Autologous T cell therapy for MAGE-A4+ solid cancers in HLA-A*02+ patients: a phase 1 trial

Abstract

Affinity-optimized T cell receptors can enhance the potency of adoptive T cell therapy. Afamitresgene autoleucel (afami-cel) is a human leukocyte antigen-restricted autologous T cell therapy targeting melanoma-associated antigen A4 (MAGE-A4), a cancer/testis antigen expressed at varying levels in multiple solid tumors. We conducted a multicenter, dose-escalation, phase 1 trial in patients with relapsed/refractory metastatic solid tumors expressing MAGE-A4, including synovial sarcoma (SS), ovarian cancer and head and neck cancer ( NCT03132922 ). The primary endpoint was safety, and the secondary efficacy endpoints included overall response rate (ORR) and duration of response. All patients (N = 38, nine tumor types) experienced Grade ≥3 hematologic toxicities; 55% of patients (90% Grade ≤2) experienced cytokine release syndrome. ORR (all partial response) was 24% (9/38), 7/16 (44%) for SS and 2/22 (9%) for all other cancers. Median duration of response was 25.6 weeks (95% confidence interval (CI): 12.286, not reached) and 28.1 weeks (95% CI: 12.286, not reached) overall and for SS, respectively. Exploratory analyses showed that afami-cel infiltrates tumors, has an interferon-γ-driven mechanism of action and triggers adaptive immune responses. In addition, afami-cel has an acceptable benefit-risk profile, with early and durable responses, especially in patients with metastatic SS. Although the small trial size limits conclusions that can be drawn, the results warrant further testing in larger studies.

Fig. 1. Study design and patient disposition.

RECIST.

Fig. 2. Response and prognostic characteristics of the mITT population and patients with SS.

a, Swimmer’s plot of patient response over time in the mITT population (N = 38). The inset is response over time in patients with SS (n = 16). b, DoR profiles show the change from baseline in target lesions using RECIST version 1.1 in the overall treatment group after first infusion for responders and non-responders. The inset shows the change from baseline in target lesions in patients with SS. The probability of DoR was 100% (95% CI: 100, 100) at ≥12 weeks and 60% (95% CI: 20.4, 80.5) at ≥24 weeks. Duration of stable disease probability at ≥24 weeks for patients with SS was 50% (95% CI: 22.5%, 75.0%). c, Kaplan–Meier curves show PFS in the mITT population. The inset shows the PFS curve for the patients with SS. The median PFS for the mITT population was 12.3 weeks (95% CI: 10.9, 19.1) and 20.4 weeks (95% CI: 10.0, 52.1) for the patients with SS. PFS events in the mITT population included 25 events (65.8%) of PD and four events (10.5%) of death. PFS probability was 60% (95% CI: 44.8, 76.6) at 12 weeks and 30% (95% CI: 14.1, 45.2) at 24 weeks. PFS probability in the patients with SS was 70% (95% CI: 0.44, 0.89) at 12 weeks and 50% (95% CI: 0.19, 0.68) at 24 weeks. d, Kaplan–Meier curves show OS in the mITT population. The inset shows the OS for patients with SS. The median OS for the mITT population was 42.9 weeks (95% CI: 20.7, not reached) and 58.1 weeks (95% CI: 36.3, not reached) for patients with SS. OS probability was 90% (95% CI: 69.8, 94.0) at 12 weeks and 70% (95% CI: 49.6, 82.1) at 24 weeks. OS probability in patients with SS was 90% (95% CI: 0.63, 0.99) at 12 weeks and remained 90% (95% CI: 0.59, 0.97) at 24 weeks.

Fig. 3. Serum IFNγ patient profiles and associations between peak and AUC concentrations of IFNγ levels and anti-tumor response.

