Single-cell dynamics of breakthrough toxicities after anakinra prophylaxis for axicabtagene ciloleucel in lymphoma

Abstract

Chimeric antigen receptor (CAR) T-cell (CAR-T) therapy is limited by cytokine release syndrome (CRS) and neurotoxicity (NT). We sought to use once-daily prophylactic anakinra, an interleukin-1 (IL-1) receptor antagonist, to prevent CRS/NT that would require hospitalization (grade ≥2) in patients receiving axicabtagene ciloleucel for large-cell lymphoma, with the goal of facilitating outpatient therapy and management. Our study, in line with others, demonstrates that once-daily prophylactic anakinra is insufficient to prevent the development of toxicities that would require hospitalization in most patients. As part of the initial study design, we prospectively incorporated single-cell RNA sequencing to gain insight into the molecular immune signaling associated with breakthrough CRS and NT despite anakinra prophylaxis. In patients who developed breakthrough CRS or NT, we found that interferon gamma (IFN-γ) pathways and ligand-receptor activities were significantly enriched, as were cytokine levels of IFN-γ and CXCL10 in CD14+ monocytes. This correlated with increased IFN-γ and other cytokines in the peripheral blood. In infused CAR-T products, IL-4 and IL-10 anti-inflammatory pathways were negatively associated with grade ≥2 toxicities, regardless of anakinra treatment. These data identify IFN-γ as a potential key mechanism in CAR-T-associated toxicities, which is not inhibited by anakinra but may be otherwise targetable. This trial was registered at www.ClinicalTrials.gov as #NCT04150913.

Graphical abstract

Figure 1.

Study schema. After consent, screening, and enrollment, patients were apheresed for CAR-T manufacturing. After release of the CAR-T product (axi-cel), patients received standard cy/flu lymphodepletion per axi-cel label. Four hours before CAR-T infusion, patients received their first dose of 200 mg anakinra subcutaneously. Dosing continued through day 7 (first 6 patients) or 14 (subsequent protocol amendment). Patients were assessed for toxicity during the first 30 days and response starting 1 month after axi-cel infusion. Cy, cyclophosphamide; flu, fludarabine; SQ, subcutaneous.

Figure 2.

Swimmers plot of positron emission tomography/computed tomography responses in patients treated with anakinra prophylaxis. Response per patient as assessed per the revised International Working Group criteria for malignant lymphoma. Response rates are reported as the number of patients per total number of patients and percent. Survival is reported per survival category as a percent of patients available to assess at that time point with the 95% CI. CI, confidence interval; CR, complete response; ORR, overall response rate; OS, overall survival; PD, progressive disease; PFS, progression-free survival; PR, partial response.

Figure 3.

CAR-T expansion after infusion as measured by flow cytometry and PCR. CAR-T expansion was measured in patient peripheral blood by (A) flow cytometry or (B) PCR for percent CAR⁺ cells among PBMCs. The number of CAR-Ts per μL of blood was calculated using the patient’s absolute lymphocyte count. Data are presented as the time course from infusion (day 0), with pre–CAR-T (day −5) values also shown, and separated by patients with (red) or without (green) breakthrough grade ≥2 CRS. (C-D) Peak CAR-T expansion for each patient, separated by those who did and did not experience grade ≥2 CRS or NT by (C) flow cytometry or (D) PCR. P values indicate significance as determined by a 2-tailed Mann-Whitney U test.

Figure 4.

Breakthrough CRS and NT course of patients treated with anakinra prophylaxis. The grade and duration of CRS and NT within the first 30 days after axi-cel treatment reported per patient based on the day of onset. Grade coloring indicates the highest grade of CRS or NT experienced during the toxicity duration. CRS graded per Lee et al and NT graded per Common Terminology Criteria for Adverse Events version 4.03 in line with historical ZUMA-1 cohort. cyto, cyclophosphamide; dex, dexamethasone; siltux, siltuximab; toci, tocilizumab.

