Absent/weak CD44 intensity and positive human papillomavirus (HPV) status in oropharyngeal squamous cell carcinoma indicates a very high survival

Abstract Patients with human papillomavirus DNA positive (HPVDNA+) oropharyngeal squamous cell carcinoma (OSCC) have better clinical outcome than those with HPV DNA negative (HPVDNA−) OSCC upon intensive oncological treatment. All HPVDNA+ OSCC patients may not require intensive treatment, however, but before potentially deintensifying treatment, additional predictive markers are needed. Here, we examined HPV, p16INK4a, and CD44 in OSCC in correlation to clinical outcome. Pretreatment tumors from 290 OSCC patients, the majority not receiving chemotherapy, were analyzed for HPV DNA by Luminex and for p16INK4a and CD44 by immunohistochemistry. 225/290 (78%) tumors were HPVDNA+ and 211/290 (73%) overexpressed p16INK4a, which correlated to presence of HPV (P < 0.0001). Presence of HPV DNA, absent/weak CD44 intensity staining correlated to favorable 3-year disease-free survival (DFS) and overall survival (OS) by univariate and multivariate analysis, and likewise for p16INK4a by univariate analysis. Upon stratification for HPV, HPVDNA+ OSCC with absent/weak CD44 intensity presented the significantly best 3-year DFS and OS, with >95% 3-year DFS and OS. Furthermore, in HPVDNA+ OSCC, p16INK4a+ overexpression correlated to a favorable 3-year OS. In conclusion, patients with HPVDNA+ and absent/weak CD44 intensity OSCC presented the best survival and this marker combination could possibly be used for selecting patients for tailored deintensified treatment in prospective clinical trials. Absence of/weak CD44 or presence of human papillomavirus (HPV) DNA was shown as a favorable prognostic factors in tonsillar and tongue base cancer. Moreover, patients with the combination of absence of/weak CD44 and presence of HPV DNA presented a very favorable outcome. Therefore, we suggest that this marker combination could potentially be used to single out patients with a high survival that could benefit from a de-escalated oncological treatment.

Furthermore, patients with HPV DNA positive (HPV DNA +) OSCC have been reported to have a better 5-year overall survival (OS) compared with those with HPV DNA À OSCC (80% and 40%, respectively), the latter similar to that of other head-neck squamous cell carcinoma (HNSCC) patients [5][6][7].
Lately, as a consequence of the low survival in HNSCC, oncologic treatment has been intensified with chemoradiotherapy and epidermal growth factor receptor (EGFR) inhibitors [7]. Many patients with HPV DNA + OSCC may not benefit from this intensified treatment, and could potentially be cured by radiotherapy (RT) alone, with possibly less severe sequele. To better identify patients with a favorable prognosis, before potentially reducing treatment, additional predictive markers are needed [5].
Expression of CD44, a cell adhesion glycoprotein participating in epithelial cell-stroma interactions and important for tumor invasion and metastasis [8], has previously been described as a prognostic marker in many cancers [9,10]. Moreover, high CD44 expression has been correlated to worse prognosis in HNSCC [11][12][13][14], and in a pilot study from the rural county of Dalarna, Sweden, we found that medium/strong CD44 intensity staining was a negative prognostic factor in tonsillar and base of tongue cancer [11]. However, we could not correlate this finding to HPV status due to limited numbers of patients.
Furthermore, in the present literature there are different standard procedures for defining HPV status. HPV status can, for example, be defined as presence of HPV DNA alone, or HPV DNA together with overexpression of p16 INK4a . However, overexpression of p16 INK4a alone has also been used as a surrogate marker of functionally active HPV [12,13].
Here, in this larger Stockholm, Sweden cohort of HPV+ and HPVÀ OSCC, where 80% of the patients did not receive chemotherapy, we have evaluated CD44 intensity staining and p16 INK4a in relationship to HPV status and in relation to OS and DFS.

Patients, Materials, and Methods
Patients 2000-2007, 385 patients were diagnosed with TSCC (ICD-10 C09.0-9) and BOTSCC (ICD-10 C01.9) in the county of Stockholm, and of these 290 with available pretreatment biopsies and treated with intention to cure were included in the study. For most patients (n = 229, 79%) treatment consisted of conventional RT (2.0 Gy/day, for 6.5-7 weeks, total dose: 68 Gy) or accelerated RT (1.1 + 2.0 Gy/day for 4.5 weeks, total dose: 68 Gy), while a minority (n = 61, 21%) also had induction chemotherapy followed by concomitant RT. Furthermore, some patients also received interstitial radiation (brachytherapy) (total dose of 78 Gy). Finally, patients with nodal disease also underwent neck dissection, 6-8 weeks after completed RT. Patients were thereafter followed up by clinical examination every 3 months the first 2 years, and every 6 months the third year.

