A, Structure of the TCR. The TCR recognizes peptide antigens presented by the major histocompatibility complex (MHC) class 1. The TCR is a heterodimer of 2 subunits, TCRα and TCRβ, with each subunit consisting of a constant region that functions to anchor the receptor to the cell membrane and a hypervariable region that functions in antigen recognition. In this study, the hypervariable regions of the TCRβ subunit were characterized. B, Heatmap of TRBV gene use. Heatmap of log counts per million (CPM) of 245 patients across 65 TRBV genes. Significantly stable clusters (top row) are identified through bootstrap resampling of hierarchical clustering. Pathologic complete response (pCR; middle row) indicates patients who had a pCR in light blue and those with residual disease (RD) in dark blue. Estrogen receptor (ER) status (bottom row) indicates patients who are ER positive in dark red and patients who are ER negative in light red.
Significant differences between groups determined by use of the Fisher exact test. Error bars indicate 95% CI.
eFigure 1. Flow Diagram of Patients and Samples Used in Analysis
eFigure 2. Comparison of Global TRBV Metrics
eFigure 3. Correlation of TRBV Usage Metrics and Prognostic Features
eFigure 4. Pairwise Correlation of Usage in Individual TRBV Genes
eFigure 5. Comparison of NMF-Based Clustering and Direct Clustering at Various Ranks
eFigure 6. Model Diagnostics of NMF Using Various Rank Representations
eFigure 7. Heatmap of TRBV Gene Use
eFigure 8. Comparison of Metagene Levels
eFigure 9. Direct Comparison of Immune-Related Features
eFigure 10. Pathologic Complete Response (pCR) Rates Compared Between Treatments Dichotomized by High or Low Immune Enrichment Signature and High and Low Use of TRBV11-3
eTable 1. Patient Characteristics in the Sequenced and Analyzed Cohorts
eTable 2. IMGT Designation of TRBV Genes Included in Each Metagene
eTable 3. Significance of TRBV Features Used in ER-Adjusted Logistic Regression Models to Predict pCR
eTable 4. Multivariable Logistic Regression of TRBV Features, Clinical Covariates, and Gene Expression Signatures to Predict pCR While Comparing Only Trastuzumab and Combined Therapy in Treatment Arm
eTable 5. Multivariable Logistic Regression of TRBV Features, Clinical Covariates, and Gene Expression Signatures to Predict pCR While Comparing Only Lapatinib and Combined Therapy in Treatment Arm
eAppendix 1. Methods
eAppendix 2. Statistical Analysis Logs
Customize your JAMA Network experience by selecting one or more topics from the list below.
Powles RL, Redmond D, Sotiriou C, et al. Association of T-Cell Receptor Repertoire Use With Response to Combined Trastuzumab-Lapatinib Treatment of HER2-Positive Breast Cancer: Secondary Analysis of the NeoALTTO Randomized Clinical Trial. JAMA Oncol. 2018;4(11):e181564. doi:10.1001/jamaoncol.2018.1564
Does the use of specific T-cell receptor β variable region (TRBV) genes add to the association of immune markers with anti-HER2 therapy in HER2-positive breast cancer?
In this biomarker study of data from the NeoALTTO trial, high levels of a TRBV metagene consisting of 3 individual TRBV genes identified a subgroup of patients who respond poorly to single HER2 blockade treatments, irrespective of overall immune signature expression, but who experience a significant benefit from dual anti-HER2 blockade.
Direct measurements of TRBV use is an independent biomarker for response to dual HER2-targeted therapy.
Dual anti-HER2 blockade increased the rate of pathologic complete response (pCR) in the Neoadjuvant Lapatinib and/or Trastuzumab Treatment Optimisation (NeoALTTO) trial, and high immune gene expression was associated with pCR in all treatment arms. So far, no marker has been identified that is specifically associated with the benefit from dual HER2 blockade.
To examine if use of the T-cell β chain variable genes adds to the potential association of immune gene signatures with response to dual HER2 blockade.
