Background
Cancer of the uterine cervix is the third most common gynecologic cancer in the United States, with approximately 13,000 new cases and 4120 cancer deaths estimated to occur in 2016 [
2]. Persistent human papillomavirus (HPV) infection, particularly with HPV16, HPV18, and other high risk types, plays a key role in the development of cervical cancer. While screening measures for cervical dysplasia have decreased the incidence of cervical cancer in the United States, cervical cancer remains a major world health problem for women [
3,
4]. Depending on the disease stage, treatment for cervical cancer may involve a combination of hysterectomy, pelvic lymph node dissection, and chemoradiation. Targeted therapies using small molecules or monoclonal antibodies are largely still in clinical trial stages [
4].
PD-1 is an immune suppressive molecule in the B7-CD28 family that regulates T-cell activation [
5]. PD-L1 is a transmembrane protein that can be expressed on tumor cells in the cancerous microenvironment [
6]. PD-L1 has been hypothesized to bind its receptor PD-1 on T-cells to downregulate anti-tumor T-cell activity and facilitate immune evasion [
7]. Expression of PD-L1 has also been found to be associated with worse survival in solid tumors, including esophageal, gastric, colorectal cancers, and pulmonary adenocarcinoma [
8,
9].
The PD-1/PD-L1 axis has emerged as a promising new target for immune checkpoint therapies. Lasting responses have been seen with anti-PD-L1 antibodies in the treatment of metastatic renal and lung carcinomas, as well as melanomas [
7]. Currently, there are also ongoing clinical phase I/II trials evaluating the effects of anti-PD-1 therapies, including pembrolizumab (NCT02054806) and nivolumab (NCT02488759), against advanced cervical cancer, though no study results have been reported [
10]. Tumor surface expression of PD-L1 has been found to correlate with objective responses to anti-PD-L1 immunotherapies, and provide a rationale for screening tumor samples to identify candidates for these targeted therapies [
11]. As data is still lacking on the expression patterns of PD-L1 in benign and malignant cervical tissues, we investigated the expression of PD-L1 in human cervical tissue and cervical tumors on tissue microarrays. Our findings provide rationale for further investigation of PD-L1 immunotherapies in the treatment of cervical cancer. Since, currently there is no unified scoring system for PD-L1 expression by immunohistochemistry as opposed to Her2/neu [
12], we are introducing a scoring algorithm which is based on the work published by Garon et al. [
13].
Discussions
This study was carried out on a TMA which is made of small size samples. Due to heterogeneity of the PD-L1 reactions, some of the negative samples may have been randomly selected from the “negative” portion of otherwise “positive” tumors. The small size of the TMA samples, however, may mimic the size of the needle biopsy specimens obtained in advanced cancer cases which may have similar PD-L1 reaction outcome. The results of the current study provide a baseline information for the future investigations.
Since there is no unified scoring system for the PD-L1 reactions by IHC, the scoring introduced in this study is based on the Garon et al. NSCLC study [
13]. using the antibody clone 22c3 which is used in the Dako platform [
21]. In this platform, the cutoff point is set at 50% for the positive cells regardless of the intensity of the reaction. The Ventana platform, based on clone SP142, uses 5% cutoff for positivity, again regardless of the intensity [
19]. Food and Drug Administration (FDA) has approved both platforms. Other investigators have used different scoring schemes. In a recently published study, the investigators have used different cutoff points for different immunotherapeutic agents, regardless of the intensity, as listed in the supplemental eTable 3 embedded with their publication [
22]. This pathology scoring system faces an issue. As new therapeutic agents are introduced and/or future clinical studies result in changes of the response rates, the scoring system needs to be accordingly adjusted. Therefore, a pathology scoring system is needed for a comprehensive and consistent evaluation of the PD-L1 reactions. Thus, it becomes incumbent upon the clinical oncologist to adjust their protocols as new therapeutic agents are introduced or therapeutic outcomes are changed. The pathologists, however, may amend their scoring reports with comments to include the current rates of the positivity in relation to responses to various immunotherapeutic agents. Although, the 50% cutoff for positivity is employed in this study, the variable rates are also incorporated as “Low-Positive” if less than 50% positive cells are encountered. In this study, we have treated the “Low-Positive” samples as “Negative” for statistical calculations. In addition, the intensity of the IHC reactions is retained as a component of the scoring system for future evaluations.
At the outset, PD-L1 is “positively” expressed in more than 34% of the uterine cervical carcinomas based on the proposed scoring. Its highest expression appears in squamous cell carcinomas although it is significantly lower in the high grade (grade-III) SCCs (Table
6). While the percentages of the “Positive” PD-L1 expression in adenosquamous carcinomas (~29%) and endocervical carcinomas (~17%) are lower than SCCs (~38%), the differences are not statistically significant (Table
6). In other words, PD-L1 is highly expressed in the uterine cervical carcinomas except for the high grade SCC. In all, more than 34% of the patients with uterine cervical cancers may become clinically eligible for the immunotherapy. In addition, about 39% of the patients, in “LoPos, 1a” and “LoPos, 1b” categories may also be considered for the therapy, though a smaller percentage of such cases may respond to the targeted therapy as it has been noted in the NSCLC clinical trial [
13].
Therapeutic targeting of the PD-1/PD-L1 immune regulatory axis has led to meaningful results in the treatment of many solid tumors, including melanoma, renal cell carcinoma, NSCLC, and head & neck squamous cell carcinoma [
7]. For the treatment of advanced cervical cancer, there are also ongoing clinical phase I/II trials evaluating the effects of anti-PD-1 therapies, including pembrolizumab (NCT02054806) and nivolumab (NCT02488759), although study results have not yet been reported [
10]. While data on the use of PD-L1 tumor expression as a screening marker to identify patients for anti-PD-L1 immunotherapies have been conflicting, several studies have found positive correlations between tumor expression of PD-L1 and response to the targeted therapies [
11,
23].
Though the use of anti-PD-1/PD-L1 immunotherapies has not been extensively explored for the treatment of cervical cancer, recent studies have evaluated biologic rationales for PD-L1 expression in the cervical tumors, including its pathogenesis through human papillomavirus (HPV), its tumor microenvironment composed of tumor-infiltrating lymphocytes, as well as its genetic basis for increased expression.
Human papilloma virus (HPV) plays a key role in the development of the two most common histologic subtypes of cervical cancer, SCC and endocervical adenocarcinoma. In particular, HPV16 has been linked to 59 and 36%, and HPV18 to 13 and 37%, of SCC and the adenocarcinoma, respectively [
24]. The two histologic subtypes have also been found to demonstrate distinct molecular profiles, though current treatment guidelines are generally not type-specific [
25]. Some studies have speculated that HPV may activate PD-1/PD-L1 immunosuppression to evade host immune responses against the virus, resulting in persistence and recurrence of cervical intraepithelial neoplasia (CIN) [
26]. A recent study by Mezache et al. examined PD-L1 expression in CIN and cervical SCC, and found PD-L1 expression to be strongly associated with HPV infection. In their study, PD-L1 was also upregulated in both the carcinoma and the surrounding inflammatory tissue cells mostly CD8+ lymphocytes [
27].
While the current study has evaluated PD-L1 expression of the tumor cells, it is unclear whether PD-L1 expression in tumor-infiltrating inflammatory cells also plays a relevant role in predicting response to anti-PD-L1 therapies [
23]. Indeed, some studies have suggested that PD-1/PD-L1 expression in the tumor microenvironment may be important for therapeutic activity [
5,
23]. The tumor microenvironment of cervical cancers has been well-studied, with some studies demonstrating the presence of tumor-infiltrating lymphocytes (TILs), and particularly CD8+ T-cells, to be strongly associated with improved clinical outcome and prognosis [
28,
29]. A study by Karim et al. found PD-L1 to be expressed in only 19% of cervical cancers, whereas more than half of tumor-infiltrating CD8+ T-cells were positive for PD-1 [
30]. Differences in expression from their study may be a result of using a different antibody clone (clone 5H1). Yang et al. also evaluated PD-1 and PD-L1 expression on cervical T-cells and dendritic cells, respectively, and found their upregulation to be associated with high risk-HPV positivity and increasing CIN grade [
31]. Together, these findings suggest a strong immunopathogenic rationale for PD-L1 expression in cervical carcinomas.
Several studies have explored a genetic basis for PD-L1 expression in various cancers, as well as its correlation with clinical outcome. In particular,
PD-L1 amplification and deletion has been found to be associated with poor overall survival amongst many major cancer types [
32]. Overexpression of PD-L1 by immunohistochemical methods has also been found to be associated with worse survival in solid tumors, including esophageal cancer, gastric cancer, colorectal cancer, and pulmonary adenocarcinoma [
8,
9]. Ironically, a higher percentage of the reaction has occurred in the lower grade carcinomas in this study. This finding may be due to a lower sample size of grade III SCCs in this series and/or the scoring which has excluded the “Low-Positive” tumors as “Positive” (Table
3). Besides, clone SP142 yields a lower number of positive cells by IHC [
22], however, it is not clear which clone is most relevant to the clinical outcome of cervical carcinomas. Further controlled studies are required to elucidate the clone-outcome relationships in uterine cervical cancers. In a study regarding uterine cervical cancers, the investigators have lumped adencarcinomas and SCCs as a sinle group with no tumor grading distiction [
10]. There is, however, a general agreement that a higher expression of PD-L1 is associated with a poorer prognosis in uterine cervix, endometrium, and head & neck caners [
10,
33‐
35].
As recent trials have suggested tumors, with a genetic basis for PD-1 ligand expression, are particularly sensitive to anti-PD-1 immunotherapies. Howitt et al. have investigated the status of genes encoding PD-1 ligands in cervical SCC. They have found co-gain or co-amplification of
CD274 and
PDCD1LG2, coding PD-L1 and PD-L2, in a significant number of cervical SCCs [
36]. In a recent study of cervical SCC, patients with diffuse expression of PD-L1 were also found to have significantly worse disease-free and disease-specific survival as compared to those with only marginal PD-L1 expression [
10].
Finally, the scoring algorithm in this study has been adopted from the non-small cell lung cancers study [
13], noting that NSCLCs and cervical carcinomas are the same entity. Our study has been carried out on a TMA with little clinical information and no treatment data. Therefore, the cutoff points for positivity for each therapeutic agent should be determined after the clinical trials for cervical cancers have been completed. Until then, the adopted scoring is merely a reporting mean for pathologists in which 50% as well as variable cutoffs are integrated into a single system with addition of the IHC intensities.