J Immunother Cancer: first published as 10.1136/jitc-2022-005320 on 28 September 2022. Downloaded from http://jitc.bmj.com/ on July 14, 2023 at Sant Paul Hospital Consortia. Protected Open access Original research Tumor microenvironment gene expression profiles associated to complete pathological response and disease progression in resectable NSCLC patients treated with neoadjuvant chemoimmunotherapy Marta Casarrubios ‍ ‍ ,1 Mariano Provencio ‍ ‍ ,1 Ernest Nadal ‍ ‍ ,2 Amelia Insa ‍ ‍ ,3 María del Rosario García-­Campelo ‍ ‍ ,4 Martín Lázaro-­Quintela ‍ ‍ ,5 Manuel Dómine ‍ ‍ ,6 Margarita Majem ‍ ‍ ,7 Delvys Rodriguez-­Abreu ‍ ‍ ,8 Alex Martinez-­Marti ‍ ‍ ,9 Javier De Castro Carpeño ‍ ‍ ,10 Manuel Cobo ‍ ‍ ,11 Guillermo López Vivanco ‍ ‍ ,12 Edel Del Barco ‍ ‍ ,13 Reyes Bernabé ‍ ‍ ,14 Nuria Viñolas ‍ ‍ ,15 Isidoro Barneto Aranda ‍ ‍ ,16 Bartomeu Massuti ‍ ‍ ,17 Belén Sierra-­Rodero ‍ ‍ ,1 Cristina Martinez-­Toledo ‍ ‍ ,1 Ismael Fernández-­Miranda ‍ ‍ ,1 Roberto Serna-­Blanco ‍ ‍ ,1 Atocha Romero ‍ ‍ ,1 Virginia Calvo ‍ ‍ ,1 Alberto Cruz-­Bermúdez ‍ ‍ 1 by copyright. To cite: Casarrubios M, Provencio M, Nadal E, et al. ABSTRACT Tumor microenvironment gene Background Neoadjuvant chemoimmunotherapy for non-­ KEY MESSAGES expression profiles associated to small cell lung cancer (NSCLC) has improved pathological ⇒ Neoadjuvant chemoimmunotherapy for non-­ small complete pathological response cell lung cancer (NSCLC) has improved clinical out- responses and survival rates compared with chemotherapy and disease progression in comes. However, in the context of the recent FDA alone, leading to Food and Drug Administration (FDA) resectable NSCLC patients approval of nivolumab plus chemotherapy for resectable approval of nivolumab plus chemotherapy combina- treated with neoadjuvant chemoimmunotherapy. Journal stage IB-­IIIA NSCLC (AJCC 7th edition) without ALK or tion in this setting, there is a lack of knowledge and for ImmunoTherapy of Cancer EGFR alterations. Unfortunately, a considerable percentage biomarkers to identify patients who achieve com- 2022;10:e005320. doi:10.1136/ of tumors do not completely respond to therapy, which plete pathologic responses, or to determine patients jitc-2022-005320 has been associated with early disease progression. So who are at higher risk of disease recurrence after far, it is impossible to predict these events due to lack surgery. ► Additional supplemental of knowledge. In this study, we characterized the gene ⇒ Our study supports that tumors achieving CPR con- material is published online only. expression profile of tumor samples to identify new stitute a distinct biological entity, with a different To view, please visit the journal biomarkers and mechanisms behind tumor responses baseline profile and undergoing greater therapy-­ online (http://​dx.d​ oi.​org/​10.​ to neoadjuvant chemoimmunotherapy and disease induced changes than non-­CPRs. Our results identi- 1136/j​ itc-​2022-0​ 05320). recurrence after surgery. fy possible key players and mechanisms to achieve Methods Tumor bulk RNA sequencing was performed in complete responses, and reinforce the use of CPR Accepted 13 September 2022 16 pretreatment and 36 post-­treatment tissue samples as an endpoint in neoadjuvant clinical trials. from 41 patients with resectable stage IIIA NSCLC treated ⇒ In addition, these differences in gene expression with neoadjuvant chemoimmunotherapy from NADIM trial. profile could be used as a source for CPR predic- A panel targeting 395 genes related to immunological tive biomarkers, with greater ability to describe the © Author(s) (or their employer(s)) 2022. Re-­use processes was used. Tumors were classified as complete immune landscape and better performance than permitted under CC BY-­NC. No pathological response (CPR) and non-­CPR, based on the PD-­L1 and TMB, as well as for the identification of commercial re-­use. See rights total absence of viable tumor cells in tumor bed and lymph patients without CPR at high risk of disease recur- and permissions. Published by nodes tested at surgery. Differential-­expressed genes rence, which would allow personalization of follow-­ BMJ. up and adjuvant treatment in these patients. between groups and pathway enrichment analysis were For numbered affiliations see assessed using DESeq2 and gene set enrichment analysis. end of article. CIBERSORTx was used to estimate the proportions of Correspondence to immune cell subtypes. macrophages, all with a significant area under the Dr Alberto Cruz-­Bermúdez; Results CPR tumors had a stronger pre-­established receiver operating characteristic curve (ROC) >0.9 for CPR ​alberto.​cruz.b​ ermudez@​gmail.​ immune infiltrate at baseline than non-­CPR, characterized prediction. A greater effect of neoadjuvant therapy was com by higher levels of IFNG, GZMB, NKG7, and M1 also seen in CPR tumors with a reduction of tumor markers Casarrubios M, et al. J Immunother Cancer 2022;10:e005320. doi:10.1136/jitc-2022-005320 1 J Immunother Cancer: first published as 10.1136/jitc-2022-005320 on 28 September 2022. Downloaded from http://jitc.bmj.com/ on July 14, 2023 at Sant Paul Hospital Consortia. Protected Open access and IFNγ signaling after treatment. Additionally, the higher expression of METHODS several genes, including AKT1, BST2, OAS3, or CD8B; or higher dendritic Study design and patient samples cells and neutrophils proportions in post-­treatment non-­CPR samples, All patients with available primary tumor tissue sample were associated with relapse after surgery. Also, high pretreatment PD-­ for molecular determinations from NADIM clinical L1 and tumor mutational burden levels influenced the post-­treatment trial (NCT03081689) were included in this study. In the immune landscape with the downregulation of proliferation markers and NADIM trial, patients with resectable stage IIIA NSCLC type I interferon signaling molecules in surgery samples. Conclusions Our results reinforce the differences between CPR and were treated with three cycles of nivolumab (360 mg) plus non-­CPR responses, describing possible response and relapse immune chemotherapy, consisting of paclitaxel (200 mg/m2) and mechanisms, opening the possibility of therapy personalization of carboplatin (area under the curve (AUC) 6; 6 mg/mL per immunotherapy-­based regimens in the neoadjuvant setting of NSCLC. min), before surgical resection. Patients received adju- vant treatment with 240 mg of nivolumab every 2 weeks for 4 months, followed by 480 mg of nivolumab every 4 BACKGROUND weeks during the next 8 months. The objective patholog- Neoadjuvant chemoimmunotherapy introduction in ical response was assessed by blinded pathology central resectable stages of lung cancer has led to improved review which determined the percentage of residual patient outcomes in terms of pathological responses and viable tumor in resected primary tumors. Tumors were showed promising survival rates,1–3 which has led to the subsequently classified into two categories according recent FDA approval of nivolumab in combination with to their pathological response: complete pathological chemotherapy for resectable stage IB-­IIIA non-­small cell response (CPR) and non-­CPR tumors. CPR tumors were lung cancer (NSCLC) (American Joint Committee on those with no viable tumor cells in the resected tumor Cancer, AJCC 7th edition) without ALK or EGFR alter- bed and lymph nodes analyzed, and non-­CPR tumors, ations.4 However, a considerable percentage of tumors do those who have any percentage of viable tumor cells in not completely respond to chemoimmunotherapy, which the resection specimens. has been associated with early disease progression. Thus, Fifty-­three tumor tissue samples of patients enrolled the identification of responder patients, as well as those in NADIM clinical trial were collected and subjected to non-­responders at higher risk of disease recurrence, is RNA extraction and sequencing. Fifty-­two of 53 samples crucial. had a good sequencing quality and were analyzed: 16 In this sense, several biomarkers related to the tumor by copyright. of them were obtained at diagnosis and 36 after tumor immunity cycle have been proposed, ranging from the resection (online supplemental table 1). Eleven of them presence of neoantigens (measured indirectly through were paired samples (pretreatment and post-­treatment) tumor mutational burden (TMB)),5 the expression levels in which RNA sequencing was done at both timepoints. of PD-­L1 (measured by immunohistochemistry in tumor The median follow-­up time of the entire cohort at data sections),6 7 to the direct characterization and quantifi- cut-­off was 38.0 months (95% CI 36.7 to 40.7), with a 94% cation of immune cells in tissue or blood through: flow maturity at 36 months. The disease progression status of cytometry,8 multiple