Clinical and prognostic implications of CD47 and PD-L1 expression in surgically resected small-cell lung cancer

Background Pharmacological inhibition of the immune-checkpoint molecule CD47 has shown promising results in preclinical small-cell lung cancer (SCLC) models, whereas anti-programmed death-ligand 1 (PD-L1) inhibitors have been recently implemented in the standard of care of advanced-stage SCLC patients. Nevertheless, the expression pattern, clinical relevance and prognostic implication of both CD47 and PD-L1 are rather controversial in surgically treated SCLC patients. Materials and methods In total, 104 Caucasian SCLC patients from two Central European thoracic centers were included in this study. CD47 and PD-L1 expression as well as the expression of the four major SCLC molecular subtype markers (ASCL1, NEUROD1, YAP1 and POU2F3) were measured by immunohistochemistry. Expression levels were independently evaluated and statistically correlated with clinicopathological data and survival. Results Positive CD47 and PD-L1 expressions were seen in 84.6% and 9.6% of the samples, respectively. Meanwhile, the tumor-associated stroma was positive for PD-L1 in 59.6% of the cases. Stromal PD-L1 expression correlated with longer overall survival (OS) (versus PD-L1-negative stroma; median OS was 42 versus 14 months, respectively, P = 0.003) and was confirmed as an independent predictor of favorable outcome upon multivariate analysis (hazard ratio 0.530, 95% confidence interval 0.298-0.943, P = 0.031). Notably, neither CD47 nor PD-L1 presence was related to a distinct molecular SCLC subtype. Conclusion CD47 shows a remarkably high expression while tumoral PD-L1 expression is generally low in surgically treated SCLC. Importantly, stromal PD-L1 expression may indicate a favorable clinical outcome and serve as a novel prognostic factor in these patients. Additional studies are warranted to further investigate the clinical impact of CD47 and PD-L1 expression in SCLC.


INTRODUCTION
Small-cell lung cancer (SCLC) is one of the most aggressive types of lung cancer (LC) and accounts for w15% of all LC cases worldwide. Historically, SCLC has been regarded as a cancer of neuroendocrine origin and is clinically characterized by rapid tumor progression and early metastatic dissemination. Although surgical resection has been shown to potentially offer favorable long-term outcomes, it is rarely carried out in SCLC 1,2 since the vast majority of patients are being diagnosed with metastatic disease when local resection is not indicated anymore. 3 Accordingly, there is a need for novel therapeutic concepts as clinical outcomes in SCLC remain extraordinarily poor with 5-year survival rates below 10%. 4 Recent advancements in the field of SCLC research might provide hope for improved outcomes in this recalcitrant type of disease. 5,6 Firstly, nationwide screening programs have been shown to effectively increase the number of limited-stage LC cases and to significantly reduce LCassociated mortality. 7,8 Consequently, the overall number of potentially resectable patients is expected to increase in the upcoming decades. It is evident that there is a need for novel therapeutically tractable molecules and prognostic biomarkers for surgically managed patients. 9 Secondly, evidence that SCLC may represent a far more heterogeneous malignancy than previously assumed is also growing. 10 Indeed, a new model of SCLC nomenclature defined by the differential expression of four key-transcription regulators (ASCL1, NEUROD1, YAP1 and POU2F3) has just been recently presented and validated in surgically resected samples. 11,12 Importantly, these distinct molecular subsets have major differences in cellular origin, neuroendocrine differentiation and growth properties. 10,13 In addition, our group previously showed that the molecular subtypes of SCLC are also linked with different prognostic outcomes since high ASCL1 expression is associated with poor survival, whereas high POU2F3 expression is linked with improved prognosis. 12 A significant number of targetable molecules have also been identified in preclinical studies and clinical trials recently. 10 Most importantly, immunotherapeutic concepts have been discussed as especially promising for future perspectives in SCLC. 14 Indeed, representing the most significant therapeutic novelties for decades, immunotherapeutic targeting of anti-programmed death-ligand 1 (PD-L1) has been recently included into the treatment guidelines for extensive-stage SCLC patients. 15,16 Moreover, immunecheckpoint-mediated blockage of CD47, a 'don't-eat-me' cell-surface signal for macrophages, has been shown to effectively suppress SCLC tumor growth in vitro and in vivo. 17 This is of clinical importance for SCLC as clinical trials targeting CD47 by systemic immunotherapy are already enrolling patients with different solid and hematologic malignancies. 18 Because only a few studies with small cohorts have aimed to examine the distribution and prognostic relevance of PD-L1 and CD47 in SCLC to date, the findings are preliminary and controversial and thus warrant further investigations. 19,20 Accordingly, in the current study we aimed to evaluate (i) the expression pattern of CD47 and PD-L1, (ii) the relation between CD47/PD-L1 expression and recently discussed major molecular SCLC subtypes and (iii) the prognostic impact of CD47 and PD-L1 expression in a large international cohort of surgically treated Caucasian SCLC patients.

