Neoadjuvant atezolizumab for resectable non-small cell lung cancer: an open-label, single-arm phase II trial

In an ongoing, open-label, single-arm phase II study (NCT02927301), 181 patients with untreated, resectable, stage IB–IIIB non-small cell lung cancer received two doses of neoadjuvant atezolizumab monotherapy. The primary end point was major pathological response (MPR; ≤10% viable malignant cells) in resected tumors without EGFR or ALK alterations. Of the 143 patients in the primary end point analysis, the MPR was 20% (95% confidence interval, 14–28%). With a minimum duration of follow-up of 3 years, the 3-year survival rate of 80% was encouraging. The most common adverse events during the neoadjuvant phase were fatigue (39%, 71 of 181) and procedural pain (29%, 53 of 181), along with expected immune-related toxicities; there were no unexpected safety signals. In exploratory analyses, MPR was predicted using the pre-treatment peripheral blood immunophenotype based on 14 immune cell subsets. Immune cell subsets predictive of MPR in the peripheral blood were also identified in the tumor microenvironment and were associated with MPR. This study of neoadjuvant atezolizumab in a large cohort of patients with resectable non-small cell lung cancer was safe and met its primary end point of MPR ≥ 15%. Data from this single-arm, non-randomized trial suggest that profiles of innate immune cells in pre-treatment peripheral blood may predict pathological response after neoadjuvant atezolizumab, but additional studies are needed to determine whether these profiles can inform patient selection and new therapeutic approaches.

T he survival of patients with resectable non-small cell lung cancer (NSCLC) has not substantially improved since the establishment of adjuvant chemotherapy as standard treatment more than 20 years ago.Inhibitors of PD-1 or PD-L1 are approved for the treatment of advanced-stage NSCLC and resected stage II-III, PD-L1-expressing NSCLC 1 .These agents have shown some benefit when given before surgery in small (n = 21-23) studies of patients with resectable NSCLC; however, the pathological response rates in these studies have wide confidence intervals (CIs) and predictive biomarkers of response remain unclear 2,3 .The phase II Lung Cancer Mutation Consortium 3 (LCMC3) study was performed to evaluate the efficacy of neoadjuvant atezolizumab, a PD-L1 inhibitor, in a large population of treatment-naive patients with resectable, stage IB-IIIB NSCLC.Prospective correlative stud-ies were performed to elucidate potential predictors of treatment response and resistance.
Atezolizumab is hypothesized to enhance antitumor immunity by restoring the function of cytotoxic T cells 4 , but the details of this mechanism in humans and the effects of atezolizumab on other immune cell populations are largely undefined.Because not all patients respond to PD-(L)1 inhibition, identifying biomarkers predictive of response and resistance may aid in treatment selection.Moreover, elucidating the mechanisms by which cancer cells evade antitumor immunity may inform rational combination therapies.To define the clinical and biological effects of neoadjuvant atezolizumab and to identify biomarkers predictive of response (or lack thereof), the LCMC3 study evaluated the immune environment pre-and post-treatment with atezolizumab and cor-Neoadjuvant atezolizumab for resectable non-small cell lung cancer: an open-label, single-arm phase II trial Jamie E. Chaft 1,2,28 , Filiz Oezkan 3,4,5,6,28 , Mark G. Kris 1,2 , Paul A. Bunn 7 , Ignacio I. Wistuba 8 , David J. Kwiatkowski 9,10 , Dwight H. Owen 3,11 , Yan Tang 10 , Bruce E. Johnson 9,10 , Jay M. Lee 12 , Gerard Lozanski 11 , Maciej Pietrzak 11 , Michal Seweryn 11,13,14 , Woo Yul Byun 11 , Katja Schulze 15 , Alan Nicholas 15 , Ann Johnson 15 , Jessica Grindheim 15  In an ongoing, open-label, single-arm phase II study (NCT02927301), 181 patients with untreated, resectable, stage IB-IIIB non-small cell lung cancer received two doses of neoadjuvant atezolizumab monotherapy.The primary end point was major pathological response (MPR; ≤10% viable malignant cells) in resected tumors without EGFR or ALK alterations.Of the 143 patients in the primary end point analysis, the MPR was 20% (95% confidence interval, 14-28%).With a minimum duration of follow-up of 3 years, the 3-year survival rate of 80% was encouraging.The most common adverse events during the neoadjuvant phase were fatigue (39%, 71 of 181) and procedural pain (29%, 53 of 181), along with expected immune-related toxicities; there were no unexpected safety signals.In exploratory analyses, MPR was predicted using the pre-treatment peripheral blood immunophenotype based on 14 immune cell subsets.Immune cell subsets predictive of MPR in the peripheral blood were also identified in the tumor microenvironment and were associated with MPR.This study of neoadjuvant atezolizumab in a large cohort of patients with resectable non-small cell lung cancer was safe and met its primary end point of MPR ≥ 15%.Data from this single-arm, non-randomized trial suggest that profiles of innate immune cells in pre-treatment peripheral blood may predict pathological response after neoadjuvant atezolizumab, but additional studies are needed to determine whether these profiles can inform patient selection and new therapeutic approaches.