Design, Setting, and Participants
In the NeoALTTO trial, HER2-positive patients recruited between January 5, 2008, and May 27, 2010, were treated with paclitaxel plus either lapatinib or trastuzumab or both as neoadjuvant therapy. In this study, RNA sequencing data from baseline tumor specimens of 245 patients in the NeoALTTO trial were analyzed and reads were aligned to TRBV gene reference sequences using a previously published Basic Local Alignment Search Tool T-cell receptor mapping pipeline. Total TRBV gene use, Shannon entropy, and gene richness were calculated for each tumor, and nonnegative matrix factorization was used to define TRBV co-use metagenes (TMGs). The association between TRBV metrics, tumor genomic metrics, and response was assessed with multivariable logistic regression. Statistical analysis was performed from January 23 to December 2, 2017.
Main Outcomes and Measures
The association between TRBV use metrics and pCR.
Among the 245 women with available data (mean [SD] age, 49  years), total TRBV use correlated positively with a gene expression signature for immune activity (Spearman ρ = 0.93; P < .001). High use of TRBV11-3 and TMG2, characterized by high use of TRBV4.3, TRBV6.3, and TRBV7.2, was associated with a higher rate of pCR to dual HER2-targeted therapy (TRBV11-3 interaction: odds ratio, 2.63 [95% CI, 1.22-6.47]; P = .02; TMG2 interaction: odds ratio, 3.39 [95% CI, 1.57-8.27]; P = .004). Immune-rich cancers with high TMG2 levels (n = 92) had significantly better response to dual HER2-targeted treatment compared with the single therapy arms (rate of pCR, 68% [95% CI, 52%-83%] vs 21% [95% CI, 10%-31%]; P < .001), whereas those with low TMG2 levels did not benefit from dual therapy. High TMG2 levels were also associated with a higher rate of pCR to the combined therapy in immune-poor tumors (n = 30; pCR, 50% [95% CI, 22%-78%] vs 6% [95% CI, 0%-16%]; P = .009).
Conclusions and Relevance
Use patterns of TRBV genes potentially provide information about the association with response to dual HER2 blockade beyond immune gene signatures. High use of TRBV11.3 or TRBV4.3, TRBV6.3, and TRBV7.2 identifies patients who have a better response to dual HER2 targeted therapy.
ClinicalTrials.gov Identifier: NCT00553358
The Neoadjuvant Lapatinib and/or Trastuzumab Treatment Optimisation (NeoALTTO) trial demonstrated that trastuzumab, a monoclonal antibody against HER2, and lapatinib, a small-molecule tyrosine kinase inhibitor of the EGFR and HER2 family, administered together with paclitaxel chemotherapy in the neoadjuvant setting increased rates of pathologic complete response (pCR) by approximately 20% in patients with early-stage HER2-positive breast cancer compared with either treatment alone.1 A higher tumor-associated lymphocyte count2 and a higher expression of immune signatures3 were associated with higher rates of pCR but not with specific benefit from dual treatments. In this study, we examine whether specific T-cell receptor (TCR) species could indicate an association with response to a specific treatment arm.
The TCR recognizes peptide antigens bound to major histocompatibility complex molecules and is composed of 2 different protein chains, α and β (Figure 1A). Each TCR possesses unique antigen specificity determined by the structure of the antigen-binding site formed by the α and β chains. T-cell receptor diversity arises from the random combinatorial joining of variable, joining, and diversity genes to generate each complete chain that defines antigen specificity. We estimate the diversity of T-cell populations in the tumor microenvironment by assessing the messenger RNA expression of the T-cell receptor β chain variable (TRBV) gene from previously published RNA sequencing data3 of pretreatment biopsies of patients in the NeoALTTO trial.
A total of 455 patients with HER2-positive early-stage breast cancer were randomized to 3 treatment arms in the NeoALTTO trial.1 Baseline biopsies were performed on 423 of these patients, and RNA was successfully extracted from 254 patients, of which 245 had matching clinical data (eFigure 1 in the Supplement). Data for multivariable analysis were available for 225 patients (eTable 1 in the Supplement). We adapted a previously published method4 to map aligned reads from bulk RNA sequencing data against a database of human TCR genes to identify use of variable region TCR sequences (the software is available at https://github.com/ElementoLab/TCRVseq). The total TRBV load, TRBV richness, and TRBV entropy were used to characterize TRBV use in tumors and to identify co-use of TRBV metagenes using nonnegative matrix factorization (eAppendix 1 in the Supplement). Features of TRBV use were evaluated for association with pCR using multivariable logistic regression adjusted for clinical covariates (eAppendix 2 in the Supplement). This biomarker study was approved by the Human Investigations Committee of Yale Cancer Center. Written informed consent was obtained from all patients at entry into the NeoALTTO trial, which also covered future biomarker research.
