PD-L1 Predicts Poor Prognosis in Surgically Resected Limited Stage Small-Cell Lung Cancer

Introduction

Small-cell lung cancer (SCLC) is an aggressive high-grade neuroendocrine tumor with limited treatment strategies. Although SCLC is sensitive to chemoradiotherapy, it has a high relapse rate with a dismal prognosis. Chemotherapy combined anti-PD-L1 has been approved for the first line treatment of extensive stage-SCLC (ES-SCLC).¹ Subgroup of ES-SCLC with positive PD-L1 expression (≥1%) seemed to have a better response rate and survival, indicating that PD-L1 may predict efficacy to immunotherapy and survival in SCLC.

DLL-3 is an inhibitory ligand to Notch pathway, which is involved in the development of pulmonary neuroendocrine cells.² DLL-3 is upregulated and overexpressed on the cell surface in SCLC and other high-grade neuroendocrine tumors^3,4 and predicted the response of Rova-T (NCT01901653).⁵

PARP is a critical enzyme involved in the process of DNA repair, and catalyzes the synthesis of poly (ADP-ribose) (PAR) chain after recognition of single-strand DNA breaks and prevents cells from apoptosis by cleaving DNA-damage and thus inhibits apoptosis induction. The safety and efficacy have been tested in several clinical trials, indicating that PARP1 inhibitors with or without conventional chemotherapy were well-tolerated and effective in ES-SCLC.^6–8

With the success in ES-SCLC, more attention needs to be paid to immunotherapy and targeted therapy in LS-SCLC. Whether PD-1/PD-L1 antibodies, DLL-3, and PAPR1 inhibitors could be used in LS-SCLC is still unknown. Understanding the expression of PD-L1, DLL-3, and PARP1 in LS-SCLC and their correction to clinicopathological features and outcomes in LS-SCLC may facilitate the development of new strategies in LS-SCLC.

Patients and Methods

Patients

We retrospectively screened 404 SCLC patients diagnosed with SCLC from January 2011 to January 2018 in the First Affiliated Hospital of Xi’an Jiaotong University. 46 of them underwent surgery, and adequate materials and histological specimens containing abundant tumor cells were available for 43 of them. All protocols were approved by the Ethics Committee of Xi’an Jiaotong University, and informed consent was obtained from all patients before surgery. Our study complied with the Declaration of Helsinki. Upon enrollment, SCLC was confirmed in all patients by biopsy or surgical pathology and graded by the American Joint Committee on Cancer (AJCC) 8th tumor lymph node metastasis (TNM) staging system. The clinicopathological characteristics of the patients were collected from the hospital medical records, and registered patients were prospectively followed up until July 1, 2020.

The clinicopathological characteristics of 43 surgically resected LS-SCLC are shown in Table 1. The medium age is 60 years, and ranges from 28–79 years. Thirty-five (81.40%) of them were male. Current or former smokers accounted for 62.80%. Twenty-seven patients (62.79%) were central SCLC, others were peripheral SCLC. NSE was overexpressed in 27 patients (62.79%) before surgery. According to TNM 8th version, 14 patients (32.56%) were at stage I, 23 patients (53.49%) were stage II, and 6 patients (13.95%) were stage III.

Table 1 The Characteristics of Forty-Three Surgically Resected LS-SCLC

Immunohistochemistry (IHC) Staining

The expression of PD-L1 (DAKO, 22C3), DLL-3 (bs-7860R, Bioss, China), ASCL-1 (bs-1155R, Bioss, China), PARP1 (bs-20764R, Bioss, China), and SLFN11 (sc-374,339, Santa Cruz, US) were detected in 5-μm sections cut from FFPE SCLC tissues by IHC as previously described.⁹ Immunostaining for PD-L1 was defined positive if it was observed in the membrane and/or cytoplasm in >5% of the tumor cells and tumor infiltrating immune cells.¹⁰

Statistical Analysis

The association between PD-L1, DLL-3, ASCL-1, PARP1, SLFN11, and clinicopathological characteristics of LS-SCLC patients was performed by χ² or student’st-test, whenever is appropriate. DFS was defined as the time from operation until the first documented progression of disease, death, or date of last follow-up. OS was defined as initial diagnosis until the date of death or last follow-up. The survival curves were calculated by the Kaplan-Meier method and analyzed by the Log rank test. Univariate and multivariate regression was performed by the Cox proportional hazards model. All variables that had P-values of less than 0.05 were included in the Cox model. All tests were two-sided, and P-values of less than 0.05 was considered statistically significant. All of the statistical analyses were conducted using SPSS22.0 (Chicago, IL, US).