Articles
Nature MediciNe related these changes with the primary efficacy measure of MPR (Extended Data Fig. 1).
Adjuvant treatment is summarized in Fig. 1.The exploratory end points of median disease-free survival (DFS) and overall survival (OS) were not reached.The 3-year DFS and OS were 72% (95% CI 62-79%) and 80% (95% CI 71-87%), respectively.Survival by MPR status is presented in Fig. 2b,c.Survival by disease stage, adjuvant atezolizumab use and lymph node status is presented in Extended Data Fig. 4. The duration of DFS in the eight patients with pCR is presented in Supplementary Table 2.

Safety.
Of the 181 patients in the safety-evaluable population, 97% experienced at least one AE during the neoadjuvant phase (up to 90 d after last dose of neoadjuvant atezolizumab), most commonly fatigue (39%) and procedural pain (29%) (Supplementary Table 4).The most frequent atezolizumab-related AE was fatigue (20%  The reasons were clinical progression (n = 3), physician did not want to delay patient surgery (n = 1), physician did not consider the patient a good surgical candidate (n = 1) and physician discontinued patient from the study because of an AE (n = 1).d One patient was determined to have pre-existing congestive heart failure, one declined surgery and one was lost to follow-up.

IMMUNOME.
In exploratory analyses, 111 pre-treatment peripheral blood samples were evaluated via ten-color 60-marker flow cytometry (IMMUNOME) (Supplementary Table 5 and Extended Data Fig. 5).Samples were split into a training set (n = 57) to develop a predictive model for MPR and a test set 1 (n = 54) (Extended Data Fig. 2).The area under the curve (AUC) was 0.987 for the training set and 0.722 for test set 1; the addition of individual clinical parameters did not significantly improve the predictive power (Fig. 3, Extended Data Fig. 6 and Supplementary Table 6).The final multiparametric GAM-LASSO (generalized additive model-least absolute shrinkage and selection operator) model consisted of 14 immune cell subsets in the pre-treatment peripheral blood that significantly correlated with MPR.Higher prevalence of non-T/non-natural killer (NK) cells expressing the immunoregulatory receptors immunoglobulin-like transcript 2 (ILT2), killer cell immunoglobulin-like receptor (KIR) 2DL1 (KIR2DL1) and KIR2DL2 and of NK group 2 member D (NKG2D) + non-T/non-NK cells positively associated with MPR (Supplementary Table 7 details immune cell descriptions and effect sizes).Higher prevalence of NK and NK-like T cell subsets in the peripheral blood (several of which express inhibitory receptors such as ILT2, NKG2A and NKG2D), subsets of γ/δ T cells, γ/δ NK-like T cells, degranulated myeloid cells and naive CD4 + /CD8 + T cells inversely associated with MPR (Supplementary Table 7).
The ability of the final model to predict the probability of MPR was evaluated in test set 2, consisting of nine study participants with radiographical PD who were not included in the training set or test set 1. The mean probability of MPR was predicted to be 0.20 in the PD cohort, 0.11 in the non-MPR cohort and 0.43 in the MPR cohort (PD versus non-MPR, P = 0.11; PD versus MPR, P = 0.0035; non-MPR versus MPR, P < 0.0001) (Fig. 3 and Supplementary Table 8).

