Background

Advances in cancer immunotherapy have revolutionized the primary therapeutic modality for advanced non-small cell lung cancer (NSCLC). Immune checkpoint inhibitors (ICIs) are the cornerstones of systemic therapies for this disease. However, they benefit only a fraction of patients.1 2 The adverse effects of ICI therapies, termed immune-related adverse effects (irAEs), are unique and are generally secondary to inflammatory tissue damage. The onset and duration of irAEs are unpredictable, and the predisposing factors for individuals to develop irAEs remain unclear.3

T lymphocytes are the central mediators of the adaptive immune system and play a crucial role in immune surveillance and cancer eradication. Effective T-cell responses rely on T cells encountering their cognate antigens, including tumor-associated antigens and neoantigens, expressed by tumor cells.4 Hence T-cell receptor (TCR) sequencing has become an invaluable tool to identify and track specific T-cell clonotypes, their dynamics longitudinally and observe how these relate to patient response and toxicity.5

Accordingly, analysis of the T-cell repertoire has successfully predicted patient outcomes and responses to various cancer therapies across many tumor types.6–14 In these studies, analysis of peripheral blood and/or tumor tissue highlights the relationship between T cell clonality, diversity, intratumor heterogeneity, and clinical outcomes in patients with metastatic melanoma, Merkel cell carcinoma, NSCLC, and others.15 In advanced NSCLC, patients with higher TCR richness in their peripheral blood during treatment had improved outcomes with prolonged progression-free and overall survival.16 However, these studies had limited sample sizes. Furthermore, changes in the T-cell repertoire have not been extensively characterized at the time of irAEs.4 17–19

Here, we systematically analyzed a set of longitudinally collected patient peripheral blood samples from the LONESTAR clinical trial to evaluate changes in T-cell repertoire, using standard metrics, to determine their temporal dynamics with ICI treatment, and to determine their association with radiologic response and irAEs on dual ICI therapy.15 16 18–21

Patients and methodsPatient cohort

Longitudinal patient plasma samples were collected from the ongoing LONESTAR clinical study (Randomized phase III trial of local consolidation therapy (LCT) after nivolumab and ipilimumab for immunotherapy-naive patients with metastatic NSCLC, ClinicalTrials.gov identifier: NCT03391869), conducted at The University of Texas MD Anderson Cancer Center. This open-label, single-center, randomized clinical study enrolled patients with histologically or cytologically confirmed metastatic NSCLC. Key exclusion criteria included prior immunotherapy or more than one prior line of chemotherapy, tumors harboring EGFR-sensitizing mutations or ALK fusions eligible for standard-of-care targeted therapies, and active, known, or suspected autoimmune disease. All patients provided written informed consent to participate in the study, including blood collection for the TCR beta sequencing analysis.

In the parent trial, eligible patients received ipilimumab at 1 mg/kg every 6 weeks and nivolumab at 3 mg/kg (I+N) every 2 weeks for 12 weeks (induction therapy); for the LONESTAR study, the patients who did not experience radiologic disease progression were then randomly assigned to LCT with radiation and surgery for residual disease versus no LCT (online supplemental figure 1). The protocol synopsis is provided in the online supplemental file.

Response assessment

The radiologic response to treatment was assessed prospectively according to the Response Evaluation Criteria in Solid Tumors (RECIST) V.1.1.22 Patients with the non-progressive disease were defined as those with complete response (CR), partial response (PR) or stable disease (SD) in a 12-week response assessment scan and termed “responders” in this manuscript; those with progressive disease were termed “non-responders”.

Sample collection

Blood samples were collected at (1) baseline (before I+N therapy, on the same day as cycle 1 of therapy) (time point A), (2) after I+N induction (time point B, 12 weeks after cycle 1, this time point samples from patients with no irAEs were also used as no toxicity “control”), and (3) at the time of a grade ≥2 irAE (time point C if applicable) (figure 1). IrAEs were prospectively collected and graded using the Common Terminology Criteria for Adverse Events V.4.03.23 IrAE samples were obtained within 14 days of symptom onset during a clinic visit. If systemic steroids were used, the duration and dose of the steroid use have been captured. The attribution of adverse events to the ICIs was made prospectively by available clinical, laboratory, and histologic data and reported according to the National Cancer Institute adverse reporting guidelines.24 If there were≤14 days between two time points, only one sample was used to analyze both time points (eg, postinduction and irAE). If patients did not experience any irAEs within at least 6 months of follow-up, their samples from the end of induction were used as no-toxicity “control” for comparative analyses.