immunofluorescence panels,1 or the patients was determined at 34.2 months, that is, the T cell repertoire analysis (by sequencing of the TCR follow-­up time of the patient without disease progres- receptor).9 Additionally, ctDNA levels, as a reflection of sion with the shortest follow-­up available. Regarding this, the patient’s tumor burden, have been shown to be useful patients with tumors with non-­ complete pathological in this context.3 10 11 There are considerable data for meta- response (post-­treatment non-­CPR tumors, n=14) were static NSCLC and immune monotherapy,12 being data in categorized as patients with disease progression (n=5) or locally advanced stages more limited, showing contradic- patients with non-­disease progression (n=9). Two patients tory results for the main biomarkers PD-­L1 and TMB.11 (32 and 35) were excluded from survival analyses and In this sense, gene expression profiles, considering analyses based on progression status since they died from hundreds of genes, can give a more comprehensive view COVID-­19 (and both presented no evidence of disease of the antitumor immune response, overcoming some status at the time of death), compromising the identi- limitations of established biomarkers.13 fication of disease progression biomarkers. However, In this study, we characterize the gene expression they were included in all the other analyses involving profile of tumor samples from locally advanced resect- the pathological response status, the effect of therapy, able NSCLC patients before and after neoadjuvant treat- and the relevance of PD-­L1 and TMB. Details of clinico- ment by bulk RNA-­seq, for the discovery of biomarkers pathological characteristics of all patients, pathological associated with complete pathological responses and the responses, survival outcomes, sequencing quality param- identification of patients with non-­complete pathological eters, samples and techniques used are summarized in response at high risk of disease progression. online supplemental table S1. PD-L1 tumor proportion score and TMB assessment As previously reported, formalin-­fixed paraffin-­embedded (FFPE) tumor diagnostic samples were used to assess the tumor mutational burden (TMB) with the Oncomine 2 Casarrubios M, et al. J Immunother Cancer 2022;10:e005320. doi:10.1136/jitc-2022-005320 J Immunother Cancer: first published as 10.1136/jitc-2022-005320 on 28 September 2022. Downloaded from http://jitc.bmj.com/ on July 14, 2023 at Sant Paul Hospital Consortia. Protected Open access Tumor Mutation Load Assay (Thermo Fisher Scientific) not consider genes individually but as a group. Thus, using an Ion S5 Sequencer (Thermo Fisher Scientific, small differences in the expression of a group of genes Palo Alto, California, USA) according to the manufac- can give rise to a significant differential expression of the turer’s instructions. The commercially available PD-­ L1 signaling pathway of which they are part, even if they did immunohistochemistry assay PD-­L1 IHC 22C3 pharmDx not differ in their expression individually. (Dako, Glostrup, Denmark) was used to assess PD-­ L1 To estimate the proportions of the immune cell subsets tumor proportion score (TPS) in FFPE tumor diagnostic present in samples using RNA-­seq data, the analytical tool samples.1 CIBERSORTx was used.17 This tool can accurately deter- A cut-­off of ≥25% for PD-­L1 TPS was used to classify mine the immune infiltration of each sample by using pretreatment tumors into PD-­ L1-­ high and PD-­L1-­low a deconvolution algorithm and a signature reference of expression, as it showed the best performance to iden- immune cells (LM22), which can define 22 subtypes of tify complete pathological responses. The median TMB immune cells based on a gene signature matrix of 547 of the entire cohort (5.89 Mut/Mb) was used to classify genes. The signature matrix used was the LM22 with tumors with TMB-­high and TMB-­low.1 100 permutations with B mode batch correction and with quantile normalization disabled. The input data for RNA extraction CIBERSORTx analysis was the TPM normalization of the Bulk RNA derived from biopsy FFPE samples at diagnosis RNA-­seq data, done by normalizing each gene length and surgery specimens were extracted with truXTRAC with the total read counts. Samples with a CIBERSORTx FFPE DNA kit (Covaris) as per manufacturer’s instruc- p value <0.05 were used for the analysis. tions. RNA quantification was done using the Qubit RNA BR Assay Kit (Thermo Fisher Scientific Cat. No. Q10210) Statistical analysis on a Qubit apparatus. Heatmaps and PCA plots were done with ggplot2 R package using as an input the variance stabilizing trans- Library construction and sequencing formation (vsd) normalization extracted with DESeq2 Only tissue samples derived from primary tumors were analysis. Volcano plots were depicted with the web app sequenced, not considering lymph node-­derived samples. VolcaNoseR18 using as an input the results of the differ- Briefly, 10 ng of input RNA was reverse transcribed into ential expressed genes extracted from DESeq2 analysis. cDNA using the SuperScript IV VILO Master Mix (Cat. In the differential expression analysis, a multiple testing by copyright. No. 11756050), and libraries were prepared using the correction was performed using the Benjamini-­Hochberg Ion AmpliSeq Library Kit 2.0 (Life Technologies Cat. procedure to reduce the number of false positives. No. 4475345) and the Oncomine Immune Response Regarding the GSEA analysis, a false discovery rate (FDR) Research Assay (Cat. No. A31930) panel that targets 395 <5% was considered to identify the significantly upregu- genes related to immunological processes (online supple- lated pathways within each group. Non-­parametric Mann-­ mental table S2). Equal volumes from eight samples at 50 Whitney U test was used to analyze differences between pM were pooled together for sequencing on an Ion 530 CPR and non-­CPR tumors and between patients with or Chip. Template preparation was performed using an Ion without disease progression. Wilcoxon rank-­sum test was Chef System (Thermo Fisher Scientific) and Ion 520 and used for pairwise analysis between pretreatment and post-­ Ion 530 Kit (Thermo Fisher Scientific). Sequencing was treatment timepoints. The AUC ROC analysis was used performed on an Ion S5 XL Sequencer (Thermo Fisher to determine the capability of DEGs and immune cell Scientific). The ImmuneResponse RNA plugin was used subsets as predictive biomarkers. Values with the highest to align the sequences to the reference genome (Immu- likelihood ratio from the ROC curve analysis were used as neResponse_v3.1) and to count the sequencing reads thresholds to categorize DEGs or immune cell subsets for with the Torrent Suite software. each sample into high or low groups in the identification of patients at high risk of relapse after surgery. Analysis RNA sequencing data analysis of progression-­ free survival (PFS) and overall survival Differential-­expressed genes (DEGs) between groups were (OS) was done with the Kaplan-­ Meier curve analysis, assessed using the DESeq2 package in R.14 Raw counts comparing groups with the log-­rank test. An adjusted p data were used as an input for DESeq2 analysis. The cut-­ value of <0.05 was considered statistically significant. IBM off criteria for assessing the DEGs was |log2FC|>1.5 and SPSS software (V.25.0) and GraphPad Prism (V.8.4) were adjusted p<0.05. used in the statistical analysis. Pathway enrichment analysis was done using gene set enrichment analysis (GSEA).15 16 The input data for GSEA was the output normalized counts from DESeq2. RESULTS The gene sets used for GSEA analysis were the 35 Onco- Differential immune landscape in pretreatment samples mine Immune Response gene function annotation sets, allows prediction of complete pathological responses at described in online supplemental table S2. The differ- surgery ence between this analysis and the one of differential Gene expression levels, using a panel of 395 immune-­ gene expression performed with DESeq2 is that it does related genes, were compared between pretreatment Casarrubios M, et al. J Immunother Cancer 2022;10:e005320. doi:10.1136/jitc-2022-005320 3 J Immunother Cancer: first published as 10.1136/jitc-2022-005320 on 28 September 2022. Downloaded from http://jitc.bmj.com/ on July 14, 2023 at Sant Paul Hospital Consortia. Protected Open access samples of patients with CPR (n=9) and non-­CPR (n=5) immune cells (p=0.012), as well as a higher proportion tumors. The unsupervised heatmap based on all genes of follicular helper CD4+ T cells (p=0.038), activated NK does not perfectly classify patients with CPR and non-­CPR cells (0.053), M1 macrophages (p=0.002) and resting tumors into two distinct groups. However, six out of nine dendritic cells (0.011) than non-­CPR