- Email: [email protected]
Diagnosis of HPV driven oropharyngeal cancers: Comparing p16 based algorithms with the RNAscope HPV-test
Oral Oncology 62 (2016) 101–108
Contents lists available at ScienceDirect
Oral Oncology journal homepage: www.elsevier.com/locate/oraloncology
Diagnosis of HPV driven oropharyngeal cancers: Comparing p16 based algorithms with the RNAscope HPV-test Haïtham Mirghani a,⇑, Odile Casiraghi b, Joanne Guerlain a, Furrat Amen c, Ming-Xiao He d, Xiao-Jun Ma d, Yuling Luo d, Céline Mourareau e, Françoise Drusch f, Aïcha Ben Lakdhar b, Antoine Melkane a, Lacau St Guily g, Cécile Badoual h, Jean Yves Scoazec b,f,i, Isabelle Borget j, Anne Aupérin j, Veronique Dalstein e, Philippe Vielh b,f,i a
Department of Otolaryngology – Head and Neck Surgery, Gustave Roussy Cancer Campus, 114 rue Edouard Vaillant, Villejuif, France Department of Biopathology, Gustave Roussy Cancer Campus, 114 rue Edouard Vaillant, Villejuif, France c Department of Otolaryngology, Peterborough City Hospital and Addenbrooke’s Hospital, Cambridge, UK d Advanced Cell Diagnostics, 3960 Point Eden Way, Hayward, CA 94545, USA e INSERM UMR-S 903, SFR CAP-Santé FED 4231, Université de Reims Champagne-Ardenne, F-51100 Reims, France f Biobank, Gustave Roussy Cancer Campus, 114 rue Edouard Vaillant, Villejuif, France g Department of Otolaryngology-Head and Neck Surgery, Hôpital Tenon, Assistance Publique-Hôpitaux de Paris, France h Department of Pathology, Hôpital Européen Georges Pompidou, Assistance Publique-Hôpitaux de Paris, France i Laboratory of Translational Research, Gustave Roussy Cancer Campus, 114 rue Edouard Vaillant, Villejuif, France j Department of Biostatistics and Epidemiology, Gustave Roussy Cancer Campus and University Paris-Sud, Villejuif, France b
a r t i c l e
i n f o
Article history: Received 23 June 2016 Received in revised form 28 September 2016 Accepted 15 October 2016 Available online 21 October 2016 Keywords: Oropharyngeal Oropharynx Head and neck cancer Human papillomavirus 16 E6/E7 mRNA/ transcripts RNAscope HPV-testÒ P16 immunostaining In situ hybridization
a b s t r a c t Background: Accurate identification of HPV-driven oropharyngeal cancer (OPC) is a major issue and none of the current diagnostic approaches is ideal. An in situ hybridization (ISH) assay that detects high-risk HPV E6/E7 mRNA, called the RNAscope HPV-test, has been recently developed. Studies have suggested that this assay may become a standard to define HPV-status. Methods: To further assess this test, we compared its performance against the strategies that are used in routine clinical practice: p16 immunohistochemistry (IHC) as a single test and algorithms combining p16-IHC with HPV-DNA identification by PCR (algorithm-1) or ISH (algorithm-2). Results: 105 OPC specimens were analyzed. The prevalence of HPV-positive samples varied considerably: 67% for p16-IHC, 54% for algorithm-1, 61% for algorithm-2 and 59% for the RNAscope HPV-test. Discrepancies between the RNAscope HPV-test and p16-IHC, algorithm-1 and 2 were noted in respectively 13.3%, 13.1%, and 8.6%. The 4 diagnostic strategies were able to identify 2 groups with different prognosis according to HPVstatus, as expected. However, the greater survival differential was observed with the RNAscope HPVtest [HR: 0.19, 95% confidence interval (CI), 0.07–0.51, p = 0.001] closely followed by algorithm-1 (HR: 0.23, 95% CI, 0.08–0.66, p = 0.006) and algorithm-2 (HR: 0.26, 95% CI, 0.1–0.65, p = 0.004). In contrast, a weaker association was found when p16-IHC was used as a single test (HR: 0.33, 95% CI, 0.13–0.81, p = 0.02). Conclusions: Our findings suggest that the RNAscope HPV-test and p16-based algorithms perform better that p16 alone to identify OPC that are truly driven by HPV-infection. The RNAscope HPV-test has the advantage of being a single test. Ó 2016 Elsevier Ltd. All rights reserved.
Introduction Accurate identification of HPV-driven head and neck cancer is of paramount importance as this may guide treatment and follow up strategies in the near future [1]. Currently, the gold standard assay ⇑ Corresponding author. E-mail address: [email protected] (H. Mirghani). http://dx.doi.org/10.1016/j.oraloncology.2016.10.009 1368-8375/Ó 2016 Elsevier Ltd. All rights reserved.
to detect these cancers is based on the identification of E6/E7 viral oncogenes mRNA by reverse-transcriptase-PCR (RT-PCR) in fresh frozen samples [2,3]. However, this technique is not appropriate for routine screening as it is technically demanding and fresh frozen tissue is typically unavailable [4]. Therefore, several alternative approaches are used in clinical practice to define HPV status. Assessment of p16-protein expression by immunohistochemistry is one of the most commonly used methods and many ongoing
102
H. Mirghani et al. / Oral Oncology 62 (2016) 101–108
clinical trials rely on this marker [1,4]. However, it is now widely acknowledged that its use as a standalone test is associated with a significant proportion of false positive and false negative results [4,5]. Indeed, several mechanisms that are not linked to HPV infection can induce p16 over expression and conversely, cancers that are clearly caused by high-risk HPV (HR-HPV) can lack p16 overexpression, although this second situation is less common [6,7]. To overcome this issue, some have advocated the use of diagnostic algorithms that combine several tests (Fig. 1) [8–10]. However, the inclusion of multiple steps to achieve an accurate and reliable HPV status is technically cumbersome, may produce discordant results, and inevitably increases costs and workload. In this context, new assays that are able simultaneously to identify HPV-infection and to demonstrate its implication in oncogenesis are highly desirable. Some are already under development. The RNAscope HPV-test (Advanced Cell Diagnostics, Hayward, CA, USA) is a recent ISH technique that detects E6/E7 mRNA from 18 HR-HPVs. It uses formalin fixed paraffin embedded (FFPE) tissue sections and the interpretation is visual with light microscopy, the transcripts are directly visualized within the tumor. These features make the RNAscope HPV-test a potentially ideal tool for routine laboratory testing. In a recent pilot study, we compared this test against the gold standard method (identification of HR-HPV E6/ E7 mRNA by RT-PCR in fresh frozen samples) and we found that it performed very well (sensitivity: 93%, specificity: 94%, positive predictive value: 96% and negative predictive value: 88%) [11], thus confirming the outcomes of two previous studies [12,13]. To further assess the RNAscope HPV-test we sought to compare its performance against p16-IHC as a single test and p16-based algorithms, which represent the most accurate means to identify HPVdriven oropharyngeal cancer in routine clinical practice. Methods Patients and tumor samples Tumor samples from 105 patients with oropharyngeal squamous cell carcinoma (OPC) were retrieved from the Gustave
Roussy Cancer Center tissue bank. Fifty of these patients had been enrolled into our previous study. All samples were collected prior to treatment. Clinical notes and pathological reports of each patient were retrospectively analyzed by the investigators. Our Institutional Review Board approved the study protocol and all patients provided written informed consent. HPV testing assays 4 different tests were performed: p16 immunohistochemistry (IHC), identification of HR-HPV DNA by PCR and in situ hybridization (ISH), and detection of HR-HPV transcripts by the RNAscope HPV-testÒ. The combination of p16-IHC and ISH to identify HPVDNA was designated as algorithm-1, whereas the combination of p16-IHC and PCR to detect HPV-DNA was designated as algorithm-2. Samples were considered to be HPV-driven with algorithm-1 if p16-IHC was positive and HPV-DNA was detected by ISH. For algorithm-2, samples were considered HPV-driven if p16-IHC was positive and HPV-DNA detected by PCR. Preparation of sections from FFPE tumor tissue blocks Paraffin sections were prepared according to the sandwich method. That is to say the first and last sections were stained by hematoxylin and eosin to check for tumor presence. The remaining 4 lm sections were used to perform p16-IHC, HPV-DNA ISH and the RNAscopeÒ HPV test. p16-IHC p16-IHC was performed on a Ventana Benchmark Ultra (Ventana Medical Systems Inc., Tucson, AZ) using the CINtec p16 Histology Kit (Ref. 9511, Roche mtm laboratories AG, Heidelberg, Germany), in accord with both manufacturers’ recommendations. The previously prepared 4-lm whole tumor FFPE tissues paraffin sections were deparaffinized and subjected to antigen retrieval using CC1 buffer for 30 min. Subsequently they were consecutively incubated in the prediluted CINtec p16 primary antibody (clone
Fig. 1. Stepwise diagnostic algorithm. p16-IHC is used as a first line assay. p16-negative samples are considered as HPV-unrelated. Conversely p16-positive samples are potentially HPV-driven, and some may consider these cases as HPV-related. To confirm this, identification of HR-HPV is necessary. To this end, 2 strategies can be used. Smeets et al. have proposed to identify HR-HPV DNA by PCR whereas the approach developed by Singhi et al. is based on ISH. In their original work, Singhi et al. recommended using probes that are specific to HPV16 and if these probes are negative to use secondarily consensus probes that detect an extended panel of HR-HPV types. As genotypespecific probes are not available for diagnostic use in Europe owing to licensing restrictions, some have proposed directly using the high-risk HPV cocktail probes that are recognized IVDs with CE marking. For these authors, samples that are p16-positive but lacking HPV-DNA by ISH require further confirmation by PCR given that the sensitivity of these ISH assays is not optimal. However, the complexity of these strategies raises questions about their relevance to routine practice. TSP: type specific probes, CP: consensus probes.