Study design and clinical data
This study included histologically confirmed SCLC patients who underwent diagnostic or therapeutic surgical resection at two Central European high-volume thoracic surgical centers (Department of Thoracic Surgery, Medical University of Vienna, Vienna, Austria and Department of Thoracic Surgery, National Koranyi Institute of Pulmonology, Budapest, Hungary) between January 2000 and December 2019. Clinical and follow-up data were retrospectively extracted from institutional medical records or obtained directly from the Central Statistical Office. Wedge resection and segmentectomy were defined as sublobar resection, whereas lobectomy or pneumonectomy as lobar resection. Staging of the patients was retrospectively carried out using the data from histological findings and medical records in accordance with the eighth edition of the TNM (tumore nodeemetastasis) classification for LC by the International Association for the Study of Lung Cancer. 21,22 Recurrent disease was termed as evidence of recurrence in either mediastinal or hilar lymph nodes or in the ipsilateral lung. Evidence of any other recurrences was classified as distant organ metastases.
If recurrence or metastasis was suspected, further clinical evaluation [e.g. bronchoscopical/transthoracic biopsy of suspect lesion, positron emission tomographyecomputed tomography (CT), magnetic resonance imaging] was carried out according to the individual scenario. Overall survival (OS) was defined as the time of surgery until the date of the last available follow-up or death in months. Diseasefree survival (DFS) was calculated as the time of surgery until the evidence of metastases or recurrent disease in months.
According to institutional protocols, routine oncological follow-up was carried out for all included patients. Notably, these follow-ups comprised regular (i.e. every 3 months in the first post-operative year and then every 6-12 months) blood tests, X-rays and/or thoracic CT scans of the chest. Distinct patients received tumor-specific adjuvant treatment based on the decision of a multidisciplinary tumor board consisting of board-certified medical, radiation and surgical oncologists in accordance with the contemporary National Comprehensive Cancer Network (NCCN) guidelines. 23 Notably, there were no relevant differences across the two host institutions. Adjuvant chemotherapy consisted of a platinum agent (cisplatin or carboplatin) combined with etoposide. In rare cases, irradiation was also applied.
IHC staining results were independently evaluated in a double-blinded fashion by two experienced pathologists specialized in pulmonary malignancies. If there were significant differences between ratings, the section was discussed and re-evaluated. CD47 expression was scored as negative (absent or <10% positivity of all tumor cells) or positive (10% positivity of all tumor cells) considering membranous and/or cytoplasmic signals. For PD-L1, tumoral and stromal expression were both evaluated and defined as either negative (absent or <1% of tumor or stromal cells) or positive (10% of tumor or stromal cells). For ASCL1, NEUROD1, YAP1 and POU2F3, the previously defined cut-off values were used. 12

Statistical analysis
Collected and generated data were evaluated with GraphPad Prism 8 (GraphPad Software, Inc., San Diego, CA), Microsoft Excel Version 16.48 (Microsoft Corporation, Redmond, WA) and SPSS Statistics Version 27.0 (IBM Corp., Armonk, NY). The chi-square test was applied for the comparison of binominal variables. If the expected frequency was below 5, Fisher's exact test was used. In order to identify risk factors for OS and DFS, all clinicopathological data and staining results were included in a univariate analysis using the Cox regression model and presented as hazard ratios (HRs) including corresponding 95% confidence intervals (CIs). Risk factors with significant results in univariate analysis were consecutively included for multivariate analysis to test for robustness and independency. For survival curve estimation, the Kaplane Meier method together with the log-rank test was applied to assess significant differences in OS and DFS. All tests were calculated in a two-sided manner and P values of <0.05 were defined as statistically significant.