Figure 1

Time points for blood collection. Time point A: Baseline (prior to ICI therapy, n=107). Time point B: After 12 weeks of ICI therapy 12-week samples (this time point also used as control for treated patients with no irAEs, n=91). Time point C: At the time of irAE (if applicable, n=77). 25 toxicity samples had an overlapping collection time point (eg, postinduction and toxicity, as described in methods). ICI, immune checkpoint inhibitors; irAEs, immune-related adverse events; LCT, local consolidation therapy.

TCR sequencing

Sequencing of the complementary determining region 3 (CDR3) of human TCRβ chains was performed using the immunoSEQ Assay (Adaptive Biotechnologies). Libraries are made using multiplex PCR primers which target the CDR3 of the human TCRβ gene following rearrangement of the variable (V), diversity (D) and joining (J) gene segments. Sequences were collapsed and filtered in order to identify and quantitate the absolute abundance of each unique TCRβ CDR3 region for further analysis, as previously described.25–27

Computational and statistical analysis

After the processed TCR data were downloaded from Adaptive, we used several metrics to characterize the diversity and clonality of each sample and compared the metrics between groups. We first estimated the T cell fraction for each sample, which is defined as the number of total productive templates divided by the total number of input cells (total amount of DNA loaded divided by 0.0066). Several diversity indices were computed, including the Chao1, diversity 50 (D50) and inverse Simpson.28 29 Our analysis of T-cell repertoire clonality integrated density and diversity metrics and assessed the degree of clonal expansion of different T cell clonotypes within a sample. Clonality is defined by the degree of clonal expansion of different T cells within a sample.14 30 Chao1 is a non-parametric abundance-based species richness estimator that has been used to estimate repertoire diversity and is particularly useful for data sets skewed toward low abundance.31 Baseline (time point A), postinduction (time point B), and toxicity (time point C) time points are further split by responders and non-responders. This resulted in a minimum sample size of 25 per group, which achieves 80% power to detect a minimum effect size of 0.808 in comparing the T cell diversity metrics using a two-sample t-test. For ratio-like outcomes such as the D50 index, a sample size of 25 per group achieved 80% power to detect a ratio of 1.71 between the responders and non-responders with a two-sided Mantel-Haenszel test.32 The significance level of the test was targeted at 0.05. The actual achieved significance level is 0.0649.

D50 index reflects the percentage of unique CDR3 sequences accounting for 50% of the total number of observed sequences and is a measure of T-cell diversity that considers both the number of T-cell clones and their relative abundance.33 The Simpson index defines sample diversity, by , where is the proportional abundance of each species and R is the total number of species in the sample, the inverse Simpson index is calculated as , with a larger value representing higher diversity. The Hill number index, another measurement of diversity, reflects the effective number of species in a population. In addition to measuring density, relative abundances were computed and categorized as small (0<x ≤0.0001), medium (0.0001<x ≤0.001), large (0.001<x ≤0.01), or hyperexpanded (0.01<x ≤1) based on the values.

The indices were then compared between the non-responder and responder groups at baseline and at the end of the induction. Non-responders were defined as patients showing progressive disease, and responders were defined as patients showing CR, PR, and SD (in 12 weeks). Moreover, we compared indexes between samples from patients showing toxicity (toxicity) at the time of their irAE and those from patients not showing toxicity (non-toxicity) at the end of the induction time point. For each comparison, statistical p values were computed using the Wilcoxon test.

We first analyzed time of toxicity samples from patients with pneumonitis or diarrhea/colitis to identify TCR clonotypes shared among different patients, that is, sequences with a prevalence larger than 1 for each given toxicity. Next, we identified the enriched sequences by comparing these sequences to the sequences from all patient samples with no irAE. We then scrutinized the public data set VDJdb, a curated TCR sequence database with known antigen species.34 35 We compared enriched sequences from our study cohort and identified similar sequences with known antigen species in VDJdb. During the query process, the max mismatch tolerance was set to three (the assumption is that the two sets share the same antigen species if the mismatch between our sequence and the sequence in VDJdb is less than three) for the overlap with TCR databases from viral infections. Since some antigen species can be more abundant than others in the database, we counted the number of queries when summarizing the antigen species prevalence among TCR sequences from patients with pneumonitis and diarrhea/colitis toxicities. Then, we performed a Fisher’s exact test for the antigen species prevalence among patients having pneumonitis and diarrhea/colitis toxicities. The resulting p values were adjusted using the Benjamini and Hochberg method to control for the type I error rate of multiple comparisons.36 Those significant antigen species were identified with adjusted p values (false discovery rate (FDR)) less than 0.05.

