Co-expression of SOX2 and HR-HPV RISH predicts poor prognosis in small cell neuroendocrine carcinoma of the uterine cervix

Background Small cell neuroendocrine carcinoma of the uterine cervix (SCNEC) is a rare cancer involving the human papilloma virus (HPV), and has few available treatments. The present work aimed to assess the feasibility of SOX2 and HPV statuses as predictive indicators of SCNEC prognosis. Methods The associations of SOX2 and/or high-risk (HR)-HPV RNA in situ hybridization (RISH) levels with clinicopathological characteristics and prognostic outcomes for 88 neuroendocrine carcinoma (NEC) cases were analyzed. Results Among these patients with SCNEC, SOX2, P16INK4A and HR-HPV RISH expression and SOX2/HR-HPV RISH co-expression were detected in 68(77.3%), 76(86.4%), 73(83.0%), and 48(54.5%), respectively. SOX2-positive and HR-HPV RISH-positive SCNEC cases were associated with poorer overall survival (OS, P = 0.0170, P = 0.0451) and disease-free survival (DFS, P = 0.0334, P = 0.0309) compared with those expressing low SOX2 and negative HR-HPV RISH. Alternatively, univariate analysis revealed that SOX2 and HR-HPV RISH expression, either separately or in combination, predicted the poor prognosis of SCNEC patients. Multivariate analysis revealed that the co-expression of SOX2 with HR-HPV RISH may be an independent factor of OS [hazard ratio = 3.597; 95% confidence interval (CI): 1.085–11.928; P = 0.036] and DFS [hazard ratio = 2.880; 95% CI: 1.199–6.919; P = 0.018] prediction in SCNEC. Conclusions Overall, the results of the present study suggest that the co-expression of SOX2 with HR-HPV RISH in SCNEC may represent a specific subgroup exhibiting remarkably poorer prognostic outcomes compared with the expression of any one marker alone. Supplementary Information The online version contains supplementary material available at 10.1186/s12885-021-08059-1.


Introduction
Small cell neuroendocrine carcinoma of the uterine cervix (SCNEC) is a highly aggressive and rare malignant cervical cancer (< 3%). The incidence of lymph node and distant metastases is high in the early stage of SCNEC, although SCNEC is usually only detected at the advanced stage [1][2][3]. SCNEC has a poor prognosis, which is closely related to the stage at diagnosis. Particularly, the 5-year survival rate for early SCNEC cases is 30-46%, while it is only 0-15% for patients at the advanced stage [4]. Despite the increase in multidisciplinary therapies, patients with advanced SCNEC still have a poor prognosis [5]. Therefore, it is important to improve SCNEC prognosis.
The sex-determining region Y-box 2 (SOX2) gene, located on chromosome 3q26.3-q27, belongs to the SOX family [6]. Notably, SOX2 has been recognized as a potent transcription factor involved in self-renewal, maintenance of stem cell properties, and pluripotency in embryonic stem cells [6,7]. SOX2 plays a vital role in tumor development, progression, and cell survival in various cancer types [8][9][10]. A few studies have reported that SOX2 is overexpressed in cervical squamous cell carcinoma (SCC), and plays an important role in the progression from squamous dysplasia to SCC. The expression of SOX2 is correlated with the degree of differentiation of SCC, and upregulation of SOX2 has been shown to enhance cervical cancer cell invasion and migration in vitro [11,12]. The small infiltrating cancer nests surrounding CIN 3 margins or the CIN 3-like SCC with deep invasion generally display a decreased SOX2 level locally, and this indicates reduced SOX2 expression during invasive growth [13]. SOX2 has been reported to be related to HPV infection in previous studies [14,15]. However, its expression, clinical significance and the association between SOX2 and HPV status in SCNEC have not been evaluated.
Numerous studies have demonstrated a closed etiopathogenetic relationship between the development of cervical cancers and high-risk (HR) human papilloma virus (HPV) infection [16]. The occurrence of SCC and adenocarcinoma is associated with HPV16 infection, while that of SCNEC is associated with HPV18 infection [17]. HPV infection is detected using a variety of approaches, such as polymerase chain reaction (PCR), immunohistochemistry (IHC), and in situ hybridization (ISH) [18,19]. Previous studies have shown that HPV mRNA detection and P16 INK4A /Ki67 IHC are valuable biomarkers for HPV oncogenic expression [20]. The detection of mRNA expression indicates changes at the molecular level, and mRNA amplification becomes a poor prognostic factor when persistently infected with a highly oncogenic type, such as HPV 18 [21]. Furthermore, HPV mRNA expression, detected using HR-HPV RISH, in SCNEC has not been investigated extensively.
The present study retrospectively examined the expression levels of SOX2 and HPV mRNA in SCNEC and investigated the relationships between the expression levels and clinicopathological characteristics in SCNEC cases.