patients. However, patients with CPR tumors are clustered together, as also only the proportion of M1 macrophages was statisti- seen in the PCA plot (online supplemental figure 1A,B). cally significant after multiple comparisons correction Specifically, 25 DEGs were identified between CPR (figure 1E). The AUC ROC for predicting the response and non-­CPR tumors (online supplemental table S3). Of in these tumors based on the proportion of M1 macro- them, 17 were upregulated and 8 were downregulated in phages in pretreatment samples was 0.9778 (95% CI CPR compared with non-­CPR tumors (figure 1A). Among 0.910 to 1.045) (p=0.004). A proportion of M1 macro- the upregulated genes in CPR tumors, some were involved phages higher than 0.04471 showed 100% sensitivity and in type II interferon signaling (FASLG, CXCL13, CXCL10, 88.9% specificity in identifying tumors that will achieve CXCL9 and IFNG); other genes were related to NK cells a complete pathological response after immunochemo- (NCR1, KIR2DL3 and GNLY); lymphocyte enrichment therapy (figure 1F). and activation (NKG7, GZMB, IL2RB, CD38, IDO1 and TNFRSF9 aka 4-­ 1BB); and tumor antigens (MAGEA10, Differential immune landscape between CPR and non-CPR MAGEA1 and GAGE1). Altogether, all these genes depict tumors in post-treatment samples a proinflammatory phenotype in pretreatment tumor Regarding surgical samples, the unsupervised heatmap samples of patients with CPR tumors. However, the eight considering all genes was unable to classify patients with genes upregulated in non-­CPR tumors were of diverse CPR and non-­ CPR tumors into two separate groups functions: the myeloid marker ARG1, the TNF receptor (online supplemental figure 3A). Similar limitation was TNFSF18, the housekeeping gene TFRC, the tumor found in the PCA (online supplemental figure 3B). antigen MAGEA12, the proliferation markers MAPK1 and Nevertheless, DEGs between CPR and non-­CPR tumors CDK1, and IL1B and MIF involved in type II interferon were identified (online supplemental table S5). Twen- signaling and innate immune response, respectively ty-­two genes were found to be upregulated in non-­CPR (figure 1A). tumors, most of them related to proliferation (CDKN3, Analyzing the individual ability of DEGs at baseline to CCNB2, KIAA0101, MKI67, BUB1, CDK1, TOP2A, FOXM1, identify those patients with tumors that achieve CPR at by copyright. MELK, MAD2L1, G6PD and DGAT2) or tumor markers surgery (figure 1B), IFNG stands out, showing an AUC (CDKN2A, KRT5, BRCA1 and TWIST1), among others of 1.000 (95% CI 1.000 to 1.000) (p=0.002). An IFNG (MAGEA3, CEACAM1, CXCL8, TNFRSF18, HMBS and TPM value higher than 27.08 seems to classify with 100% ISG15) (figure 2A). sensitivity and 100% specificity tumors that will achieve CPR (figure 1C). In addition, GZMB and NKG7, related Among the DEGs, we performed a comparison of to cytotoxic activity of T and NK cells, showed an AUC their TPM values between CPR and non-­ CPR tumors ROC of 0.933 (95% CI 0.793 to 1.000) (p=0.009) and and represented those whose p values were significant 0.977 (95% CI 0.910 to 1.000) (p=0.004), respectively, after multiple testing correction (p<0.002): the cyclin when classifying CPR and non-­CPR tumors (figure 1C). B2 CCNB2 gene and CDKN3 both involved in the cell The complete list of AUC ROC for the remaining genes, cycle, and ISG15 involved in type I interferon signaling including CD38, CXCL9 and IL2RB, which are also impli- (p=0.001, figure 2B). The AUC ROC for distinguishing cated in T cell-­mediated immune responses, are shown CPR and non-­CPR tumors based on the TPM value of in online supplemental table S4. Among the upregulated these three genes were 0.821, 0.818 and 0.814, respec- genes in non-­CPR tumors, CDK1, ARG1 and TNFSF18 tively (figure 2C). showed an AUC of 0.911, 0.888 and 0.866, respectively, Further GSEA analysis in surgical samples from patients when classifying CPR and non-­ CPR responses (online with CPR tumors showed an upregulation of pathways supplemental table S4). related to antigen processing, TCR coexpression and Functional enrichment analysis with GSEA software lymphocyte infiltrate, suggesting the presence of a higher was also performed to characterize the upregulated path- immune infiltrate in these tumors. In contrast, non-­CPR ways within each pathologic response group, revealing an tumors showed an upregulation of proliferation, tumor upregulation of pathways related to TCR coexpression, marker, interferon signaling, housekeeping and tumor lymphocyte infiltrate, type II interferon signaling and antigen pathways that may be explained by the pres- antigen processing in CPR tumors. In the case of non-­ ence of viable tumor cells in these surgical specimens CPR tumors, there was an upregulation of tumor markers, (figure 2D). housekeeping genes, proliferation and PD-­ 1 signaling When looking for differences between CPR and non-­ pathways (figure 1D). CPR tumors in the estimated immune cell infiltrate Finally, the CIBERSORTx algorithm was also used (online supplemental figure 2B), a higher proportion to estimate the relative proportion of immune cell of follicular helper CD4+ T cells in non-­CPR tumors was subsets in these tissue samples (online supplemental found, though this was not statistically significant after figure 2A). CPR tumors showed a higher score of total correcting for multiple testing (p=0.025, figure 2E). 4 Casarrubios M, et al. J Immunother Cancer 2022;10:e005320. doi:10.1136/jitc-2022-005320 J Immunother Cancer: first published as 10.1136/jitc-2022-005320 on 28 September 2022. Downloaded from http://jitc.bmj.com/ on July 14, 2023 at Sant Paul Hospital Consortia. Protected Open access Figure 1 Differential immune landscape in pretreatment samples. (A) Volcano plot showing DEGs between CPR and non-­CPR by copyright. tumors. Red dots indicate upregulated genes in CPR tumors, while blue dots indicate upregulated genes in non-­CPR tumors. (B) Expression levels in TPM of IFNG, GZMB and NKG7. (C) ROC curve for the prediction of complete pathological response using median TPM expression levels of IFNG, GZMB and NKG7, as cut-­off. (D) Gene set enrichment analysis for pathological responses. Differentially upregulated pathways in non-­CPR (left) and CPR tumors (right). (E) Absolute immune cell score and frequency of each immune cell subtype in CPR and non-­CPR tumors determined by CIBERSORTx. P<0.002 was considered statistically significant after Bonferroni’s correction for multiple tests. (F) ROC curve for the prediction of complete pathological response using median proportion of M1 macrophages as cut-­off. Each patient is represented by a black symbol. Comparisons were done between CPR (n=9) and non-­CPR (n=5) groups. CPR, complete pathological response; DEGs, differential-­expressed genes. Identification of patients at high risk of relapse after surgery in patients with non-­CPR tumors, we have analyzed non-­ Patients whose tumors do not achieve complete responses CPR surgery specimens and identified DEGs between are likely at higher risk of relapse. To identify gene tumors of patients with or without disease progression expression patterns that may affect long-­term outcomes (online supplemental table S6). Ten genes were identified Casarrubios M, et al. J Immunother Cancer 2022;10:e005320. doi:10.1136/jitc-2022-005320 5 J Immunother Cancer: first published as 10.1136/jitc-2022-005320 on 28 September 2022. Downloaded from http://jitc.bmj.com/ on July 14, 2023 at Sant Paul Hospital Consortia. Protected Open access by copyright. Figure 2 Differential immune landscape in post-­treatment samples. (A) Volcano plot showing the DEGs between CPR and non-­CPR tumors in post-­treatment samples. Red dots indicate upregulated genes in CPR tumors, while blue dots indicate upregulated genes in non-­CPR tumors. (B) Expression levels in TPM of CCNB2, CDKN3 and ISG15. (C) ROC curve for the assessment of complete pathological response using the median TPM expression levels of CCNB2, CDKN3 and ISG15 as cut-­off. (D) Gene set enrichment analysis for pathological responses. Differentially upregulated pathways in non-­CPR and CPR post-­treatment tumors. (E) Frequency of follicular helper T cells in tumors determined by CIBERSORTx analysis. P<0.002 was considered statistically significant after Bonferroni’s correction for multiple tests. Each patient is represented by a black symbol. Comparisons were done between CPR (n=22) and non-­CPR (n=14) groups. CPR, complete pathological response; DEGs, differential-­expressed genes. as differentially upregulated in tumors from patients with (PFS). In addition, higher levels of HMBS (p=0.018) disease progression: IFI6 and OAS3, both involved in inter- were associated with lower overall survival (OS) (online feron signaling; AKT and KRT7 as tumor markers; BST2, supplemental