To capture the unique information carried by the longitudinal data structure, we conducted generalized linear mixed models with diversity indices as responses, aiming to detect longitudinal effects from responses and toxicity. To estimate the average changes over time in each index for each patient, we sorted the samples by time and created longitudinal spaghetti plots, tracing each patient’s path of change for each index. Furthermore, we calculated the slopes of line segments bounded by two adjacent time points to measure the rate of changes. Χ² tests were conducted to detect differences between non-responder and responder groups, as well as between toxicity and non-toxicity groups.

ResultsPatient cohort

All patients had metastatic NSCLC and were immunotherapy-naïve at baseline. The median age of the cohort was 67 years old, 61 (51.3 %) were women, and 84 (70.6%) had adenocarcinoma histology. In total, 250 samples from 119 patients were used for the analysis including 107 drawn at baseline, 91 following 12-week exposure to ICI (postinduction), and 77 at the time of toxicity development. 25 toxicity samples had an overlapping collection time point (eg, postinduction and toxicity, as described in Methods). Eight of the toxicity samples (8.4%) were collected while patients were on systemic steroids for toxicity management. The clinicopathological characteristics of the cohort, types and grades of immune adverse events, description of supportive therapies for irAEs including systemic therapies in patients who had subsequent post tox samples are detailed in online supplemental tables S1–S3.

Increased peripheral T cell diversity at baseline is associated with response to a combination of ipilimumab and nivolumab

Standard metrics used to analyze T-cell repertoire include T-cell density, diversity, and clonality.33 T-cell density refers to the total number of T cells found within the sample.14 T-cell diversity accounts for both TCR sequence “richness” and “evenness,” which reflect the number of unique TCR sequences and the distribution of TCR sequences, respectively.33 Clonality integrates density and diversity metrics and assesses the degree of clonal expansion of different T cells within a sample.14 30 These components can show the balance between T cell expansion and contraction in response to different stimuli.27 Specific diversity measures aim to assess clonal dominance.30 In addition, defined indexes are available to aid in comparing T-cell repertoires between different samples.33

We evaluated the T-cell repertoire metrics and their relationship with clinical response in baseline and after induction samples respectively, as summarized in figure 2A,B.33 In the whole study cohort T-cell clonality (p=0.77) and the frequency of most dominant clonotypes (p=0.64) demonstrated no difference between responders and non-responders, while T-cell richness was significantly higher in the peripheral blood of responders (p=0.04, figure 3A–C). When analyzing the correlations of baseline TCR repertoire parameters among radiologic response groups of CR+PR versus SD versus progressive disease (PD) cohorts, we revealed a trend of higher T cell richness with the depth of the response (figure 3D–F). When compared CR+PR versus PD difference in baseline T cell richness was statistically significant (p=0.03) (online supplemental figure 2). In subgroup analysis, previously chemotherapy-exposed patients had statistically (p=0.04) higher T cell richness in responders, whereas chemotherapy naïve patients had a similar trend, but this did not statistical significance (p=0.1) (online supplemental figure 3–5). In responder patients (CR+PR+SD), when adenocarcinoma and non-adenocarcinoma NSCLC histologies were evaluated separately, there was a trend in both histologies for increased baseline T cell richness, although with smaller sample size the statistical findings lost their significance in the subgroup analysis (p=0.17 for adenocarcinoma histology, p=0.08 for non-adenocarcinoma histology) (online supplemental figures 6 and 7). Diversity metrics such as Chao1, D50 index, and number of unique clonotypes were also higher in responders supporting the importance of peripheral T-cell repertoire diversity and its association with response (online supplemental figure 8).

Figure 2

Overview of treatment response, clinicopathologic features and T cell indices. (A) T cell indices from the baseline peripheral blood represented in a heatmap with clinicopathologic features and treatment responses. (B) T cell indices from the post 12-week therapy peripheral blood represented in a heatmap with clinicopathological features and treatment responses. Cell.type: Histopathology of the NSCLC (Adeno, adenocarcinoma; NOS, NSCLC otherwise non-specified; SCC, squamous cell carcinoma). Radiological responses further defined as CR, complete response; PD, progressive disease; PR, partial response; SD, stable disease. NSCLC, non-small cell lung cancer.

