in different cancer types support the potential relevance of genetic alterations in HPV⁺ HNSCC and suggest that decreased TRAF3 facilitates the pathogenesis of tumors within the hematopoietic or lymphatic microenvironment.

We previously reported that alternative pathway NF-κB2 and RELB subunits are often aberrantly expressed, and localized to the nucleus in oropharyngeal HNSCC tumors linked to HPV (24), but the possible relationship to genomic alterations, other factors and to the malignant phenotype was unknown. Here, our analysis of TCGA HNSCC data indicated that TRAF3 deletion is associated with increased mRNA expression of molecules involved in the alternative NF-κB pathway, such as NIK, cIAPs, RELB, and NF-κB2. Co-occurrence and mutual exclusivity analyses for key components of the NF-κB pathways revealed a significant association of altered TRAF3 with molecules involved in an alternative NF-κB pathway in HPV⁺, but not in HPV⁻ HNSCC tissues. We further demonstrated that HPV⁺ HNSCC cell lines with decreased endogenous TRAF3 also display aberrant expression and nuclear activation of the alternative NF-κB pathway. Furthermore, this potential of decreased TRAF3 to foster expression and activation of alternative NF-κB in HPV⁺ HNSCC tumors could potentially be enhanced within the local tonsillar and lymphatic microenvironment, where alternative NF-κB signal inducers such as LTβ are abundant (32).

LTβ is a key signaling ligand that binds to surface receptor LTβR and activates the alternative NF-κB pathway (33). We originally detected aberrant activation of the alternative pathway in HPV⁻ HNSCC and demonstrated that activation could be enhanced by LTβ stimulation (34). Here, we observed increased LTβR expression compared with HPV⁻ cell lines and showed that LTβ could further enhance NF-κB activation in HPV⁺ HNSCC cells expressing lower endogenous TRAF3. TRAF3 knockdown further enhanced aberrant and LTβ-stimulated expression of the alternative NF-κB2 subunit p52 and RELB. Interestingly, in a subset of EBV-infected nasopharyngeal HNSCC that harbor TRAF3 alterations, Or and colleagues demonstrated 12p13.3 copy-number gains and increased expression of LTβR (35). Overexpression of LTβR in nasopharyngeal epithelial cells resulted in increased NF-κB activity and cell proliferation (36). Together, these data implicate TRAF3 and/or LTβR genomic alterations, along with HPV E6 and EBV LMP1 viral oncogenes that promote alternative NF-κB pathway activation, in viral pathogenesis of oro- and nasopharyngeal HNSCC. LMP-1 has been shown to bind TRAF3 and enhance NIK-dependent alternative pathway activation (36). HPV E6 is reported to promote alternative NF-κB activation via a mechanism requiring its PDZ domain implicated in binding phosphatases (8). Our data support a role for deficient TRAF3 expression in enhancing NIK, cIAP1, IKKα, NF-κB2-p52, and RELB proteins, but we have found it exceedingly difficult to find conditions to detect HPV16 E6 protein and an association with TRAF3 in HNSCC.

TRAF3 has been reported to have a unique dual role in positively regulating type-I IFNs in response to different DNA viruses in other tissues (11). IFNA1 potentially has antiviral effects through inhibition of viral protein expression, proliferation, as well as immunostimulatory properties, that link innate and adaptive immunity (37). In this study, we observed significantly lower expression of IFNA1 in HPV⁺ HNSCC lines, consistent with the lower expression of IFNA1 detected in HPV⁺ compared with HPV⁻ tumor tissues from TCGA data. Previous studies have shown that HPV infection can also induce ubiquitin carboxyl-terminal hydrolase L1 (UCHL1), which could inhibit TRAF3 K63 ubiquination, important in downstream TBK-IRF3–mediated type-I IFN expression (7). In addition, we have shown a significant increase in both type-I IFN mRNA and protein expression by overexpressing TRAF3 in HPV⁺ UMSCC cell lines with low endogenous levels or TRAF3 rescue in TRAF3-KO cell lines. TRAF3 also sensitized HPV⁺ cells to antiproliferative effects of exogenous IFNA1 that may be produced by cells in the tumor microenvironment, or IFN2A, which is used clinically.

