IL-6 receptor inhibition promotes expansion of cytolytic CD4⁺ T cells after cellular immunotherapy

Key Points

• •IL-6R inhibition during HSCT promotes type I interferon signaling in blood monocytes and CD4⁺ T cells.
• •IL-6R inhibition during cellular immunotherapy promotes the expansion of cytolytic CD4⁺ T cells, including the Eomes⁺ fraction.

Visual Abstract

Abstract

Graft-versus-host disease (GVHD) remains a major barrier to the success of allogeneic hematopoietic stem cell transplantation (HSCT). In preclinical models, the dysregulation of interleukin-6 (IL-6) in the peritransplant period promotes GVHD via STAT3-dependent T-cell differentiation and monocyte activation, but the nature of immunological effects invoked in patients remains unclear. To advance our understanding of IL-6 signaling in humans during HSCT, we performed single-cell RNA sequencing on circulating CD14⁺ monocytes and CD4⁺ T cells from patient samples within a clinical trial of IL-6 receptor (IL-6R) inhibition (tocilizumab [TCZ] or placebo) on a backbone of calcineurin inhibition and short-course methotrexate. We studied patients who did not develop acute GVHD (aGVHD) and included a placebo-treated group analyzed before the development of aGVHD. IL-6R inhibition promoted type I interferon–associated transcriptional programs in both CD4⁺ T cells and CD14⁺ monocytes. Surprisingly, IL-6R inhibition with TCZ profoundly enhanced cytolytic CD4⁺ T-cell differentiation. In experimental HSCT and chimeric antigen receptor T-cell systems, the genetic deletion of IL-6 signaling in donor T cells enhanced the expansion of cytolytic eomesodermin-positive (Eomes⁺) CD4⁺ regulatory T cell subsets and attenuated T helper 1 (Th1) and Th17 differentiation. Consistent with the promotion of this cytolytic CD4⁺ phenotype, antitumor effects were improved in the absence of IL-6 signaling. In summary, these data demonstrate that IL-6R inhibition during cell therapy imprints type I interferon programs and cytolytic CD4⁺ T-cell differentiation, including the Eomes⁺ fraction associated with favorable immunotherapy outcomes.

We still have much to learn about the in vivo function of T-cell subsets and how it is modulated, especially in the context of T-cell immunotherapy. Williams and colleagues studied patient samples and murine genetic models to describe a previously unrecognized role for interleukin-6 (IL-6) in suppressing the differentiation of eomesodermin-positive cytolytic and type 1 regulatory CD4⁺ T cells. Inhibition of IL-6 receptor signaling induces these cells and increases antitumor activity in model systems. These findings suggest strategies to optimize use of drugs such as tocilizumab in the period immediately before T-cell infusion.

Introduction

Immune toxicity, be it graft-versus-host disease (GVHD) after allogeneic hematopoietic stem cell transplantation (HSCT) or cytokine release syndrome (CRS) after chimeric antigen receptor (CAR) T cells, remains a major barrier to successful outcomes.¹ Both acute GVHD (aGVHD) and CRS are dependent on donor T-cell function and cytokine dysregulation (reviewed by Hill et al²), with interleukin-6 (IL-6) identified as a key cytokine in both humans^3,4 and murine models.5, 6, 7, 8 The complete inhibition of IL-6 receptor (IL-6R) signaling in patients receiving tocilizumab (TCZ) after HSCT allowed a unique opportunity to study the role of IL-6 in primary immune responses. Consequently, we used single-cell RNA sequencing (scRNA-seq) to define cellular transcriptomes in patient blood in the presence or absence of IL-6 and compared this to preclinical models, in which IL-6R was genetically deleted. Given that IL-6 dysregulation is primarily a feature of CRS after CAR T-cell therapy, we extended our study of IL-6 biology to this setting and demonstrate highly analogous findings with regard to T-cell differentiation. We demonstrate the unexpected effects of IL-6 inhibition in controlling eomesodermin (Eomes)-dependent cytolytic CD4⁺ T-cell expansion, together with the induction of type I interferon signatures.

