Identification of high protein kinase CK2α in HPV(+) oropharyngeal squamous cell carcinoma and correlation with clinical outcomes

Patients

The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board of Minneapolis VA Health Care System (4632-A, approved June 6, 2016). Patient consent was waived by the Institutional Review Board due to the retrospective nature of the study. Data analysis was completed under an approved Research and Development Committee protocol (VAM-20-00609, approved July 17, 2020).

For this retrospective cohort study, a database search was performed in the tumor registry for the Minneapolis Veterans Affairs Healthcare System. All patients with newly diagnosed primary OPSCC from January 2005 to December 2015 were identified. Inclusion criteria were (1) biopsy-proven primary OPSCC, and (2) adequate tissue in archived tissue bank to perform immunohistochemical (IHC) staining. Patients who had other simultaneous active malignancy at the time of diagnosis or lacked follow up were excluded. Patient demographics including age at diagnosis, gender, smoking history, and alcohol consumption history were recorded. Tumor characteristics including primary subsite, tumor stage, and treatment modalities were reviewed from patient charts. Clinical outcomes including disease progression and death were also recorded. Patients were censored at the end of this study.

p16, Ki-67 and CK2 IHC stain

Formalin-fixed paraffin-embedded (FFPE) tissue blocks were retrieved from the pathology tissue archive. IHC stain was performed with Bond Refine Polymer Detection Kit on the Bond III Automated Stainer (Leica Biosystems, Buffalo Grove, IL, USA), with the following primary antibodies: mouse monoclonal anti-p16 antibody E6H4 (Ventana Medical Systems, Oro Valley, AZ, USA), rabbit monoclonal anti-Ki-67 antibody SP6 (Cell Marque, Sigma-Aldrich, Rocklin, CA, USA), and rabbit monoclonal anti-CK2 alpha antibody ab76040 (Abcam, Cambridge, United Kingdom). Positive p16 expression was defined as greater than 70% of tumor cells with strong nuclear and cytoplasmic staining. Reactivity for Ki-67 was evaluated as percentage of viable tumor nuclei positive for this marker.

CK2α staining was verified using myometrium for the “negative” control which showed a very low level staining and using testicle as the positive control as it has a high level of CK2. For further negative control, the CK2α primary antibody was substituted with appropriate isotype control antibody ab172730 (Abcam). Specificity of CK2α staining was tested by comparing CK2α antibody to isotype control in sequential sections from five OPSCC tumors with CK2α score of 3 and five OPSCC tumors with CK2α score of 1. CK2α scoring was evaluated by semi-quantitative assessment of the relative antigen density in combined cytoplasm and nucleus of viable tumor cells (Zhou et al., 2014). Reactivity was evaluated as score 3 (intense), score 2 (moderate), score 1 (weak), and score 0 (no staining). All scoring and expression levels were evaluated and agreed upon by two staff pathologists (J.P. and J.C.M.).

E6 and E7 RNA in situ hybridization stain

E6 and E7 mRNA of HPV16/18 were detected in tumor sections using the Bond RNAscope kit and following the Leica RNAscope DAB in situ hybridization protocol (*RS_DAB, Leica Biosystems). Processing was performed using the Automated Stainer instrument.

Definition of HPV status

To obtain unambiguous information on the HPV status in these tumors, they were stained for cellular p16^INK4a, commonly used as a surrogate marker for HPV positivity, and for the HPV RNAs E6 and E7. P16 was positive in 70.3% of patients and E6/E7 RNAs were detected in 70.6% of patients. Two patient tumors were p16-positive and E6/E7 RNA-negative; conversely, two tumors were E6/E7 RNA-positive and p16-negative. As expected, the relationship between p16 and E6/E7 detection in tumors was significant (P < 0.0001). In the remainder of this study, positive E6/E7 detection in tumors was used as our operational definition of HPV(+) vs. HPV(−) status, including those patients with discordant p16 results.

