Significance of druggable targets (PD-L1, KRAS, BRAF, PIK3CA, MSI, and HPV) on curatively resected esophageal squamous cell carcinoma

Background Esophageal squamous cell carcinoma (ESCC) still remains intractable disease with few therapeutic options. Programmed death-ligand 1 (PD-L1), which is essential for immune evasion, is involved in the pathogenesis of ESCC and thus is a potential therapeutic target. PIK3CA, KRAS, and BRAF mutations, microsatellite instability (MSI) caused by deficient mismatch repair (dMMR), and human papillomavirus (HPV) can potentially upregulate PD-L1 expression, which might contribute to the clinical outcome of patients with ESCC. Methods We investigated the significance of the present druggable markers [PD-L1, PIK3CA, KRAS, and BRAF mutations, MSI caused by deficient dMMR, and HPV] in 64 curatively resected ESCCs, using immunohistochemistry (PD-L1 and MMR protein expression), direct sequencing (KRAS, BRAF, and PIK3CA mutations), real-time PCR (HPV infection), and MSI using quasi-monomorphic markers. Results PD-L1 expression, PIK3CA mutation, and MSI/dMMR were detected in 35.9, 12.5, and 17.2% of ESCCs, respectively. HPV was rarely detected (1.6%) (high-risk HPV68), whereas KRAS and BRAF mutations were not detected in ESCCs. PD-L1-positive tumors were not correlated with PIK3CA mutation or MSI/dMMR (all P > 0.05). PD-L1, PIK3CA mutation, and MSI/dMMR characterized the patients associated with light smoking, female and younger age, and younger age and well-differentiated tumors, respectively (all P < 0.05). In multivariate analysis, only PD-L1-positivity was an independent favorable prognostic factor for overall survival (OS) and disease-free survival (DFS) (P = 0.023, P = 0.014). In the PD-L1-negative ESCCs, PIK3CA mutation had a poor prognostic impact on both OS and DFS (P = 0.006, P = 0.002). Conclusions PIK3CA mutation may be an alternative prognostic biomarker in PD-L1-negative curatively resected ESCCs that can be optional to identify high-risk patients with worse clinical outcome who require more intensive therapy and follow-up.


Introduction
Globally, esophageal cancer is the seventh most prevalent cancer and the sixth most common cause of cancerrelated mortality, representing one of the most aggressive malignant tumors that are very difficult to treat [1]. The prognosis of patients with ESCC is poor with a five-year survival rate of 15-30% even for patients who undergo complete resection [2]. Among the Korean population, esophageal cancer is the eighth most common cause of cancer-related mortality [3]. The incidence and mortality rates are high among Korean males with esophageal squamous cell carcinoma (ESCC), which is the most common histological form of esophageal cancer in Korea [3]. Although there have been rapid advances in surgical techniques and multimodal chemoradiation therapies for other cancers, ESCCs still remain intractable disease with few therapeutic options. The risk factors for ESCCs are also less established in the Korean population.
There are no established prognostic and predictive markers as well as druggable markers for ESCCs as there is limited knowledge on the targetable genetic abnormalities in ESCC [4,5]. Recently, cancer immunotherapy, which exploits the patient immune system to attack cancer cells, has been gaining attention for treating esophageal cancers [6]. One of the major hallmarks of cancer is the evasion of the immune system, which enables the cancer cells to escape the attack from immune cells [7]. The cancer cells can express several immune inhibitory signaling proteins, which induce immune cell dysfunction and apoptosis [7]. Programmed death-ligand-1 (PD-L1), an immune inhibitory protein, binds to programmed death-1 (PD-1) expressed on the tumor-infiltrating lymphocytes (TILs), such as T-cells, B-cells, dendritic cells, and natural killer T-cells to suppress anti-cancer immunity and enable neoplastic growth [7]. Recent studies have demonstrated that PD-L1 expression was detected in 30-45% of ESCCs and that PD-1 blockade was effective in only a subset of patients with PD-L1-positive esophageal cancer [6]. This suggested that the PD-1/PD-L1 axis is highly related to ESCCs and that the identification of additional biomarkers can aid in predicting the clinical response of patients with ESCC. As PD-L1 expression in ESCC has clinical importance, it is important to elucidate the mechanisms underlying the regulation of PD-L1 expression.
The Cancer Genome Atlas (TCGA) reported that PIK3CA gene alterations and related genetic alterations, which were evaluated by exome sequencing, activate the phosphatidylinositol 3-kinase (PI3K) pathway in 24% of ESCCs that can be targeted by small molecules [8]. The PD-L1/PD-1 axis is reported to be driven by the oncogenic activation of the signaling pathways, such as the mitogen-activated protein kinase (MAPK) and PI3K pathways. Additionally, this axis is regulated by other factors in the tumor microenvironment, including microsatellite instability (MSI) caused by deficient mismatch repair (dMMR) and human papillomavirus (HPV) [9][10][11][12][13]. However, MSI/dMMR, RAS/BRAF/PIK3CA mutations, and HPV status have been rarely investigated in esophageal cancers. PD-L1 blockade is used to treat gastrointestinal cancers exhibiting MSI/dMMR, while a combination of targeted therapies is used to treated gastrointestinal cancers with RAS/BRAF/PIK3CA mutation [10,13]. Thus, there is a need for a comprehensive evaluation in esophageal cancers combining the present druggable and targetable biomarkers. In this study, we investigated PD-L1 expression in association with PIK3CA, KRAS, and BRAF mutations, MSI/dMMR status, and HPV infection to determine their prognostic relevance for curatively resected ESCCs.

