Prognostic and predictive significance of nuclear HIF1α expression in locally advanced HNSCC patients treated with chemoradiation with or without nimotuzumab

Background Anti-EGFR-based therapies have limited success in HNSCC patients. Predictive biomarkers are greatly needed to identify the patients likely to be benefited from these targeted therapies. Here, we present the prognostic and predictive association of biomarkers in HPV-negative locally advanced (LA) HNSCC patients. Methods Treatment-naive tumour tissue samples of 404 patients, a subset of randomised Phase 3 trial comparing cisplatin radiation (CRT) versus nimotuzumab plus cisplatin radiation (NCRT) were analysed to evaluate the expression of HIF1α, EGFR and pEGFR by immunohistochemistry and EGFR gene copy change by FISH. Progression-free survival (PFS), locoregional control (LRC) and overall survival (OS) were estimated by Kaplan–Meier method. Hazard ratios were estimated by Cox proportional hazard models. Results Baseline characteristics of the patients were balanced between two treatment groups (CRT vs NCRT) and were representative of the trial cohort. The median follow-up was of 39.13 months. Low HIF1α was associated with better PFS [HR (95% CI) = 0.62 (0.42–0.93)], LRC [HR (95% CI) = 0.56 (0.37–0.86)] and OS [HR (95% CI) = 0.63 (0.43–0.93)] in the CRT group. Multivariable analysis revealed HIF1α as an independent negative prognostic biomarker. For patients with high HIF1α, NCRT significantly improved the outcomes [PFS:HR (95% CI) = 0.55 (0.37–0.82), LRC:HR (95% CI) = 0.55 (0.36–0.85) and OS:HR (95% CI) = 0.54 (0.36–0.81)] compared to CRT. While in patients with low HIF1α, no difference in the clinical outcomes was observed between treatments. Interaction test suggested a predictive value of HIF1α for OS (P = 0.008). Conclusions High HIF1α expression is a predictor of poor clinical response to CRT in HPV-negative LA-HNSCC patients. These patients with high HIF1α significantly benefited with the addition of nimotuzumab to CRT. Clinical trial registration Registered with the Clinical Trial Registry of India (Trial registration identifier—CTRI/2014/09/004980).


BACKGROUND
Head and neck squamous cell carcinomas (HNSCCs) are the sixth most common cancers worldwide and comprise a major cancer burden in many regions of the world. 1 The common risk factors associated with the disease are tobacco and/or alcohol abuse and high-risk human papilloma virus (HPV) infection. 2 HNSCC patients are often diagnosed with locoregionally advanced (LA-HNSCC) primary disease with concurrent chemoradiation as the standard treatment of care. Anti-epidermal growth factor receptor (EGFR) therapy is the only targeted therapy approved for the treatment of LA-HNSCC patients. However, addition of anti-EGFR monoclonal antibody (mAb) to the chemoradiation regimen has largely met with limited success in these patients. 3 Nimotuzumab (h-R3) is a humanised IgG1 mAb against EGFR shown to have low toxicity as compared to other anti-EGFR mAbs. 4 6 In order to improve the clinical benefit-to-risk ratio of the given treatment, predictive biomarkers are greatly needed that can help in identifying the patients who are most likely to be benefited from the treatment. The biomarkers predictive of anti-EGFR-based therapy response are well established and are integrated into www.nature.com/bjc clinical care for colorectal (CRC) and non-small-cell lung cancer (NSCLC) patients. 7,8 However, predictive biomarkers of anti-EGFRbased treatment response in HNSCCs are completely lacking, and these treatments are offered irrespective of the molecular heterogeneity across the tumours. Even though EGFR overexpression is the principal mechanism of receptor activation in HNSCCs, at present, the role of EGFR protein expression and gene copy number for predicting the response to anti-EGFRbased treatments remains equivocal as reviewed by Bossi et al. 9 Phosphorylated EGFR dimers (pEGFR) are surrogate markers of EGFR activity; however, reports evaluating their prognostic and predictive value in LA-HNSCCs are limited. In the present study, we have analysed the expression of pEGFRY1068 and pEGFRY1173 that are among the major phosphorylation sites and are involved in the activation of important downstream pathways-PI3K-AKT and RAS-MAPK. 10 Further, hypoxic microenvironment is a common feature of solid tumours including HNSCCs and a major contributor of radiotherapy and chemotherapy resistance. 11,12 Hypoxia-inducible factor 1α (HIF1α) is a transcription factor that mediates adaptive responses to hypoxia by regulating numerous cellular processes such as angiogenesis, oxygen transport, glycolysis and pH control. 13,14 HIF1α overexpression is associated with poor prognosis and resistance to chemoradiation in HNSCCs. 15 Interestingly, several preclinical studies have demonstrated that antitumour activity of EGFR inhibitors is linked to downregulation of HIF1α expression in different cancers, including HNSCCs. [16][17][18][19][20] In addition, HNSCC cell lines have been shown to be more sensitive to cetuximab under hypoxia. 21,22 The results from these preclinical studies warrant clinical evaluation of HIF1α expression for its predictive value. In the present study, we have analysed HIF1α, EGFR and pEGFR protein expression and EGFR gene copy number in HPV-negative LA-HNSCC patients to establish a correlation between these tumour biomarkers and treatment response to cisplatin radiation and nimotuzumab plus cisplatin radiation.