The read length of standard RNA sequencing does not allow reconstructing the entire T-cell β chain; for this reason, we refer to use of a specific TRBV gene in TCRs in a tumor instead of TCR gene expression. Hierarchical clustering with bootstrap resampling identified a stable cluster of tumors that showed minimal use of most TRBV genes (Figure 1B). Patients in this immune cold cluster tended to have a lower rate of pCR compared with the remaining patients (28% [95% CI, 21%-35%] vs 38% [95% CI, 28%-50%]; P = .10 determined by use of the Fisher exact test). Overall, a median of 3.85 (range, 0.137-29.8) counts per million of TRBV genes were detected across samples, with a median of 53 different variants (range, 5-65 different variants) observed across tumors (eFigure 2 in the Supplement). The overall TRBV load was highly correlated with tumor-associated lymphocyte counts (eFigure 3 in the Supplement) and was significantly higher in patients who achieved a pCR (P = .009 determined by use of the Wilcoxon rank-sum test).
We found the use of different TRBV genes to be highly correlated (median pairwise Pearson correlation, 0.43 [range, –0.09 to 0.82]) (eFigure 4 in the Supplement) and used nonnegative matrix factorization to define 4 metagenes (TMG1-TMG4) that captured the broader TRBV use patterns across tumors (eFigures 5 and 6 in the Supplement). Each metagene contained at most 1 gene from each TRBV subgroup (eFigure 7 and eTable 2 in the Supplement), implying less than 75% sequence identity between genes of a given metagene. All 4 metagenes representing median use of the TRBV genes were significantly higher in patients with pCR (eFigure 8 in the Supplement).
Metrics for TRBV were generally correlated with other measures of immune activity but not with the rate of somatic mutation or proliferation metagene expression (eFigures 3 and 9 in the Supplement). Tumor-associated lymphocyte counts were available for 225 tumor samples in the cohort2 and found to be positively correlated with a previously reported immune signature (ρ = 0.43; P < .001), TRBV load (ρ = 0.37; P < .001), and the levels of all 4 TMGs (TMG1, ρ = 0.22; TMG2, ρ = 0.28; TMG3, ρ = 0.39; and TMG4, ρ = 0.34).
We assessed whether the global TRBV metrics, individual TRBV gene use, and TMGs, are associated with pCR in estrogen receptor–adjusted logistic regression. Only TRBV11-3 use (odds ratio, 1.67 [95% CI, 1.26-2.27]; adjusted P = .04) and metagene TMG2 (odds ratio, 1.49 [95% CI, 1.14-1.96]; adjusted P = .04) were significantly associated with pCR in the overall cohort (eTable 3 in the Supplement). In multivariable logistic regression that included clinical variables and known prognostic signatures (eAppendix 2 in the Supplement), TMG2 and TRBV11-3 had statistically significant treatment-arm interaction terms, indicating that higher use is associated with greater probability of pCR in patients treated with dual HER2 blockade compared with patients in the single treatment arms (Table). When both TMG2 and TRBV11-3 were considered together, only the TMG2 interaction term for combined therapy vs single treatments remained significant (interaction odds ratio, 5.03 [95% CI, 1.62-18.46]; P = .009) (Table).
To examine the added information on association with pCR provided by TMG2, patients were dichotomized (median split) by TMG2 level and by immune signature expression into high and low groups. Patients with high TMG2 levels and immune-rich tumors (92 of 245 [37.6%]) had a significantly higher pCR rate in the dual HER2-targeted arm compared with the other arms (68% [95% CI, 52%-83%] vs 21% [95% CI, 10%-31%]; P < .001 determined by use of the Fisher exact test) (Figure 2). A high TMG2 level was also associated with a higher rate of pCR to combined therapy in immune-poor tumors (30 of 245 [12.2%]), where the rate of pCR was 6% (95% CI, 0%-16%) with monotherapy vs 50% (95% CI, 22%-78%) with dual HER2 blockade (P = .009 determined by use of the Fisher exact test). The interaction of TMG2 by treatment was also significant for the lapatinib vs combined therapy but not for trastuzumab vs combined therapy (eTables 4 and 5 in the Supplement) in 2-marker models, suggesting that the T-cell metagene may be associated with a specific trastuzumab benefit in the context of comparison with combined therapy. High use of the single variable-gene TRBV11-3, which is not included in any of the 4 TMGs, showed the same characteristics as TMG2 in immune-rich and immune-poor cancers (eFigure 10 in the Supplement).