Patients and methods
The patient medical charts and pathology reports were retrieved for identifying surgically resected and pathologically confirmed ESCC cases. In total, 78 ESCC cases were obtained between 2000 and 2018. Based on the hematoxylin and eosin (H&E) slide review and chart review, only patients with primary esophageal cancers who were chemotherapy-or radiation therapy-naive at the time of the surgery, and whose formalin-fixed, paraffinembedded (FFPE) tumor tissue blocks and complete medical charts were available were included in this study. Nine patients treated with preoperative chemotherapy or radiotherapy that might induce any different bias in the study and five patients unavailable for specimens were excluded. In total, 64 patients with ESCC who underwent radical esophagectomy (Ivor-Lewis operation) with standard lymphadenectomy as the initial definitive treatment were included in the study.
The clinical characteristics and follow-up data were retrospectively obtained from the medical records, pathology report files, and radiological study results. Based on the clinician's judgment, the type of esophageal resection was determined and the selected patients were offered adjuvant chemotherapy with or without radiation therapy. The radiation doses ranged from 5000 cGy to 5400 cGy (23 dose fractions) over a period of 8 weeks. The patients underwent fluoropyrimidine−/taxene-based chemotherapy. Chemotherapy was prescribed based on the performance statuses, comorbidities, and toxicity profiles of individual patients.
Smoking history was measured in packs per year and the patients were classified into 2 categories using 20 packs/year as the cut-off value with heavy smoking defined as > 20 packs/year [14]. Alcohol consumption status was divided into 2 categories using 14 drinks/week as the cut-off value and heavy alcohol consumption was defined as > 14 drinks/week [14].
The pathology reports and histological slides were reviewed and were re-evaluated by an experienced gastrointestinal pathologist (MJK). The diagnosis and histological differentiation were evaluated according to the World Health Organization classification [15]. The stage of the esophageal carcinomas was determined based on the American Joint Committee on Cancer staging system (Eighth edition) [14].
The degree of infiltration of TILs was evaluated from the H&E-stained slides and scored from 0 to 3 as follows: 0, none; 1+, focal infiltration with more than mild staining intensity; 2+, moderate infiltration with more than mild staining intensity; 3+, diffuse infiltration with more than mild staining intensity.
The study protocol was approved by Hallym University Sacred Heart Hospital Institutional Review Board (IRB No. 2019-03-014-001) and performed in accordance with the relevant guidelines and regulations (Declaration of Helsinki). Informed consent was obtained from the patients and from the next of kin (deceased patients) before enrollment in the study.