Figure 3

T cell receptor repertoire (T-cell richness, T cell clonality, T cell max frequency) comparisons in baseline (prior to ipilimumab and nivolumab therapy) for responders (complete response (CR)+partial response (PR)+stable disease (SD)) versus progressive disease measured by RECIST V.1.1 (A, B and C). Comparisons of baseline T cell repertoire parameters among radiologic response groups CR+PR versus SD versus PD (D, E and F). Total of 69 responders (complete response: 1 patient; partial response: 28 patients; stable disease: 40 patients), 24 non-responder (progressive disease). PD, progressive disease; RECIST, Response Evaluation Criteria in Solid Tumors.

No major differences were observed in T-cell richness, T-cell clonality and T-cell max frequency after 12 weeks of therapy in responders compared with patients who had progressive disease (online supplemental figure 9). Next, we assessed longitudinal changes in T-cell richness, clonality and frequency from baseline to the end of induction therapy. After 12 weeks of I+N therapy in the entire study cohort, there were no differences in these metrics after induction therapy (online supplemental figure 10). Similarly, these metrics showed no significant differences after induction therapy in responders versus non-responders (online supplemental figures 11 and 12). Responders were revealed to have a higher D50 index (p=0.05) as well as a trend towards the higher expansion of hyperexpanded T-cell clonotypes (p=0.09; online supplemental figure 13). Furthermore, the longitudinal assessment revealed that responders exhibited changes of greater magnitude in their T-cell clonotypes (p=0.0057; online supplemental figure 14). Low and medium frequency clones also exhibited a similar trend but this was not significant (p=0.22 and 0.6; online supplemental figures 15 and 16).

Changes in TCR repertoire stratified by irAEs

Considering the role of T-cells in antigen-specific responses and prior associations with irAEs, we next evaluated the T-cell repertoire in patients experiencing toxicities. A total of 77 patients’ samples from the time of a grade ≥2 irAE; postinduction samples from 36 patients who did not develop irAEs were used as non-toxicity controls in this analysis. Interestingly, patients with irAEs showed lower T-cell richness in the periphery, whereas T-cell clonality, inverse Simpson index, and maximum frequency did not show any significant differences (figure 4). In patients with irAEs, when the baseline T-cell repertoire was compared with the time of an irAE, decreased T-cell richness was noted (p=0.002). Furthermore, T-cell density was lower in the toxicity group (p=0.02) when compared with the “control” group (12-week time point in I+N treated patients with no irAEs) (figure 5). There was a trend in lower T cell richness with higher grade irAE (grade 2 vs grade ≥3) but this was not statistically significant (online supplemental figure 17). Finally, longitudinal samples from the toxicity group had a higher change in absolute slope means for the relative abundance of medium and small clonotypes, top 1,000 and top 3,000 T cell clones in toxicity samples (online supplemental figure 18).

Figure 4

T cell receptor repertoire comparisons at the time of toxicity (grade ≥2 immune related adverse events) versus postinduction (after 12-week treatment with ipilimumab and nivolumab therapy) in patients with no toxicity. Total of 77 toxicity samples and 36 end of induction/no toxicity samples are included in the analysis.

Figure 5

T cell richness and density at the time of toxicity (A) Longitudinal changes of T cell richness in patients’ samples with toxicity, baseline (prior to ipilimumab and nivolumab therapy) and end of induction (after 12-week treatment with ipilimumab and nivolumab therapy). Toxicity: grade ≥2 immune related adverse events. (B) Box plots for T cell density when end of induction (after 12-week treatment with ipilimumab and nivolumab therapy) samples in no toxicity group compared with toxicity samples in patients with toxicity. Toxicity: grade ≥2 immune related adverse events.

When early toxicities (within 12 weeks of ICI therapy) were compared with delayed toxicities (>12 weeks of ICI therapy), T-cell richness was statistically higher in early toxicities, while there was a trend of higher T cell inverse Simpson and lower T-cell clonality although these differences did not reach statistical significance (p=0.06; figure 6).

Figure 6

Comparison of TCR repertoire related with the timing of irAE, Early grade ≥2 immune related adverse events (within 12 weeks of therapy; 30 samples) versus delayed grade ≥2 immune related adverse events (>12 weeks of therapy; 38 samples.).