HPV DNA analysis
DNA was extracted from 30 lm paraffin-embedded tonsillar tumor biopsies as described previously [2]. Blank control samples were treated in the same way to exclude cross-contamination between samples. Presence of HPV DNA and type and betaglobin was analyzed by the Luminex method [14].

Immunohistochemistry
In brief, tumor sections (4-5 lm) were deparaffinized, rehydrated, rinsed in water, followed by antigen retrieval in citrate buffer (pH 6) for 20 min. The slides were then left for 10 min in 0.5% H 2 O 2 in water and washed in phosphate buffered saline (PBS). Blocking was done with 1% horse serum in PBS in a moist chamber for 40 min before the sections were stained with the primary antibody (mAb CD44 [clone: DF1485, dilution 1:100, Dako, Glostrup, Denmark] and mAb p16 INKA4a [clone: JC8, dilution 1:100, Santa Cruz Biotech, Dallas, TX]) at +8°C over night. The avidin-biotin-peroxidase complex (ABC) kit (Vectastain, Vector Laboratories, Burlingame, CA) was used for antigen detection according to the manufacturer. Slides were developed in chromogen 3'-diaminobenzydine (DAB) (Vector Laboratories) with hematoxylin as a counter stain.

Evaluation of immunohistochemistry staining
The fraction of CD44 positive cells was evaluated semiquantitatively in four grades of percentages of stained malignant cells: 0 (0%), 1 (1-25%), 2 (26-75%), or 3 (76-100%) and the intensity of the staining was scored separately and evaluated as absent, weak, moderate, and strong staining [11]. The fraction and intensity of p16 INK4a positive cells was evaluated and samples with strong p16 INK4a staining in >70% positive cells were considered as p16 INK4a positive (p16 INK4a +) [15]. All evaluations were conducted by two independent researchers (A. N. and C. N.) blinded for clinical data and outcome.