Immunohistochemistry and microsatellite status determination
Immunohistochemical staining was performed on 4-μm thick tissue microarray (TMA) sections using the Bench-Mark XT automated immunostainer system (Ventana Medical Systems, Inc., Tucson, AZ, USA), according to the manufacturer's instructions. Using a TMA manufacture tool (Quick-Ray™; Unitma, Seoul, South Korea), TMAs were constructed from two 3.0-mm cores of tumor tissue from the representative areas of each case after carefully reviewing all H&E-stained slides. The primary antibody used in this analysis was anti-PD-L1 (rabbit anti-human PD-L1 monoclonal, 1:25, clone SP142; Ventana). PD-L1 expression was evaluated based on the proportion of membranous staining in the tumor cells. The expression was scored as follows: 0 for < 5% of tumor cells, 1+ for 5-10%, 2+ for 10-50%, and 3+ for > 50% of tumor cells. Consistent with several previous published reports, an IHC score of ≥1+ was considered positive [12,16]. To evaluate the expression of MMR proteins, the following primary antibodies were used in this analysis MSI status in the tumor was analyzed by immunohistochemistry of MMR proteins and by multiplex polymerase chain reaction (PCR) for five quasi-monomorphic mononucleotide repeat markers (BAT25, BAT26, NR21, NR24, and NR27). The MSI/dMMR tumors were defined as those exhibiting loss of expression of one or more MMR proteins or exhibiting high-level MSI (MSI-H), which was determined by PCR [17,18]. Microsatellite-stable/proficient MMR (MSS/pMMR) tumors exhibited intact MMR protein expression and/or MSS or low-level MSI (MSI-L) status.

Mutation and HPV detection analyses
Genomic DNA was extracted from 10-μm thick sections of 10% neutral manually dissected FFPE tumor tissue using the Maxwell® 16 FFPE Purification Kit for DNA (Promega, USA). The mutations in the KRAS (exon 2 and 3), BRAF (exon 14), and PIK3CA (exon 9 and 20) genes were analyzed by directional sequencing of PCR fragments amplified from the genomic DNA as previously described [12]. All sequences were confirmed in duplicate for replicate amplification reactions.

Statistical analysis
The clinical and pathological parameters are represented as mean ± standard deviation for continuous variables and as frequency for categorical variables. The categorical variables were analyzed using the chi-squared (χ 2 ) test or two-sided Fisher's exact test. Survival analyses were performed using the Kaplan-Meier method. The survival curves were compared using the log-rank test. Overall survival (OS) was defined as the interval from the first day of surgery until death or the end of the follow-up period. Disease-free survival (DFS) was defined as the interval from the first day of surgery until tumor progression, death, or end of the follow-up period. We also used the Cox proportional hazards model for the univariate and multivariate analyses of OS and DFS. OS and DFS were analyzed until February 2019. All statistical analyses were performed in the SPSS statistical analysis software (v18; SPSS, Chicago, IL, USA). The difference was considered statistically significant when the P-value was less than 0.05.

Patient demographic characteristics
Of the 64 patients, 60 (93.8%) were male and 4 (6.2%) were female, with the median age at diagnosis of 63 years (range 46-84 years) ( Light smoking was associated with the N0 category (P = 0.040) and high TIL (P = 0.018). Heavy alcohol drinking was associated with male patients (P = 0.024). Other clinical or pathological parameters were not associated with smoking or alcohol consumption (Supplementary Table 1).

PD-L1 expression, PIK3CA mutation, MSI, and their clinicopathological correlation
The correlation between clinicopathologic characteristics of the patients and PD-L1 expression, PIK3CA mutation, and MSI status was evaluated ( Table 2).
Among the 11 samples with dMMR, immunohistochemical analysis revealed that the expression of MSH2/ MSH6, MLH1/MSH2/PMS2/MSH6, and MSH6 proteins was not detected in 7, 2, and 2 samples, respectively. However, dMMR cases exhibited only MSI-L with allelic size variation in one of the five mononucleotide markers. The remaining tumors were classified as MSS. None of the tumors exhibited MSI-H. Among the four quasimonomorphic markers, only BAT25 marker exhibited MSI. Cases that did not exhibit MLH1, MSH2, MSH6, and PMS2 expression, which was evaluated by immunohistochemical analysis, exhibited poorly-differentiated squamous cell carcinoma with high TILs (Fig. 2a). The remaining cases lacking at least one MMR proteins    exhibited well-differentiated squamous cell carcinomas with low TILs (Fig. 2b).