Figure 5.

Serum markers associated with breakthrough grade ≥2 CRS and/or NT in patients treated with anakinra prophylaxis. Serum proteins were analyzed via V-PLEX in serum collected from patient peripheral blood prior to CAR-T infusion; on the day of CAR-T infusion (day 0); and on days 1, 3, 5, 7, 14, and 28 after CAR-T infusion. Symbols represent the peak cytokine concentration detected for each patient during that timeframe. Lines indicate mean ± standard deviation. Black symbols in the IL-6 plot indicate patients treated with tocilizumab as part of their toxicity management. Data are separated by patients with (red) or without (green) grade ≥2 CRS (circles) or NT (triangles). P values indicate significance as determined by a 2-tailed Mann-Whitney U test. Cytokines shown are significantly increased in (A) patients with grade ≥2 CRS and patients with grade ≥2 NT, (B) only patients with grade ≥2 CRS, or (C) only patients with grade ≥2 NT. CRP, C-reactive protein; GM-CSF, granulocyte-monocyte colony-stimulating factor; ICAM, intercellular adhesion molecule; IP-10, INF-γ inducible protein (also known as CXCL10); MIP, macrophage inflammatory protein; PD-L, programmed death-ligand; TNF, tumor necrosis factor; VCAM, vascular cell adhesion molecule.

Figure 6.

Transcriptomic landscape of CAR-T under treatment with anakinra prophylaxis. (A) Schematic of sample collection from anakinra prophylaxis and nonanakinra prophylaxis cohorts along with data analysis design. Samples from patient IP or PBMCs collected on day 7 after CAR-T infusion underwent scRNA-sequencing and were analyzed for signatures of CRS development and differential gene expression and differential cell-cell interactions between patients receiving anakinra prophylaxis vs no prophylaxis. (B) Illustration of the different samples and comparisons being used in this study. Associations with breakthrough CRS/NT under anakinra prophylaxis in this cohort are identified and compared with CRS/NT associations seen in absence of anakinra prophylaxis using samples published in Haradhvala et al. An additional n = 4 samples from patients without anakinra prophylaxis were generated as part of this study to control for potential between-study differences in sample processing. (C) Harmony batch-corrected based uniform manifold approximation and projection (UMAP) representation for all cells analyzed in the study, colored by cell type. The cells from the IPs cluster together and that cluster is labeled “Infusion.” (D) UMAP of IP and day 7 CAR⁺ and CAR⁻ cells colored by assigned T-cell subtype. (E) Scatterplot of normalized enrichment score for gene sets associated with CRS (enriched in patients with grade ≥2 CRS) in CD4⁺ and CD8⁺ IPs in the nonanakinra prophylaxis cohort from Haradhvala et al. (F) Average expression of IL-4, IL-10, and IL-13 genes and their receptors across different cell subsets and time points. (G) Volcano plot of DE analysis in day 7 CAR⁺ CD4 CTL cells from patients in this study, without grade ≥2 CRS, with (n = 5) vs without (n = 4) anakinra prophylaxis. (H) Distribution of the log₂ pseudobulk IL-10 expression in day 7 CAR⁺ CD4 CTL cells from panel G. (I) Fraction of cells in the cell cycle across different T-cell subsets in patients with and without anakinra prophylaxis. False discover rate (FDR)–corrected Mann-Whitney U q values are shown comparing patients with grade ≥2 CRS or NT to those with grade ≤1 CRS or NT. CM, Central Memory; CTL, cytotoxic T lymphocyte; EM, Effector Memory; MK, megakaryocyte; NK, natural killer cell; pDC, plasmacytoid dendritic cell; T-reg, T-regulatory; TEMRA, Terminal effector memory cells that re-express the CD45RA molecule.

Figure 6.