Results

The Expression of PD-L1, DLL-3, PARP1, ASCL-1, SLFN11, and Correlation to Clinicopathological Factors

PD-L1, DLL-3, PARP1, ASCL-1, and SLFN11 were examined in 43 samples and the representative images were shown in Figure 1 and Supplementary Figure 1. Among 43 surgically resected LS-SCLC specimen, PD-L1 was positive (>5%) in 58.14% (25 out of 43) cases. PAPR1 was not significantly different between tumor and matched non-cancerous tissues (Figure 1B), which was only positive in nine patients (20.93%). DLL-3 was significantly overexpressed in tumor tissues, compared with matched para-noncancerous tissues (Figure 1C). And the positive rate was 65.12% (28/43), similar to previously reported. No significant relationship was found between PD-L1, PARP1, or DLL-3 and clinicopathological factors (Table 2).

Table 2 Correlation Between PD-L1, DLL-3, PARP1 and Clinical Pathological Factors

Figure 1 The distribution of PD-L1 (A), DDL-3 (B), and PARP1 (C).

In the PARP1 pathway, SLFN11 is actively recruited to sites of DNA damage, inhibits homologous recombination,¹¹ and activates a cellular replication-stress response. SLFN11 expression correlates with sensitivity to DNA-damaging agents.¹² In our cohort, SLFN11 was not significantly different between tumor and matched non-cancerous tissues, and 20 patients (46.51%) positively expressed SLFN11 (Supplementary Figure 1A). Achaete-scute homolog-1 (ASCL-1) transcriptionally activates DLL-3 and is also upregulated in neuroendocrine tumor, including SCLC, and has been shown to be positively related to DLL-3 in surgically resected SCLC samples.¹³ ASCL-1 was also upregulated in SCLC (Supplementary Figure 1B), compared with matched para-noncancerous tissues, and was positively expressed in 29 patients (67.44%). Moreover, no significant relationship was found between ASCL-1, SLFN11, and clinicopathological factors (Supplementary Table 1).

Relationship Between PD-L1, DLL-3, PARP1, ASCL-1, and SLFN11

A positive correlation has been demonstrated in previous studies, however, we did not find a correlation between DLL-3 and ASCL-1 in our cohort (r=0.12, P=0.23). But we noticed a positive correlation between PARP1 and SLFN11 (r=0.21, P=0.09), though it was not statistically significant.

We also testified the relationship between PD-L1, DLL-3, and PAPR1, but no correlation was found between PD-L1, DLL-3, and PAPR1: r=0.01, P=0.94 for PD-L1 and DLL-3; r=−0.01, P=0.96 for PD-L1 and PAPR1; r=−0.10, P=0.50 for DLL-3 and PARP1.

Gender, Age, Smoking Status, Adjuvant Treatment, TNM Stage, and PD-L1 Expression Were Correlated with DFS

The median duration of follow-up was 30 months (range from 0.2–87 months). Among clinicopathological characters, we found that male patients had poorer DFS than female patients (mDFS: 22.00±2.90 vs 29.00±33.94 months in male and female patients, respectively, P=0.031, Figure 2A). Elder patients (elder than 60 years old) performed poorer DFS than those who were less than 60 years old (mDFS: 12.00±5.59 vs 33.00±11.80 months in elder and younger than 60 years old patients, respectively, P=0.018, Figure 2B). Adjuvant therapy (either chemotherapy or radiotherapy) improved DFS (mDFS: 29.00±8.90 vs 21.00±5.03 months in patients with or without adjuvant therapy, respectively, P=0.018, Figure 2C). Advanced TNM stage also correlated with poor DFS (mDFS: 29.00±8.42, 22.00±0.79, and 1.00±0.00 months in TNM stage I–III, respectively, P=0.000, Figure 2D). Current or former smokers seemed to perform shorter DFS (mDFS: 21.00±7.75 vs 29.00±11.00 months in current or former smokers and never smokers respectively, P=0.053, Figure 2E). We further analyzed the relationship between smoking index and prognosis, and noticed that moderate smokers had the worst DFS (mDFS: 29.00±11.00, 6.00±3.00, and 21.00±4.79 months in never smokers, moderate smokers, and heavy smokers, respectively, P=0.021, Figure 2F). While central or peripheral tumor, comorbidities, onset symptoms, or NSE level did not have significant effects on LS-SCLC prognosis (Supplementary Figure 2A-D).

Figure 2 Clinicalpathological factors including gender (A), age (B), adjuvant treatment (C), TNM stage (D), smoking status (E), smoking index (F) and DFS.

Patients with positive PD-L1 expression had significantly worse DFS than those with negative PD-L1 (mDFS: 21.00±3.50 vs 34.00±29.70 months in PD-L1 positive and negative groups respectively, P=0.049, Figure 3A). Neither PARP1, DLL-3, ASCL-1, nor SLFN11 preformed any difference in DFS (Figure 3B and C, Supplementary Figure 3A and C).