Patients and samples
In the present retrospective study, we enrolled 88 patients with histologically confirmed SCNEC, who had under surgical resection at Sun Yat-sen University Cancer Center between January 2010 and December 2014. Patients were enrolled when they were diagnosed with primary SCNEC, with available clinical information. The last follow-up was conducted in June 2020. The study protocol was approved by the Institutional Ethical Board of Sun Yat-sen University Cancer Center. The raw data relevant to the study were imported into the Research Data Deposit public platform (www.researchdata.org.cn; RDD approval number: RDDA2020001710).

Statistical analysis
The SPSS 19.0 software was used for data analyses. The expression of SOX2 in different SCNEC subgroups was compared using an unpaired T-test. Correlations of SOX2 and P16 INK4A with the HPV mRNA expression levels and clinicopathological parameters in patients with SCNEC were analyzed using the chi-square test. Kaplan-Meier analysis was used to analyze overall survival (OS) and disease-free survival (DFS), while the logrank test was used for comparison. The correlation of prognosis was analyzed using univariate and multivariate Cox regression analyses. P < 0.05 (two-sided) indicated statistical significance. To construct a nomogram, the predictive power of each variable for OS and DFS was evaluated using univariate Cox regression. Thereafter, the significant variables were used in multivariate Cox analysis. To determine independent prognostic variables, the Akaike information criterion (AIC) score was used for backward selection for suitable variables. Finally, the variables were enrolled for the nomogram construction, with 1-, 3-, and 5-year OS and DFS selected as the primary endpoints. To evaluate the nomogram predicting power, a concordance index (C-index) with receiver operating characteristic (ROC) curve analysis was applied [25]. The discrimination of the predicted values from the actual values was visualized by generating calibration curves for 1-and 3-year OS data [26].

Patient characteristics
The characteristics of 88 patients with SCNEC are summarized in Table S1. Sixty-six patients (SCNEC-alone, 75%) presented with only small cell carcinoma components, and Twenty-two patients (SCNEC-mix, 25%) presented with small cell carcinoma mixed with other epithelium-derived tumors (SCC, adenocarcinoma, and others) were noted. There were 79 (89.8%) and 9 (10.2%) cases at FIGO stages I-IIA and IIB-IV, respectively, and a median 30.6-month follow-up was conducted to examine OS and DFS.

Expression of HPV mRNA in SCNEC was detected using the RNAscope technique
In situ expression of HPV mRNA was detected using RNAscope and scored accordingly. HPV mRNA was detected predominantly in the cytoplasm of cancer cells with variable staining intensity. The scores of 0 to 4 and the proportions of T1-4 are shown in Fig. 1c and d. The HPV mRNA-positive rate was 86.4% (76/88) in patients with SCNEC, and the expression rates in SCNEC-alone and SCNEC-mix were 81.8% (54/66) and 100% (22/22), respectively ( Fig. S1A and S1C). T-test analyses indicated (P = 0.2118) no significant difference between the groups (Fig. S1C). Another TMA contained 37 SCNEC cases with lymph node metastases. There was no significant difference in the HPV mRNA expression between primary tumor and lymph node metastases (n = 37; P = 0.1134) ( Fig. S1D and S1E). Furthermore, the The association of SOX2, P16 INK4A , and HPV mRNA expression with the clinicopathological characteristics in SCNEC is summarized in Table 1. SOX2 expression within tumors was significantly related to vascular invasion (P = 0.023) and relapse (P = 0.023), whereas SOX2 expression showed no significant correlation with other clinicopathological features of patients with SCNEC. However, P16 INK4A expression was only significantly related to Ki67 (P = 0.005). Notably, HPV mRNA levels were significantly correlated with FIGO staging (P = 0.021), pre-operative chemotherapy (P = 0.007), relapse (P = 0.027), neuroendocrine markers (P = 0.021), and pathological classification (P = 0.014). These data suggest that the overexpression of SOX2 and HPV mRNA potentially facilitates tumorigenesis and development of SCNEC.
SOX2 and HPV mRNA influenced the prognosis of patients with SCNEC Tables 2 and 3 (Fig. 2). In conformance to the above results, Kaplan-Meier curves for OS and DFS based on SOX2, P16 INK4A , HR-HPV RISH, SOX2/P16 INK4A , and SOX2 /HR-HPV RISH expression showed significant differences, which were verified through log-rank tests (Fig. 3). Survival analysis was also conducted, which revealed that SOX2 may be adopted to predict prognosis. Moreover, the statistical analyses indicated that SOX2 and HPV mRNA expression were associated with a series of pathological parameters related to OS (Figs. 4 and 5). Therefore, our data further suggest that SOX2, alone and in combination with HPV mRNA, is an independent prognostic marker for patients with SCNEC.