a, Comparison of serum levels in patients with SS (no CRS n = 2, Grade 1 n = 5, Grade 2 n = 7, Grade 3 n = 1, Grade 4 n = 1) and those with other indications (no CRS n = 15, Grade 1 n = 5, Grade 2 n = 2) across CRS groups. Peak IFNγ levels (pg ml⁻¹) were significantly greater in patients with CRS (all grades, n = 21) compared with non-CRS (n = 17) (median, 270.8 pg ml⁻¹ and 47.7 pg ml⁻¹, respectively; P = 0.00012). Calculated IFNγ AUC from days 0 to 21 was significantly greater in patients with CRS compared with non-CRS (median, 1,455.0 pg ml⁻¹ and 467.0 pg ml⁻¹, respectively; P = 0.00061) (Wilcoxon rank-sum test, two-sided). b, Serum IFNγ concentration (pg ml⁻¹) measured across sample sets collected on days 0–21 after infusion. Magnitude of maximum percent change in SLD was positively correlated with reduction in target tumor lesion (Spearmanʼs r = −0.64; P = 0.000057). c, Peak serum IFNγ levels were significantly greater in patients with best overall responses of PR (n = 9) compared with PD (n = 7; P = 0.0052) and SD (n = 19; P = 0.025) (Kruskal–Wallis test, two-sided). d, Patients ranked by best percent change in SLD. Non-SS subset included three NE patients. The asterisk in superscript (^*) indicates CRS; number indicates dose-escalation cohort. AUC, area under the curve; CRTAM, class I-restricted T cell-associated molecule; CXCL, chemokine (C-X-C motif) ligand; FASLG, FAS ligand; IQR, interquartile range; KLRD1, killer cell lectin like receptor D1; NCR1, natural cytotoxicity triggering receptor 1; PTN, pleiotrophin. Box plots depict median as horizontal lines within boxes, with box bounds as the first and third quartiles. Dots represent individual data points. Lower whiskers are minimum values within 1.5 times the IQR below the 25th percentile. Upper whiskers are maximum values within 1.5 times the IQR above the 75th percentile.

Fig. 4. Detection of afami-cel and T cell phenotypes in patient tumor biopsies.

a, Digital image quantification of T cell markers (total CD3⁺), CD4 (CD3⁺CD4⁺), cytotoxic (CD3⁺CD8⁺), regulatory (CD3⁺CD4⁺FoxP3⁺) and co-staining for phenotypes (proliferating, Ki67⁺; activated, GrazB⁺; PD-L1⁺) and combination of these phenotypes for the four patients referenced in Extended Data Fig. 10a. Baseline biopsies were taken 2–5 weeks before afami-cel infusion. b, A multiparametric analysis of T cell infiltration in tumor biopsies using IHC, MAGE-A4 SPEAR RNAscope (Advanced Cell Diagnostics) and multiplex immunofluorescence. Images of post-infusion biopsy (liver) from Patient 4: A, spatial plot generated using spatial analysis module in HALO (Indica Labs) showing malignant cells (PanCK⁺ (red), regulatory (CD3⁺CD4⁺FoxP3⁺) T cells (yellow) and cytotoxic (CD3⁺CD8⁺) T cells (blue/cyan)). Scale bar is not applicable as this is not a raw image; B, CD3 IHC/SPEAR⁺ T cell RNAscope duplex (CD3 IHC staining (blue); MAGE-A4 SPEAR T cell staining (purple)); C, Ultivue 8-plex multiplex dataset showing CD4 (orange), PD-L1 (red), CD8 (green), Ki67 (purple), FoxP3 (blue), GrazB (white), CD3 (yellow) and PanCK (teal); D, PanCK immunofluorescence staining from Ultivue 8-plex panel (displayed in absorption mode in HALO for clarity); E, MAGE-A4 IHC stain (DAB (brown)); and F, hematoxylin (purple) and eosin (pink) stain. c, Heat map of log₂-transformed normalized counts of genes associated with ‘T cell exhaustion’ and ‘Negative regulation of T cell-mediated immunity’ in baseline and post-infusion biopsies from Patients 1–4. Patients 1–3 were patients with SS; Patient 4 was a patient with ovarian cancer. DAB, 3,3′-diaminobenzidine; FoxP3, forkhead box protein 3; GrazB, granzyme B; Ov, ovarian; PanCK, pancytokeratin; PD-L1, programmed death ligand 1.

Extended Data Fig. 1. Study design.

HLA, human leukocyte antigen; IHC, immunohistochemistry; MAGE-A4, melanoma-associated antigen A4.

Extended Data Fig. 2. MAGE-A4 antigen H-score in the modified intent-to-treat population by tumor type.

Individual MAGE-A4 antigen H-scores per tumor type. The overall median MAGE-A4 H-score across all nine tumor types was 189 (range: 15–300). Median MAGE-A4 H-score was highest at 249 in synovial sarcoma (range: 60–300); melanoma sub-group excluded due to a single evaluable patient. Box plots depict median as horizontal lines within boxes, with box bounds as the first and third quartiles. Dots represent individual data points. Lower whisker is the minimum value of the data within 1.5 times the interquartile range below the 25^th percentile. Upper whisker is the maximum value of the data within 1.5 times the interquartile range above the 75^th percentile. MAGE-A4, melanoma-associated antigen A4.

Extended Data Fig. 3. Maximal change in target lesion from baseline using RECIST v1.1 (modified intent-to-treat population; N = 38).