HPV infection in esophageal cancers
Only one patient (1.6%) harbored high-risk HPV68 infection in the ESCC tissue, which exhibited PD-L1negative status, wild-type PIK3CA gene expression, and MSS status (Fig. 2c). The patient was a 68-year-old male with stage IIIA cancer in the lower esophagus and lymph node metastasis. The patient was a never-alcohol drinker but a heavy smoker smoking 25 pack/year. The patient did not exhibit synchronous or metachronous malignancies. The analysis of H&E slide of this patient revealed a large, exophytic mass (3.5 × 1.8 cm) comprising poorlydifferentiated tumor with markedly high TILs. The tumor cells were relatively non-cohesive. In addition to the main tumor mass, mild to moderate dysplasia with koilocytosis was observed. The patient did not receive any postoperative adjuvant therapy and was alive without any recurrence for 32 months.

Clinicopathological correlation with survival
We performed Kaplan-Meier survival analyses to determine whether PD-L1 expression, PIK3CA mutation, and MSI/ dMMR were correlated with OS or DFS in patients with ESCC. Kaplan-Meier curves revealed that the PD-L1positive tumors exhibited a favorable OS when compared to the PD-L1-negative tumors (median 54 vs. 16.5 months) with borderline statistical significance (log-rank test: P = 0.078). Patients with PD-L1-positive tumors exhibited better DFS rate than those with PD-L1-negative tumors (median 117 vs. 11 months) (log-rank test: P = 0.029) (Fig. 3a-b). There were significant differences in the OS (median 6 vs. 27 months) and DFS (median 5 vs. 24 months) between patients with PIK3CA-mutated and those with PIK3CA-wild-type tumors (log-rank test: P = 0.154 and P = 0.171, respectively) ( Fig. 3c-d).
To investigate the prognostic factors for OS and DFS in patients with ESCC, univariate and multivariate Cox proportional hazard regression analyses were performed ( Table 3). The univariate analysis revealed that vascular invasion was significantly associated with poor OS (P = 0.047). Additionally, PD-L1 positive status and the  In the PD-L1-negative esophageal cancers, patients with PIK3CA-mutated tumors exhibited poorer OS and DFS than patients with PIK3CA-wild-type tumors (P = 0.006 and P = 0.002, respectively) ( Fig. 3g-h). However, there were no significant differences in the OS or DFS of patients with PD-L1-positive esophageal cancers with or without PIK3CA mutation (P = 0.239 and P = 0.227, respectively).
There were no significant differences in the OS or DFS for both PD-L1-positive and PD-L1-negative esophageal cancers with or without MSI/dMMR (P = 0.936 and P = 0.883 for PD-L1 positivity; P = 0.141 and P = 0.070 for PD-L1 negativity).

Discussion
In this study, we screened 64 ESCC tissue specimens for PD-L1 expression, HPV infection, MSI/dMMR, and mutations in the KRAS, BRAF, or PIK3CA genes to determine their prognostic value in esophageal cancers. We confirmed that Korean ESCC cases were associated with male patients (94%) with a median age of 63 years, heavy smoking (54.7%) or alcohol drinking (50%), and poor 5-year OS (23.9%). The multivariate analysis revealed that PD-L1 expression, which was detected in one-third of ESCC cases, was an independent favorable prognostic factor for OS and DFS. The combined analysis revealed that PIK3CA mutation in the PD-L1-negative ESCCs was a poor prognostic factor for both OS and DFS. The incidence of HPV was very rare, while the KRAS and BRAF mutations were not detected in ESCCs. To the best of our knowledge, currently this is the first study evaluating combined the present druggable and targetable biomarkers in ESCCs. Previously, several studies have reported that the frequencies of tumor cells exhibiting PD-L1 expression in ESCC range from 18.4 to 79.7%, which might be attributable to the differences in the antibodies used, interpretation criteria, specimen types affected by neoadjuvant therapy, and the geographic or racial characteristics [19][20][21][22][23][24][25]. In this study, PD-L1-positive tumors were observed in 35.9% of cases, which is highly consistent with a previous Korean population-based study (33.5%) [16]. As approximately one-third of the cells exhibited PD-L1 expression in ESCCs, PD-L1 can serve as prognostic and predictive biomarkers in esophageal cancer cases associated with limited chemotherapeutic options. To facilitate the clinical application of PD-L1, this study specifically utilized the SP142 anti-PD-L1 antibody, a U.S. FDA-approved antibody for a validated immunohistochemistry protocol, and the criterion of ≥5% of PD-L1, which were used for identifying patients with advanced urothelial cancers for atezolizumab treatment [26]. In the present study, PD-L1 positivity was an independent prognostic factor for favorable DFS in the patients with curatively resected, adjuvant therapy-naive ESCC. PD-L1 positivity also exhibited better OS than those with PD-L1 negativity in the multivariate analysis, although the univariate analysis did not reach statistical significance (P = 0.078), of which result might be confounded by the factors that included or excluded in the analyses for OS. Our finding relevant to favorable DFS concurred with that of previous studies that demonstrated favorable prognosis of PD-L1 expression in surgically resected ESCCs without preoperative or neoadjuvant treatment [22][23][24]27]. In this study, high PD-L1 expression was prevalent among light or nonsmoker patients without lymph node metastasis and with high TILs. This may be because the non-smoking or light smoking patients with PD-L1-positive tumors had highly immunogenic tumors [25], which indicated that a strong adaptive immune response may attack the tumor microenvironment and ultimately influence the favorable clinical outcomes of patients [11,23]. Some studies have reported a poor prognostic correlation of PD-L1 expression in patients treated with neoadjuvant concurrent chemoradiotherapy [19,20], while others have demonstrated no correlation between PD-L1 expression and survival in the patients undergoing postoperative chemotherapy [21]. The discrepancy in the prognostic values of PD-L1 may be because the neoadjuvant or adjuvant treatment agents may serve as confounders for the PD-1/PD-L1 axis and because the PD-L1 expression protects tumor cells from pro-apoptotic agents [10]. Lim et al. [19] also reported PD-L1 expression significantly increased after concurrent chemoradiation therapy. Likewise, Ng et al. [28] also described that PD-L1 expression was significantly induced in ESCC cell lines after standard chemotherapy treatments, suggesting that the potential benefit of combined conventional chemotherapy together with anti-PD-L1 immunotherapy to achieve better treatment outcome. Some studies have reported that the PD-1/PD-L1 axis mediates resistance to radiotherapy and anti-CTLA-4 antibody immunotherapy [29]. Therefore, an optimal treatment combination of immune checkpoint inhibitors and other therapies is required for the clinical treatment of patients with ESCC.