Transcriptomic landscape of CAR-T under treatment with anakinra prophylaxis. (A) Schematic of sample collection from anakinra prophylaxis and nonanakinra prophylaxis cohorts along with data analysis design. Samples from patient IP or PBMCs collected on day 7 after CAR-T infusion underwent scRNA-sequencing and were analyzed for signatures of CRS development and differential gene expression and differential cell-cell interactions between patients receiving anakinra prophylaxis vs no prophylaxis. (B) Illustration of the different samples and comparisons being used in this study. Associations with breakthrough CRS/NT under anakinra prophylaxis in this cohort are identified and compared with CRS/NT associations seen in absence of anakinra prophylaxis using samples published in Haradhvala et al. An additional n = 4 samples from patients without anakinra prophylaxis were generated as part of this study to control for potential between-study differences in sample processing. (C) Harmony batch-corrected based uniform manifold approximation and projection (UMAP) representation for all cells analyzed in the study, colored by cell type. The cells from the IPs cluster together and that cluster is labeled “Infusion.” (D) UMAP of IP and day 7 CAR⁺ and CAR⁻ cells colored by assigned T-cell subtype. (E) Scatterplot of normalized enrichment score for gene sets associated with CRS (enriched in patients with grade ≥2 CRS) in CD4⁺ and CD8⁺ IPs in the nonanakinra prophylaxis cohort from Haradhvala et al. (F) Average expression of IL-4, IL-10, and IL-13 genes and their receptors across different cell subsets and time points. (G) Volcano plot of DE analysis in day 7 CAR⁺ CD4 CTL cells from patients in this study, without grade ≥2 CRS, with (n = 5) vs without (n = 4) anakinra prophylaxis. (H) Distribution of the log₂ pseudobulk IL-10 expression in day 7 CAR⁺ CD4 CTL cells from panel G. (I) Fraction of cells in the cell cycle across different T-cell subsets in patients with and without anakinra prophylaxis. False discover rate (FDR)–corrected Mann-Whitney U q values are shown comparing patients with grade ≥2 CRS or NT to those with grade ≤1 CRS or NT. CM, Central Memory; CTL, cytotoxic T lymphocyte; EM, Effector Memory; MK, megakaryocyte; NK, natural killer cell; pDC, plasmacytoid dendritic cell; T-reg, T-regulatory; TEMRA, Terminal effector memory cells that re-express the CD45RA molecule.

Figure 7.

INF-γ signaling pathway and RETN ligand-receptor interaction upregulated in CD14 monocytes in patients with breakthrough CRS to anakinra. (A) Comparisons between the log₂-fold change of differentially expressed genes associated with high- (grade ≥2) vs low-grade CRS (grade ≤1) in patients with anakinra prophylaxis (x-axis, n = 10 patients from this study: 6 grade ≤1 and 4 grade ≥2) and without anakinra prophylaxis (y-axis, n = 4 low-grade patients from our biobanked samples and n = 16 patients from Haradhvala et al: 6 grade ≤1 and 10 grade ≥2) among day 7 CD14⁺ monocytes. Colors indicate DE significance levels, setting Benjamini-Hochberg-corrected P value <.1 as significant. Triangles represent genes that had a significant contrast (Benjamini-Hochberg-corrected P value <.1) between the effects observed with or without anakinra prophylaxis. (B) IFN-γ and (C) CXCL10 cytokine levels between patients with grade ≥2 and grade ≤1 CRS and NT who received anakinra prophylaxis (n = 10 from this study) across 5 time points after CAR-T infusion as measured by Luminex. Colors indicate the grade of CRS, and shape indicates the grade of NT for each patient. (D) Dot plot of significantly interacting ligand-receptor (LR) pairs with corresponding cell type combinations. Each dot is colored by log₂ change in interaction score and sized by −log₁₀U test FDR-adjusted P value. (E) Distribution of interaction scores for each LR pair between samples from patients in this study who did or did not develop grade ≥2 CRS and/or NT in the presence and absence of anakinra. Colors indicate the grade of CRS, and shape indicates the grade of NT for each patient.