figure 4). Of special interest is AKT, in ISG15 and IFI27 involved in type I interferon signaling as which higher levels in post-­treatment samples were asso- well as CD8B, HMBS and OAS1 (figure 3A). Importantly, ciated with both lower PFS and OS (p=0.033 and p=0.003, of these genes associated with disease progression, AKT1, respectively). The PFS and OS of patients whose tumors BST2, OAS3, CD8B, IF27 and KRT7, were not associated expressed low levels of AKT were 80% (95% CI 40.87 to with the percentage of viable tumor cells in the surgical 94.58) and 100% (95% CI not estimable) at 36.0 months, specimen (online supplemental table S7). Additionally, respectively, compared with 25% (95% CI 0.90 to 66.53) the expression of these genes, categorized as high and in patients with high AKT tumors for both PFS and OS low expression, was further correlated with survival. We (figure 3B). There were no differentially regulated path- have found that higher levels of IFI6 (p=0.010), BST2 ways between patients with or without disease progression. (p=0.010), CD8B (p=0.019), OAS3 (p=0.010) and IFI27 Regarding the estimated cell proportions, no differ- (p=0.010) in post-­treatment samples of non-­CPR tumors ences were observed between tumors of patients with or were associated with lower progression-­ free survival without disease progression (online supplemental figure 6 Casarrubios M, et al. J Immunother Cancer 2022;10:e005320. doi:10.1136/jitc-2022-005320 J Immunother Cancer: first published as 10.1136/jitc-2022-005320 on 28 September 2022. Downloaded from http://jitc.bmj.com/ on July 14, 2023 at Sant Paul Hospital Consortia. Protected Open access by copyright. Figure 3 Immune expression signature associated with disease progression in patients with non-­CPR tumors. (A) Volcano plot showing the DEGs between post-­treatment tumors from patients with disease progression (n=5) versus no disease progression (n=9). Red dots indicate upregulated genes in tumors from patients with disease progression, while blue dots indicate upregulated genes in tumors from patients with non-­disease progression. (B) Kaplan-­Meier plots of progression-­free survival and overall survival for patients with high (n=4) and low (n=10) expression of AKT (p=0.033 and p=0.003, respectively). (C) Kaplan-­Meier plots of progression-­free survival and overall survival for patients with high (n=4) and low (n=19) proportion of activated dendritic cells (p=0.019 and p=0.052). (D) Kaplan-­Meier plots of progression-­free survival and overall survival for patients with high (n=4) and low (n=10) proportion of neutrophils (p=0.033 and p=0.003). CPR, complete pathological response; DEGs, differential-­expressed genes. 2B). However, a higher proportion of activated dendritic them in two different clusters (online supplemental cells or neutrophils in non-­CPR tumors were associated figure 5A). This can also be seen in the PCA, where two with lower PFS and OS (figure 3C and D). The PFS and large clusters corresponding to pretreatment and post-­ OS at 36.0 months of patients whose tumors had a higher treatment samples are easily identified. Interestingly, all proportion of activated dendritic cells were 25% (95% CI the surgery samples that appear in the PCA pretreatment 0.90 to 66.53) and 50% (95% CI 5.79 to 84.48), respec- cluster were from non-­CPR tumors (online supplemental tively, compared with 80% (95% CI 40.87 to 94.58) and figure 5B). 85.7% (95% CI 33.40 to 97.84) in those who presented Patients with CPR tumors showed differential expres- lower levels (p=0.019 for PFS, and p=0.052 for OS). sion of up to 74 genes between their paired pretreatment Similarly, PFS and OS at 36.0 months for patients with and post-­treatment tissue samples (online supplemental a higher proportion of estimated neutrophils was 25% table S8). Up to 47 genes were found to be downregu- (95% CI 0.90 to 66.53) for both parameters; meanwhile, lated after neoadjuvant chemoimmunotherapy in post-­ it was 80% (95% CI 40.87 to 94.58) and 100% (95% CI treatment tumor samples of CPR patients: 17 of them not estimable) for patients with a lower proportion of were related to proliferation (MAD2L1, CDK1, MKI67, neutrophils (p=0.033 for PFS and p=0.003 for OS). KIAA0101, CDKN3, BUB1, MELK, TOP2A, FOXM1 and Neoadjuvant chemoimmunotherapy induces a stronger CCNB2) and tumor markers (BRCA2, PGF, BRCA1, immune infiltrate in CPR compared with non-CPR tumors CDKN2A, TRIM29, KRT5 and TWIST1), while 12 were The unsupervised clustering of all samples (diagnostic attributed to the interferon signaling (IFI6, MX1, OAS3, samples and resection specimens) seems to clearly group IFIT3, ISG15, TAP1, OAS1, STAT1, CXCL9, IFNG, CXCL10 Casarrubios M, et al. J Immunother Cancer 2022;10:e005320. doi:10.1136/jitc-2022-005320 7 J Immunother Cancer: first published as 10.1136/jitc-2022-005320 on 28 September 2022. Downloaded from http://jitc.bmj.com/ on July 14, 2023 at Sant Paul Hospital Consortia. Protected Open access by copyright. Figure 4 Changes in the immune landscape during treatment. (A) Volcano plot showing the DEGs between paired post-­ treatment and pretreatment CPR tumor samples (n=7). Red dots indicate upregulated genes, while blue dots indicate downregulated genes in post-­treatment samples. (B) Differentially upregulated (upper panel) and downregulated pathways (lower panel) in post-­treatment CPR tumor samples. (C) Frequency of immune cell subtypes obtained with CIBERSORTx analysis in pretreatment and post-­treatment timepoints from CPR tumors. (D) Volcano plot showing the DEGs between paired post- and pretreatment non-­CPR tumor samples (n=4). Red dots indicate upregulated genes, while blue dots indicate downregulated genes in post-­treatment samples. (E) Differentially upregulated pathways in post-­treatment non-­CPR tumor samples. (F) Frequency of immune cell subtypes in non-­CPR tumors and comparison between pretreatment and post-­treatment timepoints. CPR, complete pathological response; DEGs, differential-­expressed genes. and CXCL11). Also, IFNG and GZMB were upregulated antigen receptor complex-­associated CD79A and CD79B in pretreatment tumor samples of patients with CPR with genes (figure 4A). respect to their post-­treatment counterparts. In addition, Further analysis with GSEA software confirmed the the highest reductions belonged to the proinflammatory downregulation of type I and II interferon signaling, IL17A (FC=10.58) and IL4 (FC=9.10) genes (figure 4A). tumor antigen, and proliferation pathways in post-­ Concerning upregulated genes after treatment in CPR treatment samples of these patients. Moreover, post-­ tumors, 27 genes were increased with respect to their treatment samples revealed an upregulation of pathways pretreatment paired samples: some of them reflecting related to lymphocyte infiltrate, antigen processing and a lymphocyte infiltrate (CCL21, CXCR4, GZMK, CD52, TCR coexpression (figure 4B). IL7R, LAMP3 and PTPRC), neutrophils (CA4) or macro- In addition, the CIBERSORTx analysis showed a phages (MRC1, ALOX15B) among others. Of note is the higher proportion of CD8+ T cells (p=0.018), memory upregulation of genes related to B cells, such as the B resting CD4+ T cells (p=0.018) and resting dendritic cells cell markers JCHAIN, FCRLA and CD22, and the B cell (p=0.028); and lower proportions of follicular helper 8 Casarrubios M, et al. J Immunother Cancer 2022;10:e005320. doi:10.1136/jitc-2022-005320 J Immunother Cancer: first published as 10.1136/jitc-2022-005320 on 28 September 2022. Downloaded from http://jitc.bmj.com/ on July 14, 2023 at Sant Paul Hospital Consortia. Protected Open access CD4+T cells (p=0.028) and M1 macrophages (p=0.018) of high PD-­L1 expression at pretreatment seem to have in post-­treatment samples of CPR patients, showing lower proportions of follicular helper T cells (p=0.004) no differences in the absolute score of immune cells and M2 macrophages (p=0.040), though it was not signif- (figure 4C). icant after multiple testing correction (figure 5I). Regarding non-­ CPR tumors, only eight genes were When looking at the influence of baseline TMB status differentially expressed between pretreatment and in post-­treatment samples, we found that 14 genes were post-­ treatment samples (online supplemental table downregulated after chemoimmunotherapy in post-­ S9). TRIM29, ILI1A, KRT5, MMP9, S100A9 and CD44 treatment samples whose pretreatment samples had a were downregulated in post-­treatment samples, whereas higher TMB: the proliferation markers FOXM1, TOP2A, JCHAIN and DMBT1 were upregulated (figure 4D). GSEA BUB1, MKI67 and KIAA0101; the housekeeping genes analysis of pathways revealed an upregulation of lympho- G6PD and HMBS; the neutrophil marker KREMEN1; the cyte infiltrate and TCR coexpression pathways in post-­ interferon signaling genes