Gene expression analysis of tumor tissue.
To determine how the IMMUNOME observations from pre-treatment peripheral blood translate to the tumor microenvironment, we analyzed gene expression at the single-cell level in tumor tissue.Single-cell RNA-sequencing (scRNA-seq) data from 15 surgical tumor samples were analyzed in an exploratory fashion to assess cells expressing the markers of interest.In tumor tissue, NKG2A and KIRs, including KIR2DL1, were predominantly expressed on NK cells (Fig. 4 and Extended Data Fig. 8); these receptors were more highly expressed on a greater percentage of NK cells in patients with lesser pathological regression (Fig. 4).ILT2 in tumor tissue, by contrast, was expressed to only a small extent on NK cells and was instead predominantly expressed on dendritic cells (DCs), macrophages and monocytes.PD-L1 was mainly expressed on DCs and common myeloid progenitor cells and to a lesser extent in tumor tissue (Fig. 4 and Extended Data Fig. 9).
Bulk RNA-seq data were available for 54 patients at baseline and 44 patients at surgery.In exploratory analyses, ILT2 and PD-L1 transcripts in the tumor sample assessed via bulk RNA-seq were significantly associated with pathological response in non-squamous tumors at baseline and surgery (Extended Data Fig. 10).

Discussion
This phase II LCMC3 study of neoadjuvant atezolizumab, the largest study of preoperative checkpoint inhibitor monotherapy in early-stage NSCLC to date, met its primary end point with an MPR rate of 20% (6% pCR) in primary tumors from patients with resectable stage IB-IIIB NSCLC.Neoadjuvant atezolizumab was well tolerated, with a low incidence of treatment-related grade ≥3 AEs.
There was only one treatment-related death (immune-mediated pneumonitis), the onset of which occurred 1 month after surgery; the patient died despite optimal medical management.The composite perioperative mortality rate in LCMC3 was equivalent to that of neoadjuvant chemotherapy and of surgery without chemotherapy 5 .Overall, the safety profile was consistent with that observed in advanced disease [6][7][8] .Biomarkers predictive of drug toxicity remain an unmet need.Despite a high-risk population, including approxi-mately half of patients with clinical stage III disease, 88% of patients had planned surgery.Median DFS and OS were not reached, with an encouraging 3-year OS rate of 80%.MPR was selected as the primary end point for this study to provide reasonable comparison to historical studies of neoadjuvant chemotherapy, in which MPR rates of 15-22% have been reported, rather than single-digit pCR rates 9,10 .These response rates have recently been substantiated in the randomized chemotherapy arm of CheckMate 816, reporting an MPR rate of 9% and pCR rate of 2% 11 .Small studies of neoadjuvant nivolumab monotherapy have demonstrated MPR rates of 22-45% 2,3 .Although MPR rates of 57-83% following neoadjuvant chemoimmunotherapy have recently been reported 12,13 , the addition of chemotherapy may confound the interpretation of immune predictors.Moreover, the results of these small studies have large confidence intervals.Larger