Ethical statement
This study has been approved by the responsible ethics committees (Medical University of Vienna EK#: 2196/2019 and National Koranyi Institute of Pulmonology ETT-TUKEB 23636e2/2018, 23 636/10/2018/EÜIG) and was conducted with regard to the Declaration of Helsinki and the institutional Good Scientific Practice guidelines.

Characterization of the study cohort
Altogether, 104 SCLC patients undergoing surgical resection between January 2000 and December 2019 were included. The median age at the time of surgery was 64 years (range 41-83 years). Forty-nine patients (47.1%) were female and 87 patients (83.7%) had tobacco exposure according to the medical records. The most common comorbidities were arterial hypertension (51%), chronic obstructive pulmonary disease (COPD) (41.3%) and diabetes mellitus (16.3%).
Regarding the type of surgery, the most common procedures were lobectomy (51%), segmentectomy (13.5%), wedge resection (11.5%) and pneumonectomy (10.6%). Fifty-seven patients (54.8%) had limited-stage disease (stage I þ II) and 62 patients (59.6%) received adjuvant chemotherapy according to local protocols. Clinicopathological data of the study population with regard to CD47 and PD-L1 expression are shown in Table 1  High presence of CD47 in surgically resected SCLC Positive CD47 expression was detected in 88 out of 100 (84.6%) patients. However, CD47 expression did not correlate with clinicopathological patient characteristics. There was additionally no association between the expression of  Figure 1A and B.

Prognostic implications of CD47 and PD-L1 expression
We lastly aimed to assess the prognostic value of CD47 and PD-L1 status for the clinical outcome regarding postoperative OS and DFS of patients.