OAS3, ISG15 and IFI27; the treatment samples of non-­CPR patients (figure 4E). In negative regulator of T cell mediated immune response contrast with CPR tumors, no differences were seen in the VTCN1 and the genes involved in the innate immune proportions of immune cells subtypes analyzed between response LCN2 and S100A8 (figure 5F). Additionally, type pretreatment and post-­treatment samples for non-­CPR I interferon signaling and proliferation pathways were tumors (figure 4F). shown to be downregulated in post-­treatment samples of high TMB status pretreatment (figure 5H). Finally, no Pretreatment PD-L1 and TMB status is not associated differences were seen in the proportions of immune cell with distinct basal immune landscape but impacts it after subtypes analyzed (online supplemental figure 2A). neoadjuvant treatment The immune landscape between high and low PD-­ L1 TPS tumors and between TMB high and low tumors was DISCUSSION characterized. Concerning PD-­L1 expression, only two Neoadjuvant chemoimmunotherapy has demonstrated differentially expressed genes were found upregulated in safety, antitumor activity, and superiority, in terms of PD-­L1 high tumors: the tumor antigen GAGE1, and the pathological responses and survival, compared with gene that encodes the PD-­L1 protein, CD274 (figure 5A). chemotherapy in different phase II and phase III GSEA analysis showed an upregulation of the TCR coex- studies.1–3 11 19 These results have defined a new standard by copyright. pression and drug target pathways in high PD-­L1 tumors of care, leading to a new FDA approval of nivolumab plus (figure 5B), whereas no differences were seen in the chemotherapy for resectable stage IB-­ IIIA (AJCC 7th proportions of immune cell subtypes (online supple- edition) without ALK or EGFR alterations. In the new mental figure 2A). The comparison between TMB high scenario to come, many questions remain to be answered, and low tumors revealed an upregulation of the tumor including: the value of the established biomarkers PD-­L1 antigens MAGEA1 and MAGEA10 in TMB high tumors and TMB, the role of new putative biomarkers, the main (figure 5C), with an upregulation of the lymphocyte infil- mechanisms behind tumor responses, or the personal- trate pathway (figure 5B). The frequency of regulatory ization of adjuvant therapy based on individual risk of T cells seemed to be higher in patients with high TMB, relapse. Specifically, CPRs will be a relevant clinical entity though it was not significant after multiple testing correc- in terms of their number (near 25%–63% with chemo- tion (p=0.038, figure 5D). immunotherapy1 2 19 vs 0%–9.5% with chemotherapy We also evaluated how the PD-­L1 and TMB status in alone20–24) and impact on patient survival.25 However, a pretreatment tumor samples influenced the immune considerable percentage of patients (40%–75%) still do landscape during treatment, affecting the gene expres- not achieve CPR with chemoimmunotherapy, presenting sion in post-­treatment samples. Pretreatment PD-­L1 TPS a higher risk of relapse.1–3 Thus, the search for predictive status was associated with the differential expression biomarkers of CPR, as well as informed treatment deci- of 18 genes in post-­treatment samples (online supple- sion for adjuvant therapy are relevant and novel fields in mental table S10). High pretreatment PD-­L1 TPS was resectable stages of NSCLC. associated with downregulation of 17 genes: the prolifer- We herein described that a proinflammatory gene ation markers CDK1, FOXM1, MKI67, KIAA0101, BUB1, expression profile measured in the pretreatment tissue by TOP2A, CDKN3, CCNB2 and MELK; the housekeeping RNA sequencing is associated with CPR after neoadjuvant genes HMBS and G6PD; the tumor markers CDKN2A, chemoimmunotherapy. Specifically, high levels of IFNG, KRT5 and BRCA1; the type I interferon signaling mole- GZMB, and NKG7 as well as the high estimated frequency cules ISG15 and IFI27 and the TNF receptor superfamily of M1 macrophages could potentially outperform PD-­L1 member TNFRSF18 (figure 5E). Additionally, GSEA iden- or TMB regarding CPR prediction.1 9 Reinforcing these tified the upregulation of the antigen processing pathway results, we have previously shown that CPR patients and downregulation of proliferation pathways in post-­ exhibited a stronger pre-­established immune response treatment tumor samples with higher PD-­L1 status than before treatment in both tissue and blood. Indeed, a their pretreatment counterparts (figure 5G). Regarding higher TCR repertoire clonality in pretreatment tumor the estimated cell proportions, post-­treatment samples tissue, reflecting higher tumor immunogenicity, was Casarrubios M, et al. J Immunother Cancer 2022;10:e005320. doi:10.1136/jitc-2022-005320 9 J Immunother Cancer: first published as 10.1136/jitc-2022-005320 on 28 September 2022. Downloaded from http://jitc.bmj.com/ on July 14, 2023 at Sant Paul Hospital Consortia. Protected Open access by copyright. Figure 5 Immune landscape regarding PD-­L1 and TMB status in pretreatment samples. (A) Volcano plot of DEGs in pretreatment samples between PD-­L1 high (≥25%, n=9) and PD-­L1 low (<25%, n=6) tumors. Red dots indicate upregulated genes, while blue dots indicate downregulated genes in PD-­L1 high samples. (B) Differentially upregulated pathways in pretreatment samples of TMB high and PD-­L1 high tumors. (C) Volcano plot of DEGs in pretreatment samples between TMB high (≥5.89, n=7) and TMB low (<5.89, n=7) tumors. Red dots indicate upregulated genes, while blue dots indicate downregulated genes in TMB high samples. (D) Frequency of regulatory T cells measured with CIBERSORTx in TMB high compared with TMB low tumors. (E) Volcano plot of DEGs in post-­treatment samples whose pretreatment tissue showed PD-­ L1 high (≥25%, n=11) or PD-­L1 low (<25%, n=11) expression. Red dots indicate upregulated genes, while blue dots indicate downregulated genes in PD-­L1 high samples. (F) Volcano plot of DEGs in post-­treatment samples whose pretreatment tissue had TMB high (≥5.89, n=12) or TMB low (<5.89, n=10). Red dots indicate upregulated genes, while blue dots indicate downregulated genes in TMB high samples. (G) Differentially upregulated pathways in post-­treatment samples of pretreatment PD-­L1 high tumors. (H) Differentially upregulated pathways in post-­treatment samples of pretreatment TMB low tumors. (I) Frequency of follicular helper T cells and M2 macrophages in post-­treatment samples of high or low PD-­L1 expression at pretreatment. Each patient is represented by a black dot. DEGs, differentially expressed genes; TMB, tumor mutational burden. associated with CPR.9 Additionally, this was also mirrored presentation; activates NK, T, and dendritic cells; and in blood, being CPR patients characterized by a higher switches tumor-­associated macrophages to M1 macro- baseline cytotoxic profile, with lower levels of inhibitory phages, promoting tumor elimination.26 27 A higher IFNγ cytokines and cells.8 The role of IFNγ in antitumoral signature has been shown to be associated with improved response has been exhaustively reviewed previously. best overall response and PFS after immunotherapy in This cytokine is produced by dendritic cells, helper and advanced stages of melanoma, head and neck carcinoma cytotoxic T cells and NK cells. Its signaling increases the and gastric cancer patients,28 29 but it has not been fully expression of MHC class molecules, enhancing antigen demonstrated in NSCLC.30 31 Our results suggest that 10 Casarrubios M, et al. J Immunother Cancer 2022;10:e005320. doi:10.1136/jitc-2022-005320 J Immunother Cancer: first published as 10.1136/jitc-2022-005320 on 28 September 2022. Downloaded from http://jitc.bmj.com/ on July 14, 2023 at Sant Paul Hospital Consortia. Protected Open access an IFNγ signature in pretreatment tissue, character- chemotherapy diminishes CD8+ T cell tumor infiltration ized by higher differential expression of the interferon-­ and constrains their antitumoral activity.33 responsive genes IFNG, GZMB, NKG7, IDO1, CXCL9, and It is important to identify the patients that will benefit CXCL10, could distinguish patients whose tumors will the most from chemoimmunotherapy achieving complete achieve CPR after neoadjuvant chemoimmunotherapy. pathological responses and patients with non-­CPR tumors We hypothesize that an immune response was previously that present a higher risk of relapse after surgery. In this orchestrated in pretreatment samples, where high levels regard, we found for the first time a relationship between of CXCL9 and CXCL10 regulate immune cell infiltration higher levels of AKT1 in post-­treatment tumor samples and activation and NKG7 and GZMB induce degranu- and a higher risk of progression and death. Remark- lation and target cell apoptosis in CTL and NK