Our analysis indicates that use of a very small number of TRBV genes (TRBV4.3, TRBV6.3, TRBV7.2, and TRBV11.3) appears to be associated with selective response to dual treatment with trastuzumab and lapatinib, both in immune-rich and immune-poor cancers. The small number of TRBV genes involved suggests the possibility that only a small number of genes, and possibly a small number of TCRs, are needed to estimate the immunogenic activity associated with trastuzumab response and subsequent sensitivity to dual HER2 blockade, even when overall immune infiltration is modest. These TRBV genes have very high use in normal CD4-positive follicular helper T cells5 that have been associated with better outcomes after therapy,6 which may suggest as a possible mechanism that previous infections in some patients could prime the TFH effector cells for responses that augment the outcome of HER2-targeted therapy regimens.7
One limitation of our study is that, owing to the insert size of the RNA sequencing complementary DNA library and the limits on read length in current sequencing technology, we were unable to fully reconstruct the T-cell β chain to identify specific rearranged TCRs from bulk RNA sequencing data. Moreover, these findings are exploratory and will need to be confirmed in additional cohorts, and potentially with dual regimens combining trastuzumab with pertuzumab. Finally, this study was not designed to assess whether the pCR benefit translates to better survival outcomes in these patients.
Use of a small number of TRBV genes potentially provides information about the association with response to dual HER2 blockade beyond immune gene signatures. If confirmed in additional cohorts, the TRBV signature could help identify patients who do poorly with single anti-HER2 treatments, but benefit from dual blockade.
Accepted for Publication: March 20, 2018.
Corresponding Author: Christos Hatzis, PhD, Breast Medical Oncology, Yale Cancer Center, Yale School of Medicine, 333 Cedar St, PO Box 208032, New Haven, CT 06520 (email@example.com).
Published Online: June 14, 2018. doi:10.1001/jamaoncol.2018.1564
Author Contributions: Dr Hatzis had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Mr Powles and Dr Redmond contributed equally to the study.
Study concept and design: Powles, Redmond, Di Cosimo, Elemento, Pusztai, Hatzis.
Acquisition, analysis, or interpretation of data: Powles, Redmond, Sotiriou, Loi, Fumagalli, Nuciforo, Harbeck, de Azambuja, Sarp, Di Cosimo, Huober, Baselga, Piccart-Gebhart, Elemento, Hatzis.
Drafting of the manuscript: Powles, Redmond, Nuciforo, Hatzis.
Critical revision of the manuscript for important intellectual content: Powles, Redmond, Sotiriou, Loi, Fumagalli, Harbeck, de Azambuja, Sarp, Di Cosimo, Huober, Baselga, Piccart-Gebhart, Elemento, Pusztai, Hatzis.
Statistical analysis: Powles, Redmond, Elemento, Hatzis.
Obtained funding: Sotiriou, Pusztai, Hatzis.
Administrative, technical, or material support: Fumagalli, Harbeck, de Azambuja, Huober, Pusztai.
Study supervision: Loi, Di Cosimo, Baselga, Piccart-Gebhart, Elemento, Hatzis.
Conflict of Interest Disclosures: Dr de Azambuja reported receiving travel grants from GlaxoSmithKline and Roche and honoraria from Roche, outside this work. Dr Sarp reported being an employee of Novartis. Dr Baselga reported receiving personal fees from Roche. Dr Piccart-Gebhart reported receiving personal fees from Roche. Drs Sotiriou and Piccart-Gebhart reported being co-inventors of the Genomic Grade Index. No other disclosures were reported.
Funding/Support: This project was supported by Investigator Awards from the Breast Cancer Research Foundation and by a Yale Cancer Center Core Grant (NCI P30 CA16359-34) to Drs Hatzis and Pusztai and by a Susan G. Komen Foundation grant to Dr Pusztai.
Role of the Funder/Sponsor: The funders/sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Create a personal account or sign in to: