A phase I/Ib trial and biological correlate analysis of neoadjuvant SBRT with single-dose durvalumab in HPV-unrelated locally advanced HNSCC

Abstract

Five-year survival for human papilloma virus-unrelated head and neck squamous cell carcinomas remain below 50%. We assessed the safety of administering combination hypofractionated stereotactic body radiation therapy with single-dose durvalumab (anti-PD-L1) neoadjuvantly (n = 21) ( NCT03635164 ). The primary endpoint of the study was safety, which was met. Secondary endpoints included radiographic, pathologic and objective response; locoregional control; progression-free survival; and overall survival. Among evaluable patients at an early median follow-up of 16 months (448 d or 64 weeks), overall survival was 80.1% with 95% confidence interval (95% CI) (62.0%, 100.0%), locoregional control and progression-free survival were 75.8% with 95% CI (57.5%, 99.8%), and major pathological response or complete response was 75% with 95% exact CI (51.6%, 100.0%). For patients treated with 24 Gy, 89% with 95% CI (57.1%, 100.0%) had MPR or CR. Using high-dimensional multi-omics and spatial data as well as biological correlatives, we show that responders had: (1) an increase in effector T cells; (2) a decrease in immunosuppressive cells; and (3) an increase in antigen presentation post-treatment.

Conflict of interest statement

S.D.K. is funded by Genentech for the ongoing phase II portion of this work, but there has been no overlap in the research. She is also funded by Roche and Ionis for work unrelated to the content of this manuscript. Though unrelated to the contents of this manuscript, the authors declare that A.D. is a founder of Omix Technologies Inc. A.D.A. is also a consultant for Altis Biosciences LLC, Rubius Inc. and Forma Inc. A.D.A. is a consultant for Hemanext Inc. A.J. has research support from NCI R01CA149456, R01DE024371, and P50CA261605; and stock/options ownership in Suvica and Champions Oncology. A.J.’s institution has contracts with Cantargia, DebioPharm, Genentech, Iovance, Khar Biopharma, Merck, Moderna, Pfizer, Sanofi, and SQZ for trials where A.J. is the local P.I. All other authors declare no competing interests.

© 2022. The Author(s).

Figures

Fig. 1. Summary of clinical outcomes.

a, Diagram of trial design. b, Summary of pathological outcomes of all evaluable patients (n = 19 patients). c, Summary of clinical outcomes of patients treated at MTD (n = 18). d, Kaplan–Meier survival curves for overall survival, PFS and local PFS survival for patients treated at MTD (n = 18 patients). e, A plot of the relationship between radiation dose and pathologic outcome at time of surgery (P = 0.05), spearman correlation coefficient 0.45, CI (−0.01909, 0.7573). 12 Gy n = 3 patients, 18 Gy n = 7 patients and 24 Gy n = 9 patients. f, Analysis of the relationship between time to surgery and pathological response (mean ± s.e.m.). Pathologic tumor response (pTR) pTR0–3 n = 6 patients and MPR/CR n = 13 patients. g, Representative images of a patient’s tumor response to treatment. h, Representative standardized uptake value (SUV) images showing a decrease in signal intensity post-treatment. A two-sided Fisher’s exact test was used in e. Statistical significance was determined by an unpaired two-sided Student’s t-test for f. Significance was concluded if P < 0.05. #For the last 9 patients the GTV was dosimetrically calculated at 24 Gy. Source data

Fig. 2. Responders had an increase in effector T cells within the TME.

a, viSNE plot of CD45+ cells in the TME of patients pre- and post-treatment stratified by response and colored by cell type. Density plots of two clusters are included below the viSNE plot (nonresponders n = 6 patients, responders n = 10 patients). b, Histograms representing the proportion of cells, within the cytokine-producing T cell cluster, producing IFN-γ and expressing TCF1. c, Histograms representing the proportion of cells, within the CD103+CD39+ CD8 T cell cluster, expressing Ki-67 and TCF1. d, t-Distributed stochastic neighbor embedding (t-SNE) plots of CD45+ cells in a responder and two nonresponders. Blue represents TME samples taken before treatment, and pink represents TME samples taken after treatment. e, Magnification of the large cluster of pathways increased in responders pre-treatment represented in Extended Data Fig. 3b (nonresponders n = 5 patients, responders n = 8 patients). f, Significant genes determined using the GSEAPreranked module, based on an adjusted P value to account for multiple comparisons, which were used in the gene mapping identified pre-treatment (nonresponders n = 5 patients, responders n = 8 patients). The gray bars denote pathological response. g, Average of the top 5 TCR sequences (nonresponders n = 3 patients, responders n = 5 patients). h, Scatterplot with annotations depicting clones with more than 8 transcripts before and after treatment for patients 01-010 and 01-014. Red dots were clones significantly increased pre-treatment and blue dots were clones that were significantly increased post-treatment. i, Quantification of CD8+ T cells (CD3+CD8+) within the TME pre- and post-treatment (nonresponders n = 5 patients, responders n = 7 patients). j, Quantification of VECTRA images of CD4 T cells (CD3+CD8−Foxp3−) within the TME pre- and post-treatment (nonresponders n = 5 patients, responders n = 7 patients). k, CPS (no. PD-L1-expressing cells/no. CK+ cells) calculated from VECTRA images (nonresponders n = 6 patients, responders n = 12 patients). l, Quantification of VECTRA images of PD-1-expressing T cells (CD3+PD-1+) within the TME pre- and post-treatment (nonresponders n = 6 patients, responders n = 12 patients). m, The percentage of cells expressing PD-L1 from VECTRA images (mean ± s.e.m.) (nonresponders n = 6 patients, responders n = 12 patients). n, Quantification of the median proximity of a CD8+ cell to a CK+ cell from VECTRA images (nonresponders n = 6 patients, responders n = 12 patients). o, Representative images of the distance between cancer cells (CK+) and the nearest CD8 T cells (highlighted in red). Statistical significance was determined using a two-tailed paired Student’s t-test. *P < 0.05, **P < 0.01, ***P < 0.001. NR, non-responder; R, responder. Source data

Fig. 3. Responders increase antigen presentation and TCR expansion.

a, Density plots of the antigen-presenting cluster from Fig. 2a (nonresponders n = 6 patients, responders n = 10 patients). b, Histogram representing the amount of CD86 expressed by cells within the antigen-presenting cluster. c, t-SNEs of CD45+ cells in a responder and two nonresponders. Blue represents TME samples taken before treatment, and pink represents TME samples taken after treatment. d, Left, representative VECTRA images of HLA-DR expression (left images with CK expression, right images without CK expression). Right, quantification of VECTRA images of HLA-DR expression on CK+ cells within the TME pre-treatment (mean ± s.e.m.) (nonresponders n = 4 patients, responders n = 12 patients). e, Representative VECTRA images of DLNs collected at time of surgery. DC–T cell interactions are highlighted with white arrows. f, Density of CD8+ T cells (CD3+CD8+) in the DLNs (nonresponders n = 3 patients, responders n = 3 patients). g, Density of activated CD8+ T cells (CD3+CD8+IFN-γ+) and replicating CD8+ T cells (CD3+CD8+Ki67+) in the DLNs (nonresponders n = 3 patients, responders n = 3 patients). h, Density of activated CD4+ T cells (CD3+CD8−Foxp3−IFN-γ+) in the DLNs (nonresponders n = 3 patients, responders n = 3 patients). i, Percentages of CD3+ cells that are CD8+ T cells, activated CD8+ T cells and activated CD4+ T cells (nonresponders n = 3 patients, responders n = 3 patients). j, Quantification of how many T cells were within 15 μm of APCs (CD3−HLA−DR+) (mean ± s.e.m.) (nonresponders n = 3 patients, responders n = 3 patients). k, Average of the top 5 TCR sequences pre- and post-treatment in the blood (nonresponders n = 3 patients, responders n = 5 patients). l, Scatterplot with annotations depicting clones with more than 8 transcripts after treatment in the TME and blood for patients 01-010 and 01-014. Red dots are clones significantly increased in the TME and blue dots are clones that were significantly increased in the blood. Dots along the y and x axes are clones not present in the TME or blood post-treatment, respectively. Statistical significance was determined using a two-tailed paired Student’s t-test. *P < 0.05, **P < 0.01, ***P < 0.001. Source data

Fig. 4. Responders decrease immunosuppressive cells in the TME.

a, VECTRA image quantification of Treg cells (CD3+CD8−Foxp3+) in patient tumors pre- and post-treatment (nonresponders n = 5, responders n = 7). Nonresponders P = 0.84 and responders P = 0.03. b, Ratio of CD8+ T cells (CD3+CD8+) to Treg cells in the TME pre- and post-treatment (nonresponders n = 5 patients, responders n = 7 patients). Nonresponders P = 0.28 and responders P = 0.0032. c, Representative VECTRA image of the TME pre- and post-treatment. Patients 01-002 and 01-007 representing two different kinds of non-responders to treatment. To the right, CD8+ T cells (red) and Treg cells (blue) are highlighted. d, VECTRA quantification of the distance between Treg cells and DCs (CD209+) in the TME of responders and nonresponders pre- and post-treatment (nonresponders n = 5 patients, responders n = 7 patients). Nonresponders P = 0.38 and responders P = 0.0024. e, MultiPLIER quantification of LV57 (nonresponders n = 5, responders n = 8). Nonresponders P = 0.0006 and responders P = 0.99. Statistical significance was determined using a two-tailed paired student’s t-test. *P < 0.05, **P < 0.01, ***P < 0.001. Source data

Fig. 5. Circulating lymphocytes and metabolites correlate with response.

a, viSNE depicting cell populations identified in the blood (nonresponders n = 3 patients, responders n = 9 patients). b, Expression of IFN-γ, c, TNF-α, d, IL-2, e, Tbet, f, TGF-β, g, IL-17A and h, DNAM in different populations in the blood. B cells are in pink, T cells are in blue and myeloids are in brown. i, Quantification of CD4+ effector memory T cells (CD3+CD19−CD56−CD14−CD4+CD8−Foxp3−CD45RA−CD27−) at baseline, time of surgery, EOT and 6-month follow-up (mean ± s.e.m.). C1D1 P = 0.57, surgery P = 0.016 and EOT P = 0.31. j, CD8+ EMRA T cells (CD3+CD19−CD56−CD14−CD4−CD8+CD45RA+, CD27−) (C1D1 P = 0.22, surgery P = 0.18, EOT P = 0.12 and 6-month follow-up P = 0.026) and CD8+ effector memory T cells (CD3+CD19−CD56−CD14−CD4−CD8+CD45RA−, CD27−) (C1D1 P = 0.37, surgery P = 0.28 and EOT P = 0.44) at baseline, time of surgery, EOT and 6-month follow-up (mean ± s.e.m.). For blood analysis at various time points in both i and j, C1D1 (nonresponders n = 3 patients, responders n = 10 patients), surgery (nonresponders n = 3 patients, responders n = 11 patients), EOT (nonresponders n = 3 patients, responders n = 8 patients) and for 6-month follow-up (nonresponders n = 4 patients, responders n = 2 patients). A two-tailed Student’s t-test was used to determine statistical significance, *P < 0.05. Source data

Fig. 6. Summary of mechanisms underlying treatment failure.

a, Diagram of the steps involved in a successful T cell-mediated anti-tumor response: (1) Initial TIL infiltration, antigen presentation and clonal expansion. (2) Antigen presentation in the lymph nodes and activation/replication of T cells. (3) Clonal expansion and T cell activation in the blood post-treatment. (4) Immune suppression by Treg cells in the TME after treatment. b, Summary of each patient who failed therapy compared with the average responder. No evidence of a step in a patient is depicted with ‘−’, some evidence of a step is indicated with ‘+’ and a lot of evidence of a step is indicated with ‘++’. TDLN, tumor draining lymph node.

Extended Data Fig. 1. Representative images of SBRT planning and response to treatment.

(A) Representative image of volumetric contouring of gross tumor only to determine the specific doses of SBRT delivered. (B) A multiple logistic regression analysis was used to determine if time to surgery or dose of radiation can independently account for pathological response at time of surgery (n = 19 patients). P = 0.07 (coefficient estimate per Gy of RT = 0.2253, s.e.m. = 0.1429, CI (−0.03648,0.5582); intercept for the logistic regression was −6.207). (C) Representative MRI of a patient pre- and post-treatment with tumor volume measurements magnified. (D) Representative CT image of a patient pre- and post-treatment with tumor volume measurements magnified. (E) White blood cell (WBC) count for each patient at four time points (C1D1, surgery, end of treatment (EOT) and 6-month follow-up). C1D1 n = 21 patients, surgery n = 20 patients, EOT n = 12 patients and 6-month follow-up n = 8 patients (P = 0.0006*** and P = 0.0019**). A multiple linear regression test was used to determine interdependence of time to surgery and dose of radiation for B. Statistical significance was determined by a paired two-sided Student’s t-test for E. Significance was concluded if P<0.05 (*P<0.05, **P<0.01, ***P<0.001). These are representative images of 21 patients.

Extended Data Fig. 2. Longitudinal Assessment of Quality of Life.

QOL data for each patient was collected using the FACT-H&N Version 4 Questionnaires. Sub scale scores and total scores were derived following the FACT-H&N Scoring Guidelines. This data is depicting information from 21 patients. A linear mixed model with random patient effect were used to test if the QOL scores changed by event timepoint for each of the sub scale scores and the total scores, and the Wald test results were reported. Multiple comparisons were adjusted using Benjamini and Hochberg’s method. Source data

Extended Data Fig. 3. Responders have an increase in genes and pathways associated with inflammation.

(A) Significant HALLMARK pathways that differed between responders and non-responders at baseline (non-responders n = 5 patients, responders n = 8 patients). (B) Gene mapping analysis of significant pathways increased in responders and non-responders at baseline using significant genes identified in GO biological and KEGG pathway analysis. Red clusters are increased in responders at baseline and blue clusters are increased in non-responders at baseline (non-responders n = 5 patients, responders n = 8 patients). (C) Significant HALLMARK pathways that differed between responders and non-responders post-treatment (non-responders n = 5 patients, responders n = 8 patients). (D) MultiPLIER cell type analysis of the RNA sequencing data (non-responders n = 5 patients, responders n = 8 patients) (LV964 non-responders p = 0.63 and responders p = 0.0024; LV823 non-responders p = 0.07 and responders p = 0.0059; LV31 non-responders p = 0.88 and responders p = 0.007; LV765 non-responders p = 0.28 and responders p = 0.001; LV96 non-responders p = 0.59 and responders = 0.0025). (E) Representative graphs depicting TCR clone expansion of the top 10 clones for two patients, a responder (01-009) and a non-responder (01-007). (F) Scatterplot with annotations depicting clones with more than 8 transcripts before and after treatment for patients 01-002 and 01-007. Light grey dots were not included in the analysis because they had less than 8 sequences. Dark grey dots are clones that were not significantly different between pre- and post-samples. Red dots were clones significantly increased pre-treatment and blue dots were clones that were significantly increased post-treatment. Dots along the Y and X axis are clones not present in the pre-sample or post-sample, respectively. Analysis was conducted using the ImmunoSEQ analyzer. (G) Quantification of PD-L1 expressing cancer cells (CK+) within the TME pre- and post-treatment (non-responders n = 6 patients, responders n = 12 patients). Non-responders p = 0.36 and responders p = 0.26. (H) Quantification of CPS score (12 Gy n = 2 patients, 18 Gy n = 3 patients, and 24 Gy n = 3 patients) and PD-L1+ cells (12 Gy n = 2 patients, 18 Gy n = 5 patients, 24 Gy n = 6 patients) in the TME post-treatment by VECTRA categorized by dose of SBRT given to the patient. Significance was determined by a two-way paired student’s t-test, *p < 0.05, **p < 0.01, and ***p < 0.001. The error bars represent the standard error of the mean (± SEM). Source data

Extended Data Fig. 4. CD68+ Multinucleated giant cells surround keratin pearls post-treatment in responders’ TME.