cells.32 ably, AKT1 levels were not associated with percentage of Unfortunately, the lower number of events in our cohort residual viable tumor cells, reinforcing its role beyond of patients limited the survival analysis with the IFNG tumor marker gene with implications in resistance to expression signature. chemoimmunotherapy. This gene functions as a regu- This established pretreatment immune response, which lator of cell proliferation and survival, and it could be seems to be associated with pathological responses to promoting tumor growth regardless of the antitumoral neoadjuvant chemoimmunotherapy, was poorly reflected immune response. An alteration in the PI3K-­AKT-­mTOR by PD-­L1 or TMB levels. In this sense, only CD274 (PD-­ pathway was also found in NSCLC patients with disease L1), GAGE1, and MAGEA1 genes were differentially progression on EGFR-­TKI therapy.34 Also, high numbers expressed between tumors with high or low PD-­L1 or of neutrophils in post-­ treatment tumor samples seem TMB, reinforcing the limitation of these markers in to be associated with lower progression-­free and overall predicting responses. Moreover, CD274 was not differ- survival in our cohort of patients. This could be explained entially expressed between CPR and non-­CPR tumors. by the presence of neutrophil extracellular traps known However, it appears that the pretreatment levels of PD-­L1 to be involved in tumor metastasis,35 36 and present in and TMB influence more the composition of the immune tumor tissue of lung cancer patients.37 38 landscape after treatment than before treatment. In this Importantly, by characterizing the expression levels of regard, high pretreatment PD-­L1 TPS or TMB led to a the primary tumor, we were able to predict relapses at the higher expression of antigen processing genes or type I systemic level. This supports that the possible ongoing micrometastases, which are frequent at this tumor stage, by copyright. interferon signaling but lower expression of proliferation markers in post-­ treatment samples. Pathways probably respond similarly enough to what occurs in the primary associated with a better neoadjuvant response, as we have tumor, reinforcing the value of the surgical specimens as already seen in the CPR versus non-­CPR comparison. a valid source for the personalization of adjuvant therapy. Regarding post-­ treatment tissue samples, we found The greatest changes in the immune expression profile that non-­CPR tumors had upregulation of genes related during neoadjuvant treatment were observed in CPR to proliferation and tumor marker genes in comparison tumors reflecting a stronger immune response. Tumor with CPR tumors, which is consistent with the presence of and proliferation markers, as well as genes involved viable tumor cells in these surgical specimens. Conversely, in the IFNγ signaling pathway, appear to be downregu- CPR tumors showed upregulation of antigen processing, lated at post-­treatment as compared with their pretreat- TCR coexpression and lymphocyte infiltration pathways, ment counterparts, reflecting tumor resolution in CPR supporting that platinum-­ based chemotherapy allows patients. The downregulation of proliferation and tumor a proficient antitumoral immune response in combina- markers and IFNγ signaling in post-­treatment samples tion with anti-­PD1 immunotherapy. Similar results were of CPR tumors is contrasted with the upregulation of recently published using a smaller subset of 21 cases from lymphocyte infiltration, antigen processing, and B cell NADIM trial through a more comprehensive gene panel, markers, possibly reflecting a reduction of tumor cells in which tumors achieving major pathological response with an evolution to a more mature immune response after chemoimmunotherapy had higher numbers of in which the B cells may acquire a greater role, with a CD8+ T cells and genes related to innate immune drop in the proportion of M1 macrophages and follicular response, as well as an upregulation of genes related to T cells. This seems to be consistent with previous studies DNA replication and cell cycle in incomplete patholog- from our group, in which we found lymphoid aggregates, ical response tumors.13 Moreover, increased abundance compatible with tertiary lymphoid structures (TLSs), of B and plasma cells, as well as, CD4+ and CD8+ T cells present in post-­treatment samples of CPR tumors, though were found in post-­ treatment chemoimmunotherapy we had no markers to characterize them.1 Additionally, tumors compared with treatment-­naïve or neoadjuvant the presence of TLS and B cell infiltration has been chemotherapy-­treated tumors.13 However, opposite associated with immune checkpoint blockade response results have been described in bladder cancer, with nega- in different solid tumors.39–41 Moreover, the presence of tive chemoimmunotherapy phase III trials, supported by these TLSs seems to generate a protective immunity in studies in animals models showing that platinum-­based NSCLC patients,42–44 and it is characteristic of the regres- sion tumor bed in resected NSCLC patients treated with neoadjuvant immunotherapy.41 Nonetheless, we did not Casarrubios M, et al. J Immunother Cancer 2022;10:e005320. doi:10.1136/jitc-2022-005320 11 J Immunother Cancer: first published as 10.1136/jitc-2022-005320 on 28 September 2022. Downloaded from http://jitc.bmj.com/ on July 14, 2023 at Sant Paul Hospital Consortia. Protected Open access 3 see a higher estimated relative proportion of B cells in Medical Oncology, Fundación INCLIVA, Hospital Clínico Universitario de Valencia, post-­ treatment samples of CPR patients, but a higher Valencia, Spain 4 Medical Oncology, Hospital Universitario A Coruña, A Coruña, Spain, A Coruña, estimated proportion of CD8+ T cells and CD4+ memory Spain resting T cells, suggesting a protective immunity against 5 Medical Oncology, Hospital Universitario de Vigo, Pontevedra, Spain the tumor. This also matches with the hypothetical role of 6 Medical Oncology, Hospital Universitario Fundación Jiménez Díaz, Madrid, Spain TLS in sustaining an immune-­responsive microenviron- 7 Medical Oncology, Hospital de la Santa Creu i Sant Pau Servei de Oncologia ment, proposed by other authors, because of their asso- Medica, Barcelona, Spain 8 ciation with CD8+ T cells infiltration.40 Concerning our Medical Oncology, Hospital Universitario Insular de Gran Canaria, Las Palmas, Canarias, Spain samples, it would be of interest to study these lymphoid 9 Medical Oncology, Vall d’Hebron Institute of Oncology (VHIO), Hospital Universitari aggregates to determine if they are actually TLS and the Vall d’Hebron, Barcelona, Spain, Barcelona, Barcelona, Spain role exerted in the response to chemoimmunotherapy. 10 Medical Oncology, Hospital Universitario La Paz, Madrid, Spain 11 However, there were few differences between paired Medical Oncology Intercenter Unit, Regional and Virgen de la Victoria University tumor samples of non-­CPR patients that denotes ineffec- Hospitals, IBIMA, Málaga, Spain 12 Medical Oncology, Hospital Universitario Cruces, Barakaldo, Spain tive immune stimulation after treatment. Similarly, no 13 Medical Oncology, Hospital Universitario de Salamanca, Salamanca, Spain estimated cell changes were observed between pretreat- 14 Medical Oncology, Hospital U. Virgen Rocio, Seville, Spain ment and post-­treatment samples, indicating a calmer 15 Medical Oncology, Hospital Clínic, Barcelona, Spain scenario in non-­CPR tumors. 16 Medical Oncology, Hospital Universitario Reina Sofia, Cordoba, Spain 17 Medical Oncology, Hospital General de Alicante, Alicante, Spain CONCLUSIONS Twitter Marta Casarrubios @casarrubiosmc, Martín Lázaro-­Quintela @mlazqui, Margarita Majem @margamajem, Alex Martinez-­Marti @AlxMtzMarti, Atocha In this study, we performed RNA sequencing in pretreat- Romero @LiquidBiopsyLab and Alberto Cruz-­Bermúdez @CruzAPhD ment and post-­treatment tissue samples of patients with Acknowledgements We would like to thank the patients, their families, all the resectable stage IIIA NSCLC treated with neoadjuvant participating clinical teams, and all the Spanish Lung Cancer Group, BMS and chemoimmunotherapy from the NADIM trial using a ThermoFisher staff, for making this study possible. panel of 395 genes related to immunological processes. Contributors AC-­B and MP conceived and design the study. MP, EN, AI, MdRG-­C, Limitations of the study include: the reduced number ML-­Q, MD, MM, DR-­A, AM-­M, JDCC, MCo, GLV, EDB, RB, NV, IBA, BM, and VC of cases that impedes the analysis of specific mutations; recruited and treated patients. MCa, BS-­R, CM-­T, IF-­M, RS-­B, and AC-­B carried out the experiments and analyzed the data. MCa, MP, BS-­R, CM-­T, IF-­M, RS-­B, AR, and the reduced number of events to study the prognostic by copyright. AC-­B