Prognostic nomograms were created to predict OS and DFS
Variables obtained based on Cox proportional analysis were then applied to build the respective OS and DFS prognostic nomograms (Fig. 6). The factors that were incorporated into the OS nomogram included stromal, parametrium, and nerve invasion, SOX2, HR-HPV RISH, and SOX2/HR-HPV RISH. Additionally, four risk factors-nerve invasion, SOX2, HR-HPV RISH, and SOX2/ HR-HPV RISH-were enrolled in the DFS nomogram. One point was assigned to the prognostic factor in the as-constructed nomograms. Then, the total points were summed up to determine outcome probability by plotting a perpendicular line to the axis of "1-, 3-, and 5-year OS/ DFS probabilities". Figure 7 shows the calibration plots used to predict OS and DFS at 1-, 3-, and 5-year intervals, which reveal the accurate predictive power.

Discussion
The present work examined the significance of SOX2, HR-HPV RISH, and clinicopathological features in SCNEC cases. The respective nomograms were constructed according to Cox hazards analysis to predict OS and DFS for SCNEC cases. Thereafter, each point was assessed for prognostic risk, and individualized posttreatment was provided. To our knowledge, the present work is the first retrospective analysis of the value of    SOX2 and HR-HPV RISH in predicting the prognosis for SCNEC. The dysregulated OCT4/SOX2 complex has been detected in various human malignant tumors [27], and thus, SOX2 plays a critical role in cancer development [7]. Overexpression of SOX2 has been detected in human cancers, and therefore, it may serve as an oncogene [28]. Additionally, previous studies have examined the effect of SOX2 levels on small cell neuroendocrine carcinomas (NECs) in certain organs, and found that SOX2 possibly plays a vital role in small cell NEC progression in the endometrium, esophagus, and lung [29][30][31]. Nonetheless, little research has focused on SOX2 expression and its clinical value in cancer. In the present study, SOX2 independently predicted the poor prognosis of SCNEC, similar to the results of prior studied [32]. Hence, SOX2 plays an important role in SCNEC development.
The role of HPV in the etiology of SCNEC is well established, and HR-HPV can be detected in the majority of the patients [33]. The presence of P16 INK4a / Ki-67 can serve as a candidate marker for HR-HPV infection in HPV-associated endocervical neoplasia [34,35]. However, the scoring system of 16 INK4a is currently controversial, often leading to a misinterpretation of the staining results [36], and the diagnostic value of Ki-67 in SCNEC remains ambiguous. HR-HPV RISH is a robust technique for HR-HPV diagnosis [37,38] and detects the full-length or fragments of E6 and E7 transcripts using cascade signal amplification [38]. Studies have shown that persistent infection with HR-HPVs results in integration of the viral genome fragments into the host chromosomes, thus facilitating the transcription of typespecific E6/7 genes and protein overexpression, which eventually leads to the activation of the downstream carcinogenetic signaling pathways [37]. Recent studies have shown that HR-HPV RISH effectively diagnoses endocervical adenocarcinoma and endocervical glandular neoplasia [39,40]. Therefore, a high specificity of HR-HPV RISH for HPV-driven cervical neoplasia is expected. In our study, HR-HPV RISH showed higher sensitivity and specificity for SCNEC, compared to P16 INK4a and Ki-67 IHC. Multivariate analysis demonstrated that SOX2/HR-HPV RISH co-expression served as an independent factor in predicting the OS and DFS in SCNEC cases. Further studies, using larger cohorts, should be conducted to validate our findings. Consistent with our results, SOX2 was proven to be a potential marker to predict overall survival and recurrence in p16+ oropharyngeal cancer [41]. Recent studies have shown that SOX2 was related to HPV infection. Interestingly, HPV infection drives switches in SOX2 expression in the transformation zone in the uterine cervix [15], and SOX2 locus amplification was related with HPV mRNA positivity in vulvar carcinoma [14]. Furthermore, SOX2 was reported to be regulator of HPV16 at the transcriptional level in cervical squamous cell carcinoma [42]. That may explain the possible molecular mechanism between them. In our study, both SOX2 and HR-HPV RISH were independent prognostic factors for SCNEC. Unfortunately, there is no significant difference between the expression of SOX2 and HR-HPV RISH in the correlation analysis, which may also be related to the sample size. Therefore, the possible molecular mechanism between  SOX2 and HPV infection in SCNEC remains to be further studied. There were several limitations to this study. First, its retrospective nature may lead to inevitable selection bias. Second, this study was conducted with a small sample size from a single center. Third, the present work only focused on the significance of SOX2 and HR-HPV RISH in predicting prognosis, while other prognostic factors, such as molecular biomarkers or inflammatory prognostic markers, were not included. Therefore, the results of this work should be further validated in multi-center studies with a larger sample size.