Articles
Nature MediciNe studies, such as LCMC3 and CheckMate 816, provide more accurate MPR rates following neoadjuvant treatment, with an MPR rate in the tumor post-atezolizumab monotherapy of 20% and in the tumor and lymph nodes post-nivolumab plus chemotherapy of 37% 11 .The ability to select patients for the most effective systemic regimen-single-agent immunotherapy, chemotherapy or chemoimmunotherapy-remains a major unmet clinical need.The phase III IMpower030 study of neoadjuvant atezolizumab and chemotherapy is ongoing.
LCMC3 is also the largest analysis of pre-and post-treatment samples in patients with NSCLC treated with single-agent immunotherapy, enabling the robust study of predictive biomarkers.Consistent with findings in the metastatic setting, MPR was associated with high PD-L1 TPS, but as PD-L1 status was unknown for 19% of patients, this outcome should be interpreted with caution.We show in a rigorous training-testing analysis that multi-lineage immunophenotyping of a pre-treatment peripheral blood sample provides information that may predict the probability of pathological response.Consistent with previous studies of patients with NSCLC treated with PD-(L)1 blockade, we observed significant expansion of peripheral blood-activated CD8 + T cells in patients with tumors demonstrating MPR [14][15][16] .Unexpectedly, we also found significant associations between NK and NK-like T cell markers in both the peripheral blood and tumor tissue and response.Specifically, high pre-treatment levels of NK-like T cells and NK cells, including subsets expressing ILT2 + NKG2A + , in the peripheral blood were significantly associated with lack of response, a result verified in two test cohorts.Expression of ILT2, which binds to human leukocyte antigen (HLA)-G, on NK cells and invariant NK T cells contributes to tumor tolerance by reducing the proliferative and cytotoxic activities of these cells [17][18][19][20] .NKG2A is expressed by immature NK cells 21,22 and binds to HLA-E 23,24 , the expression of which is increased in various solid tumors [25][26][27] .NKG2A can suppress the proliferation of NK cells and NK cell-mediated cytotoxicity 22,24,26,27 .Antibodies directed against both ILT2/HLA-G and NKG2A/HLA-E are being investigated as therapeutic agents.
In our scRNA-seq data, NK cells in tumor tissue showed higher expression of NKG2A and KIRs, including KIR2DL1, on a larger percentage of NK cells in patients with less pathological regression, suggesting that these subsets impair responses to anti-PD-L1.Differential expression of KIR receptors on tumor-infiltrating lymphocytes of patients with NSCLC treated with neoadjuvant nivolumab has been shown to be associated with MPR 28 .ILT2 in tumor tissue was instead predominantly expressed on DCs, macrophages and monocytes and was positively associated with MPR.The two non-T/non-NK cell subsets that we found to positively associate with MPR in the peripheral blood immunophenotyping analysis express ILT2 and the NKG2A-related molecule NKG2D and therefore are possibly of myeloid lineage 29 .ILT2 and PD-L1 transcripts in tumor samples assessed via bulk RNA-seq were significantly associated with pathological regression in non-squamous tumors at baseline and surgery.Notably, we found PD-L1 to be mainly expressed on DCs and common myeloid progenitor cells and to a lesser extent in tumor tissue.Recent literature suggests that ILT2 and PD-L1 are upregulated on DCs following antigen stimulation and may protect activated DCs from CD8 + T cell attack 30,31 .Thus, our data suggest that not only the adaptive immune system but also the innate immune system in the circulation and tumor tissue play a role in mediating antitumor immune responses to neoadjuvant anti-PD-L1 therapy.
We recognize that interpretation of our predictive biomarker analyses is limited because of the single-arm design of this study.Additional studies are needed to validate the significantly associated cell subsets and further establish the role of peripheral NK cells and NK-like T cells, as well as DCs in tumor tissue, in antitumor responses to immune-checkpoint inhibitors in NSCLC.Following external validation, the value of these cell subsets in predicting outcomes in the clinical practice setting must be determined.Each

Nature MediciNe
IMMUNOME panel subset was limited to ten markers; therefore, cells with markers of interest may overlap between subsets (Supplementary Tables 5, 7 and 9).We also acknowledge as a limitation that MPR (as defined by Pataer) 10 was assessed only on the primary tumor; data on the lymph nodes were exploratory and will be presented elsewhere.
In conclusion, neoadjuvant treatment with single-agent atezolizumab yielded a 20% MPR rate in patients with stage IB-III NSCLC, with no new safety signals and encouraging survival.These data confirm that anti-PD-L1 monotherapy is effective in a subset of patients and begins to address two major unmet needs: understanding which biomarkers are predictive of immunotherapy response and identifying patients who may not need chemotherapy.Our biomarker analyses showed that pre-treatment peripheral blood immune cell profiles may predict MPR in atezolizumab-treated patients with resectable NSCLC.Although confirmatory and functional studies are needed, the insights from this analysis also suggest an important role for the innate immune system in the context of PD-(L)1 inhibition and the potential for new treatment regimens involving agents that modulate ILT2/HLA-G and NKG2A/HLA-E.a, The expression of different NK cell surface receptors was determined by scRNA-seq.b, Tumor samples collected at resection were classified into three groups of five samples each on the basis of the percentage of viable tumor cells by pathological analysis: low (≤25% viable tumor cells), middle (26-50%) and high (>50%).Dot size represents the percentage of NK cells in the group expressing the gene.Color represents the scaled average normalized expression.NK cells were downsampled to have the same number of cells in each group.ILT2 is also known as LILRB1, ILT4 as LILRB2, KIR2DL1 as CD158a, NKG2D as CD314 and KLRK1 and PD-L1 as CD274.CD, cluster of differentiation; CMP, common myeloid progenitor; ILT, immunoglobulin-like transcript; KIR2DL1, killer cell immunoglobulin-like receptor DL1; LILRB, leukocyte immunoglobulin-like receptor subfamily B; NKG2, natural killer group protein 2.
Batch effects due to sequencing date and library preparation kit were removed using the R package limma 43 .After discarding genes not present at ≥0.5 c.p.m. in ≥10% of samples (due to low abundance), 17,729 genes remained for analysis.xCell (v.1.3)was used to identify enriched cell subsets in pre-and post-treatment samples on the basis of the relative abundance of their transcriptomes 44 .scRNA-seq.Fresh tumor samples were collected from patients enrolled after 4 January 2019.In total, 15 surgical samples from six patients with NSCLC and nine with squamous NSCLC were analyzed via scRNA-seq.Briefly, fresh tumor and matched-normal samples were dissociated into single-cell suspensions using the Human Tumor Dissociation kit (Miltenyi Biotec) and gentleMACS Dissociator (Miltenyi Biotec).Erythrocytes were removed using the Red Blood Cell Lysis Solution kit (Miltenyi Biotec).After washing in cold PBS, sample viability (>70%) was confirmed using Trypan blue staining.Viable samples were then loaded onto a Chromium Controller (10x Genomics).Droplet emulsions were immediately recovered for reverse transcription via a Bio-Rad thermocycler.Single-cell expression libraries were constructed using the Chromium Single Cell 5′ Feature Barcode Library kit (v.1) (10x Genomics), the quality of which was assessed using the BioAnalyzer High Sensitivity DNA kit (Agilent).The resulting libraries were then sequenced using NextSeq 500 (Illumina).
Raw sequencing data were aligned to the GRCh38 reference genome using Cell Ranger pipeline (10x Genomics) to generate gene-cell count matrices.Data normalization and integration were performed using the Seurat R package (v.4.0.2).Cells were filtered from the downstream analysis using the following criteria: <200 or >6,000 genes detected and >0.1 fraction of mitochondrial genes.The integrated Seurat object was further scaled by regressing out unique molecular identifier count and the fraction of mitochondrial genes.The optimal principal component for dimensionality reduction was determined empirically for each analysis by the drop-off in principal component variance.