Statistical analyses
Patient characteristics were analyzed with the Chi 2 -test and independent T-test when appropriate. Two-sided P-values were reported for all analyses.
Disease-free survival (DFS) was defined from the date of diagnosis to the date of the last known occasion that the patient was disease free, or the date of disease recurrence (local, regional, or distant recurrence). Death without documented recurrence was censored at the date of death. OS was defined as time from the date of diagnosis until the date of death of any reason. Cumulative survival was calculated and presented with the Kaplan-Meier method and analyzed with the log-rank test. Univariate and multivariate Cox proportional models were used to calculate the Hazard ratio (HR). All above described analyses were performed in SPSS (IBM SPSS Statistics, version 20, Stockholm, Sweden).
CD44 expression without or with HPV DNA status and clinical parameters CD44 expression was evaluated both by staining intensity and by the fraction of positive cells (Fig. S1). In all, 268/ 291 (92%) of the tumors expressed CD44, and the majority had a strong intensity staining (53%), while 26% and 21%, respectively, stained intermediately or weakly (Table 1). Patients with CD44+ tumors (defined by dichotomization at intensity >weak or by >26% positive cells), presented significantly more often with larger, higher differentiated tumors, but no significant differences were observed in tumor localization (tonsil or tongue base), stage, or age between patients with CD44+ and CD44À tumors (data not shown).
However, if CD44 expression was grouped as absent or present, no significant difference was observed between HPV DNA + and HPV DNA À tumors, where 205/225, 91% of the HPV+ and 63/65, 97% of the HPV DNA À tumors expressed CD44 ( Table 1).
Additionally, age was divided in percentiles and the effect on outcome was analyzed. A lower age was significantly correlated to a better DFS and OS in the univariate analysis (Table 2).
A multivariate including HPV status, CD44 intensity staining, age, sex, stage, and tumor site was performed, but here p16 INK4a was excluded due to the high correlation between HPV DNA and p16 INK4a overexpression. Patients with HPV DNA + tumors had a favorable DFS and OS (HR 0.31, 95% CI: 0.16-0.62, P = 0.001 and 0.27, 95% CI: 0.16-0.44, P < 0.001, respectively) ( Table 2). Similar trends in DFS and OS were observed for patients with tumors expressing CD44 when assaying for staining intensity (CD44 absent/weak intensity vs.   Chi-square test. 3 Independent T-test.   Table 2). Finally, a lower age was also significantly correlated to a better DFS and OS in the multivariate analysis.
Effects on survival combining HPV DNA /CD44 or HPV DNA /p16INK4a, as well as HPV DNA /age Strong evidence suggests that "HPV positive" OSCC and "HPV negative" OSCC are different disease entities with different characteristics and should be separated when analyzed. Hence, the cohort was divided into an HPV DNA + and HPV DNA À cohort and the influence of CD44 and p16 INK4a as prognostic factors was also analyzed (Fig. 2, and Table 3A and B). As shown in Figure 2A and B, patients with HPV DNA + tumors and an absent/weak CD44 intensity had a significantly better DFS (96% vs. 86%) and OS (95% vs. 80%) compared with patients with HPV DNA + and medium/strong CD44 intensity tumor expression (P = 0.034 and P = 0.060, respectively). This was the case irrespective if the patients were treated with RT alone or induction chemotherapy followed by RT (data not shown). In the HPV DNA À cohort absent/weak CD44 as compared to medium/strong CD44 intensity staining showed a similar tendency with 75% versus 63% DFS and 67% versus 47% OS, respectively, but these differences were not statistically significant ( Fig. 2C and D).
Similarly, when analyzed separately for HPV DNA + and HPV DNA À tumors, in the multivariate analyses adjusted for sex, stage, age, and tumor site, CD44 expression was still correlated to a favorable DFS and OS for the HPV DNA + group, but not the HPV DNA À group (Table 3A and B).
Only a minority of patients with HPV DNA + tumors were p16 INK4a negative and vice versa. When DFS and OS were analyzed in the HPV DNA + and HPV DNA À groups, a statistically significant difference was observed only in OS between patients with HPV DNA +/p16 INK4a + and patients with HPV DNA +/p16 INK4a À tumors (univariate DFS: 0.59 and OS: P < 0.0001, respectively) (Table 3A). Likewise, when p16 INK4a overexpression was analyzed separately for HPV DNA + and HPV DNA À tumors in the multivariate analyses adjusted for sex, stage, age, and tumor site, there was only a significant correlation to OS in the HPV DNA + group and not in the HPV DNA À group (multivariate DFS: P = 0.28 and OS: P < 0.0001, respectively) (Table 3A and B).
Finally, a lower age upon diagnosis correlated significantly to a more favorable DFS and OS in the HPV DNA + group both in the univariate and the multivariate analysis (Table 3A). A similar tendency was observed in the HPV DNA À patient group, but did not reach statistical significance (Table 3B).
Effects on survival combining HPV DNA with p16 INK4a expression and CD44 expression As described in the introduction, combining HPV DNA and p16 INK4a expression as criteria of active HPV infection was proposed. Hence, CD44 intensity staining was examined separately also for patients with HPV DNA +/ p16 INK4a + and patients with HPV DNA À/p16 INK4a À tumors ( Fig. 3 and Tables S1 and S2).
Notably, no additional increase in DFS or OS could be observed in the absent/weak CD44 group when subtracting patients with HPV DNA +/p16 INK4a À tumors (Fig. 3A and B) and DFS and OS were similar to that observed for patients with absent/weak intensity staining in Figure 2. However, the cumulative survival rate increased in the medium/ strong intensity group in Figure 3A and B, most probably due to the exclusion of all patients with HPV DNA +/ p16 INK4a À tumors. Hence, the observed survival difference between absent/weak and medium/strong CD44 intensity staining among HPV DNA +/p16 INK4a + tumors in Figure 3A and B did not reach statistical significancealthough absent/weak CD44 expression in HPV DNA +/p16 INK4a + OSCC patients showed >95% DFS and OS.