H. Mirghani et al. / Oral Oncology 62 (2016) 101–108
E6H4) for 20 min at room temperature, and revealed with Ultra View Universal DAB Detection Kit. Slides were finally counterstained with Hematoxylin II for 8 min (Ventana) and Bluing reagent for 4 min and then washed. A tonsil squamous cell carcinoma with high p16 expression was used as a positive control. The primary antibody was omitted from negative controls. p16 IHC was scored as positive if there was strong, homogeneous and diffuse nuclear and cytoplasmic staining present in greater than 70% of the malignant cells. All other staining patterns were scored as negative. Detection of high-risk HPV DNA by ISH DNA ISH was performed on representative 4-lm whole tumor FFPE sections from each case using the ISH I View Blue Plus Detection Kit (Ventana Medical System, Inc., Tucson, AZ) according to the manufacturer’s instructions. The assay utilized the Ventana HPV III Family 16, Probe B, a cocktail recognizing the high-risk (HR) HPV types 16, 18, 31, 33, 35, 45, 51, 52, 56, 58, 59, 68, and 70. Ventana Red Counterstain II (Ventana Medical System, Inc., Tucson, AZ) was used. Controls in each run included a known HPV 16-positive OPSCC case (positive control) and a section of normal tonsil (negative control). Positive staining was identified as blue nuclear dots. Any definitive nuclear staining in the tumor cells was considered positive. Cases were classified in a binary manner as either positive or negative. Detection of high-risk HPV E6/E7 mRNA by RNA ISH Detection of HR-HPV E6/E7 mRNA was performed using the RNAscope 2.0 BROWN assay kit and the HPV-HR18 probe cocktail (Advanced Cell Diagnostics Inc.) in accordance with the manufacturer’s instructions. Briefly, 4 lm sections were deparaffinized and pretreated with heat and protease before hybridization with target-specific probes for the E6 and E7 genes of 18 HR-HPV genotypes (HPV18, 26, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73 and 82). Ubiquitin C (UBC, a constitutively expressed endogenous gene) and the bacterial gene, dapB, were used as positive and negative controls, respectively. The UBC test was used to assess the presence of hybridizable RNA in order to confirm adequate RNA quality and was defined as adequate if there were at least 5 punctate signal dots in the majority of tumor cells in the section. This is especially important when the HPV probe signal is negative to avoid a false-negative result. Cases with weak UBC staining were retested with the RNAscope 2.5 BROWN Assay Kit (ACD). The dapB test was used to assess nonspecific staining, only those cases that were negative or weakly stained were considered for HPV scoring. A positive HPV test result was defined as punctate staining that localized to the cytoplasm and/or nucleus of any of the malignant cells, and where staining was present in the negative control, it was at least three times as strong as the dapB staining.
103
DNA extraction and HPV genotyping 2–30 10-lm serial sections (according to the size of the specimen) were prepared from FFPE tissue samples in which tumor presence was checked. DNA was extracted by the MaxwellÒ 16 FFPE Tissue LEV DNA Kit (Promega) according to the manufacturer’s instructions. DNA was eluted win 50 lL of Elution Buffer. HPV genotyping was then performed using INNO-LiPA HPV Genotyping Extra II kit (Fujirebio) according to the manufacturer’s instructions. This assay is based on a PCR amplification of the L1 gene by the SPF10 Plus primer system. Positive controls (DNA from SiHa and HeLa cell lines) and no template controls were included in each set of reactions. All hybridization steps up to color development were then performed using an Auto-LiPA system (Fujirebio). Hybridization patterns were analyzed using the INNO-LiPA HPV genotyping Extra II Reading Card. The Extra II version of the assay allows the individual detection of 32 HPV genotypes: 13 high-risk (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68), 6 probable/possible high-risk (26, 53, 66, 70, 73, 82) and 13 low-risk or not classified (6, 11, 40, 42, 43, 44, 54, 61, 62, 67, 81, 83, 89).
Statistical analysis Overall survival was defined as the time from the date of registration to the date of death or to the date of censorship (ie, the last date of follow-up). Disease specific survival was defined as the time from the date of registration to the date of death from cancer or to the date of censorship. The distributions of overall survival and disease specific survival were estimated with the KaplanMeier method. A log-rank test was performed to test the difference in survival between groups in univariate analysis. In multivariable analyses, a Cox proportional hazards model was used to adjust for covariates of statistical significance in univariate analysis. The Wald test was used to estimate the 95% confidence intervals (CIs) of hazard ratios. The 4 diagnostic strategies (p16 IHC alone, the RNAscope HPV testÒ and the 2 algorithms) were compared using the Akaike information criterion. All statistical tests were two-sided, and a P value of 0.05 or less was considered statistically significant.
Results Population description 73 men and 32 women were included in our study. Location, TNM staging, age, and tobacco consumption are summarized in Table 1.
Global analysis of the tests Review of the miscroscopic slides The RNAscopeÒ HPV-test slides were scored independently by 2 authors (MX. H and O. C) and the p16 IHC and HR-HPV DNA ISH tests were also assessed independently by 2 authors (O. C and A. BL). Scoring systems were based on a binary classification (positive vs. negative). All the people involved in microscopic slides review were blinded from the results of all the different assays that were tested (including HPV genotyping by PCR, described in the next paragraphs) and to the scoring of the other reviewers. The results were collated, and discordant scores were resolved at a meeting between the pathologists to establish a consensus interpretation.
Data for the RNAscopeÒ HPV-test was available for the whole cohort whereas data for p16, HPV-DNA ISH and HPV DNA genotyping were available for 104/105, 94/105 and 104/105 patients respectively. The scoring of the RNAscope HPV-test and p16 slides were completely concordant between the pathologists. Discordant cases were only seen with the HR-HPV DNA-ISH assay and were resolved at a meeting to establish a consensus interpretation. The outcomes of all the tests that were carried out including the 2 p16-based algorithms are summarized in Table 2 and the complete dataset is presented in the supplementary data section (supplementary data, Table A).
104
H. Mirghani et al. / Oral Oncology 62 (2016) 101–108
between p16-IHC and the RNAscope HPV-test are presented in the supplementary data section (Table B).
Table 1 Population description. Sex
M: 73 (69.5%) F: 32 (30.5%)
Age
Min: 27 Max:83 Mean: 56 Median: 58
Outcomes comparison between algorithm-1 and the RNAscope HPVtest
Stage
T1: 10 (9.5%) T2: 42 (40%) T3: 30 (28.6%) T4: 23 (21.9%) N0: 28 (26.6%) N1: 9 (8.6%) N2a: 7 (6.6%) N2b: 34 (32.4%) N2c: 15 (14.2%) N3: 9 (8.7%) Nx: 3 (2.9%)
Tobacco
Yes: 55 (52.4%) No: 50 (47.6%)
Location
Tonsil: 62 (59%) Tongue base: 37 (35.3%) Other oropharyngeal subsites: 6 (5.7%)
Table 2 Results summary.
RNAscope HPV test p16 IHCb HPV-DNA ISH HPV-DNA PCR Algorithm 1c Algorithm 2d
Number of casesa
Positive
Negative
105/105 (100%) 104/105 (99%) 94/105 (89.5%) 104/105 (99%) 99/105 (94.3%) 103/105 (98%)
62/105 (59%) 70/104 (67%) 69/94 (73%) 80/104 (77%) 54/99 (54.5%) 63/103 (61.2%)
43/105 (41%) 34/104 (32%) 25/94 (26.6%) 24/104 (23%) 45/99 (45.5%) 40,103 (38.2%)
a The RNAscope HPV test was performed on the whole cohort (n = 105) whereas p16 IHC, HPV-DNA ISH and HPV-DNA PCR were performed in 104/105, 94/105 and 104/105 patients respectively. b all the tumoral samples that were considered as p16-positive (>70% nuclear and cytoplasmic staining) had a strong staining intensity. c Algorithm 1: p16 followed by HPV DNA detection by ISH. d Algorithm 2: p16 followed by HPV DNA detection by PCR.
HPV-genotypes distribution Distribution of HPV genotypes was as follows: 69 (87%) HPV16, 4 (5%) HPV18, 4 (5%) HPV33, 1 (1%) HPV35, 1 (1%) HPV16/33 and 1 (1%) HPV16/35 (supplementary data, Table A).