interpreted the data. All the authors have read and contributed to the final value of pretreatment IFNγ signature, and the lack of version of the manuscript and approve its submission for publication. AC-­B is the immunotherapy or chemotherapy cohorts to deter- guarantor of this study. mine the prognostic or predictive value of described Funding Work in the authors’ laboratories was supported by '‘Instituto de Salud biomarkers. However, to date, this is the largest RNAseq Carlos III’' (ISCIII) PI19/01652 grant cofunded by European Regional Development effort in chemoimmunotherapy samples from neoadju- Fund (ERDF), Bristol-­Myers Squibb (BMS), Ministry of Science and Innovation RTC2017-­6502-­1 'INmunoSIGHT', RTC2019-­007359-­1 'BLI-­O' and European vant NSCLC patients. Union’s Horizon 2020 research and innovation programme, CLARIFY 875160 grant, We conducted various in-­ depth analyses, using well-­ to MP. ThermoFisher provided reagents for RNA sequencing. AC-­B received a established software packages, revealing key differences in Spanish Lung Cancer Group (SLCG) grant and is supported by a ISCIII-“Sara Borrell” the biology of complete pathological responses, including contract CD19/00170. MCa is supported by PEJD-­2019-­PRE/BMD-­17006 contract a pre-­established immune infiltrate at baseline (which granted to AC-­B. could be used as a source for response biomarkers) and Competing interests MP reports non-­financial support (reagents for TCR sequencing) from ThermoFisher; grants, personal fees and non-­financial support a reduction of tumor markers and IFNγ signaling after from BMS; grants, personal fees and non-­financial support from ROCHE; grants, neoadjuvant treatment (showing a greater impact of personal fees and non-­financial support from ASTRAZENECA; personal fees from treatment on gene expression compared with non-­CPR). MSD; personal fees from TAKEDA; outside the submitted work. EN grants or We also identify genes and estimated cell populations contracts: Roche, Pfizer, Merck-­Serono, BMS, and Nanostring. Consulting fees: Roche, BMS, MSD, Merck-­Serono, Pfizer, Lilly, Amgen, Boehringer, AstraZeneca, associated with increased risk of progression in non-­ Takeda, Bayer, and Sanofi. Payment honoraria: Roche, BMs, MSD, Merck-­Serono, complete pathological responders. Finally, we describe Pfizer, Lilly, Amgen, Boehringer, Astrazeneca, Takeda, Bayer, and Sanofi. Support that the pretreatment PD-­L1 and TMB levels are poorly For attending meetings and travel: MSD, Roche, and BMS. AI consulting fees: Pfizer, associated with distinct basal immune landscape but Amgen, and AstraZeneca. Payment for expert testimony: Bristol, Pfizer, Takeda, Roche, and AstraZeneca. Support for attending meetings or travels: Roche and impact it considerably more after chemoimmunotherapy. Bristol. MRG-­C reports personal fees from BMS, personal fees from MSD, personal Our results confirm mechanistic differences between fees from Roche, personal fees from Pfizer, personal fees from AstraZeneca, outside CPR and non-­CPR responses and open the possibility of the submitted work; ML-­Q: payment for honoraria: BMS, MSD, Ipsen, Astella, personalization of immunotherapy-­based regimens in the AstraZeneca, and Pfizer. Support for attendings meetings/travels: MSD, Takeda, Janssen, Pfizer, and BMS. MD reports personal fees from Astra-­Zeneca, personal neoadjuvant setting. fees from BMS, personal fees from Boehringer Ingelheim, personal fees from MSD, personal fees from Pfizer, personal fees from Roche, outside the submitted work; Author affiliations MM reports grants and personal fees from BMS, personal fees and non-­financial 1 Medical Oncology, Instituto de Investigación Sanitaria Puerta de Hierro-­Segovia support from MSD, personal fees and non-­financial support from Boehringer de Arana (IDIPHISA), Hospital Universitario Puerta de Hierro-­Majadahonda, Ingelheim; personal fees, non-­financial support, and other from AstraZeneca; Majadahonda-­Madrid, Spain personal fees, non-­financial support, and other from Roche; personal fees from 2 Medical Oncology, Catalan Institute of Oncology, Oncobell Program, IDIBELL, Kyowa Kyrin; personal fees from Pierre Fabre, outside the submitted work; DR-­A: L’Hospitalet de Llobregat, L'Hospitalet, Barcelona, Spain consulting fees: Roche, AstraZeneca, BMS, MSD, Lilly, Pfizer, and Novartis. Payment 12 Casarrubios M, et al. J Immunother Cancer 2022;10:e005320. doi:10.1136/jitc-2022-005320 J Immunother Cancer: first published as 10.1136/jitc-2022-005320 on 28 September 2022. Downloaded from http://jitc.bmj.com/ on July 14, 2023 at Sant Paul Hospital Consortia. Protected Open access for honoraria: Roche, Astrazeneca, BMS, MSD, Lilly, Pfizer, and Novartis. Support REFERENCES for attending meetings/travels: Roche, MSD, and Novartis; AM-­M: advisory board: 1 Provencio M, Nadal E, Insa A, et al. Neoadjuvant chemotherapy AstraZeneca, BMS, Roche, MSD, and Pfizer. Speaker's bureau: AstraZeneca, BMS, and nivolumab in resectable non-­small-­cell lung cancer (NADIM): Roche, MSD, and Pfizer. Support for attending meetings or travel: AstraZeneca, an open-­label, multicentre, single-­arm, phase 2 trial. Lancet Oncol BMS, Roche, MSD, Pfizer, and Lilly. Payment for expert testimony: AstraZeneca. 2020;21:1413–22. 2 Forde PM, Spicer J, Lu S, et al. Neoadjuvant nivolumab Steering committee member: AstraZeneca. JDCC reports personal fees from Astra plus chemotherapy in resectable lung cancer. N Engl J Med Zeneca, personal fees from Boehringer Ingelheim, personal fees from Merck Sharp 2022;386:1973–85. and Dohme, personal fees from Hoffmann-­la Roche, personal fees from Bristol-­ 3 Provencio M, Serna-­Blasco R, Nadal E, et al. Overall survival and Myers Squibb, personal fees from Takeda, personal fees from Pfizer, personal biomarker analysis of neoadjuvant nivolumab plus chemotherapy fees from Novartis, outside the submitted work; IBA reports consulting or advisory in operable stage IIIA Non–Small-­Cell lung cancer (NADIM phase II board for Bristol Myers, Takeda, Roche, AstraZeneca, and Behringer Inngelheim; trial). JCO 2022;40:2924–33. BM reports grants and personal fees from Roche, personal fees and other from 4 FDA. Opdivo label 2022, 2022. BMS, personal fees from Takeda, other from MSD, personal fees from Boehringer, 5 McGrail DJ, Pilié PG, Rashid NU, et al. High tumor mutation burden fails to predict immune checkpoint blockade response across all other from Takeda, outside the submitted work; AR: consulting fees: AstraZeneca. cancer types. Ann Oncol 2021;32:661–72. Payment or honoraria: Vivo diagnóstico. Advisory board: Takeda. VC: payment or 6 IASLC. The IASLC atlas of PD-­L1 testing in lung cancer. Journal of honoraria: Roche, BMS, MSD, AstraZeneca, and Boehringer. Thoracic Oncology 2017. Patient consent for publication Not applicable. 7 Davis AA, Patel VG. The role of PD-­L1 expression as a predictive biomarker: an analysis of all US food and drug administration (FDA) Ethics approval Informed consent for the collection of research samples and approvals of immune checkpoint inhibitors. J Immunother Cancer study protocol were approved by the clinical research ethics committee of Hospital 2019;7:278. Puerta de Hierro (ref. 20.16) and the Spanish Lung Cancer Group (SLCG) Board (ref. 8 Laza-­Briviesca R, Cruz-­Bermúdez A, Nadal E, et al. Blood N/A) in accordance with the International Conference on Harmonization Guidelines biomarkers associated to complete pathological response on NSCLC patients treated with neoadjuvant chemoimmunotherapy included in on Good Clinical Practice and the Declaration of Helsinki. Participants gave informed NADIM clinical trial. Clin Transl Med 2021;11:e491. consent to participate in the study before taking part. 9 Casarrubios M, Cruz-­Bermúdez A, Nadal E, et al. Pretreatment Provenance and peer review Not commissioned; externally peer reviewed. tissue TCR repertoire evenness is associated with complete pathologic response in patients with NSCLC receiving neoadjuvant Data availability statement Data are available on reasonable request. The chemoimmunotherapy. Clin Cancer Res 2021;27:5878–90. datasets supporting this study are available from the corresponding author upon 10 Romero A, Nadal E, Serna R, et al. OA20.02 pre-­treatment levels justified demand. of ctDNA for long-­term survival prediction in stage IIIA NSCLC treated with neoadjuvant chemo-­immunotherapy. J Thorac Oncol Supplemental material This content has been supplied by the author(s). It has 2021;16:S883–4. not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been 11 Provencio M, Calvo V, Romero A, et al. Treatment sequencing peer-­reviewed. Any opinions or recommendations discussed are solely those in resectable lung cancer: the good and the bad of adjuvant of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and versus