(A) Representative VECTRA image of the TME of a responder showing keratin pearls surrounded by MNGCs and CD3+ cells (highlighted in red). To the right, a zoom in on the image highlights MNGCs with white arrows. (B) Representative VECTRA image of the TME of a non-responder post-treatment showing Tregs surrounding a keratin pearl. Several Tregs highlighted by pink arrows. (C) Quantification of Keratin pearl area, MNGC area, and the combined area of both keratin pearls and MNGC (non-responder n = 5 patients, responder n = 7 patients). Image J was used for quantification and a paired two-sided student’s t-test was used to determine significance. Source data

Extended Data Fig. 5. Antigen presentation is increased in responders while Tregs are decreased.

(A) RNA expression of genes associated with MHC II expression (CIITA) and MHC II genes (HLA-DRA and HLA-DMA) (non-responders n = 5 patients, responders n = 8 patients) (HLA-DRA non-responders p = 0.25 and responders p = 0.02; HLA-DMA non-responders p = 0.07 and responders p = 0.0014; CIITA non-responders p = 0.19 and responders p = 0.06). (B) RNA expression of MHC I genes (HLA-A and HLA-B) (non-responders n = 5 patients, responders n = 8 patients). (C) Venn diagrams depicting TCR sequences shared between the blood and the TME pre- and post-treatment in representative responders and non-responders (non-responders n = 3 patients, responders n = 5 patients). (D) Scatterplot with annotations depicting clones with more than 8 transcripts after treatment in the TME and blood for patients 01-016 and 01-007. Light grey dots were not included in the analysis because they had less than 8 sequences. Dark grey dots are clones that were not significantly different between post-treatment TME and blood samples. Red dots were clones significantly increased the TME and blue dots were clones that were significantly increased in the blood. Dots along the Y and X axis are clones not present in the TME or blood post-treatment, respectively. (E) Dot plot of template expansion of the top 5 TCR amino acid sequences shared between the most samples. Any patient sample was required to have at least two templates to be included. The top 5 amino acid sequences found are on the right. Analysis was conducted using the ImmunoSEQ analyzer. (F) Top 25 genes for LV57 identified by MultiPLIER (non-responders n = 5 patients, responders n = 8 patients). (G) Quantification of LV57 by dose of RT (12 Gy n = 3 patients, 18 Gy n = 5 patients, 24 Gy n = 3 patients) and quantification of the CD8:Treg Ratio in tumor samples post-treatment by VECTRA (p = 0.04) (18 Gy n = 5 patients, 24 Gy n = 4 patients). Statistical significance was determined using a two-tailed unpaired student’s t-test, *p < 0.05 and **p < 0.01. The error bars represent the standard error of the mean (± SEM). Source data

Extended Data Fig. 6. Clustering of cell populations in the blood.

(A) Defining CD4+ and CD8+ T cells withing the activated T cell cluster (non-responders n = 3 patients, responders n = 9 patients). (B) CITRUS populations identified using differentiating markers (CD3, CD4, CD8, CD14, CD56). (C) Gating strategy for CD45+ cells used for clustering analysis. (D) Gating strategy for memory T cells. (E) Quantification of Effector Memory CD4 T cells post treatment by dose of radiation, at time of surgery (18 Gy n = 6 patients, 24 Gy n = 8 patients). Statistical significance was determined using a two-tailed unpaired student’s t-test. The error bars represent the standard error of the mean (± SEM). Source data

Extended Data Fig. 7. An increase in free-fatty acid metabolism correlated with response while decreased acyl-carnitines correlated with poor treatment response.

(A) Quantification of circulating free-fatty acids from serum collected from patients pre- and post-treatment (mean+/− SEM) (pre non-responders n = 4 patients, post non-responders n = 5 patients, pre responders n = 13 patients, post responders n = 13 patients). (B) Quantification of circulating metabolites associated with the Krebs cycle pre- and post-treatment (mean +/− SEM) (pre non-responders n = 4 patients, post non-responders n = 5 patients, pre responders n = 13 patients, post responders n = 13 patients). Source data

Extended Data Fig. 8. Predictor importance plot for the random forest model averaged across the random initializations.

(A) Diagram depicting how the random forest model was trained. (B) Diagram depicting an example of how the model will determine if a patient responds to treatment. (C) Predictor importance is computed using the mean decrease in Gini index and plotted relative to the CD4+ effector T-cell importance, which had the maximum mean decrease in Gini index among the predictors. This figure was made with Python version 3.6.3 and Matplotlib version 3.2.2. Source data