The PIK3CA mutation activates many downstream pathways that regulate critical cellular functions involved in the development of cancer [8]. In this study, PIK3CA mutation was detected in 12.5% of esophageal cancers and was associated with female or younger patients (≤60 years) lacking mutations in the KRAS and BRAF genes. There are limited studies that have evaluated PIK3CA mutation, which has an incidence of 1.5-21%, in esophageal cancers [30][31][32][33]. Additionally, mutations in the KRAS and BRAF genes are reported to be extremely rare (0.5-1%) in ESCCs [30,34]. The detection of PIK3CA mutation in ESCCs may suggest that PIK3CA mutation may be involved in the development and progression of esophageal cancers, especially in female and younger patients (≤60 years) who are not commonly affected by ESCCs. Consistent with this finding, PI3K inhibitors are reported to inhibit the growth of an esophageal cancer cell line with PIK3CA mutation in vitro [32]. Additionally, mutations in the KRAS, BRAF, and PIK3CA genes have emerged as an important predictive marker of resistance to epidermal growth factor receptor (EGFR)-targeting monoclonal antibodies or tyrosine kinase inhibitors in colorectal and lung cancers [10]. However, the effectiveness of these therapies is unclear in esophageal cancers. In this study, we demonstrated that approximately one-tenth of ESCC cases exhibited PIK3CA mutation. However, none of the ESCC cases exhibited mutations in the KRAS and BRAF genes. This suggested that a subset of these patients may respond to PI3K inhibitors but may not respond to EGFR-targeting treatment.
The prognostic significance of PIK3CA mutation has been controversial. PIK3CA mutation has been associated with favorable prognosis in patients with curatively resected ESCCs in a Japanese population-based study [31], whereas a Korean population-based study reported no correlation between PIK3CA mutation and survival [33]. We also observed that PIK3CA mutation alone does not affect survival. However, the combined analyses revealed that the PIK3CA mutation in PD-L1-negative esophageal cancers was significantly correlated with poor survival. This indicated that PIK3CA mutation may be potentially important and may be an alternative prognostic biomarker in PD-L1-negative tumors that can be used to identify high-risk patients with worse clinical outcome who require more intensive therapy and follow-up.
There have been limited studies on MSI/dMMR in ESCC, of which results appear to different from conventional colorectal or gastric adenocarcinomas [17,35,36]. Previous studies have demonstrated that the frequency of MSI in ESCC varies from 2 to 66.7% [17,35,36]. MSI/dMMR in esophageal cancer may predict the clinical response to PD-L1 [13]. In this study, we detected MSI/dMMR in 17.2% of ESCCs, among which one-fourth was PD-L1-positive. We observed that MSI/ dMMR represented younger patients (≤60 years) and well-differentiated tumors but did not have any survival impact in esophageal cancers. In this study, we evaluated the MSI/dMMR status using immunohistochemistry and quasi-monomorphic mononucleotide markers. We observed that the dMMR cases exhibited only MSI-L but not MSI-H. There was no perfect concordance between dMMR and MSI. Similarly, Hayashi et al. [17] reported that MSI at one or more loci was observed in 40% of ESCCs with only one classified as MSI-H. Additionally, the loss of MLH1 expression, which was evaluated by immunohistochemistry, did not correspond to MSI. The MSI and MMR protein expression patterns in the ESCCs seem to be different from those in the colorectal or gastric cancers. MSI-L observed in ESCC could result from defects in MMR proteins other than MSH2 and MLH1 or from defects in other genes not involved in MMR, or it could simply represent a background level of genetic instability that may be present in all tumors [17,35]. The squamous cell histology in esophageal cancers might contribute to the discrepancy between MSI and dMMR because other intestinal MSI cancers are related to intestinal-type adenocarcinomas with mucinous and lymphoid-rich histology [17,36].
In this study, only 1 tumor (1.6%) carried the high-risk HPV68 infection. This indicated that the HPV infection rate in ESCC samples is very low and that HPV is not the etiological cause of ESCC, which concurred with findings of a previous study [8]. Globally, there is high variability in the detection rates of HPV in ESCCs. The HPV detection rate in the tumor tissues in the high-ESCC-incidence countries is significantly higher than that in the low-ESCC-incidence countries (up to 64% in China vs. up to 17% in North America) [37]. In Korea, the incidence of esophageal cancers is low with 2457 new cases reported in 2017, which ranked 19 out of 23 malignancies [3]. The low prevalence of esophageal cancers in Korea may result in a low HPV detection rate in ESCCs. This is the first study investigating the incidence of HPV in ESCCs of Korea.
The limitations of this study include the retrospective analysis of archival samples, the relatively low number of samples, and the possible selection bias of current study. Considering the low incidence of esophageal cancers, the enrolled number of 64 patients is not small sample size. The results of this study provide valuable insights into the associations between surgically resected ESCCs and druggable targets to date.