A novel finding of this study supporting the importance of decreased TRAF3 in viral pathogenesis is that overexpression of TRAF3 enhanced the expression of TP53 and RB, which are critical tumor suppressor proteins inactivated by HPV E6 and E7 oncoproteins (21, 22, 38, 39). Proliferation was enhanced by knockdown of TP53 and TP53-regulated CDKN1A(p21) in cells selected to express TRAF3, but not in control cells. However, the exact mechanism by which TRAF3 modulates these tumor suppressors and the relationship of these effects to E6 and E7 is complicated by the difficulty in detecting these proteins. The minor reduction in E6 observed with TRAF3 was not accompanied by appearance of increased or decreased MW bands, and/or modulated by proteasome inhibitors, that would support a ubiquitin or proteasome-dependent mechanism (38). Further, the effects of TRAF3 on decreased E6 were detected at later time points, compared with effects on TP53 and RB expression or NF-κB activation. The different kinetics suggest that TRAF3 modulates E6, TP53, and RB expression by different transcriptional or posttranslational mechanisms.

Of potential therapeutic relevance, TRAF3 re-expression in HPV⁺ HNSCC cells with low endogenous expression sensitized them to inhibition by cisplatin chemotherapy and TNFα, an important cell death ligand induced by radiation and immune therapy (40, 41). Cisplatin is the most active chemotherapeutic agent in HNSCC (42), leading to DNA damage and subsequently promoting cell death (43). TNFα is known to bind to TNF receptors and induce apoptotic cell death in normal cells, but HNSCC are relatively resistant to TNFα (28, 44). NF-κB and cIAPs have been implicated in resistance to TNFα and chemotherapy (20, 45). Our data suggest that TRAF3 expression can inhibit cIAP1 expression and activation of the alternative NF-κB pathway, and sensitize HPV⁺ HNSCC to TNFα and cisplatin-induced cell death. We have recently reported that IAP1 inhibitors can sensitize HPV⁻ HNSCC with gene copy gain and expression of IAP1 to TNF, chemotherapy, and radiotherapy (28, 46, 47), supporting future investigation of these agents in HPV⁺ HNSCCs expressing deficient TRAF3 and/or cIAP1.

Although both HPV⁺ HNSCCs and NPCs develop within the adenotonsillar tissue and exhibit high rates of spread to regional lymph nodes, the mechanisms for this predilection are not well understood. Intriguingly, TRAF3 re-expression inhibited cell migration in vitro, suggesting that deficient TRAF3 could contribute to enhanced migration, underlying the high rates of regional spread of both HPV⁺ HNSCC and NPC in vivo. The alternative of NF-κB pathway could have a broader role in migration in HNSCC, as we found that IKKα and RELB knockdown decreased cell migration and proliferation in HPV⁻ HNSCC (34, 48). Together, the unique genomic alterations in TRAF3 and aberrantly activated alternative NF-κB pathway components such as cIAPs or NIK, could serve as potential markers and targets for novel therapies of HPV⁺ HNSCC.

C. Van Waes reports receiving a commercial research grant from Astex Pharamaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Zhang, C.M. Annunziata, Z. Chen, C. Van Waes

Development of methodology: J. Zhang, X. Yang, P.E. Clavijo, Z. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Zhang, T. Chen, X. Yang, S.S. Späth, C. Silvin, N. Issaeva, Z. Chen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Zhang, T. Chen, X. Yang, H. Cheng, S.S. Späth, P.E. Clavijo, J. Chen, N. Issaeva, W.G. Yarbrough, Z. Chen, C. Van Waes

Writing, review, and/or revision of the manuscript: J. Zhang, T. Chen, S.S. Späth, P.E. Clavijo, J. Chen, W.G. Yarbrough, C.M. Annunziata, Z. Chen, C. Van Waes

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Yang, Z. Chen

Study supervision: Z. Chen, C. Van Waes

We thank Dr. Cheng-Ming Chiang and S.-Y. Wu (UT Southwestern Medical Center) for providing HPV16E6 expression plasmid and their helpful suggestions. We thank Drs. Liu Yang (University of Maryland School of Medicine) and WanJun Chen (NIDCR/NIH) for reading the manuscript and comments.

This work was supported by the Intramural Research Program of The National Institute on Deafness and Other Communication Disorders. J. Zhang, T. Chen, H. Cheng, X. Yang, P. Clavijo, J. Coupar, C. Silvin, Z. Chen, and C. Van Waes are supported by NIDCD intramural projects ZIA-DC-000016, 73, 74. C.M. Annunziata is supported by the NCI.

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.