Human HNSCC TCGA data analysis

Analysis of survival associated with CSNK2A1 mRNA levels in HNSCC HPV(−) patient samples was performed via cBioPortal using the PanCancer Atlas data set from The Cancer Genome Atlas (n = 415 HPV(−) samples for OS and PFS; n = 394 for DSS; 23 September 2021). For mRNA expression data, TCGA typically computes the relative expression of an individual gene in a tumor sample to the gene’s expression distribution in a reference population of samples. In this analysis, the reference set was all samples that are diploid for CSNK2A1. The returned value indicates the number of standard deviations away from the mean of expression in the reference population (Z-score) Z-score cut-off was set at >+1.5 higher expression of CSNK2A1 compared to unaltered or lower levels of CSNK2A1.

Statistical analysis

In Table 1, we utilize counts and proportions for categorical variables to describe the sample. We present descriptive data overall and examine the relationships between CK2 score category and the variables HPV status or Ki-67. Analyses for Table 2 included Pearson’s chi-square test for categorical variables or Fisher’s exact test where any cell size was less than 5. The chi-square test for independence was used to test the relationship between positive detection of E6/E7 RNA and of p16. We defined statistical significance by P value < 0.05. The analyses were performed in SAS 9.4^®software.

Survival analysis

Utilizing the Veterans cohort dataset, univariate Cox regression was performed for clinical variables CK2α and HPV status (predictors of interest) and for co-variates Ki-67, treatment, age, alcohol, smoking, T and N classifications, and clinical stage (7th edition of the AJCC cancer staging manual) to explore if any of these should be controlled for in our main two survival analyses of interest (Table 3). As our main analysis of interest, OS was stratified by HPV and separately by CK2α. Survival curves for OS according to CK2α score and separately for HPV status were fitted using the Kaplan–Meier method. Log rank P-values are presented. Since smoking status (current vs. never or former) was significant by univariate analysis, we additionally controlled for this in multivariate Cox regressions. OS time was defined as from treatment initiation to death. PFS time was defined as from treatment initiation to disease progression or death. These analyses were performed in SAS 9.4^®software. Analyses using the TCGA cohort dataset were performed using cBioPortal. A P-value of 0.05 was utilized to denote statistical significance.

Patient and tumor characteristics overall

A total of 147 patients at the Minneapolis VA hospital between 2005 and 2015 met the inclusion criteria, and sufficient tumor tissue was available for analysis in 119 patients. The mean age of diagnosis for all patients, as well as the HPV(+) and HPV(−) subgroups, was 63 years. The diagnosis age range for HPV(+) was 44 to 88 years, and for HPV(−) was 53 to 82 years. As shown in Table 1, the majority of the patients had a history of smoking (94.1%) and alcohol use (84.0%). The most common subsite of OPSCC was tonsil (58.6%), followed by base of tongue (35.3%). The majority of the patients presented with stage IV disease (70.6%) according to the 7th edition of the AJCC cancer staging manual. All tumors were conventional type squamous cell carcinoma histology.

All patients received treatment compliant with NCCN guidelines and with curative intent. Our data show that 70% of patients completed treatment for all modalities. Breaking down the different treatment groups shown in Table 1, the percentages of patients that completed the full treatment regimen are as follows: Surgery 100%, Radiation 75%, CCRT 68%, Surgery/Radiation 83%, and Surgery/CCRT 50%.

CK2α staining and association with patient characteristics

Tumors were stained for CK2α protein, and CK2α scoring was evaluated by semi-quantitative assessment of the relative antigen density in combined cytoplasm and nucleus of viable tumor cells. Reactivity was evaluated as score 3 (intense), score 2 (moderate), score 1 (weak), and score 0 (no staining). As expected, due to ubiquitous expression of CK2α in most cancer cells, no scores of 0 were observed for CK2α. Representative images of CK2α score 1, 2 and 3 with corresponding H&E and Ki-67 stains are depicted in Fig. 1. The percentage of patients with CK2α tumor score of 1, 2 and 3 were 23.5%, 50.4% and 26.1%, respectively. Age at diagnosis, alcohol use, T or N classification, or treatment modality did not differ significantly between CK2α score categories. The majority of stage I tumors had CK2α score of 1.