Statistical analyses.
An MPR rate of ≥15% was selected as evidence of clinical efficacy based on a previous study 45 .To provide 95% power to detect a 10% difference (null hypothesis 5%) at a one-sided significance level of 0.05, we targeted 180 patients for enrollment.Tumors from patients with incomplete surgical resection were considered to not have MPR.Patients who did not undergo surgery following neoadjuvant atezolizumab were not evaluable for MPR.All atezolizumab-treated patients with NSCLC were included in the safety population.The Kaplan-Meier method was used for the survival analyses, which were performed in the subset of patients in the efficacy population with R0 resection.The data cutoff date was 15 October 2021.Statistical analyses were performed using SAS Proprietary Software (v.9.4; SAS Institute), R v. The reasons were as follows: clinical progression (n = 3), physician did not want to delay patient surgery (n = 1), physician did not consider the patient a good surgical candidate (n = 1), and physician discontinued patient from the study because of an adverse event (n = 1).d One patient was determined to have preexisting congestive heart failure, 1 declined surgery, and 1 was lost to follow-up.e Eighty patients provided a total of 98 samples for the bulk RNA-seq analyses (54 at baseline and 44 at surgery).ALK, anaplastic lymphoma kinase; EGFR, epidermal growth factor receptor; ES, exome sequencing; MPR, major pathologic response; NSCLC, non-small cell lung cancer; PD, progressive disease; RECIST, Response Evaluation Criteria in Solid Tumors; RNA-seq, ribonucleic acid sequencing; scRNA-seq, single-cell ribonucleic acid sequencing; TMB, tumor mutational burden.

Fig. 1 |
Fig. 1 | Patient disposition.Primary efficacy population is bolded.a Two patients were determined to have hemangioma and solitary fibrous tumor at resection despite initial pathology consistent with NSCLC.b Includes one EGFR-positive patient.cThe reasons were clinical progression (n = 3), physician did not want to delay patient surgery (n = 1), physician did not consider the patient a good surgical candidate (n = 1) and physician discontinued patient from the study because of an AE (n = 1).d One patient was determined to have pre-existing congestive heart failure, one declined surgery and one was lost to follow-up.

Fig. 2 |
Fig. 2 | Clinical outcomes in patients who had surgical resection and whose tumors did not have known EGFR or ALK alterations.a, Pathological response (n = 143).Pathological regression is defined as percentage viable tumor cells - 100%.b, DFS by MPR status in patients with R0 resections (n = 137).c, OS by MPR status in patients with R0 resections (n = 137).HR, hazard ratio.