Methods

Patient samples

Patients enrolled in a phase 3 clinical trial of IL-6R inhibition in HSCT⁴ were selected for scRNA-seq analysis, including patients without any development of aGVHD after bone marrow (BM) transplantation (BMT) who had received either placebo (n = 5) or TCZ (n = 5) and patients who subsequently developed grade II to IV aGVHD from the placebo arm of the study (n = 6). Patient characteristics are shown in supplemental Table 1, available on the Blood website. Peripheral blood mononuclear cells (PBMC) on days +21 or +30 after HSCT were analyzed. This study was approved by the human research ethics committees of the Royal Brisbane and Women’s Hospital and QIMR Berghofer (ethics approval number HREC/14/QRBW/35). This study was conducted in accordance with the Declaration of Helsinki. Patients were chosen based on (1) samples available before day 30 (because 1 dose of TCZ was given at day −1 and the drug is cleared beyond day 30),^3,4 (2) the absence of GVHD at any time (for TCZ vs placebo no GVHD groups) and aGVHD developing within 3 weeks of the blood sample (for the placebo GVHD group), (3) the absence of cytomegalovirus reactivation, and (4) sample availability with >20 million cells.

Cellular indexing of transcriptomes and epitopes by sequencing and scRNA-seq

Antibody-stained human PBMC (supplemental Table 2) were sorted by fluorescence-activated cell sorting to isolate viable CD4⁺ T cells and CD14⁺ monocytes (supplemental Figure 1A). These were pooled at a 2:1 ratio, respectively, and stained with TotalSeq A anti-human antibodies (supplemental Table 2) to generate gel beads in emulsions. Sequencing libraries for gene expression and antibody-derived tags were prepared using the 10x Genomics Single-Cell 3′ Reagent Kit version 3 and the MGIEasy Universal Library Conversion Kit and sequenced on an MGISEQ-2000. Raw FASTQ files were analyzed using Cell Ranger version 3.1.0 and the 10x Genomics human reference genome version 3.0. Data are available in the Gene Expression Omnibus (accession number GSE304894). Computational analysis was conducted using Seurat⁹ version 3.1.4 in R version 3.6.2, as described elsewhere¹⁰ and outlined subsequently. Processed scRNA-seq data from Li et al¹¹ were accessed from the Gene Expression Omnibus (accession number GSE241783). Data were analyzed as uploaded except for TYROBP⁺ cytotoxic CD4⁺ T cells, which were grouped with all cytotoxic CD4⁺ T cells. Differential composition testing between patients with CRS grade <2 vs CRS grade ≥2 was conducted with scComp version 2.2.0.¹²

Analysis of monocyte and T-cell scRNA-seq data sets

Monocytes were selected on CD14 expression. T-cell clusters were identified based on CD4 and CD3E expression and the lack of CD14 expression. Doublets identified by DoubletDecon¹³ version 1.1.4 were removed. scVI¹⁴ version 0.60 in Python version 3.7.3 was used for dimensionality reduction (n_latent = 10; n_epochs = 100; patient as batch), followed by uniform manifold approximation and projection (UMAP) (RunUMAP function; method = “umap-learn”, metric = “correlation”) and unbiased clustering (FindClusters; resolution = 0.5; supplemental Figure 1B-F). A monocyte cluster with low total unique molecular identifier counts (supplemental Figure 1D) and a T-cell cluster with high mitochondrial gene counts (supplemental Figure 1F) were removed. Differential gene expression analysis (DGEA) was performed with Seurat’s FindMarkers function using the Wilcoxon rank-sum test with no batch correction. The average log-fold change cutoff before testing was log-fold change > 0.1 or log-fold change < -0.1, and results were filtered after testing for Bonferroni-adjusted P ≤ .01. Gene Ontology (GO) enrichment analysis was performed with Gorilla (default settings), with differentially expressed genes as input and all expressed genes as background.

Murine BMT

Female B6D2F1 (H-2b/d) and B6.Cd4-Cre × Il6ra^f/f mice were purchased from The Jackson Laboratory or bred in the Fred Hutchinson Cancer Center Animal Facility. Note that the Cd4cre allele deletes the Il6ra gene from both CD4⁺ and CD8⁺ T-cell lineages during thymic development.¹⁵ BMT was performed as described previously. Briefly, B6D2F1 mice received 1100 cGy total-body irradiation on day −1 and received allografts comprised of BM and purified splenic CD3 T cells on day 0. Posttransplant cyclophosphamide (PTCy; 50 mg/kg per day) was administered intraperitoneally on days +3 and +4. Tacrolimus (FK506, 15 mg/kg per day) was administered intraperitoneally daily from day 0 up until day +13 after BMT (14 days maximum).¹⁶ GVHD was monitored using an established clinical scoring system.¹⁷ This study was approved by the Fred Hutchinson Cancer Center Animal Ethics Committee.