Study design
This study included participants of a previously reported randomised Phase 3 clinical trial conducted at Tata Memorial Center, Mumbai, India (registered with the Clinical Trial Registry of India, trial registration identifier-CTRI/2014/09/004980). 6 Briefly, 536 LA-HNSCC patients were blindly randomised 1:1 to receive radical radiotherapy (66-70 greys) with concurrent weekly cisplatin (30 mg/m 2 ) (CRT arm) or the same schedule of cisplatin radiation with weekly nimotuzumab (200 mg) (NCRT arm). The primary endpoint of the trial was PFS. The present study was approved by the institutional ethics committee of Tata Memorial Center (IEC approval 50 of 2011) and was performed in accordance with the Declaration of Helsinki. This study was an independent biomarker study and not a part of the parental trial; therefore, a separate informed consent was obtained from all the participants.
Sample collection and human papilloma virus (HPV) screening Treatment-naive formalin-fixed paraffin-embedded (FFPE) tumour biopsy tissues and saliva samples were collected prospectively and were subjected to HPV screening. Detailed methodology for HPV screening is previously reported 23 and is briefly described in Supplementary Methods. Out of 536 patients, biopsy tissue with adequate tumour content was available for 432 patients (80%), of which saliva samples were available for 349 patients. All 432 tumour samples were analysed for p16 protein expression by immunohistochemistry (IHC). Both saliva and tumour tissue were screened in 221 cases for HPV-DNA by PCR; for 128 cases, only saliva sample and for 54 cases only tumour tissue was analysed for HPV-DNA by PCR. A HPV-positive sample is characterised by p16-positive IHC staining and/or the presence of HPV-DNA in either tumour or saliva, along with a subsequent positive HPV RNA in situ hybridisation (RNA-ISH) test. 24 HPV-negative tumour samples were subjected to pre-specified biomarker analysis, which was performed blinded to treatment allocation and patient's outcomes.