Discussion
In 290 OSCC patients, we show that those with combined HPV DNA + OSCC and absent/weak CD44 intensity staining presented the best 3-year DFS and OS, with >95% of the patients surviving >3 years after treatment. In addition, HPV DNA + status and absent/weak CD44 intensity staining and p16 INK4a + overexpression were also found as independent favorable prognostic markers in OSCC. However, while absent/weak CD44 intensity staining was a positive prognostic marker for both DFS and OS in HPV DNA + OSCC patients, p16 INK4a overexpression was only a marker of a favorable OS in these patients.
As mentioned above and shown previously, roughly 80% of all patients with HPV DNA + OSCC may not need the intensified oncological treatment regimes used today [5][6][7]. Nonetheless, additional treatment regimes have already been implemented in many clinics and better stratification markers are therefore needed before a tailored and more deintensified treatment can be introduced. Here, patients with the novel combination of CD44 expression and HPV DNA status showed 95% DFS and 95% OS despite that the majority of the patients were treated only with RT.
In contrast to absent/weak CD44 intensity staining, medium/high CD44 intensity staining correlated to worse prognosis for OSCC in general and for HPV DNA + OSCC. These results are in line with previous reports in HNSCC and other malignancies [9,10,[16][17][18][19][20]. Furthermore, it was proposed that CD44 characterize cancer stem cells in HNSCC [21]. In accordance with this, Chen and colleagues demonstrated that CD44-expressing cells displayed cancer stem like properties and had higher RT-resistance in HNSCC [22]. Moreover, CD44-expressing cells in HNSCC were demonstrated to have an increased metastatic potential and increased proliferation index [18]. Finally, CD44 signaling has also been reported to increase resistance to chemotherapy in HNSCC [23], which together with previously published data and our study all imply CD44 as a negative prognostic factor [8][9][10][16][17][18][19].
In our large OSCC cohort, both HPV DNA status and p16 INK4a expression correlated independently to a favorable prognosis and overexpression of p16 INK4a was significantly correlated to presence of HPV DNA , consistent with many previous reports [15,[24][25][26][27][28]. Nonetheless, to our knowledge, the significance of p16 INK4a overexpression has not been described before separately for HPV DNA + and HPV DNA À OSCC. Here, we show that p16 INK4a overexpression only influences OS in the HPV DNA + cohort. However, whether the correlation between the absence of p16 INK4a expression and a poorer OS in HPV DNA + OSCC is due to the high sensitivity, and to a possibly lower specificity of our HPV DNA detection method, or to actual absence of p16 INK4a expression in truly HPV DNA + tumors remains to be elucidated. In patients, where HPV positive status was defined as HPV DNA + and p16 INK4a +, CD44 absent/weak staining intensity resulted in a DFS and OS similar to that obtained for patients with HPV DNA + tumors with absent/ weak CD44 intensity staining. Furthermore a significant difference between absent/weak and medium high CD44 intensity staining in the HPV DNA + and p16 INK4a + patient group was still obtained for DFS, but not for OS. The latter could, however, partly be due to that patients with the poorest outcome (HPV DNA + and p16 INK4a À) were excluded.
Finally, similar to other studies, a lower age at diagnosis correlated to a favorable DFS and OS, in the whole cohort as well as in patients with HPV DNA + OSCC, while a lower stage correlated to favorable OS in the HPV DNA À OSCC patient group [29,30].
Taken together the data suggest that absent/weak CD44 staining in patients with HPV+ OSCC is a strong positive indicator for better clinical outcome irrespective of treatment. Previously, we demonstrated that absent major histocompatibility complex (MHC) class I staining or a high number of CD8 tumor-infiltrating T-lymphocytes were strong prognostic indicators for better clinical outcome in HPV+ OSCC [30,31]. It is possible that combining CD44 with these markers may be of even greater benefit for future selection of patients with a favorable outcome.
There are limitations in our study. First, the study was retrospective and the number of patients was limited. Furthermore, treatment was not standardized to a study protocol and the whole patient cohort was not randomized into different treatment arms. Secondly, biopsies were not available for research from all patients; however, we consider the loss of biopsies random as there was, to our knowledge, no systematic loss of specific biopsies. Thirdly, most patients were randomized into two different RT protocols (conventional/hyperfractionated). Nevertheless, improved survival effect was not observed in any of the treatment arms in a study including these patients [32]. Finally, here we have only included TSCC and BOTSCC as OSCC and not all other OSCC, but this can also be a benefit as we excluded other OSCC sites where the correlation to HPV and prognosis is more ambiguous [33].
In summary, patients with OSCC with HPV DNA + and absent/weak CD44 intensity staining presented a very high DFS and OS and could potentially be selected as candidates for tailored deintensified treatment. However, our data should be confirmed in a prospective multicentre randomized clinical trial with a larger patient sample, and with other biomarkers, before applied clinically.