Murine CAR T-cell generation and adoptive transfer

The production of CD19 CAR retrovirus and the transduction of T cells have been previously described.^18,19 For retrovirus production, Platinum-E packaging cells were transfected with the CD19 CAR plasmid using the standard calcium phosphate method. For retroviral transduction, mouse splenic CD4⁺ and CD8⁺ T cells enriched by negative selection (EasySep Mouse, Stemcell Technologies) were stimulated with mouse Dynabeads T-Activator CD3/CD28 (Thermo Fisher Scientific) and transduced with retroviral supernatant, as previously described.^8,20 T cells were cultured in Iscove modified Dulbecco medium complete medium²¹ in the presence of rhIL-2 (PeproTech, 50 IU/mL) for the first 3 days and in the presence of recombinant human (rh)IL-15 (PeproTech, 5 ng/mL) for further amplification. CAR⁺ cells were selected with biotinylated anti-human epidermal growth factor truncated (EGFRt) (clone AY13, BioLegend) and MojoSort Streptavidin Nanobeads (BioLegend). B6 (CD45.1⁺) T-cell depleted BM and acute lymphocytic leukemia cells (B6 background, H-2b, CD45.2⁺, B220⁺, CD19⁺, and green flourescent protein (GFP⁺) harboring the transgenes Arf^−/− MSCV-BCR-ABL1-ires-luciferase and MSCV-IK6-ires-GFP²² were injected IV into lethally irradiated (1000 cGy, split into 2 doses) B6 CD45.2⁺ recipients on day 0, followed by the adoptive transfer of IL-6R^+/+ or IL-6R^−/− CAR T cells on day 3. Mice were monitored according to GVHD monitoring protocols and culled when clinical scores were ≥6 or weight loss was ≥30%.²¹ BM and blood were analyzed using flow cytometry, and CAR T cells were identified by CD3 expression in combination with EGFRt staining.

High-parameter flow cytometry

Patient PBMC or mouse samples were incubated with Live/Dead Fixable Aqua Dead Cell Stain, followed by antibodies (supplemental Tables 2 and 3) and BD fixation/permeabilization or eBioscience staining kits as required. Samples were acquired on a BD LSRFortessa using FACSDiva and analyzed using FlowJo version 9.9.6. OMIQ was used to perform dimensionality reduction, followed by FlowSOM clustering.

Statistical analysis

Differences in monocyte and T-cell scRNA-seq clusters were analyzed using the Wilcoxon rank-sum test using R version 3.6.2. Flow cytometry data sets were analyzed using GraphPad Prism. Statistical significance was determined using appropriate tests, as detailed in the figure legends.

Further details and additional data are provided in supplemental Tables 1-3 and supplemental Figures 1-4.

Results

We studied patient blood samples from a phase 3, double-blind, placebo-controlled, randomized study of IL-6R inhibition with TCZ administered at day −1, which blocked IL-6 signaling for the first 4 weeks after BMT.^3,4 The study failed to meet its primary end point; grade II to IV aGVHD occurred in 27% of TCZ-treated patients vs 36% in the placebo group at day 100.⁴ IL-6 is principally derived from CD11c⁺ recipient cells (ie, antigen-presenting cells) early after BMT,⁶ whereas the principal responders to IL-6 in blood are monocytes and CD4⁺ T cells.^3,4,23 We consequently focused our study on these 2 populations after BMT in the presence or absence of IL-6R inhibition with TCZ. Given that IL-6 blockade did not continue beyond 1 month in this study, we did not look at later time points.

Cell-surface protein and gene expression analysis of CD14⁺ monocytes sorted by fluorescence-activated cell sorting (Figure 1A; supplemental Figure 1A-D) revealed that most expressed surface CD14, and there was little/no CD16 expression, consistent with classical monocytes. A small number of cells showed increased CD16 expression, consistent with intermediate nonclassical monocytes (Figure 1B-C). Unsupervised clustering of single-cell transcriptomes revealed 4 transcriptomic clusters of CD14-expressing monocytes and a fifth cluster expressing FCGR3A (encoding CD16), with greater expression of complement genes C1QA, C1QB, and C1QC and CSF1R and CX3CR1, consistent with intermediate monocytes (Figure 1D). No significant differences in monocyte proportions were observed across groups (Figure 1E). Within each cluster, DGEA comparing placebo vs TCZ (GVHD free) groups revealed ∼200 to 300 upregulated genes and ∼100 to 200 downregulated genes (Figure 1F). The aGVHD group also exhibited differential gene expression in monocytes. Of the 526 upregulated genes, 270 (51%) were cluster specific (Figure 1G), suggesting heterogeneity in transcriptional responses of monocytes to TCZ. In patients who developed aGVHD, of the 535 upregulated genes, 67% were cluster specific. An investigation of common biological processes triggered by TCZ revealed GO enrichment for type I interferon–associated genes (IFI27, IFI6, ISG15, GBP1, GBP2, GBP5, IFITM3, IFIT3, and STAT1), which was not evident in patients who developed aGVHD (Figure 1G-H). Effects in patients who went on to develop aGVHD were mainly seen in genes involved in protein initiation, demethylation, translation, and degradation (Figure 1G-H). Consequently, IL-6R blockade upregulated type I interferon–associated genes in circulating CD14⁺ classical and intermediate monocytes.

Figure 1.

Posttransplant peripheral blood monocytes upregulate type I interferon–associated genes in response to IL-6R inhibition. (A) Experimental design. (B) Uniform manifold approximation and projection (UMAP) depicting the gene expression of CD14 and FCGR3A (encoding CD16) in monocytes. (C) Monocyte clusters in UMAP dimensionality reduction. (D) Curated dot plot of genes and their average expression level in monocytic clusters. The circle size indicates the percentage of cells in each cluster in which a gene was detected. (E) Proportion of cells assigned to each cluster for each patient across the groups. (F) Number of differentially expressed genes upregulated or downregulated in the same cluster in patients. (G) Venn diagrams showing the number of distinct and overlapping genes upregulated in each individual classical monocyte cluster in TCZ-treated patients compared with placebo treated patients without GVHD. (H) Top 5 enriched GO terms identified by GO term enrichment analysis of genes upregulated or downregulated in classical monocyte clusters in TCZ-treated patients compared to placebo treated patients without GVHD. CITE-seq, cellular indexing of transcriptomes and epitopes by sequencing; ER, endoplasmic reticulum; FACS, fluorescence-activated cell sorting; FDR, false discovery rate; SRP, signal recognition particle.

To determine the TCZ-mediated effects on CD4⁺ T cells, we identified naive and antigen-experienced populations using cell-surface CD45RO, CD45RA, CD62L, and CCR7 expression (Figure 2A; supplemental Figure 1E-F). Unsupervised clustering revealed 8 transcriptomic clusters, which were annotated as naive, early activated, effector, cycling effector, effector memory like, or regulatory T-cell (Treg) populations (Figure 2B). We noted an effector memory CD4⁺ T-cell population that expressed high levels of cytolytic genes (PRF1, NKG7, GNLY, and GZMs) together with T helper 1 (Th1; IFNG, TBX21, and EOMES) but not Th17 genes (IL17A and RORC; Figure 2C), which also appeared numerically dominant in TCZ-treated patients (Figure 2D). DGEA revealed upregulation of ∼300 to 400 genes in the TCZ-treated vs placebo (GVHD free) groups, specifically in cycling effector and effector-like cells. In contrast, substantial transcriptional changes were not observed in other subsets, including naive and Tregs, or in the aGVHD group (Figure 2E). The effects in patients who subsequently developed aGVHD were limited. Of the 539 upregulated genes in cycling effector and effector-like cells in TCZ-treated patients, 191 (35%) were upregulated, and 36 were downregulated in TCZ vs placebo (GVHD-free groups), suggesting common stimulatory effects as a result of IL-6R inhibition on CD4⁺ T cells. These data suggest that TCZ upregulated genes primarily in antigen-experienced, nonregulatory CD4⁺ T cells (Figure 2F). GO term enrichment analysis of the 191 genes upregulated by TCZ in both cycling effectors and effectors (Figure 2G; supplemental Figure 1G) revealed the enrichment of genes associated with type I interferon, cytokine signaling, and viral infection, whereas downregulated genes were associated with protein translation and trafficking. In the TCZ-treated group, the upregulation of type I interferon–associated genes, including IRF7, MX1, OAS1/2, IF16, OASL, IFI27, IFIT3, ISG15, ISG20, IFI44, IFI35, and IFITM2 (Figure 2H), and cytolytic genes, including granulysin (GNLY), granzymes (GZM-A, -B, -H, -K, and -M), NKG7, and perforin (PRF1; Figure 2I; supplemental Figure 1H), was identified. Using these genes to form a cytolytic gene signature, we observed the upregulation of cycling effectors and effector CD4⁺ T cells in TCZ-treated patients (Figure 2J). To validate the gene expression signatures, we examined protein expression in PBMC from placebo vs TCZ-treated patients using flow cytometry. Unsupervised clustering analysis using FlowSOM identified several cytotoxic molecule–expressing T-cell effector memory (TEM) subsets (supplemental Figure 2A-B). TEM lacking cytotoxic mediators (TEM cytotoxic T lymphocyte (CTL) low) were reduced in TCZ-treated patients compared to placebo (∗P < .05; supplemental Figure 2C), whereas, conversely, granulysin⁺ TEM, which also expressed granzyme A, constituted a significantly greater proportion of CD4⁺ T cells in TCZ-treated patients than the placebo (∗P < .05; supplemental Figure 2C). Consequently, IL-6R inhibition after HSCT upregulated cytolytic pathways in CD4⁺ T-cell effectors in patient blood.