Figure 3 The correlation between PD-L1 (A), PARP1 (B), DDL-3 (C) and DFS.

Univariate analysis demonstrated that gender (HR=0.36. 95% CI=0.14–0.96, P=0.004), age (HR=2.11, 95% CI=1.11–4.03, P=0.024), TNM stage (HR=2.22, 95% CI=1.21–4.07, P=0.010), and adjuvant treatment (HR=0.44, 95% CI=.21–0.90, P=0.025) are associated with DFS, and further multivariate analysis identified TNM stage (HR=2.51, 95% CI=1.31–4.78, P=0.005) as an individual prognostic factor for DFS in LS-SCLC (Table 3).

Table 3 Univariate and Multivariate Analyses of Relationship Between DFS and PD-L1 and Clinicopathological Factors

Adjuvant Treatment, TNM Stage, and PD-L1 Expression Were Associated with OS

Patients who did not receive adjuvant therapy (either chemotherapy or radiotherapy) showed worse OS than those who did (mOS: 37.00±9.14 vs 22.00±11.27 months in patients with or without adjuvant therapy, respectively, P=0.002, Figure 4A). Advanced TNM stage also correlated with poor prognosis (mOS: 39.00±0.00, 32.00±6.39, and 6.00±2.89 months in TNM stage I–III, respectively, P=0.003, Figure 4B). Other clinicopathological factors, such as age, gender, smoking status, central or peripheral tumor, comorbidities, onset symptoms, and NSE level did show a significant difference in OS (Supplementary Figure 4A-G).

Figure 4 The correlation between Adjuvant treatment (A), TNM stage (B), PD-L1 (C), PARP1 (D), DDL-3 (E) and OS.

Patients with positive PD-L1 expression had significantly worse OS than those with negative PD-L1 (mOS: 22.00±7.49 vs NR, P=0.000; Figure 4C). However, neither DLL-3, PAPR1, ASCL-1, or SLFN11 had a significant effect on OS (Figure 4D and E, Supplementary Figure 3B and D).

Univariate analysis revealed that PD-L1 and TNM stage were significantly associated with poor OS (HR=4.40, 95% CI=1.87–10.37, P=0.001; HR=2.38, 95% CI=1.30–4.33, P=0.005 for PD-L1 and TNM stage, respectively), whereas adjuvant treatment was significantly correlated with favorable OS (HR=0.33, 95% CI=0.16–0.70, P=0.004). Further multivariate analysis indicated that PD-L1 and TNM stage were independent and significant negative predictive factors for OS (HR=2.89, 95% CI=1.21–6.93, P=0.017; HR=2.49, 95% CI=1.25–4.94, P=0.009 for PD-L1 and TNM stage, respectively), and adjuvant treatment was an independent positive prognostic factor for OS (HR=0.37, 95% CI=0.17–0.81, P=0.012, Table 4).

Table 4 Univariate and Multivariate Analyses of Relationship Between OS and PD-L1and Clinicopathological Factors

Discussion

Recently, novel targets, such as PD-1/PD-L1 antibodies,¹⁴ PARP, and DLL-3 inhibitors have revolutionized the treatment of ES-SCLC, and the expression of PD-L1, DLL-3/ASCL-1, and PARP1/SLFN11 have been recognized as predictive markers for clinical response. However, the studies of immunotherapy or target-immunotherapy in LS-SCLC are still on-going and the results are immature. The expression and prognostic values of PD-L1, PARP1, and DLL-3 are elusive in surgically resected LS-SCLC.

In this study, we enrolled 43 surgically resected LS-SCLC patients and defined PD-L1>5% on both tumor cells and stromal cells as positive. We firstly examined the expression pattern of PD-L1, PARP1, and DLL-3 in specimen from surgically resected LS-SCLC. PD-L1 expression in SCLC varied from one study to another.^10,15,16 After scrutinizing the data, we suppose reasons that led to different prevalence of PD-L1 are as follows: firstly, there is no standard for positivity of PD-L1 expression, and a cutoff of at least 1% or 5% cells staining were both defined as positive. Early in 2016, Takada et al¹⁷ concluded that 5% cut-off showed similar rates of expression among antibodies clones E1L3N, 28–8, and SP142. Herein, we decided to use 5% cut-off in this study. Secondly, PD-L1 is low-expressed in tumor cells but highly expressed in tumor infiltrating immune cells. When taking the expression of PD-L1 in tumor infiltrating immune cells into account, the prevalence of PD-L1 increased from 14.7% to 57.1%. Our results were similar to Ishii et al’s,¹⁰ in which PD-L1 expression on both tumor cells and tumor infiltrating immune cells were measured. Thirdly, different antibodies used in these studies also led to varied prevalence of PD-L1. Antibodies from Abcam achieved the highest positive rate, whereas none of the tumor cells were positive for PD-L1 staining by antibody 5H1/Chen’s lab.¹⁸ The positivity of other antibodies such as 28–8,¹⁹ 22C3/Dako,^20,21 E1 L3N/CST,¹⁸ and MAB1561/R&D system²² varied from 10–34.4% in tumor cells, and from 2–87.1% in both tumor and stromal cells. Most recently, Takayama et al²³ compared different PD-L1 antibodies again and came to the conclusion that anti-PD-L1 antibody clones (28–8, 22C3, and SP263) exhibited similar positive rate and with similar clinical correlates. Last, but not least, LS-SCLC and ES-SCLC exhibited different PD-L1 prevalence, and LS-SCLC seemed to have higher PD-L1 expression,^10,24 which explains the higher PD-L1 proportion in our cohort.