Evaluation of CK2α protein levels in tumors and correlations with HPV status and Ki-67 index

HPV(+) tumors, as defined by positive detection of E6/E7 RNA, represented 21.4% of CK2α score 1 tumors, 83.3% of CK2α score 2 tumors, and 90.3% of CK2α score 3 tumors. We examined whether CK2α expression levels and HPV positivity were associated. As shown in Table 2, CK2α score strongly corresponded with HPV(+) OPSCC (P < 0.0001).

Given the different mutational and HPV statuses of these malignancies, we were also interested in the proliferative state of these tumors and thus performed Ki-67 staining and analysis. We found that the tumors in this entire cohort were highly proliferative, with 72.3% of tumor cells more than 20% positive for Ki-67. In analysis of CK2α scores with the Ki-67 index in tumors, there was a positive relation of higher CK2α score with higher percentage of Ki-67 positivity (P < 0.0003; Table 2).

Survival analysis

PFS rates and OS rates among the CK2α scores were very similar, and we focused on analysis of OS. Univariate Cox regression analysis for OS was performed for the two independent variables of interest CK2α and HPV and the potential co-variates Ki-67, treatment, age, alcohol, smoking, TNM stage, and clinical classification (7th edition of the AJCC staging classification) to investigate their association with survival time of patients and potential to be entered into multivariate Cox regression (Table 3). The relationships CK2α (high score of 2 or 3 vs. low score of 1) and HPV status (independent variables of interest) and co-variate smoking (current vs. former or never) were all statistically significant. Kaplan–Meier analysis comparing the three CK2α scores in all patients showed that OS was significantly different with improved OS as CK2α score increased (Fig. 2A). Kaplan–Meier analysis by HPV status (positive vs. negative) in all patients demonstrated significantly better OS for HPV(+) OPSCC (Fig. 2B).

Because the co-variate smoking was significant in univariate analysis, we controlled for smoking in two separate multivariate Cox regressions for the predictors CK2α or HPV. CK2α level (high score of 2 or 3 vs. low score of 1) or HPV status (negative vs. positive) remained statistically significant after controlling for smoking (CK2α HR 0.461 [0.246, 0.867], P = 0.0162; HPV HR 2.567 [1.329, 4.959], P = 0.0050).

Survival analysis in TCGA HPV(−) HNSCC

Our group of HPV(−) OPSCC patients is limited, but previous studies suggest that high CK2 levels (α, α′, β) in the context of HPV(−) disease are a risk for worse outcomes (Chua et al., 2017; Faust et al., 1996; Gapany et al., 1995). Because the number of patients with HPV(−) disease was low in our cohort, we analyzed the publicly available TCGA HNSCC PanCancer RNA-seq data using cBioportal (Cerami et al., 2012; Gao et al., 2013). Disease-specific survival analysis was performed in the HPV(−) cohort comparing high CSNK2A1 (CK2α) mRNA levels (Z-score cut-off > 1.5, n = 117) to unaltered or low CSNK2A1 mRNA (n = 277). In HPV(−) HNSCC, high CSNK2A1 mRNA levels associated with significantly worse survival (logrank P = 0.0279; q = 0.0478; Fig. 3). PFS and OS were also significantly worse for high CSNK2A1 (logrank P = 0.0314 and P = 0.0359, respectively; q = 0.0478 for both PFS and OS). In the HPV(+) cohort, the subset of patients with high CSNK2A1 mRNA tumors consisted of only 7 tumors relative to 56 tumors with unaltered CSNK2A1; there was insufficient survival data in the CSNK2A1 high group for analysis.