Molecular clusters reveal opportunities for personalised small cell lung cancer immunotherapy

Small cell lung cancer (SCLC), which accounts for 15% of lung cancers, originated from the neuroendocrine (NE) lineage.1, 2 Relapse of the disease often occurs after traditional therapy, and chemo-immunotherapy is the new first line of treatment.2 Currently, there is a lack of robust indicators to predict immunotherapy for SCLC.2 SCLC usually has higher tumour mutation burden (TMB) due to smoking and may benefit from immunotherapy.1, 2 As known to all, high immunogenicity is reflected in both the tumour microenvironments (TME) (e.g., TMB and immunogenic cell death [ICD]).2, 3 A recent research has shown ICD genes were suppressed in SCLC, we are interested in whether there is a highly immunogenic SCLC group to receive immunotherapy.4 With 45 immune-associated genes, consensus clusters of prognostic importance could be discovered (Table S1). Importantly, the SCL2 cluster with a poorer prognosis in the George cohort showed the consistent trend in the immunotherapy cohorts (Figure 1A). Although our clusters did not show significant prognostic value in the Asian cohort, there was a difference in TMB in both the George cohort and the Asian cohort (Figure 1B,C). Our clusters provided additional information as clusters and clinical parameters (e.g., age and sex) were not correlated and the prognosis of clusters was significant by univariate and multivariate Cox analyses (Table S2). A previous study has shown that tumour purity affects clustering and immune gene expression.5 Although we did not find a correlation between tumour purity and clusters and TMB, the combination of cluster and tumour purity is of prognostic significance (Figure S1A–C). After investigating the distribution of mutations, the SCL2 cluster had a larger proportion of TP53 and CSMD1 mutations than the SCL1 cluster (Figure 1D). We aim to build SCLC clusters of potential benefits from immunotherapy to test indexes and prioritise targets. The SCL2 cluster had higher TNF, im-10, RB1-loss and LCAM, but lower NE scores compared with the SCL1 cluster (Figure 1E). Although verified indicators reflect different biology, there are still indicators that are highly correlated, particularly NE and LCAM scores (Table S3). Previous studies proved RB1 deletion induces NE; however, we found NE and RB1-loss scores were almost completely unrelated (Table S3).1 Taken together, our immune-related cluster is a predictive platform for testing different indexes. Furthermore, we analysed the composition of the TME between two clusters. Compared with the SCL1 cluster, the SCL2 cluster showed higher enrichment of CD8+ T, macrophages, NK, T helper, T helper 1, T helper 2 and cytotoxic cells, but lower CD56 bright NK and effector memory T cells (Figure 1F). Taken together, SCL1 had a lower cellular infiltration than SCL2 and was immunosuppressed probably with cells in a memory state. Meanwhile, higher abundance of immune checkpoints (ICPs) and ICD genes, except TNFRSF25, were in SCL2 but not the SCL1 cluster (Figure 1G,H). Previous studies have demonstrated the lack of ICPs in SCLC and the absence of robust indicators to predict immunotherapy.2 Twenty-six molecules were identified after differential expression gene (DEG) analysis in samples and cell lines (Table S4). We successfully revealed previously reported drivers, including MYC and WWTR1, and had some overlap with the non-NE features (5/15, Table S1).2 It implies that our results are highly reliable and conservative. Noting the example of the RB1-loss and NE scores, we should be cautious about defining the state of NE. There may be other neurosecretory pathways, based on our finding that activated CALCRL in SCL2 cluster. We suggest that further exploration of CALCRL is necessary for NE state.2 Given the TMB as a measure of immunogenicity could be reflected in the clusters regardless of either Asian or European ancestry.2 Conceivably, markers enriched in SCL2 may be effective targets or biomarkers for immunotherapy. Perturbation of pathways was then explored between SCL1 and SCL2 clusters, upregulation of development and downregulation of apoptosis and inflammation in SCL1 but not SCL2 cluster (Figure S2A; Table S5). Meanwhile, disrupted lung development was associated with poor prognosis (Figure S2B,C). Our results did not refute the studies of Wagner et al. and Roper et al. in that treatment may cause pathway changes, but rather we have identified three important genes directly associated with immunotherapy, including B2M, CXCL13 and CTLA4(Figure S2D).6-8 After we performed DEG and prognostic analyses for pathways and genes, only IL4I1 and LAG3 were related to anti-tumour immune responses (Figure S3A–C; Table S6). In conclusion, our results prove that the SCL2 is highly immunogenic, based on high TMB and ICD, but may benefit from combined immunotherapy (e.g., PARP inhibitors with ICP inhibitors), while less resistance to immunotherapy was found in the SCL1 cluster. Counter-intuitively, patients with immune hot may require more active ICI combined therapies rather than monotherapy. This finding has striking parallels to NCT02484404 trial.9 To our knowledge, SCL2 have a poor prognosis due to the immune-escape mechanism.10 Limitations of our work are the lack of Asian cohorts and retrospective analysis. Overall, we established the paradigm of immunotherapy to determine the priority of targets. This study supports the application of immunotherapy in SCLC. We appreciate and acknowledge the data received from the George, Pender and Mariathasan cohorts and Polley, Cai and Jiang et al. The authors declare that there is no conflict of interest. National Natural Science Foundation of China, Grant Numbers: 81773245, 81972858, 82172670; Technology Innovation and Application Development Project of Chongqing, Grant Number: cstccxljrc201910; Cultivation Program for Clinical Research Talents of Army Medical University, Grant Number: 2018XLC1010 Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.