DISCUSSION
In this study, we aimed to investigate the presence and prognostic value of the immune-checkpoint molecules CD47 and PD-L1 in a comparatively large international cohort of surgically resected SCLC patients. We identified distinct clinicopathological characteristics and the therapeutic approaches that significantly affect the outcomes of surgically resected SCLC patients. Moreover, we showed that CD47 is strongly expressed while PD-L1 presence is generally low in SCLC tissue. Ultimately, we demonstrated that stromal PD-L1 expression significantly correlates with improved outcome and represents an independent prognostic biomarker in resected SCLC.
SCLC represents one of the most aggressive types of LC since clinically it is characterized by extraordinarily poor survival rates. Although the overall clinical outcome of SCLC patients did not undergo significant improvements over the last decades, distinct demographics and treatment modalities associate with remarkable survival differences. [24][25][26] Indeed, we identified the application of adjuvant chemotherapy as one of the strongest predictors of longer OS in our study. Moreover, we also found a clear association between increasing tumor size (T2) and less-radical surgical resection (sublobar versus lobar resection), and worse DFS. Accordingly, our findings confirm comparable results from previous reports and should be considered when applying surgery in SCLC patients. [27][28][29][30] Regarding the relevance of CD47 expression in SCLC, we detected positive expression in 88% (88/100) of all evaluated patients. Clearly, CD47 expression was relatively high in our surgical SCLC cohort. In a recent study, Orozco-Morales et al. evaluated SCLC biopsy specimens and found that 7 out of 29 were positive for CD47. 19 Although this positivity rate is lower compared to our findings, SCLC may harbor a significant intratumoral heterogeneity, which is highly challenging when assessing expression signatures of various proteins of interest from small biopsy specimens. 12,31,32 Of note, our findings on the relatively high CD47 expression in malignant lung tissue are in line with other studies evaluating CD47 expression in NSCLC. Indeed, Arrieta et al. detected a CD47 positivity of 84% in tumor biopsies with a median expression of 80%. 33 In a separate study, Zhao et al. found CD47 overexpressed in 65% of the samples when evaluating primary NSCLC specimens; however, their cut-off values were significantly higher compared to our study. 34 Notably, a further study evaluated the levels of CD47 in NSCLC and pulmonary infectious disease specimens and found significantly higher CD47 expression in malignant than in infectious control tissue. 35 Accordingly, CD47 represents an immune-checkpoint molecule that is mainly overexpressed by malignant cells. Considering the prognostic value of CD47, we did not identify CD47 as a significant and independent predictor of clinical outcome although higher CD47 expression showed a clear tendency toward shortened OS (31 versus 44 months, P ¼ 0.647) and DFS (21 versus 35 months, P ¼ 0.165). This trend is consistent with data from other studies showing a correlation between high CD47 expression and worse clinical outcome in several malignancies including acute myeloid leukemia, gastric cancer and ovarian cancer. [36][37][38] Given the high expression of CD47 in our study cohort and based on the promising findings from preclinical SCLC models underlining the potency of anti-CD47 therapy in vivo, further studies are warranted to evaluate CD47 as a promising target in SCLC. 17 Conversely, we found remarkably lower tumor cell PD-L1 expression than CD47 expression with an overall positivity rate of 9.6% (10/102). Of note, however, other studies reported even lower PD-L1 expression rates in SCLC. Yasuda et al. identified only one PD-L1-expressing tumor from all 39 investigated individuals, referring to an overall positivity rate of 2.5%. 39 In another study, Schultheis et al. analyzed 94 SCLC patients and none of their specimens showed positive PD-L1 expression in tumor cells. 40 Importantly, in the current study, we also report that tumoral PD-L1 expression is more frequent among limited-stage (stage I þ II) cases compared to advanced-stage patients. Similarly, Fu et al. also reported significantly higher PD-L1 expression in limited-stage SCLC with a positivity rate of 63% among 43 surgically resected patients. 41 Accordingly, studies evaluating anti-PD-L1 blockage therapies in limited-stage SCLC patients are to be endorsed in the near future.
Similarly to our findings, Schultheis et al. found a higher presence of PD-L1 in tumor stroma although 46% of their investigated samples were small biopsies. Here, tumor stroma can substantially differ from specimens provided by surgical resection. 42 Nevertheless, 18.5% of the study cohort showed profound PD-L1 expression in tumorinfiltrating lymphocytes (TILs). 40 Regarding its prognostic value, we found a clear association between stromal PD-L1 presence and favorable OS upon surgical resection for SCLC. Other studies have found the same trends when evaluating ovarian, colorectal or breast cancer patient cohorts as we found in SCLC. [43][44][45] At first sight, our results that high PD-L1 expression correlates with better oncological outcome might seem counterintuitive. Indeed, PD-L1 is one of the best-described 'don't-kill-me' signals that malignant cells use to suppress inhibition by tumor-specific T cells and thus to evade the physiological antitumor immunity in humans. Accordingly, one might anticipate that high PD-L1 presence automatically associates with worse oncological outcome. However, there are several findings that might serve as a rational explanation for the positive prognostic value of stromal PD-L1 presence. Firstly, Ali et al. described a positive correlation between stromal PD-L1 expression and higher presence of TILs that are known to play a dominant role in the elimination of malignant cells. 46 Wang et al. also reported on significant PD-L1 up-regulation during the proliferation phase of T cells while Pulko et al. suggested that activated TILs use PD-L1 as a protective signal for their own survival. 47,48 Furthermore, up-regulation of PD-L1 signaling has been also shown to be directly promoted by cytokines involved in the antitumor immune response. [49][50][51] Given these findings, stromal PD-L1 expression might reflect an immunogenic tumor microenvironment that is able to suppress cancer growth and thus explains the association of improved clinical outcome found in our study. 52,53 Our group has recently evaluated the clinical relevance of previously presented major molecular SCLC subtypes in a large cohort of surgically resected patients and cell lines. 11,12 Beyond prognostic implications, we identified significant associations between distinct molecular subtypes and therapeutically relevant molecules. However, in the current study, neither CD47 nor PD-L1 expression showed an association with any of the major molecular SCLC subtypes.
Notably, our study has limitations that must be acknowledged. Firstly, despite that IHC is a well-established and is a routinely used method for the determination of protein-level alterations in cancer specimens, there are no standardized protocols regarding CD47 and PD-L1 evaluation. Clearly, different antibodies, staining protocols, selected cut-off values and scoring methods could lead to widely differing findings. Secondly, although our study evaluated a comparatively large cohort of surgically resected SCLC cases, the overall number of included patients still remained small. Finally, given its retrospective design, the characterization of the individual survival data has been fully extracted from the available medical records.

Conclusion
This study aimed to determine the expression pattern, clinicopathological relevance and prognostic importance of CD47 and PD-L1 expression in one of the largest international cohorts of surgically resected SCLC patients to date. We demonstrated that CD47 is highly present in SCLC samples and stromal PD-L1 expression could serve as a prognostic biomarker for a favorable outcome in SCLC. In addition, distinct demographics and therapeutic modalities were identified as relevant factors for optimal survival outcome in surgically managed SCLC patients. Given the need for improved therapeutic concepts in SCLC, further studies are indicated to verify our findings.