We identified 27 patient samples that were obtained following the resolution of the irAE. When irAE recovery samples were compared with samples collected at the time of toxicity, the T cell richness trended to increase in this subgroup, but this was not statistically significant (online supplemental figure 19). In this subgroup seven patients received systemic steroids for the management of irAE, when toxicity samples compared with the samples obtained after the recovery of toxicity following systemic steroid use did not show a statistical difference in the longitudinal assessment of T cell repertoire although this analysis was limited due to sample size (online supplemental figure 20). Lastly, we sought to leverage large catalogs of TCR-epitope pairings to gain insight into the potential specificities of T-cell clonotypes. This analysis using the VDJdb TCR database showed an increased overlap in patients with pneumonitis with TCR databases from viral infections, such as Cytomegalovirus and Epstein-Barr virus. Similar findings were observed for diarrhea/colitis with Epstein-Barr virus, Influenza A, and Cytomegalovirus (figure 7). We also examined the TCR sequences in two response groups (CR+PR+SD vs PD) against the VDJdb TCR database. We performed a χ² statistical test for each antigen overlap, considering the sequence number in each response group and the VDJdb antigen categories. Then, we ranked the antigens by their significance and plotted the top-ranked antigens for CR+PR+SD and PD patients; differences were observed in the top-ranked antigens (online supplemental figure 21).

Figure 7

UB enrichment analysis using TCR database and assessment of overlaps with irAE pneumonitis and diarrhea. Enrichment analysis using VDJdb and assessment of overlaps with samples from at the time of grade ≥2 irAE for pneumonitis (A), diarrhea/colitis (B). Significantly expressed sequences among patients with pneumonitis and diarrhea/colitis were extracted and then tested for enriched antigen species referring to VDJdb. A higher value of “−log10(FDR)” shows more significance. CMV, Cytomegalovirus; DENV, Dengue virus type 1; EBV, Epstein-Barr virus; FDR, false discovery rate; HCV, hepatitis C virus; irAE, immune-related adverse event; LCMV, lymphocytic choriomeningitis virus; MCMV, murine Cytomegalovirus; RSV, respiratory syncytial virus; SIV, simian immunodeficiency virus; sTCR, T-cell receptor; VDJdb, variable, diversity, joining database; YFV, yellow fever virus.

Discussion

In this study, we systematically evaluated the associations between peripheral blood TCR repertoire metrics and clinical outcome including response and irAEs in patients with metastatic NSCLC treated with ipilimumab+nivolumab combinations. We found that patients who experienced a response had a higher T-cell diversity using various parameters including Chao1 index, D50 diversity index, and clonotype distribution compared with patients who had intrinsic resistance to dual checkpoint inhibition. In addition, patients who experienced irAEs had lower T-cell richness at the time of toxicity compared with the non-toxicity group. Furthermore, increased TCR overlaps were observed in patients with pneumonitis when compared with TCR databases from viral infections, such as Cytomegalovirus and Epstein-Barr virus. Similar findings were also observed for diarrhea/colitis with Cytomegalovirus, Influenza A and Ebstein-Barr virus.

The D50 diversity index is a measure of T-cell diversity that considers both the number of T-cell clones and their relative abundance.33 A higher percentage is indicative of higher diversity, which may correlate with a more even distribution of T-cell clones, and therefore may be relevant for generating an antitumor immune response with ICI therapy. Last the measurement of clonotype distribution as a function of CDR3 length may be indicative of the antigen specificity of the T-cell response, which may provide a broader and more effective immune response.37

Overall, most studies have reported that high TCR diversity is often associated with better response, whereas low diversity correlates with more aggressive phenotypes.38 In our study, the T cell richness was higher at baseline in responders, compared with non-responders, consistent with the literature.15 In addition, diversity metrics such as the Chao1 index, number of clonotypes, and D50 index were higher in responders than in non-responders. Perhaps the increased diversity at baseline allows for a better selection of T-cell clones that can be active toward tumor antigens. In a previous report, it was reported that cytotoxic T-lymphocytes-associated protein 4 (CTLA-4) blockade broadens the TCR repertoire and lowers the incidence of clonotype loss after therapy, correlating with better clinical outcomes.39 This may explain some of the attributes of durable responses with anti-CTLA-4 therapies.40 41 We also noted that while baseline T cell richness was higher in responders than non-responders in both previously chemotherapy-exposed and chemotherapy-naïve patient subsets, the difference was only significant in previously chemotherapy-exposed patients in subgroup analysis. While this observation may be due to limited sample size, an alternative explanation is that cytotoxic cell death from chemotherapy may lead to a more significant number of T-cell clones with unique TCRs (richness) in a subset of patients, potentially resulting in better clinical outcomes.