neoadjuvant therapy. Am Soc Clin Oncol Educ Book responsibility arising from any reliance placed on the content. Where the content 2022;42:711–28. by copyright. includes any translated material, BMJ does not warrant the accuracy and reliability 12 Bodor JN, Boumber Y, Borghaei H. Biomarkers for immune checkpoint inhibition in non-­small cell lung cancer (NSCLC). Cancer of the translations (including but not limited to local regulations, clinical guidelines, 2020;126:260–70. terminology, drug names and drug dosages), and is not responsible for any error 13 Rocha P, Zhang J, Laza-­Briviesca R, et al. Distinct immune gene and/or omissions arising from translation and adaptation or otherwise. programs associated with host tumor immunity, neoadjuvant Open access This is an open access article distributed in accordance with the chemotherapy, and chemoimmunotherapy in resectable NSCLC. Clin Cancer Res 2022;28:2461–73. Creative Commons Attribution Non Commercial (CC BY-­NC 4.0) license, which 14 Love MI, Huber W, Anders S. Moderated estimation of fold change permits others to distribute, remix, adapt, build upon this work non-­commercially, and dispersion for RNA-­seq data with DESeq2. Genome Biol and license their derivative works on different terms, provided the original work is 2014;15:1–21. properly cited, appropriate credit is given, any changes made indicated, and the use 15 Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment is non-­commercial. See http://creativecommons.org/licenses/by-nc/4.0/. analysis: a knowledge-­based approach for interpreting genome-­wide expression profiles. Proc Natl Acad Sci U S A ORCID iDs 2005;102:15545–50. Marta Casarrubios http://orcid.org/0000-0002-2872-9552 16 Mootha VK, Lindgren CM, Eriksson K-­F, et al. PGC-­1alpha-­ responsive genes involved in oxidative phosphorylation are Mariano Provencio http://orcid.org/0000-0001-6315-7919 coordinately downregulated in human diabetes. Nat Genet Ernest Nadal http://orcid.org/0000-0002-9674-5554 2003;34:267–73. Amelia Insa http://orcid.org/0000-0002-3438-6170 17 Newman AM, Steen CB, Liu CL, et al. Determining cell type María del Rosario García-­Campelo http://orcid.org/0000-0003-2113-1504 abundance and expression from bulk tissues with digital cytometry. Martín Lázaro-Q­ uintela http://orcid.org/0000-0002-0657-4713 Nat Biotechnol 2019;37:773–82. Manuel Dómine http://orcid.org/0000-0003-1634-9832 18 Goedhart J, Luijsterburg MS. VolcaNoseR is a web APP for creating, exploring, labeling and sharing volcano plots. Sci Rep Margarita Majem http://orcid.org/0000-0002-9919-7485 2020;10:20560. Delvys Rodriguez-­Abreu http://orcid.org/0000-0003-0506-1366 19 Shu CA, Gainor JF, Awad MM, et al. Neoadjuvant atezolizumab and Alex Martinez-­Marti http://orcid.org/0000-0002-6714-9526 chemotherapy in patients with resectable non-­small-­cell lung cancer: Javier De Castro Carpeño http://orcid.org/0000-0002-3622-6306 an open-­label, multicentre, single-­arm, phase 2 trial. Lancet Oncol Manuel Cobo http://orcid.org/0000-0003-3402-1144 2020;21:786–95. Guillermo López Vivanco http://orcid.org/0000-0003-3800-6224 20 Hellmann MD, Chaft JE, William WN, et al. Pathological response Edel Del Barco http://orcid.org/0000-0001-6443-0585 after neoadjuvant chemotherapy in resectable non-­small-­cell lung cancers: proposal for the use of major pathological response as a Reyes Bernabé http://orcid.org/0000-0003-0312-9195 surrogate endpoint. Lancet Oncol 2014;15:e42–50. Nuria Viñolas http://orcid.org/0000-0002-8287-2739 21 Stewart LA, Pignon JP. Chemotherapy in non-­small cell lung cancer: Isidoro Barneto Aranda http://orcid.org/0000-0001-6302-1034 a meta-­analysis using updated data on individual patients from 52 Bartomeu Massuti http://orcid.org/0000-0002-6247-4493 randomised clinical trials. non-­small cell lung cancer Collaborative Belén Sierra-­Rodero http://orcid.org/0000-0003-0276-6088 group. BMJ 1995;311:899–909. Cristina Martinez-T­ oledo http://orcid.org/0000-0002-0663-0674 22 Burdett S, Stewart LA, Rydzewska L. A systematic review and meta-­ Ismael Fernández-­Miranda http://orcid.org/0000-0002-5136-7270 analysis of the literature: chemotherapy and surgery versus surgery alone in non-­small cell lung cancer. J Thorac Oncol 2006;1:611–21. Roberto Serna-B ­ lanco http://orcid.org/0000-0002-7563-9613 23 NSCLC Meta-­analysis Collaborative Group. Preoperative Atocha Romero http://orcid.org/0000-0002-1634-7397 chemotherapy for non-­small-­cell lung cancer: a systematic Virginia Calvo http://orcid.org/0000-0002-3503-4847 review and meta-­analysis of individual participant data. Lancet Alberto Cruz-­Bermúdez http://orcid.org/0000-0001-5136-7011 2014;383:1561–71. Casarrubios M, et al. J Immunother Cancer 2022;10:e005320. doi:10.1136/jitc-2022-005320 13 J Immunother Cancer: first published as 10.1136/jitc-2022-005320 on 28 September 2022. Downloaded from http://jitc.bmj.com/ on July 14, 2023 at Sant Paul Hospital Consortia. Protected Open access 24 Pignon J-­P, Tribodet H, Scagliotti GV, et al. Lung adjuvant cisplatin clinical response to EGFR-­TKI plus everolimus combination therapy. evaluation: a pooled analysis by the lace Collaborative group. J Clin Transl Lung Cancer Res 2020;9:1258–67. Oncol 2008;26:3552–9. 35 Papayannopoulos V. Neutrophil extracellular traps in immunity and 25 Pataer A, Kalhor N, Correa AM, et al. Histopathologic response disease. Nat Rev Immunol 2018;18:134–47. criteria predict survival of patients with resected lung cancer after 36 Wang W, Zhang J, Zheng N, et al. The role of neutrophil neoadjuvant chemotherapy. J Thorac Oncol 2012;7:825–32. extracellular traps in cancer metastasis. Clin Transl Med 26 Martínez-­Sabadell A, Arenas EJ, Arribas J. Ifnγ signaling in natural 2020;10:2–5. and therapy-­induced antitumor responses. Clin Cancer Res 37 Fang C, Liu F, Wang Y, et al. A innovative prognostic symbol based 2022;28:1243–9. on neutrophil extracellular traps (NETs)-­related lncRNA signature in 27 Ni L, Lu J. Interferon gamma in cancer immunotherapy. Cancer Med non-­small-­cell lung cancer. Aging 2021;13:17864–79. 2018;7:4509–16. 38 de Andrea CE, Ochoa MC, Villalba-­Esparza M, et al. Heterogenous 28 Liu R, Yang F, Yin J-­Y, et al. Influence of tumor immune infiltration on presence of neutrophil extracellular traps in human solid tumours is immune checkpoint inhibitor therapeutic efficacy: a computational partially dependent on IL-­8. J Pathol 2021;255:190–201. retrospective study. Front Immunol 2021;12:685370. 39 Vanhersecke L, Brunet M, Guégan J-­P, et al. Mature tertiary 29 Ayers M, Lunceford J, Nebozhyn M, et al. IFN-γ-related mRNA lymphoid structures predict immune checkpoint inhibitor efficacy profile predicts clinical response to PD-­1 blockade. J Clin Invest in solid tumors independently of PD-­L1 expression. Nat Cancer 2017;127:2930–40. 2021;2:794–802. 30 Dizier B, Callegaro A, Debois M, et al. A Th1/IFNγ gene signature 40 Cabrita R, Lauss M, Sanna A, et al. Tertiary lymphoid structures is prognostic in the adjuvant setting of resectable high-­risk improve immunotherapy and survival in melanoma. Nature melanoma but not in non-­small cell lung cancer. Clin Cancer Res 2020;577:561–5. 2020;26:1725–35. 41 Cottrell TR, Thompson ED, Forde PM, et al. Pathologic features of 31 Karachaliou N, Gonzalez-­Cao M, Crespo G, et al. Interferon gamma, response to neoadjuvant anti-­PD-­1 in resected non-­small-­cell lung an important marker of response to immune checkpoint blockade in carcinoma: a proposal for quantitative immune-­related pathologic non-­small cell lung cancer and melanoma patients. Ther Adv Med response criteria (irPRC). Ann Oncol 2018;29:1853–60. Oncol 2018;10:175883401774974. 42 Feng H, Yang F, Qiao L, et al. Prognostic significance of gene 32 Ng SS, De Labastida Rivera F, Yan J, et al. The NK cell granule signature of tertiary lymphoid structures in patients with lung protein NKG7 regulates cytotoxic granule exocytosis and adenocarcinoma. Front Oncol 2021;11:1–12. inflammation. Nat Immunol 2020;21:1205–18. 43 Germain C, Gnjatic S, Tamzalit F, et al. Presence of B cells in 33 Nikolos F, Hayashi K, Hoi XP, et al. Cell death-­induced tertiary lymphoid structures is associated with a protective immunogenicity enhances chemoimmunotherapeutic response by immunity in patients with lung cancer. Am J Respir Crit Care Med converting immune-­excluded into T-­cell inflamed bladder tumors. Nat 2014;189:832–44. Commun 2022;13:1487. 44 Germain C, Devi-­Marulkar P, Knockaert S, et al. Tertiary lymphoid 34 Fang W, Huang Y, Gu W, et al. Pi3K-­Akt-­mTOR pathway alterations Structure-­B cells narrow regulatory T cells impact in lung cancer in advanced NSCLC patients after progression on EGFR-­TKI and patients. Front Immunol 2021;12:1–14. by copyright. 14 Casarrubios M, et al. J Immunother Cancer 2022;10:e005320. doi:10.1136/jitc-2022-005320