Biomarker analysis
Fluorescence in situ hybridisation (FISH). EGFR gene copy number was assessed by FISH using EGFR/CEP7 FISH probe (Abbott Vysis, CA, USA). A detailed protocol is provided in Supplementary Methods. FISH signals were counted in at least 100 tumour cells under ×63 magnification. EGFR gene copy status was classified into five categories, depending on the percentage of tumour cells showing different copies of EGFR gene locus and centromere as disomy (≤2 copies in >90% of cells), trisomy (3 copies in ≥10% of cells or ≥4 copies in <10% of cells), low polysomy (≥4 copies in 10-40% of cells), high polysomy (≥4 copies in ≥40% of cells) and gene amplification (ratio of the EGFR gene to chromosome 7 of ≥2 or ≥15 copies of EGFR per cell in ≥10% of cells). On the basis of EGFR gene copy status, patients were grouped as FISH-negative (disomy, trisomy and low polysomy) or FISH-positive (high polysomy and/or EGFR gene amplification). 25 Immunohistochemistry. Protein expression of HIF1α, EGFR, pEG-FRY1068 and pEGFRY1173 was analysed by IHC using VECTASTA-TIN® Elite® ABC kit (Vector Laboratories, CA, USA). A detailed protocol is provided in Supplementary Methods, and details of primary antibodies and positive controls are tabulated in Supplementary Table 1. IHC staining was evaluated semiquantitatively by the pathologists who were blinded to treatment and patient's outcomes. Expression of HIF1α (nuclear), EGFR (membrane and cytoplasmic), pEGFRY1068 (membrane) and pEGFRY1173 (membrane) was assessed by deriving the H-score (scale: 1-300) using the formula H-score = Σ Pi (i + 1), where Pi is the percentage (0-100%) of stained tumour cells at each intensity and i is the intensity: i = 1 (weak), 2 (moderate) and 3 (strong). 9,15 Biomarkers were analysed as dichotomised variables. Due to unavailability of consensus regarding H-score cut point to be used for dichotomisation of these biomarkers, the respective median H-score values were used for HIF1α (H-score = 90) and EGFR (membrane, H-score = 100; cytoplasm, H-score = 140). 26,27 For pEGFRY1068 and pEGFRY1173, patients with H-score = 0 were categorised as negative and patients with H-score > 0 were positive. IHC staining of HIF1α was independently evaluated by a second pathologist.

Statistical analysis
Categorical data are presented as frequency and percentage; continuous data are expressed by median and range or interquartile range (IQR). Bivariate association between different biomarkers and between biomarkers and clinicopathological parameters was analysed by Pearson's χ 2 test. PFS, LRC and overall survival (OS) were as defined earlier 6 and were estimated using Kaplan-Meier method and compared by log-rank tests. Cox proportional hazard models were used to derive hazard ratios (HR) and 95% confidence intervals (CI). The definition used for prognostic and predictive biomarkers was as proposed by Clark et al. 28 For assessing the prognostic significance of each biomarker, only patients from the CRT arm were included in the analysis. In addition, however, we have also studied the association of biomarkers with clinical outcomes in the NCRT arm. Univariate Cox models were applied to select the most promising biomarkers (threshold P < 0.20). A multivariate Cox model using backward likelihood ratio (LR) method was then applied to adjust for potential confounders (clinical characteristics associated with PFS, LRC or OS at P < 0.20). Reported HRs (95% CI) are for low or negative biomarker expression relative to high or positive biomarker expression. For assessing the predictive significance of each biomarker, all patients with biomarker data, irrespective of the treatment group, were included in the analyses. Cox models were fit, which included treatments (NCRT vs CRT), biomarker status (low/negative vs high/positive) and the interaction between treatment effect and biomarker status. 28,29 Internal validation of prognostic and predictive models was achieved by bootstrap-resampling method (1000 samples). Concordance indexes (c indexes) were also calculated.
Agreement between IHC scoring of HIF1α by two pathologists (SR and NM) was assessed using the Bland-Altman plot, and the concordance correlation coefficient was derived. 30,31 Scoring of SR was used for analysis after obtaining consensus in the cases with H-score difference of >100, which were jointly reviewed by both the pathologists. Statistical analyses were performed using IBM SPSS software version 21 (SPSS Inc., IL, USA); STATA version 14 (StataCorp, TX, USA) was used for the bootstrap procedure and for generating forest plots; all reported P values are two-sided and P value of 0.05 or less was considered statistically significant. The study followed the REMARK guidelines for reporting. 32,33

RESULTS
Patients and HPV screening Out of 432 cases screened for HPV, 25 (5.8%) cases showed the presence of transcriptionally active high-risk HPV (Supplementary Fig. 1) and the results were inconclusive in 3 (0.7%) cases. We excluded these 28 cases and carried out biomarker analysis in the remaining 404 HPV-negative cases out of which 206 received CRT and 198 received NCRT treatment. The workflow of the study is outlined in Fig. 1. Baseline characteristics of the patients were balanced between the two treatment groups, and were representative of the total trial population ( Table 1). Kaplan-Meier plots showing the treatment outcomes in the biomarker subgroup (n = 404) are provided in Supplementary Fig. 2. A total of 241 patients (45%) had died at the time of analysis, and the median follow-up of patients still alive was 39.13 months; 4-year survival rates are reported.
Expression of biomarkers Expression of total EGFR, pEGFRY1068, pEGFRY1173 and HIF1α was assessed by IHC staining, and EGFR gene copy status was evaluated by FISH ( Supplementary Figs. 3 and 4). The frequency distribution of protein biomarker expression ( Supplementary Fig. 5) and EGFR-FISH status (Supplementary Table 2) was comparable between two treatment groups. Overall, the expression of pEGFRY1068 and pEGFRY1173 showed a skewed distribution as >80% and >70% of the cases respectively were negative (H-score = 0) in both treatment groups. We did not find any strong correlation among the studied biomarkers (Supplementary Table 3). However, moderate correlation was observed between membrane and cytoplasmic EGFR (R = 0.40), as well as between pEGFRY1068 and pEGFRY1173 (R = 0.57). Both membrane and cytoplasmic EGFR expression showed weak correlation with pEGFR dimers. A weak correlation was also observed between HIF1α and membrane EGFR expression (R = 0.15). No statistically significant association was observed between biomarkers and patient's clinical characteristics, except for the cytoplasmic EGFR that was associated with disease stage (P = 0.027, Supplementary Table 4).
Prognostic significance Univariate Cox regression analysis performed at different HIF1α H-score cut points indicated that low HIF1α expression was numerically associated with better PFS, LRC and OS in the CRT    Table 8).  Supplementary Fig. 6. In addition, analysis carried out at different cut points revealed that overall high HIF1α expression was associated with better outcomes in NCRT as compared to CRT, with minimum-interaction P value observed at the median cut point (Supplementary Table 9). Immunostaining of HIF1α was independently evaluated by a second pathologist (NM); a good agreement was observed between scoring of two pathologists (S.   Similar benefits in PFS, LRC or OS were not observed in FISHpositive patients; the interaction between treatment and EGFR-FISH status was found to be nonsignificant (Figs. 3a-c and 5j-l). Taken together, these results suggest that the treatment effect of NCRT is independent of EGFR protein or gene copy status in these patients. Furthermore, we carried out combined predictive analysis of HIF1α and membrane EGFR (Supplementary Table 11 In the remaining two subgroups that include a subgroup with low expression of both biomarkers and another with low HIF1α along with high EGFR, we did not find any significant difference in PFS, LRC or OS between the treatment groups. Overall, combined analysis of HIF1α-EGFR did not show any additional predictive value over HIF1α alone.

DISCUSSION
Prognostic biomarkers are extensively studied in HNSCCs, but they have limited utility in patients' treatment decisions. While the identification of predictive biomarkers is a pressing need to enable selection of patients for a specific treatment. In the present study, we have evaluated prognostic and predictive significance of HIF1α, EGFR, pEGFR protein expression and EGFR gene copy number in HPV-negative LA-HNSCC patients treated either with CRT or NCRT in a Phase 3-randomised study. Here we have shown high HIF1α as an independent negative prognostic factor for PFS, LRC and OS in patients treated with CRT. Interestingly, addition of nimotuzumab to CRT significantly improved the clinical outcomes in patients expressing high HIF1α, with 45% less risk of progression, 45% less risk of locoregional failure and 46% less risk of death compared to patients receiving only CRT (Fig. 3a-c). We observed statistically significant qualitative interaction between treatment and HIF1α status for OS, which was validated by bootstrap-resampling method. We did not observe any prognostic and/or predictive association of EGFR, pEGFR dimers or EGFR gene copy number. Ours is the first study demonstrating both prognostic and predictive roles of HIF1α in HPV-negative LA-HNSCC patients.