86/99 (87%) samples were concordant between algorithm-1 and the RNAscope HPV-test. Among the 13 discordant cases, 9 were negative with algorithm-1 and positive with the RNAscope HPVtest. Among these 9 samples, 6 were p16-positive (no. 4, 6, 46, 47, 69 and 97) and were classified as non HPV-related by algorithm-1 because HPV-DNA was not detected by the ISHassay. Interestingly HPV-DNA was identified by PCR in 5 out of these 6 cases (no. 6, 46, 47, 69 and 97) that were subsequently considered as HPV-driven by algorithm-2. Finally, 4 cases were positive with algorithm-1 and negative with the RNAscope HPV-test. Two of these cases were also positive with algorithm-2 (no. 54, 91). Discrepancies between algorithm-1 and the RNAscope HPV-test are presented in the supplementary data section (Table C). Outcomes comparison between algorithm-2 and the RNAscope HPVtest 94/103 (91%) samples were concordant between algorithm-2 and the RNAscope HPV-test. Among the 9 discordant cases, 4 were negative with algorithm-2 and positive with the RNAscope HPVtest. All these cases were also negative with algorithm-1 (no. 4, 29, 49, 76). Conversely, 5 samples were positive with algorithm-2 and negative with the RNAscope HPV-test. For 3 of these cases (no. 14, 34, 50) outcomes of algorithm-1 were negatives like those obtained with the RNAscope HPV-test. Discrepancies between algorithm-2 and the RNAscope HPV-test are presented in the supplementary data section (Table D). Outcomes comparison between the 2 diagnostic algorithms 89 (90%) cases were concordant between the 2 algorithms. Among the 10 discordant cases, algorithm-1 was negative and algorithm-2 was positive in 9 cases. 6 out of the 9 cases were positive with the RNAscope HPV-test. Conversely, algorithm-1 was positive and algorithm-2 negative in 1 case. This case was negative with the RNAscope HPV-test. Discrepancies between the 2 algorithms are presented in the supplementary data section (Table E). Oncological outcomes
Outcomes comparison between p16-IHC and the RNAscope HPV-test 90 (86.5%) cases were concordant between p16-IHC and the RNAscope HPV-test. Among the 14 discordant cases 3 were p16negative but positive with the RNAscope HPV-test. For 2 out of these 3 samples, the results obtained with the other assays were consistent with the RNAscope HPV-test. Indeed, HPV-DNA was identified by both PCR and ISH in 2 cases (no. 29 and 76). Conversely, 11 cases were p16-positive but negative with the RNAscope HPV-test. Two of these 11 cases did not contain HPV-DNA when tested with both PCR and ISH (case no. 88 and 99) suggesting that p16-positivity was not HPV-induced. Conversely HPV-DNA was detected by both PCR and ISH in 2 other cases (no. 54 and 91) suggesting that the RNAscope HPV-test outcomes were false negatives. For the other 7 cases (no. 14, 34, 50, 59, 81, 90 and 92) identification of HPV DNA by ISH or by PCR provided discordant results or was not performed. Discrepancies
Overall survival, disease specific survival and loco-regional control of HPV-positive patients were significantly better than their HPV-negative counterparts whatever the definition of HPV-status (p16 IHC, RNAscope HPV-test, algorithm-1 or 2). Results are summarized in Table 3. However, the disease specific survival difference was less pronounced when HPV status was defined by p16 alone (HR = 0.33 [0.13–0.81], p = 0.02) compared to the RNAscope HPV-testÒ (HR = 0.19 [0.07–0.51], p = 0.001), algorithm-1 (HR = 0.23 [0.08–0.66], p = 0.006), and to algorithm-2 (HR = 0.26 [0.10–0.65], p = 0.004). Disease specific survival curves are presented in Fig. 2. Akaike information criterion Comparison of the 4 diagnostic strategies with Akaike information criterion showed that the RNAscope HPV-testÒ had the lowest score (see supplementary data section, Table F).
105
H. Mirghani et al. / Oral Oncology 62 (2016) 101–108 Table 3 Oncological outcomes summary. Overall survival RNAscope HPV test p16 IHC Algorithm 1a Algorithm 2b a b
HR = 0.24 HR = 0.22 HR = 0.19 HR = 0.21
[0.11–0.50], [0.10–0.45], [0.08–0.44], [0.10–0.44],
p = 0.002 p = 0.001 p < 0.001 p < 0.001
Disease specific survival
Locoregional control
HR = 0.19 HR = 0.33 HR = 0.23 HR = 0.26
HR = 0.15 HR = 0.17 HR = 0.17 HR = 0.13
[0.07–0.51], [0.13–0.81], [0.08–0.66], [0.10–0.65],
p = 0.001 p = 0.02 p = 0.006 p = 0.004
[0.06–0.39], [0.07–0.40], [0.07–0.44], [0.05–0.33],
p < 0.001 p < 0.001 p = 0.003 p < 0.001
Algorithm 1: p16 followed by HPV DNA detection by ISH. Algorithm 2: p16 followed by HPV DNA detection by PCR.
Fig. 2. Disease specific survival curves.
Discussion In the near future, medical care of oropharyngeal cancer (OPC) may vary according to HPV status. This highlights the importance of accurately identifying HPV-driven cancers from those that are not. p16-IHC is widely used to define the HPV status of OPCs because of its numerous advantages (ease of use, high sensitivity and low cost, etc.) [4]. However, even when very stringent criteria are adopted, this method may result in 8–20% of false positives, which is unacceptable when the HPV status is used to alter treatment or follow-up [4,5]. To decrease this risk, 2 diagnostic algorithms that combine different assays have been developed in the last few years [8,9]. Both use p16-IHC as a screening tool and p16-positive samples are further analyzed to detect HR-HPV DNA by PCR or ISH. Although highly effective, the combination of p16IHC and detection of viral DNA by PCR is cumbersome because it
requires pathological and molecular biology facilities. On the other hand, the combination of p16-IHC and ISH to identify HPV-DNA is advantageous because it based solely upon pathological assessment. However, most available ISH assays to identify HPV-DNA are characterized by a notable lack of sensitivity (Fig. 1) [4,11,14]. To address this issue, several expert working groups have recently suggested that the addition of some pathological features may increase the specificity of p16-based screening. For instance, the College of American Pathologists have suggested that OPC samples with a strong and diffuse p16-staining (>70% of cancer cells) that are non-keratinizing or predominantly non-keratinizing can be considered as HPV-driven without further confirmation [15]. For Cancer Care Ontario, OPC samples with a moderate to strong diffuse p16-staining (>50% of cancer cells) and that display basaloid or non-keratinizing morphology are considered as HPVinduced [16]. Although relevant, these recommendations are based
106
H. Mirghani et al. / Oral Oncology 62 (2016) 101–108
on significantly different thresholds to define p16-positivity, which presents an obvious problem. Additionally, the definition of the levels of keratinization and differentiation may cause discrepancies due to intra and inter observer variation as recently highlighted by Larsen et al. [17]. The difficulties outlined above clearly illustrate the need to develop new assays to define HPV status accurately and easily. In this context, the RNAscope HPV-testÒ has several obvious advantages. It allows direct visualization of E6/E7 mRNA within tumor cells, these transcripts being considered as the most relevant marker to assert the link between HPV-infection and cancer [1,2]. Moreover this assay is based on in situ hybridization, which is a technique that is well known by pathologists and that can be used routinely in FFPE samples (see Figs. 3 and 4). In this work, our aim was to assess the performance of the RNAscope HPV-test against the most commonly used strategies to
identify HPV-driven OPC in clinical practice. Our study clearly shows that the number of samples considered as HPV-positive varies significantly, within a same cohort, depending on the method used to define HPV status. Indeed, the rates of HPV-positive samples, based on p16 expression alone, p16 expression combined to HPV-DNA detection by PCR or ISH and the RNAscope HPV-testÒ, were 67%, 61%, 54%, and 59% respectively. These variations are not surprising. It was expected that p16, which is a surrogate marker of HPV-status, would classify a high number of samples as HPV-related given that its use as standalone test is characterized by a lack of specificity [7,14,18,19]. Conversely, the diagnostic algorithm that combines p16 and HPV-DNA detection by ISH has classified fewer samples as HPV-related because of the moderate sensitivity of the available DNA ISH assay [4,11,14,20]. We found 14 discrepant cases between p16-IHC used as a single test and the RNAscope HPV-testÒ. Among these cases, 3 were
Fig. 3. Photomicrographs of an oropharyngeal squamous cell carcinoma with positive RNA in situ hybridization and positive p16 immunohistochemistry. (A) Hematoxylin and eosin morphology, (B) p16 positive immunohistochemistry, (C) Equivocal HPV-DNA in situ hybridization due to non-specific background staining, (D) positive RNA in situ hybridization with punctate brown staining, (E) Ubiquitin C (positive control to confirm that RNA quality is sufficient) and (F) dapB (diaminopimelate B) (negative control).
Fig. 4. Photomicrographs of an oropharyngeal squamous cell carcinoma with negative RNA in situ hybridization and positive p16 immunohistochemistry. (A) Hematoxylin and eosin morphology, (B) positive p16 immunohistochemistry, (C) positive HPV-DNA in situ hybridization, (D) negative RNA in situ, (E) weakly positive Ubiquitin C (positive control to confirm that RNA quality is sufficient) and (F) Negative dapB (diaminopimelate B) (negative control). For this case the PCR was also positive. This suggests that the RNAscope HPV-test provided a false negative result as p16 IHC was positive and HPV DNA was identified by both PCR and ISH.
H. Mirghani et al. / Oral Oncology 62 (2016) 101–108
p16-negative and RNAscope HPV-test positiveÒ. Interestingly, HPV-DNA was identified in 2 of these cases by both PCR and ISH. Thus, it is possible that at least 2 of these 3 cases represent p16 false-negative results. Although rare, this situation has been reported by others [6,7]. One potential explanation is the existence of genetic alterations affecting the CDKN2A gene that encode for p16. Such abnormalities are observed in smokers [21], which was the case for 2 of these 3 patients. The remaining 11 cases were positive with p16-IHC and negative with the RNAscope HPV-testÒ. In 2 of these cases, p16 over-expression was not related to HPV infection given that none of these samples contained HPV-DNA. These 2 samples were therefore p16 false-positives. Conversely, in 2 other cases, the RNAscope HPV-testÒ was probably wrong as HPV-DNA was detected by both PCR and ISH. For the 7 last cases it was not possible to determine whether the RNAscope HPV-test or p16-IHC was correct given that the results of the other assays conflicted (or only positive PCR data were available). With respect to the diagnostic algorithms, we found that the correlation between the RNAscope HPV-testÒ and algorithm-2 was very good with less than 9% discrepant cases, whereas the comparison with algorithm-1 showed a 13% discrepancy. Among the 13 conflicting cases between algorithm-1 and the RNAscope HPV-testÒ, 9 cases were considered as non HPVrelated by algorithm-1 whereas the RNAscope HPV-test was positiveÒ. For 7 of these 9 samples, HPV-DNA was detected by PCR and 5 of these cases would have been classified as HPV-driven by algorithm-2 because they were also p16-positive. This highlights once again the lack of sensitivity of the ISH-assay to identify HPV-DNA. Conversely, algorithm-1 was positive in 4 samples that tested negative with the RNAscope HPV-test. HPV-DNA detection with PCR was performed in 3 of these 4 cases. In 2 of them, the presence of HPV-DNA was confirmed suggesting that the RNAscope HPV-test results were false-negatives. The comparison of these different tests is hampered by the fact that none of them is the gold standard. Indeed, one of the main limitations of our study is the lack of matched fresh frozen samples to look for viral oncogene transcripts by qRT-PCR. Indeed fresh frozen tissue is a scarce resource (this assay was only available for 43 patients and outcomes are presented in the supplementary data section). Therefore, we used an alternative strategy. Starting from the principle that HPV-status is an independent prognostic marker, that outperforms T and N staging, we assumed that the method that better identifies HPV-related samples from their HPVunrelated counterparts would be able to identify 2 groups with the greater survival differential. We found that each of these 4 approaches were able to identify 2 groups with significantly differing prognoses. This was expected as all these different methods were designed to detect HPV related cancers. Interestingly, the disease specific survival differential was greater with the RNAscope HPV-testÒ and the 2 diagnostic algorithms compared to p16 as a standalone test. This could suggest that these 3 approaches are more relevant in identifying OPC that are truly driven by HPV-infection. However as the Hazard Ratio’s confidence intervals overlap, definitive conclusion could only be drawn from larger studies with enough power. To go further in the comparison of these 4 diagnostic approaches, we used the Akaike information criterion, which is a measure of the relative quality of statistical models for a given set of data. This statistical tool was in favor of the RNAscope HPV-testÒ (supplementary data, Table F).
Conclusion This study highlights that none of the available tests is perfect. Our results suggests that the RNAscope HPV-testÒ and the two
107
combination algorithms were all superior to p16-IHC alone in predicting prognosis but this remains to be confirmed. These observations might reflect their better ability to identify cancers that are truly driven by HPV-infection. The algorithm that associates p16-IHC with the identification of HPV-DNA by PCR had a better correlation with the RNAscope HPVtestÒ than the one that associates p16-IHC with the identification of HPV-DNA by ISH (91% vs 87%). Although, the RNAscope HPV-testÒ has the advantage of being a single test, the standard version is not yet fully automated which can be considered as a disadvantage. A fully automated version of the test is already available for research use and will be available for routine screening in the near future (pricing is expected to be in line with existing ISH-based tests). Further studies are needed to address the issue of HPV-testing in OPC and more generally in head and neck cancers. Conflict of interest The RNA in situ hybridization assays (the RNAscopeÒ HPV-test) were funded and performed by the coauthors from Advanced Cell Diagnostics Inc. (M.-X. He, X.-J. Ma and Y. Luo) who have stocks in this company and may profit by use of this test. All the other tests and the final analyses were performed at the Gustave Roussy Cancer Center or at INSERM UMR-S-903. There are no conflicts of interest for any of the other authors. Funding This work has no specific funding. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.oraloncology. 2016.10.009. References [1] Mirghani H, Amen F, Blanchard P, et al. Treatment de-escalation in HPVpositive oropharyngeal carcinoma: ongoing rials, critical issues and perspectives. Int J Cancer 2015;136:1494–503. [2] Van Houten VM, Snijders PJ, van den Brekel MW, et al. Biological evidence that human papillomaviruses are etiologically involved in a subgroup of head and neck squamous cell carcinomas. Int J Cancer 2001;93:232–5. [3] Marur S, D’Souza G, Westra WH, et al. HPV-associated head and neck cancer: a virus-related cancer epidemic. Lancet Oncol 2010;11:781–9. [4] Mirghani H, Amen F, Moreau F, et al. Human papilloma virus testing in oropharyngeal squamous cell carcinoma: what the clinician should know. Oral Oncol 2014;50:1–9. [5] Boscolo-Rizzo P, Pawlita M, Holzinger D. From HPV-positive towards HPVdriven oropharyngeal squamous cell carcinomas. Cancer Treat Rev 2015. doi: http://dx.doi.org/10.1016/j.ctrv.2015.10.009. 2015 Oct 31. pii:S0305-7372(15) 00194-2. [6] Holzinger D, Schmitt M, Dyckhoff G, et al. Viral RNA patterns and high viral load reliably define oropharynx carcinomas with active HPV16 involvement. Cancer Res 2012;72:4993–5003. [7] Robinson M, Sloan P, Shaw R. Refining the diagnosis of oropharyngeal squamous cell carcinoma using human papillomavirus testing. Oral Oncol 2010;46:492–6. [8] Smeets SJ, Hesselink AT, Speel EJ, et al. A novel algorithm for reliable detection of human papillomavirus in paraffin embedded head and neck cancer specimen. Int J Cancer 2007;121:2465–72. [9] Singhi AD, Westra WH. Comparison of human papillomavirus in situ hybridization and p16 immunohistochemistry in the detection of human papillomavirus-associated head and neck cancer based on a prospective clinical experience. Cancer 2010;116:2166–73. [10] Thavaraj S, Stokes A, Guerra E, et al. Evaluation of human papillomavirus testing for squamous cell carcinoma of the tonsil in clinical practice. J Clin Pathol 2011;64:308–12. [11] Mirghani H, Casiraghi O, Amen F, et al. Diagnosis of HPV-driven head and neck cancer with a single test in routine clinical practice. Mod Pathol 2015. doi: http://dx.doi.org/10.1038/modpathol.2015.113. Sep 25.
108
H. Mirghani et al. / Oral Oncology 62 (2016) 101–108
[12] Schache AG, Liloglou T, Risk JM, et al. Validation of a novel diagnostic standard in HPV-positive oropharyngeal squamous cell carcinoma. Br J Cancer 2013;108:1332–9. [13] Gao G, Chernock RD, Gay HA, et al. A novel RT-PCR method for quantification of human papillomavirus transcripts in archived tissues and its application in oropharyngeal cancer prognosis. Int J Cancer 2013;132:882–90. [14] Schache AG, Liloglou T, Risk JM, et al. Evaluation of human papillomavirus diagnostic testing in oropharyngeal squamous cell carcinoma: sensitivity, specificity, and prognostic discrimination. Clin Cancer Res 2011;17:6262–71. [15] Carlson DL, Barnes L, Chan J, et al. Protocol for examination of specimens from patients with carcinomas of the pharynx. Based on AJCC/UICTNM 7th ed. College of American Pathologist; 2012. Available at: http://www.cap.org. 2012. [16] www.cancercare.on.ca/common/pages/UserFile.aspx?fileId=279836.
[17] Larsen CG, Gyldenløve M, Therkildsen MH, et al. Tumor classification of human papilloma virus-related oropharyngeal squamous cell carcinomas is inconsistent. Oral Oncol 2015;51. e63-4. [18] Wasylyk B, Abecassis J, Jung AC. Identification of clinically relevant HPVrelated HNSCC: in p16 should we trust? Oral Oncol 2013;49:33–7. [19] Bishop JA, Ma XJ, Wang H, et al. Detection of transcriptionally active high-risk HPV in patients with head and neck squamous cell carcinoma as visualized by a novel E6/E7 mRNA in situ hybridization method. Am J Surg Pathol 2012;36:1874–82. [20] Ukpo OC, Flanagan JJ, Ma XJ, Luo Y, Thorstad WL, Lewis Jr JS. High-risk human papillomavirus E6/E7 mRNA detection by a novel in situ hybridization assay strongly correlates with p16 expression and patient outcomes in oropharyngeal squamous cell carcinoma. Am J Surg Pathol 2011;35:1343–50. [21] TCGA Releases Head and Neck Cancer Data. Cancer Discov 2015;5:340–1.
Contents lists available at ScienceDirect
Oral Oncology journal homepage: www.elsevier.com/locate/oraloncology
Diagnosis of HPV driven oropharyngeal cancers: Comparing p16 based algorithms with the RNAscope HPV-test Haïtham Mirghani a,⇑, Odile Casiraghi b, Joanne Guerlain a, Furrat Amen c, Ming-Xiao He d, Xiao-Jun Ma d, Yuling Luo d, Céline Mourareau e, Françoise Drusch f, Aïcha Ben Lakdhar b, Antoine Melkane a, Lacau St Guily g, Cécile Badoual h, Jean Yves Scoazec b,f,i, Isabelle Borget j, Anne Aupérin j, Veronique Dalstein e, Philippe Vielh b,f,i a
Department of Otolaryngology – Head and Neck Surgery, Gustave Roussy Cancer Campus, 114 rue Edouard Vaillant, Villejuif, France Department of Biopathology, Gustave Roussy Cancer Campus, 114 rue Edouard Vaillant, Villejuif, France c Department of Otolaryngology, Peterborough City Hospital and Addenbrooke’s Hospital, Cambridge, UK d Advanced Cell Diagnostics, 3960 Point Eden Way, Hayward, CA 94545, USA e INSERM UMR-S 903, SFR CAP-Santé FED 4231, Université de Reims Champagne-Ardenne, F-51100 Reims, France f Biobank, Gustave Roussy Cancer Campus, 114 rue Edouard Vaillant, Villejuif, France g Department of Otolaryngology-Head and Neck Surgery, Hôpital Tenon, Assistance Publique-Hôpitaux de Paris, France h Department of Pathology, Hôpital Européen Georges Pompidou, Assistance Publique-Hôpitaux de Paris, France i Laboratory of Translational Research, Gustave Roussy Cancer Campus, 114 rue Edouard Vaillant, Villejuif, France j Department of Biostatistics and Epidemiology, Gustave Roussy Cancer Campus and University Paris-Sud, Villejuif, France b
a r t i c l e
i n f o
Article history: Received 23 June 2016 Received in revised form 28 September 2016 Accepted 15 October 2016 Available online 21 October 2016 Keywords: Oropharyngeal Oropharynx Head and neck cancer Human papillomavirus 16 E6/E7 mRNA/ transcripts RNAscope HPV-testÒ P16 immunostaining In situ hybridization
a b s t r a c t Background: Accurate identification of HPV-driven oropharyngeal cancer (OPC) is a major issue and none of the current diagnostic approaches is ideal. An in situ hybridization (ISH) assay that detects high-risk HPV E6/E7 mRNA, called the RNAscope HPV-test, has been recently developed. Studies have suggested that this assay may become a standard to define HPV-status. Methods: To further assess this test, we compared its performance against the strategies that are used in routine clinical practice: p16 immunohistochemistry (IHC) as a single test and algorithms combining p16-IHC with HPV-DNA identification by PCR (algorithm-1) or ISH (algorithm-2). Results: 105 OPC specimens were analyzed. The prevalence of HPV-positive samples varied considerably: 67% for p16-IHC, 54% for algorithm-1, 61% for algorithm-2 and 59% for the RNAscope HPV-test. Discrepancies between the RNAscope HPV-test and p16-IHC, algorithm-1 and 2 were noted in respectively 13.3%, 13.1%, and 8.6%. The 4 diagnostic strategies were able to identify 2 groups with different prognosis according to HPVstatus, as expected. However, the greater survival differential was observed with the RNAscope HPVtest [HR: 0.19, 95% confidence interval (CI), 0.07–0.51, p = 0.001] closely followed by algorithm-1 (HR: 0.23, 95% CI, 0.08–0.66, p = 0.006) and algorithm-2 (HR: 0.26, 95% CI, 0.1–0.65, p = 0.004). In contrast, a weaker association was found when p16-IHC was used as a single test (HR: 0.33, 95% CI, 0.13–0.81, p = 0.02). Conclusions: Our findings suggest that the RNAscope HPV-test and p16-based algorithms perform better that p16 alone to identify OPC that are truly driven by HPV-infection. The RNAscope HPV-test has the advantage of being a single test. Ó 2016 Elsevier Ltd. All rights reserved.
Introduction Accurate identification of HPV-driven head and neck cancer is of paramount importance as this may guide treatment and follow up strategies in the near future [1]. Currently, the gold standard assay ⇑ Corresponding author. E-mail address: [email protected] (H. Mirghani). http://dx.doi.org/10.1016/j.oraloncology.2016.10.009 1368-8375/Ó 2016 Elsevier Ltd. All rights reserved.
to detect these cancers is based on the identification of E6/E7 viral oncogenes mRNA by reverse-transcriptase-PCR (RT-PCR) in fresh frozen samples [2,3]. However, this technique is not appropriate for routine screening as it is technically demanding and fresh frozen tissue is typically unavailable [4]. Therefore, several alternative approaches are used in clinical practice to define HPV status. Assessment of p16-protein expression by immunohistochemistry is one of the most commonly used methods and many ongoing
102
H. Mirghani et al. / Oral Oncology 62 (2016) 101–108
clinical trials rely on this marker [1,4]. However, it is now widely acknowledged that its use as a standalone test is associated with a significant proportion of false positive and false negative results [4,5]. Indeed, several mechanisms that are not linked to HPV infection can induce p16 over expression and conversely, cancers that are clearly caused by high-risk HPV (HR-HPV) can lack p16 overexpression, although this second situation is less common [6,7]. To overcome this issue, some have advocated the use of diagnostic algorithms that combine several tests (Fig. 1) [8–10]. However, the inclusion of multiple steps to achieve an accurate and reliable HPV status is technically cumbersome, may produce discordant results, and inevitably increases costs and workload. In this context, new assays that are able simultaneously to identify HPV-infection and to demonstrate its implication in oncogenesis are highly desirable. Some are already under development. The RNAscope HPV-test (Advanced Cell Diagnostics, Hayward, CA, USA) is a recent ISH technique that detects E6/E7 mRNA from 18 HR-HPVs. It uses formalin fixed paraffin embedded (FFPE) tissue sections and the interpretation is visual with light microscopy, the transcripts are directly visualized within the tumor. These features make the RNAscope HPV-test a potentially ideal tool for routine laboratory testing. In a recent pilot study, we compared this test against the gold standard method (identification of HR-HPV E6/ E7 mRNA by RT-PCR in fresh frozen samples) and we found that it performed very well (sensitivity: 93%, specificity: 94%, positive predictive value: 96% and negative predictive value: 88%) [11], thus confirming the outcomes of two previous studies [12,13]. To further assess the RNAscope HPV-test we sought to compare its performance against p16-IHC as a single test and p16-based algorithms, which represent the most accurate means to identify HPVdriven oropharyngeal cancer in routine clinical practice. Methods Patients and tumor samples Tumor samples from 105 patients with oropharyngeal squamous cell carcinoma (OPC) were retrieved from the Gustave
Roussy Cancer Center tissue bank. Fifty of these patients had been enrolled into our previous study. All samples were collected prior to treatment. Clinical notes and pathological reports of each patient were retrospectively analyzed by the investigators. Our Institutional Review Board approved the study protocol and all patients provided written informed consent. HPV testing assays 4 different tests were performed: p16 immunohistochemistry (IHC), identification of HR-HPV DNA by PCR and in situ hybridization (ISH), and detection of HR-HPV transcripts by the RNAscope HPV-testÒ. The combination of p16-IHC and ISH to identify HPVDNA was designated as algorithm-1, whereas the combination of p16-IHC and PCR to detect HPV-DNA was designated as algorithm-2. Samples were considered to be HPV-driven with algorithm-1 if p16-IHC was positive and HPV-DNA was detected by ISH. For algorithm-2, samples were considered HPV-driven if p16-IHC was positive and HPV-DNA detected by PCR. Preparation of sections from FFPE tumor tissue blocks Paraffin sections were prepared according to the sandwich method. That is to say the first and last sections were stained by hematoxylin and eosin to check for tumor presence. The remaining 4 lm sections were used to perform p16-IHC, HPV-DNA ISH and the RNAscopeÒ HPV test. p16-IHC p16-IHC was performed on a Ventana Benchmark Ultra (Ventana Medical Systems Inc., Tucson, AZ) using the CINtec p16 Histology Kit (Ref. 9511, Roche mtm laboratories AG, Heidelberg, Germany), in accord with both manufacturers’ recommendations. The previously prepared 4-lm whole tumor FFPE tissues paraffin sections were deparaffinized and subjected to antigen retrieval using CC1 buffer for 30 min. Subsequently they were consecutively incubated in the prediluted CINtec p16 primary antibody (clone
Fig. 1. Stepwise diagnostic algorithm. p16-IHC is used as a first line assay. p16-negative samples are considered as HPV-unrelated. Conversely p16-positive samples are potentially HPV-driven, and some may consider these cases as HPV-related. To confirm this, identification of HR-HPV is necessary. To this end, 2 strategies can be used. Smeets et al. have proposed to identify HR-HPV DNA by PCR whereas the approach developed by Singhi et al. is based on ISH. In their original work, Singhi et al. recommended using probes that are specific to HPV16 and if these probes are negative to use secondarily consensus probes that detect an extended panel of HR-HPV types. As genotypespecific probes are not available for diagnostic use in Europe owing to licensing restrictions, some have proposed directly using the high-risk HPV cocktail probes that are recognized IVDs with CE marking. For these authors, samples that are p16-positive but lacking HPV-DNA by ISH require further confirmation by PCR given that the sensitivity of these ISH assays is not optimal. However, the complexity of these strategies raises questions about their relevance to routine practice. TSP: type specific probes, CP: consensus probes.
H. Mirghani et al. / Oral Oncology 62 (2016) 101–108
E6H4) for 20 min at room temperature, and revealed with Ultra View Universal DAB Detection Kit. Slides were finally counterstained with Hematoxylin II for 8 min (Ventana) and Bluing reagent for 4 min and then washed. A tonsil squamous cell carcinoma with high p16 expression was used as a positive control. The primary antibody was omitted from negative controls. p16 IHC was scored as positive if there was strong, homogeneous and diffuse nuclear and cytoplasmic staining present in greater than 70% of the malignant cells. All other staining patterns were scored as negative. Detection of high-risk HPV DNA by ISH DNA ISH was performed on representative 4-lm whole tumor FFPE sections from each case using the ISH I View Blue Plus Detection Kit (Ventana Medical System, Inc., Tucson, AZ) according to the manufacturer’s instructions. The assay utilized the Ventana HPV III Family 16, Probe B, a cocktail recognizing the high-risk (HR) HPV types 16, 18, 31, 33, 35, 45, 51, 52, 56, 58, 59, 68, and 70. Ventana Red Counterstain II (Ventana Medical System, Inc., Tucson, AZ) was used. Controls in each run included a known HPV 16-positive OPSCC case (positive control) and a section of normal tonsil (negative control). Positive staining was identified as blue nuclear dots. Any definitive nuclear staining in the tumor cells was considered positive. Cases were classified in a binary manner as either positive or negative. Detection of high-risk HPV E6/E7 mRNA by RNA ISH Detection of HR-HPV E6/E7 mRNA was performed using the RNAscope 2.0 BROWN assay kit and the HPV-HR18 probe cocktail (Advanced Cell Diagnostics Inc.) in accordance with the manufacturer’s instructions. Briefly, 4 lm sections were deparaffinized and pretreated with heat and protease before hybridization with target-specific probes for the E6 and E7 genes of 18 HR-HPV genotypes (HPV18, 26, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73 and 82). Ubiquitin C (UBC, a constitutively expressed endogenous gene) and the bacterial gene, dapB, were used as positive and negative controls, respectively. The UBC test was used to assess the presence of hybridizable RNA in order to confirm adequate RNA quality and was defined as adequate if there were at least 5 punctate signal dots in the majority of tumor cells in the section. This is especially important when the HPV probe signal is negative to avoid a false-negative result. Cases with weak UBC staining were retested with the RNAscope 2.5 BROWN Assay Kit (ACD). The dapB test was used to assess nonspecific staining, only those cases that were negative or weakly stained were considered for HPV scoring. A positive HPV test result was defined as punctate staining that localized to the cytoplasm and/or nucleus of any of the malignant cells, and where staining was present in the negative control, it was at least three times as strong as the dapB staining.
103
DNA extraction and HPV genotyping 2–30 10-lm serial sections (according to the size of the specimen) were prepared from FFPE tissue samples in which tumor presence was checked. DNA was extracted by the MaxwellÒ 16 FFPE Tissue LEV DNA Kit (Promega) according to the manufacturer’s instructions. DNA was eluted win 50 lL of Elution Buffer. HPV genotyping was then performed using INNO-LiPA HPV Genotyping Extra II kit (Fujirebio) according to the manufacturer’s instructions. This assay is based on a PCR amplification of the L1 gene by the SPF10 Plus primer system. Positive controls (DNA from SiHa and HeLa cell lines) and no template controls were included in each set of reactions. All hybridization steps up to color development were then performed using an Auto-LiPA system (Fujirebio). Hybridization patterns were analyzed using the INNO-LiPA HPV genotyping Extra II Reading Card. The Extra II version of the assay allows the individual detection of 32 HPV genotypes: 13 high-risk (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68), 6 probable/possible high-risk (26, 53, 66, 70, 73, 82) and 13 low-risk or not classified (6, 11, 40, 42, 43, 44, 54, 61, 62, 67, 81, 83, 89).
Statistical analysis Overall survival was defined as the time from the date of registration to the date of death or to the date of censorship (ie, the last date of follow-up). Disease specific survival was defined as the time from the date of registration to the date of death from cancer or to the date of censorship. The distributions of overall survival and disease specific survival were estimated with the KaplanMeier method. A log-rank test was performed to test the difference in survival between groups in univariate analysis. In multivariable analyses, a Cox proportional hazards model was used to adjust for covariates of statistical significance in univariate analysis. The Wald test was used to estimate the 95% confidence intervals (CIs) of hazard ratios. The 4 diagnostic strategies (p16 IHC alone, the RNAscope HPV testÒ and the 2 algorithms) were compared using the Akaike information criterion. All statistical tests were two-sided, and a P value of 0.05 or less was considered statistically significant.
Results Population description 73 men and 32 women were included in our study. Location, TNM staging, age, and tobacco consumption are summarized in Table 1.
Global analysis of the tests Review of the miscroscopic slides The RNAscopeÒ HPV-test slides were scored independently by 2 authors (MX. H and O. C) and the p16 IHC and HR-HPV DNA ISH tests were also assessed independently by 2 authors (O. C and A. BL). Scoring systems were based on a binary classification (positive vs. negative). All the people involved in microscopic slides review were blinded from the results of all the different assays that were tested (including HPV genotyping by PCR, described in the next paragraphs) and to the scoring of the other reviewers. The results were collated, and discordant scores were resolved at a meeting between the pathologists to establish a consensus interpretation.
Data for the RNAscopeÒ HPV-test was available for the whole cohort whereas data for p16, HPV-DNA ISH and HPV DNA genotyping were available for 104/105, 94/105 and 104/105 patients respectively. The scoring of the RNAscope HPV-test and p16 slides were completely concordant between the pathologists. Discordant cases were only seen with the HR-HPV DNA-ISH assay and were resolved at a meeting to establish a consensus interpretation. The outcomes of all the tests that were carried out including the 2 p16-based algorithms are summarized in Table 2 and the complete dataset is presented in the supplementary data section (supplementary data, Table A).
104
H. Mirghani et al. / Oral Oncology 62 (2016) 101–108
between p16-IHC and the RNAscope HPV-test are presented in the supplementary data section (Table B).
Table 1 Population description. Sex
M: 73 (69.5%) F: 32 (30.5%)
Age
Min: 27 Max:83 Mean: 56 Median: 58
Outcomes comparison between algorithm-1 and the RNAscope HPVtest
Stage
T1: 10 (9.5%) T2: 42 (40%) T3: 30 (28.6%) T4: 23 (21.9%) N0: 28 (26.6%) N1: 9 (8.6%) N2a: 7 (6.6%) N2b: 34 (32.4%) N2c: 15 (14.2%) N3: 9 (8.7%) Nx: 3 (2.9%)
Tobacco
Yes: 55 (52.4%) No: 50 (47.6%)
Location
Tonsil: 62 (59%) Tongue base: 37 (35.3%) Other oropharyngeal subsites: 6 (5.7%)
Table 2 Results summary.
RNAscope HPV test p16 IHCb HPV-DNA ISH HPV-DNA PCR Algorithm 1c Algorithm 2d
Number of casesa
Positive
Negative
105/105 (100%) 104/105 (99%) 94/105 (89.5%) 104/105 (99%) 99/105 (94.3%) 103/105 (98%)
62/105 (59%) 70/104 (67%) 69/94 (73%) 80/104 (77%) 54/99 (54.5%) 63/103 (61.2%)
43/105 (41%) 34/104 (32%) 25/94 (26.6%) 24/104 (23%) 45/99 (45.5%) 40,103 (38.2%)
a The RNAscope HPV test was performed on the whole cohort (n = 105) whereas p16 IHC, HPV-DNA ISH and HPV-DNA PCR were performed in 104/105, 94/105 and 104/105 patients respectively. b all the tumoral samples that were considered as p16-positive (>70% nuclear and cytoplasmic staining) had a strong staining intensity. c Algorithm 1: p16 followed by HPV DNA detection by ISH. d Algorithm 2: p16 followed by HPV DNA detection by PCR.
HPV-genotypes distribution Distribution of HPV genotypes was as follows: 69 (87%) HPV16, 4 (5%) HPV18, 4 (5%) HPV33, 1 (1%) HPV35, 1 (1%) HPV16/33 and 1 (1%) HPV16/35 (supplementary data, Table A).
86/99 (87%) samples were concordant between algorithm-1 and the RNAscope HPV-test. Among the 13 discordant cases, 9 were negative with algorithm-1 and positive with the RNAscope HPVtest. Among these 9 samples, 6 were p16-positive (no. 4, 6, 46, 47, 69 and 97) and were classified as non HPV-related by algorithm-1 because HPV-DNA was not detected by the ISHassay. Interestingly HPV-DNA was identified by PCR in 5 out of these 6 cases (no. 6, 46, 47, 69 and 97) that were subsequently considered as HPV-driven by algorithm-2. Finally, 4 cases were positive with algorithm-1 and negative with the RNAscope HPV-test. Two of these cases were also positive with algorithm-2 (no. 54, 91). Discrepancies between algorithm-1 and the RNAscope HPV-test are presented in the supplementary data section (Table C). Outcomes comparison between algorithm-2 and the RNAscope HPVtest 94/103 (91%) samples were concordant between algorithm-2 and the RNAscope HPV-test. Among the 9 discordant cases, 4 were negative with algorithm-2 and positive with the RNAscope HPVtest. All these cases were also negative with algorithm-1 (no. 4, 29, 49, 76). Conversely, 5 samples were positive with algorithm-2 and negative with the RNAscope HPV-test. For 3 of these cases (no. 14, 34, 50) outcomes of algorithm-1 were negatives like those obtained with the RNAscope HPV-test. Discrepancies between algorithm-2 and the RNAscope HPV-test are presented in the supplementary data section (Table D). Outcomes comparison between the 2 diagnostic algorithms 89 (90%) cases were concordant between the 2 algorithms. Among the 10 discordant cases, algorithm-1 was negative and algorithm-2 was positive in 9 cases. 6 out of the 9 cases were positive with the RNAscope HPV-test. Conversely, algorithm-1 was positive and algorithm-2 negative in 1 case. This case was negative with the RNAscope HPV-test. Discrepancies between the 2 algorithms are presented in the supplementary data section (Table E). Oncological outcomes
Outcomes comparison between p16-IHC and the RNAscope HPV-test 90 (86.5%) cases were concordant between p16-IHC and the RNAscope HPV-test. Among the 14 discordant cases 3 were p16negative but positive with the RNAscope HPV-test. For 2 out of these 3 samples, the results obtained with the other assays were consistent with the RNAscope HPV-test. Indeed, HPV-DNA was identified by both PCR and ISH in 2 cases (no. 29 and 76). Conversely, 11 cases were p16-positive but negative with the RNAscope HPV-test. Two of these 11 cases did not contain HPV-DNA when tested with both PCR and ISH (case no. 88 and 99) suggesting that p16-positivity was not HPV-induced. Conversely HPV-DNA was detected by both PCR and ISH in 2 other cases (no. 54 and 91) suggesting that the RNAscope HPV-test outcomes were false negatives. For the other 7 cases (no. 14, 34, 50, 59, 81, 90 and 92) identification of HPV DNA by ISH or by PCR provided discordant results or was not performed. Discrepancies
Overall survival, disease specific survival and loco-regional control of HPV-positive patients were significantly better than their HPV-negative counterparts whatever the definition of HPV-status (p16 IHC, RNAscope HPV-test, algorithm-1 or 2). Results are summarized in Table 3. However, the disease specific survival difference was less pronounced when HPV status was defined by p16 alone (HR = 0.33 [0.13–0.81], p = 0.02) compared to the RNAscope HPV-testÒ (HR = 0.19 [0.07–0.51], p = 0.001), algorithm-1 (HR = 0.23 [0.08–0.66], p = 0.006), and to algorithm-2 (HR = 0.26 [0.10–0.65], p = 0.004). Disease specific survival curves are presented in Fig. 2. Akaike information criterion Comparison of the 4 diagnostic strategies with Akaike information criterion showed that the RNAscope HPV-testÒ had the lowest score (see supplementary data section, Table F).
105
H. Mirghani et al. / Oral Oncology 62 (2016) 101–108 Table 3 Oncological outcomes summary. Overall survival RNAscope HPV test p16 IHC Algorithm 1a Algorithm 2b a b
HR = 0.24 HR = 0.22 HR = 0.19 HR = 0.21
[0.11–0.50], [0.10–0.45], [0.08–0.44], [0.10–0.44],
p = 0.002 p = 0.001 p < 0.001 p < 0.001
Disease specific survival
Locoregional control
HR = 0.19 HR = 0.33 HR = 0.23 HR = 0.26
HR = 0.15 HR = 0.17 HR = 0.17 HR = 0.13
[0.07–0.51], [0.13–0.81], [0.08–0.66], [0.10–0.65],
p = 0.001 p = 0.02 p = 0.006 p = 0.004
[0.06–0.39], [0.07–0.40], [0.07–0.44], [0.05–0.33],
p < 0.001 p < 0.001 p = 0.003 p < 0.001
Algorithm 1: p16 followed by HPV DNA detection by ISH. Algorithm 2: p16 followed by HPV DNA detection by PCR.
Fig. 2. Disease specific survival curves.
Discussion In the near future, medical care of oropharyngeal cancer (OPC) may vary according to HPV status. This highlights the importance of accurately identifying HPV-driven cancers from those that are not. p16-IHC is widely used to define the HPV status of OPCs because of its numerous advantages (ease of use, high sensitivity and low cost, etc.) [4]. However, even when very stringent criteria are adopted, this method may result in 8–20% of false positives, which is unacceptable when the HPV status is used to alter treatment or follow-up [4,5]. To decrease this risk, 2 diagnostic algorithms that combine different assays have been developed in the last few years [8,9]. Both use p16-IHC as a screening tool and p16-positive samples are further analyzed to detect HR-HPV DNA by PCR or ISH. Although highly effective, the combination of p16IHC and detection of viral DNA by PCR is cumbersome because it
requires pathological and molecular biology facilities. On the other hand, the combination of p16-IHC and ISH to identify HPV-DNA is advantageous because it based solely upon pathological assessment. However, most available ISH assays to identify HPV-DNA are characterized by a notable lack of sensitivity (Fig. 1) [4,11,14]. To address this issue, several expert working groups have recently suggested that the addition of some pathological features may increase the specificity of p16-based screening. For instance, the College of American Pathologists have suggested that OPC samples with a strong and diffuse p16-staining (>70% of cancer cells) that are non-keratinizing or predominantly non-keratinizing can be considered as HPV-driven without further confirmation [15]. For Cancer Care Ontario, OPC samples with a moderate to strong diffuse p16-staining (>50% of cancer cells) and that display basaloid or non-keratinizing morphology are considered as HPVinduced [16]. Although relevant, these recommendations are based
106
H. Mirghani et al. / Oral Oncology 62 (2016) 101–108
on significantly different thresholds to define p16-positivity, which presents an obvious problem. Additionally, the definition of the levels of keratinization and differentiation may cause discrepancies due to intra and inter observer variation as recently highlighted by Larsen et al. [17]. The difficulties outlined above clearly illustrate the need to develop new assays to define HPV status accurately and easily. In this context, the RNAscope HPV-testÒ has several obvious advantages. It allows direct visualization of E6/E7 mRNA within tumor cells, these transcripts being considered as the most relevant marker to assert the link between HPV-infection and cancer [1,2]. Moreover this assay is based on in situ hybridization, which is a technique that is well known by pathologists and that can be used routinely in FFPE samples (see Figs. 3 and 4). In this work, our aim was to assess the performance of the RNAscope HPV-test against the most commonly used strategies to
identify HPV-driven OPC in clinical practice. Our study clearly shows that the number of samples considered as HPV-positive varies significantly, within a same cohort, depending on the method used to define HPV status. Indeed, the rates of HPV-positive samples, based on p16 expression alone, p16 expression combined to HPV-DNA detection by PCR or ISH and the RNAscope HPV-testÒ, were 67%, 61%, 54%, and 59% respectively. These variations are not surprising. It was expected that p16, which is a surrogate marker of HPV-status, would classify a high number of samples as HPV-related given that its use as standalone test is characterized by a lack of specificity [7,14,18,19]. Conversely, the diagnostic algorithm that combines p16 and HPV-DNA detection by ISH has classified fewer samples as HPV-related because of the moderate sensitivity of the available DNA ISH assay [4,11,14,20]. We found 14 discrepant cases between p16-IHC used as a single test and the RNAscope HPV-testÒ. Among these cases, 3 were
Fig. 3. Photomicrographs of an oropharyngeal squamous cell carcinoma with positive RNA in situ hybridization and positive p16 immunohistochemistry. (A) Hematoxylin and eosin morphology, (B) p16 positive immunohistochemistry, (C) Equivocal HPV-DNA in situ hybridization due to non-specific background staining, (D) positive RNA in situ hybridization with punctate brown staining, (E) Ubiquitin C (positive control to confirm that RNA quality is sufficient) and (F) dapB (diaminopimelate B) (negative control).
Fig. 4. Photomicrographs of an oropharyngeal squamous cell carcinoma with negative RNA in situ hybridization and positive p16 immunohistochemistry. (A) Hematoxylin and eosin morphology, (B) positive p16 immunohistochemistry, (C) positive HPV-DNA in situ hybridization, (D) negative RNA in situ, (E) weakly positive Ubiquitin C (positive control to confirm that RNA quality is sufficient) and (F) Negative dapB (diaminopimelate B) (negative control). For this case the PCR was also positive. This suggests that the RNAscope HPV-test provided a false negative result as p16 IHC was positive and HPV DNA was identified by both PCR and ISH.
H. Mirghani et al. / Oral Oncology 62 (2016) 101–108
p16-negative and RNAscope HPV-test positiveÒ. Interestingly, HPV-DNA was identified in 2 of these cases by both PCR and ISH. Thus, it is possible that at least 2 of these 3 cases represent p16 false-negative results. Although rare, this situation has been reported by others [6,7]. One potential explanation is the existence of genetic alterations affecting the CDKN2A gene that encode for p16. Such abnormalities are observed in smokers [21], which was the case for 2 of these 3 patients. The remaining 11 cases were positive with p16-IHC and negative with the RNAscope HPV-testÒ. In 2 of these cases, p16 over-expression was not related to HPV infection given that none of these samples contained HPV-DNA. These 2 samples were therefore p16 false-positives. Conversely, in 2 other cases, the RNAscope HPV-testÒ was probably wrong as HPV-DNA was detected by both PCR and ISH. For the 7 last cases it was not possible to determine whether the RNAscope HPV-test or p16-IHC was correct given that the results of the other assays conflicted (or only positive PCR data were available). With respect to the diagnostic algorithms, we found that the correlation between the RNAscope HPV-testÒ and algorithm-2 was very good with less than 9% discrepant cases, whereas the comparison with algorithm-1 showed a 13% discrepancy. Among the 13 conflicting cases between algorithm-1 and the RNAscope HPV-testÒ, 9 cases were considered as non HPVrelated by algorithm-1 whereas the RNAscope HPV-test was positiveÒ. For 7 of these 9 samples, HPV-DNA was detected by PCR and 5 of these cases would have been classified as HPV-driven by algorithm-2 because they were also p16-positive. This highlights once again the lack of sensitivity of the ISH-assay to identify HPV-DNA. Conversely, algorithm-1 was positive in 4 samples that tested negative with the RNAscope HPV-test. HPV-DNA detection with PCR was performed in 3 of these 4 cases. In 2 of them, the presence of HPV-DNA was confirmed suggesting that the RNAscope HPV-test results were false-negatives. The comparison of these different tests is hampered by the fact that none of them is the gold standard. Indeed, one of the main limitations of our study is the lack of matched fresh frozen samples to look for viral oncogene transcripts by qRT-PCR. Indeed fresh frozen tissue is a scarce resource (this assay was only available for 43 patients and outcomes are presented in the supplementary data section). Therefore, we used an alternative strategy. Starting from the principle that HPV-status is an independent prognostic marker, that outperforms T and N staging, we assumed that the method that better identifies HPV-related samples from their HPVunrelated counterparts would be able to identify 2 groups with the greater survival differential. We found that each of these 4 approaches were able to identify 2 groups with significantly differing prognoses. This was expected as all these different methods were designed to detect HPV related cancers. Interestingly, the disease specific survival differential was greater with the RNAscope HPV-testÒ and the 2 diagnostic algorithms compared to p16 as a standalone test. This could suggest that these 3 approaches are more relevant in identifying OPC that are truly driven by HPV-infection. However as the Hazard Ratio’s confidence intervals overlap, definitive conclusion could only be drawn from larger studies with enough power. To go further in the comparison of these 4 diagnostic approaches, we used the Akaike information criterion, which is a measure of the relative quality of statistical models for a given set of data. This statistical tool was in favor of the RNAscope HPV-testÒ (supplementary data, Table F).
Conclusion This study highlights that none of the available tests is perfect. Our results suggests that the RNAscope HPV-testÒ and the two
107
combination algorithms were all superior to p16-IHC alone in predicting prognosis but this remains to be confirmed. These observations might reflect their better ability to identify cancers that are truly driven by HPV-infection. The algorithm that associates p16-IHC with the identification of HPV-DNA by PCR had a better correlation with the RNAscope HPVtestÒ than the one that associates p16-IHC with the identification of HPV-DNA by ISH (91% vs 87%). Although, the RNAscope HPV-testÒ has the advantage of being a single test, the standard version is not yet fully automated which can be considered as a disadvantage. A fully automated version of the test is already available for research use and will be available for routine screening in the near future (pricing is expected to be in line with existing ISH-based tests). Further studies are needed to address the issue of HPV-testing in OPC and more generally in head and neck cancers. Conflict of interest The RNA in situ hybridization assays (the RNAscopeÒ HPV-test) were funded and performed by the coauthors from Advanced Cell Diagnostics Inc. (M.-X. He, X.-J. Ma and Y. Luo) who have stocks in this company and may profit by use of this test. All the other tests and the final analyses were performed at the Gustave Roussy Cancer Center or at INSERM UMR-S-903. There are no conflicts of interest for any of the other authors. Funding This work has no specific funding. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.oraloncology. 2016.10.009. References [1] Mirghani H, Amen F, Blanchard P, et al. Treatment de-escalation in HPVpositive oropharyngeal carcinoma: ongoing rials, critical issues and perspectives. Int J Cancer 2015;136:1494–503. [2] Van Houten VM, Snijders PJ, van den Brekel MW, et al. Biological evidence that human papillomaviruses are etiologically involved in a subgroup of head and neck squamous cell carcinomas. Int J Cancer 2001;93:232–5. [3] Marur S, D’Souza G, Westra WH, et al. HPV-associated head and neck cancer: a virus-related cancer epidemic. Lancet Oncol 2010;11:781–9. [4] Mirghani H, Amen F, Moreau F, et al. Human papilloma virus testing in oropharyngeal squamous cell carcinoma: what the clinician should know. Oral Oncol 2014;50:1–9. [5] Boscolo-Rizzo P, Pawlita M, Holzinger D. From HPV-positive towards HPVdriven oropharyngeal squamous cell carcinomas. Cancer Treat Rev 2015. doi: http://dx.doi.org/10.1016/j.ctrv.2015.10.009. 2015 Oct 31. pii:S0305-7372(15) 00194-2. [6] Holzinger D, Schmitt M, Dyckhoff G, et al. Viral RNA patterns and high viral load reliably define oropharynx carcinomas with active HPV16 involvement. Cancer Res 2012;72:4993–5003. [7] Robinson M, Sloan P, Shaw R. Refining the diagnosis of oropharyngeal squamous cell carcinoma using human papillomavirus testing. Oral Oncol 2010;46:492–6. [8] Smeets SJ, Hesselink AT, Speel EJ, et al. A novel algorithm for reliable detection of human papillomavirus in paraffin embedded head and neck cancer specimen. Int J Cancer 2007;121:2465–72. [9] Singhi AD, Westra WH. Comparison of human papillomavirus in situ hybridization and p16 immunohistochemistry in the detection of human papillomavirus-associated head and neck cancer based on a prospective clinical experience. Cancer 2010;116:2166–73. [10] Thavaraj S, Stokes A, Guerra E, et al. Evaluation of human papillomavirus testing for squamous cell carcinoma of the tonsil in clinical practice. J Clin Pathol 2011;64:308–12. [11] Mirghani H, Casiraghi O, Amen F, et al. Diagnosis of HPV-driven head and neck cancer with a single test in routine clinical practice. Mod Pathol 2015. doi: http://dx.doi.org/10.1038/modpathol.2015.113. Sep 25.
108
H. Mirghani et al. / Oral Oncology 62 (2016) 101–108
[12] Schache AG, Liloglou T, Risk JM, et al. Validation of a novel diagnostic standard in HPV-positive oropharyngeal squamous cell carcinoma. Br J Cancer 2013;108:1332–9. [13] Gao G, Chernock RD, Gay HA, et al. A novel RT-PCR method for quantification of human papillomavirus transcripts in archived tissues and its application in oropharyngeal cancer prognosis. Int J Cancer 2013;132:882–90. [14] Schache AG, Liloglou T, Risk JM, et al. Evaluation of human papillomavirus diagnostic testing in oropharyngeal squamous cell carcinoma: sensitivity, specificity, and prognostic discrimination. Clin Cancer Res 2011;17:6262–71. [15] Carlson DL, Barnes L, Chan J, et al. Protocol for examination of specimens from patients with carcinomas of the pharynx. Based on AJCC/UICTNM 7th ed. College of American Pathologist; 2012. Available at: http://www.cap.org. 2012. [16] www.cancercare.on.ca/common/pages/UserFile.aspx?fileId=279836.
[17] Larsen CG, Gyldenløve M, Therkildsen MH, et al. Tumor classification of human papilloma virus-related oropharyngeal squamous cell carcinomas is inconsistent. Oral Oncol 2015;51. e63-4. [18] Wasylyk B, Abecassis J, Jung AC. Identification of clinically relevant HPVrelated HNSCC: in p16 should we trust? Oral Oncol 2013;49:33–7. [19] Bishop JA, Ma XJ, Wang H, et al. Detection of transcriptionally active high-risk HPV in patients with head and neck squamous cell carcinoma as visualized by a novel E6/E7 mRNA in situ hybridization method. Am J Surg Pathol 2012;36:1874–82. [20] Ukpo OC, Flanagan JJ, Ma XJ, Luo Y, Thorstad WL, Lewis Jr JS. High-risk human papillomavirus E6/E7 mRNA detection by a novel in situ hybridization assay strongly correlates with p16 expression and patient outcomes in oropharyngeal squamous cell carcinoma. Am J Surg Pathol 2011;35:1343–50. [21] TCGA Releases Head and Neck Cancer Data. Cancer Discov 2015;5:340–1.