Fig. 3 |
Fig. 3 | Performance of GAM-LASSO MPR predictive models.a, Use of the GAM-LASSO model to predict MPR in test set 2, which consisted of patients within LCMC3 who were not included in either the training set (n = 57) or test set 1 (n = 54).MPR was not assessed in these nine patients because of no resection.The MPR and non-MPR cohorts derived from the merge of the model's training set and test set 1. The maximum and minimum values of the boxes denote the IQR.The line within the IQR denotes the median.The extremities of the dashed lines represent the minimum and maximum values of the data, which are 1.5× below the first quartile and 1.5× above the third quartile.The parameters for null hypothesis testing via analysis of variance (ANOVA) were as follows: d.f.= 2, total sum of squares = 1.976, mean squares = 0.988, F-value = 32.799 and Pr(>F) = 4.914 × 10 −12 .The statistical details for the comparison of MPR and non-MPR were t = −5.47,d.f.= 27.02,two-sided P = 8.6 × 10 −6 and 95% CI = −0.439 to −0.200.The statistical details for the comparison of MPR and PD were t = −3.18,d.f.= 28.45,two-sided P = 0.0035 and 95% CI = −0.383 to −0.083.The statistical details for the comparison of non-MPR and PD were t = −1.77,d.f.= 9.52, two-sided P = 0.11 and 95% CI = −0.195 to 0.023.No adjustment was made for multiplicity.b, ROC curves for the training set and test set 1. The dashed y = x line, which represents random assignment, is included for reference.a Immunophenotyping via flow cytometry.IQR, interquartile range; ROC, receiver operating characteristic.

Fig. 4 |
Fig.4| scRNA-seq analysis of selected genes expressed in tumor tissue from 15 patients following treatment with neoadjuvant atezolizumab.a, The expression of different NK cell surface receptors was determined by scRNA-seq.b, Tumor samples collected at resection were classified into three groups of five samples each on the basis of the percentage of viable tumor cells by pathological analysis: low (≤25% viable tumor cells), middle (26-50%) and high (>50%).Dot size represents the percentage of NK cells in the group expressing the gene.Color represents the scaled average normalized expression.NK cells were downsampled to have the same number of cells in each group.ILT2 is also known as LILRB1, ILT4 as LILRB2, KIR2DL1 as CD158a, NKG2D as CD314 and KLRK1 and PD-L1 as CD274.CD, cluster of differentiation; CMP, common myeloid progenitor; ILT, immunoglobulin-like transcript; KIR2DL1, killer cell immunoglobulin-like receptor DL1; LILRB, leukocyte immunoglobulin-like receptor subfamily B; NKG2, natural killer group protein 2.
4.1.0,ggplot2_3.3.5, ggpubr_0.4.0 and R v.4.0.5 (2021-03-31).Reporting summary.Further information on research design is available in the Nature Research Reporting Summary linked to this article.Extended Data Fig. 2 | CONSORT diagram for the various biomarker analyses.a Two patients were determined to have hemangioma and solitary fibrous tumor at resection despite initial pathology consistent with NSCLC.b Includes 1 EGFR-positive patient.c

5 |
(b) Pathologic response by the overlapping mutational status of KRAS and STK11 in patients with non-squamous NSCLC (n = 78).The maximum and minimum values of the boxes denote the IQR.The line within the IQR denotes the median.The extremities of the dashed lines represent the 5 th to 95 th percentiles.*TMB was only determined for the subset of patients with ES data from baseline and/or surgery with tumor purity ≥15%.P values for TMB in Panel a were determined via linear correlation test (Pearson).P values for mutation status in Panels a and b were determined via two-sided Wilcoxon rank sum test.ES, exome sequencing; IQR, interquartile range; KEAP1, Kelch-like ECH-associated protein 1; KRAS, Kirsten rat sarcoma viral oncogene homolog; MPR, major pathologic response; NS, non-significant; NSCLC, non-small cell lung cancer; STK11, serine/threonine kinase 11; TMB, tumor mutational burden.NATuRE MEDICINE | www.nature.com/naturemedicineExtended Data Fig. 4 | DFS (a) and OS (b) by disease stage, use of adjuvant atezolizumab, and lymph node status (n = 143).The MPR rate in patients who did and did not receive adjuvant atezolizumab was 41% (22/54) and 8% (7/89), respectively.DFS, disease-free survival; OS, overall survival.NATuRE MEDICINE | www.nature.com/naturemedicineExtended Data Fig. 5 | See next page for caption.NATuRE MEDICINE | www.nature.com/naturemedicineExtended Data Fig.Gating strategy for the IMMuNOME tubes described in Supplementary Table