(a) Maximal change in target lesion from baseline in the modified intent-to-treat population. Data for three patients were not available as of data cut-off on 1 September 2020. Each bar represents an individual patient. Nine patients responded to first infusion as determined by RECIST v1.1 criteria. All nine responders (23.7%) had a partial response with an overall response rate of 23.7% (95% CI: 11.4, 40.2). (b) Synovial sarcoma subgroup: Maximal change in target lesion from baseline in the synovial sarcoma subgroup. Each bar represents an individual patient. Seven of the nine responders were in the synovial sarcoma subgroup (n = 16) with a best overall response of partial response and an overall response rate of 43.8% (95% CI: 19.8, 70.1). Horizontal reference lines at -30% and 20%. CI, confidence interval; MRCLS, myxoid/round cell liposarcoma; NSCLS, non-small cell lung cancer; RECIST, Response Evaluation Criteria in Solid Tumors; SS, synovial sarcoma.

Extended Data Fig. 4. Afami-cel therapy leads to regression of large pleural-based thoracic metastasis in synovial sarcoma.

In Patient A, afami-cel was associated with reductions in metastatic disease in hemithorax, including regression of a large left lung pleural metastasis that crossed the midline at baseline (Patient A, upper right and upper left panels, respectively) and re-expansion of the right lung, as shown on a computed tomography scan at Week 12 post−afami-cel (Patient A, lower right panel, arrow), associated with patient-reported improvement in exertional dyspnea. In Patient B, afami-cel was associated with overall reduction in left lung pleural metastases including complete resolution of one pleural metastasis (Patient B, lower right panel, arrow). BOR, best overall response; MAGE-A4, melanoma-associated antigen A4; PR, partial response; SLD, sum of longest diameter; TTR, time to response.

Extended Data Fig. 5. Ratio of transduced CD4⁺ to CD8⁺ cells and memory sub-populations in afami-cel manufactured product.

(a) Ratio of CD4⁺ to CD8⁺ cell subsets within the transduced cells infused, derived from immunophenotyping: Ratios greater than 1 denote a CD4 bias in the transduced cells. Horizontal lines denote median values. Samples are grouped and color coded by BOR resulting from the infusion (blue PR, gray SD, red PD, black NE). Ratio of PR range 0.19–16.16 (median 1.30, n = 9) vs. SD range 0.03–11.58 (median 1.26, n = 16) vs. PD range 0.76–12.52 (median 3.94, n = 6). (b) Memory subset distribution by immunophenotyping within the transduced cells infused: The proportion of live single transduced afami-cel within the respective memory subsets is indicated by box plots for CD8⁺ afami-cel on the left and for CD4⁺ on the right. Samples are grouped and color coded by BOR resulting from the infusion (blue PR, gray SD, red PD, black NE). CM = central memory, CD45RA^-CCR7⁺; EM = effector memory, CD45RA^-CCR7^-; EMRA = effector memory RA⁺, CD45RA⁺ CCR7^-; SCM = stem cell memory, CD45RA⁺ CCR7⁺. Two-sided paired Wilcoxon test P values are shown linking compared cell types. Box plots depict median as horizontal lines within boxes, with box bounds as the first and third quartiles. Dots represent individual data points. Lower whisker is the minimum value of the data within 1.5 times the interquartile range below the 25^th percentile. Upper whisker is the maximum value of the data within 1.5 times the interquartile range above the 75^th percentile. BOR, best overall response; DEX, dextramer; NE, not evaluable; PD, progressive disease; PR, partial response; SD, stable disease.

Extended Data Fig. 6. Longitudinal immunophenotyping of memory subtypes within circulating afami-cel.

Percentage of parent population (a) CD4⁺ Dextramer⁺ and (b) CD8⁺ Dextramer⁺ within the respective sub-population flow gate in MP (manufactured product) and peripheral blood mononuclear cell samples at post-infusion timepoints (D, day; W, week; M, month, EOT, end of treatment). CM (central memory CD45RA^-CCR7⁺); EM (effector memory CD45RA^-CCR7^-); EMRA (effector memory RA⁺, CD45RA⁺ CCR7^-); SCM (stem cell memory, CD45RA⁺ CCR7⁺). Data are grouped by best overall response resulting from the infusion. PR (partial response), n = 9 patients; SD (stable disease), n = 17 patients; PD (progressive disease), n = 6 patients.

Extended Data Fig. 7. Serum cytokine profiles associated with cytokine release syndrome.

Peak IL-6 levels pg ml^-1 (a, left panel), calculated IL-6 AUC Day 0–21 (a, right panel), peak GMCSF levels pg ml^-1 (b, left panel) and calculated GMCSF AUC Day 0–21 (b, right panel) were significantly greater in patients with CRS (all grades, n = 21) compared with non-CRS (n = 17) in the mITT population (Supplementary Table 8). Box plots show comparison of serum levels in SS patients (no CRS n = 2, Grade 1 n = 5, Grade 2 n = 7, Grade 3 n = 1, Grade 4 n = 1) and those with other indications (no CRS n = 15, Grade 1 n = 5, Grade 2 n = 2) across CRS groups. Box plots depict median as horizontal lines within boxes, with box bounds as the first and third quartiles. Dots represent individual data points. Lower whisker is the minimum value of the data within 1.5 times the interquartile range below the 25^th percentile. Upper whisker is the maximum value of the data within 1.5 times the interquartile range above the 75^th percentile. AUC, area under the curve; CRS, cytokine release syndrome; GMCSF, granulocyte-macrophage colony-stimulating factor; IL, interleukin; SS, synovial sarcoma.

Extended Data Fig. 8. IL-15 serum concentrations in patients treated with high dose (3600 mg/m²) or low dose (1800 mg/m²) lymphodepleting chemotherapy pre- and post-lymphodepletion.

Serum IL-15 levels pg ml^-1 were measured pre- and post-LD in Group 3/Expansion patients treated with ‘high’ LD (n = 7) or ‘low’ LD (n = 24*) regimen (*excludes one patient with an unevaluable baseline sample). Box plots depict median as horizontal lines within boxes, with box bounds as the first and third quartiles. Dots represent individual data points. Lower whisker is the minimum value of the data within 1.5 times the interquartile range below the 25^th percentile. Upper whisker is the maximum value of the data within 1.5 times the interquartile range above the 75^th percentile. IL, interleukin; LD, lymphodepletion.

Extended Data Fig. 9. Post-/pre-lymphodepleting chemotherapy serum IL-15 ratio in patients treated with high dose (3600 mg/m²) or low dose (1800 mg/m²) lymphodepleting chemotherapy.

Serum IL-15 concentration increased in post-LD versus pre-LD samples, and there was no difference in IL-15 concentrations between the two LD regimens in patients treated in Group 3/Expansion (median of ‘high’ LD 18.25 (n = 7), median of ‘low’ LD 12.87 (n = 24); P = 0.13, Two-sided Wilcoxon test). Box plots depict median as horizontal lines within boxes, with box bounds as the first and third quartiles. Dots represent individual data points. Lower whisker is the minimum value of the data within 1.5 times the interquartile range below the 25^th percentile. Upper whisker is the maximum value of the data within 1.5 times the interquartile range above the 75^th percentile. IL, interleukin; LD, lymphodepletion.

Extended Data Fig. 10. Detection of afami-cel and characterization of phenotypes in patient tumor biopsies.

(a) Quantification of CD3⁺ T cells and afami-cel in post-infusion tumor biopsies from dose expansion: Four patients with contrasting profiles were prioritized for multiplex immunofluorescence analyses (summarized in Fig. 4a). Patient 1, SS, on-treatment biopsy taken 8 weeks post-infusion; Patient 2, SS, at completion biopsy taken 17 weeks post-infusion; Patient 3, SS, on-treatment and at-completion biopsies taken 7 and 25 weeks post-infusion, respectively; Patient 4, Ov tumor, on-treatment biopsy taken 6 weeks post-infusion. Table summarizes key baseline tumor characteristics, afami-cel dose, post-infusion pharmacokinetic-pharmacodynamic findings, and BOR status for these four patients. Patients 1-3 were patients with SS; Patient 4 was a patient with Ov cancer. Box plots depict median as horizontal lines within boxes, with box bounds as the first and third quartiles. Dots represent individual data points. Lower whisker is the minimum value of the data within 1.5 times the interquartile range below the 25^th percentile. Upper whisker is the maximum value of the data within 1.5 times the interquartile range above the 75^th percentile. (b) Heatmap of GSVA scores of cell type specific and immune response categories gene lists in baseline and post-infusion biopsies from Patients 1‒4. AUC, area under the curve; BOR, best overall response; C_max, maximum serum concentration; CT, computed tomography; GSVA, gene set variation analysis; IFN, interferon; MAGE-A4, melanoma-associated antigen A4; NK, natural killer; Ov, ovarian; PR, partial response; Pt, patient; SD, stable disease; SLD, sum of longest diameter; SS, synovial sarcoma; TCR, T-cell receptor; TLR, toll-like receptor; TNF, tumor necrosis factor.