HNSCCs are characterised by EGFR overexpression that is the principal mechanism of receptor activation; however, protein expression or gene copy number of EGFR have not emerged as a strong predicting biomarker for anti-EGFR-based treatment response. 9 In this study, we found high EGFR expression to be associated with improved outcomes with NCRT versus CRT; however, the treatment interaction test was nonsignificant. Lack of correlation between EGFR-based biomarkers and sensitivity of EGFR inhibitors can be due to complex biology of the EGFR signalling pathways in which different intrinsic and extrinsic or acquired resistance mechanisms can alter EGFR downstream signalling. Potential mechanisms proposed for anti-EGFR therapy resistance are overexpression of ligands, activation of alternative pathways like ERBB2 and MET and/or alterations in downstream pathways due to mutations in PI3KCA, PTEN, RAS and CCND1 gene amplification. Resistance mechanisms of anti-EGFR treatment are well established in CRC and NSCLC, but are poorly understood and are not well established in HNSCCs. 34,35 Interestingly, several preclinical studies on different cancer cell lines have repeatedly demonstrated that the response of tumour cells to EGFR inhibitors is linked to the downregulation of HIF1α. [16][17][18][19]36,37 In vivo studies have further shown that this downregulation of HIF1α upon treatment with EGFR inhibitors decreases the levels of its downstream target vascular endothelial growth factor (VEGF), a strong pre-angiogenic marker, which in turn causes vascular normalisation and improved blood flow leading to enhanced chemoradiation efficacy. 17,19 Nevertheless, the predictive impact of HIF1α or VEGF in response to anti-EGFRbased treatments in HNSCC patients has not been studied. A retrospective study by Ou et al. has reported an independent prognostic role of combined expression of low CD34 and high CA9 in predicting poor LCR; however, no predictive effect of these hypoxia-based biomarkers was observed in HPV-negative LA-HNSCC patients. This study was, however, carried out in a small number of patients with an unbalanced distribution of patients between the two treatment groups. 38 Ours is the first study demonstrating the role of high nuclear HIF1α expression in predicting poor response to cisplatin radiation and significant better treatment response in high HIF1α-expressing patients upon addition of nimotuzumab to cisplatin radiation. In addition, a study by Boeckx et al. showed increased sensitivity of HNSCC cells to cetuximab under hypoxia. 21 Similar observations were also reported by Wiechec et al., they further showed that the sensitivity to cetuximab was efficiently reversed by knockdown of HIF1α in HNSCC cells. 22 However, the underlying mechanism by which hypoxia or HIF1α mediates sensitisation towards anti-EGFR mAbs is not yet clearly understood. Our combined analysis of EGFR and HIF1α revealed that improved treatment response to NCRT was independent of EGFR expression status. In this study, we have used RNA-ISH as a confirmatory test for detecting transcriptionally active HPV, unlike the majority of the biomarker studies in which HPV detection is solely done by p16 IHC that is a surrogate marker and not specific for detecting biologically active HPV. 24 HNSCC tumours with HPV-positive status are genetically and biologically distinct from HPV-negative tumours, [39][40][41][42] and are associated with better outcomes, irrespective of the treatment modalities. [43][44][45] To maintain the homogeneity of our study group, we excluded these HPV-positive cases from the current analysis. Also, due to low HPV prevalence in our cohort, we could not perform an independent prognostic and predictive biomarker analysis in the HPV-positive subgroup. There are however few limitations of this study, which need to be considered. IHC staining was assessed semi-quantitatively; evaluation of membrane-staining intensity and quantification is inherently subjective. In addition, ours is a single-centre study, and therefore the results need multicentric external validation.
Since hypoxia is a dynamic feature of tumour microenvironment, assessing biomarker expression in biopsy specimens might not be representative of the whole tumour. In addition, integrating functional imaging and serum-based biomarker analysis can offer complementary information on development of robust predictive biomarkers. However, very few reports have studied correlations between tissues or serum-based biomarkers and information obtained from functional imaging. Recently, Nicolay et al. in a prospective study have shown the association of tumour hypoxia markers-HIF1α and CA9-studied by IHC in pre-treatment biopsies with the hypoxia dynamics during chemoradiation assessed by 18F-FMISO PET/CT imaging in LA-HNSCC patients. 46 In addition to hypoxia and angiogenic markers, other frequently altered downstream molecules of EGFR signalling, including the PI3K-AKT-mTOR pathway, need to be evaluated in combination for their predictive potential in HNSCCs. 47,48 In conclusion, our results suggest that high nuclear HIF1α expression is associated with poor clinical outcomes in CRTtreated patients. Addition of nimotuzumab to CRT significantly improves the outcomes in high HIF1α-expressing patients. In addition to nimotuzumab, anti-angiogenic drugs can be explored for high HIF1α-expressing patients. 49 These targeted therapies are frequently associated with different levels of toxicity and often expensive; therefore, it is required to identify the patients upfront who are most likely to be benefited from these treatments.