Besides PD-1 and PD-L1, DLL-3 and PARP1 inhibitors also gained promising data in Phase I and II clinical trials. We also observed high prevalence of DLL-3 and ASCL-1 in surgically resected LS-SCLC, as has been demonstrated in a Japanese cohort and a Chinese study.²⁵ Contrary to previous studies, the positive expression rate was relatively low in our cohort. We suppose it is because we only included surgically resected LS-SCLC in our study.

We next identify the correlation between clincal pathlogical factors and PD-L1, PARP1, DLL-3, as well as the prognostic effects of clincal pathlogical factors. As cigarette smoking has been proved to lead to tumor spread,^26,27 we noticed that current or former smokers did perform shorter DFS than never smokers, though it was not significantly different. Additionally, moderate smokers were proved to have the worst survival, which may be due to a limited number of moderate smokers in our cohort.

Although systemic chemotherapy using cisplatin/etoposide with concurrent thoracic radiation is a standard treatment for LS-SCLC,²⁸ locoregional relapse is still frequent (26–63%).^29,30 Masayoshi Inoue et al³¹ reviewed several clinical trials and concluded that surgery reduced the locoregional relapse rate in LS-SCLC. Moreover, perioperative chemotherapy has also been proved to prolong survival.³² In accordance with these studies, we also demonstrated that adjuvant treatment prolonged LS-SCLC patients’ DFS and OS.

Interestingly, we identified high PD-L1 correlated with shorter DFS and OS and was an individual negative prognostic factor in surgically resected LS-SCLC for OS, in accordance with two studies from China.^33,34 However, the prognostic value of PD-L1 in SCLC is controversial: in contrast to our findings, two Japanese cohorts^10,35 demonstrated that PD-L1 correlated with better prognosis. In resectable SCLC, positive PD-L1 expression predicted improved DFS and a lower risk of brain metastasis.³⁶ Moreover, another study conducted among 66 stage I–III SCLC patients and 38 metastatic cases showed no effect of PD-L1 expression on OS.²⁴ The prognostic role of PD-L1 still needs to be explored, especially in LS-SCLC.

Cell intrinsic PD1/PD-L1 signaling has been demonstrated as a predictor for poor efficacy of cisplatin treatment,³⁷ and it explains that PD-L1 predicted poor prognosis in our study, in which we enrolled 30 patients receiving adjuvant treatment. Thus, targeting the cellular PD1/PD-L1 axis may improve chemosensitization of aggressive SCLC.

In conclusion, we found that PD-L1 expression, TNM stage, and adjuvant treatment were independent prognostic factors for OS in surgically resected LS-SCLC. Our findings provide evidence for adjuvant treatment, whether chemotherapy or radiotherapy, in surgically resected LS-SCLC. Additionally, PD-1/PD-L1 antibodies may improve survival for surgically resected LS-SCLC patients, and further perspective studies are needed.

Here are several limitations of our study: firstly, the number of cases is limited. Since SCLC is a highly aggressive neuroendocrine tumor, characterized by rapid growth and early tendency to widespread metastasis, over 70% of new diagnosed cases were already metastatic. In our study, the percentage of LS-SCLC patients is relatively low (13.86%; 56/404). Secondly, none of the patients received immunotherapy or DLL-3/PAPR1 inhibitors in our cohort, so we could not analyze the correlation between PD-L1, DLL-3, PARP1, and immunotherapy or targeted therapy efficacy. Thirdly, since we did not obtain adequate tumor tissues from ES-SCLC biopsy samples, we could not compare the expression of PD-L1, PARP1, and DLL-3 between LS-SCLC and ES-SCLC, and we would love to conduct a perspective study in the near future.

Conclusion

PD-L1 and DLL-3 were highly-expressed in SCLC. PD-L1, adjuvant treatment, and TNM stage were individual prognostic factors for OS in surgically resected LS-SCLC.

Grant Support

International cooperation project in science and technology of Shaanxi province (No. 2019KW-074).