In this study, we observed that T-cell richness was lower in the periphery during irAE events, there was a trend in lower T-cell richness and clonality with higher grade irAE (grade 2 vs grade ≥3), but this was not statistically significant (online supplemental figure 17). Lower T-cell richness in the periphery during an irAE event is perhaps due to T cell trafficking to affected organs, compared with that in patients who were exposed to ICI and did not experience any irAEs, but it is impossible to confirm without assessing tissue. A similar finding was reported for human chronic viral infections, such as Epstein-Barr virus, Cytomegalovirus, and Hepatitis C virus infections, and the narrow TCR repertoires reported in these patient subsets have been thought to be secondary to the constant exposure to viral antigens leading to clonal expansion of T cells.5 42

While we did not observe any correlations between irAEs and baseline peripheral blood TCR parameters, it is important to note that in a previous study using single-cell sequencing paired with single-cell V(D)J sequencing of TCR reported that single-cell TCR clonotype diversity (Shannon entropy) of activated CD4 effector memory T cells were higher in patients who experienced severe irAEs. Therefore, the T cell subpopulation, which could not be captured by the bulk TCR sequencing immunoSEQ applied in the current study, may be contributing to response dynamics at baseline.43

We hypothesized that early and delayed toxicities may reveal different T cell repertoire changes; therefore, we compared T cell repertoire changes in early toxicities (within 12 weeks of ICI therapy) to delayed toxicities (>12 weeks of ICI therapy). In our cohort, T cell richness was higher in early toxicities and there was a trend of lower T cell clonality, but this did not reach statistical significance. This suggests that some T cell dynamics may be different, and further studies may provide further insight into the mechanistic significance of the timing of toxicity and T cell repertoire changes.

This hypothesis-generating study has several strengths. In this study, the samples were collected from a prospective clinical trial, where all patients received the same immunotherapy regimen in a controlled schedule and toxicity attributions were defined prospectively, less than 10% of toxicity samples were collected while patients were on systemic steroid therapy. All baseline samples were collected from immunotherapy-naïve patients and serial samples were available for comparative analysis, and 12-week “control” samples were selected from patients who were observed for six subsequent months without the development of irAEs. There were differences in clinicopathologic features, response dynamics and the duration of follow-up between the “toxicity group” and “control group” summarized in online supplemental tables 1 and 2, which may cause biological imbalances between these two groups, although the duration of follow-up was similar between both groups, and patients with follow-up less than 6 months were only 9% in the toxicity group. The unique serial collection of samples from a prospective trial has better controlled for confounding factors that may impact correlative analysis that are often unavoidable in retrospective studies. However, our study does exhibit several important limitations. First, as a prospective study, we were not able to obtain an independent validation cohort. Second, as the parent protocol, the LONESTAR trial has not been unblinded yet, so we were not able to correlate our findings with patient survival. Immune-based treatments may result in delayed clinical responses and unconventional response patterns due to increased intratumoral immune cell infiltration, which has been reported in patients with NSCLC at a rate ranging between 0.6% and 5.8%, where radiologic findings suggest tumor growth without actual progressive disease.44 45 In the LONESTAR study, the initial radiologic response assessment for systemic therapy with ipilimumab and nivolumab was obtained in 12 weeks to minimize the risk of mislabeling delayed clinical responders and pseudoprogression as PD. Nevertheless, RECIST V.1.1 was used for radiologic response assessment for trial purposes, and potentially, a subset of patients with delayed responses may be missed in this assessment. The trial enrolled patients with metastatic disease. While the trial allowed surgery as a local consolidative modality, few patients had tumor resection; therefore, tissue-based TCR analysis was not performed. Finally, as is the case in all studies of peripheral blood, T cells analyzed include diverse T cell specificities, most of which are not relevant to antitumor immune response, which may limit interpretation of findings associated with immunotherapy efficacy. Nevertheless, this study suggests that the peripheral blood T-cell repertoire may reflect the overall immune response to a certain degree and may have a potential impact on clinical response and toxicities to immunotherapy. As blood-based tests are non-invasive and real-time, these findings suggest that analysis of the peripheral blood T-cell repertoire may have the potential to help guide patient selection for immunotherapy.