Figure 2.

Posttransplant peripheral blood CD4⁺ T cells upregulate cytotoxicity and type I interferon–associated genes in response to IL-6R inhibition. (A) UMAP depicting antibody-derived transcript expression of CD45RA, CD45RO, CD62L, and CD197 in CD4⁺ T cells. (B) T-cell clusters in UMAP dimensionality reduction. (C) Curated dot plot of genes and their average expression levels in T-cell clusters. Circle size indicates the percentage of cells in each cluster in which a gene was detected. (D) Proportion of cells assigned to each cluster for each patient across the groups. (E) Number of differentially expressed genes upregulated or downregulated in the same cluster in patients. (F) Venn diagram showing the number of distinct and overlapping genes upregulated or downregulated in effector and cycling effector clusters in TCZ-treated patients compared to placebo treated patients without GVHD. (G) Top 5 enriched GO terms identified by GO term enrichment analysis of genes upregulated or downregulated in TCZ-treated patients from both the effector and cycling effector clusters. (H) Dot plot of type I interferon gene expression among cells from the effector and cycling effector clusters across patient groups. (I) Dot plot of cytotoxic T-lymphocyte gene expression among cells from the effector and cycling effector clusters across patient groups. (J) Violin plots of cytotoxic T-lymphocyte gene set score among effector and cycling effector clusters across patient groups. Donor CMV serostatus is indicated for each matched transplant recipient. ADT, antibody-derived tag; CMV, cytomegalovirus; ER, endoplasmic reticulum; FDR, false discovery rate; SRP, signal recognition particle.

To determine whether (1) IL-6 signaled directly to CD4⁺ T cells to limit type I interferon responses and cytotoxicity and (2) transcriptomic signatures identified in blood mirrored changes seen in tissue, we examined blood and splenic CD4⁺ T cells in an experimental haploidentical BMT model (B6 → B6D2F1), in which recipients were lethally irradiated (1100 cGy) and then received BM and T cells from B6.Cd4-Cre⁺ × Il6ra^f/f donors (IL-6R^−/−) or littermate controls (B6.Cd4-Cre^neg × Il6ra^f/f). It is noteworthy that Cd4-Cre deletes the IL-6R in all αβTCR⁺ donor T cells (this occurs during their thymic development), which attenuates experimental aGVHD, principally attributed to the inhibition of Th17/Tc17 and Th22 responses.^6,24,25 This revealed a mixture of transcriptomic states among CD4⁺ T cells represented by 10 distinct clusters (Figure 3A). Although T-cell effector and effector memory clusters dominated in both spleen and blood, these displayed distinct transcriptomes between tissue types. A highly upregulated interferon-stimulated gene product 15 (ISG15) expressing T-cell cluster was unique to spleen, whereas a naive T-cell cluster was unique to blood (Figure 3A), further highlighting the diversity in CD4⁺ T-cell transcriptomes within tissue compared to circulation. To identify the commonalities of T-cell transcriptomes across species, we projected the mouse scRNA-seq data set onto the human data (Figure 3B). This revealed clear concurrent clusters of effectors, effector memory, cycling effectors, and Treg subsets within CD4⁺ T cells. Analyses of cell proportions across clusters revealed that IL-6 inhibition expanded effector memory CD4⁺ T cells (cluster 5) and impaired effector CD4⁺ T cells (cluster 3) in the blood (Figure 3C), consistent with the findings in the human data set. We generated a CD4⁺ cytotoxicity gene signature module from the human RNA-seq data set and used this to compare transcriptomes between species. This demonstrated a predominant cytolytic gene signature in mouse effector memory and cycling effector CD4⁺ T cells within both blood and spleen (Figure 3D), which expressed high levels of Eomes (Figure 3E). Next, we examined the cytotoxic/effector molecule expression in blood and spleen at the protein level. In blood, an increased frequency of perforin- and granzyme A–expressing donor CD4⁺ T cells was seen in recipients of Cd4-Cre⁺ × Il6ra^f/f grafts (IL-6R^−/−; Figure 3F). In the spleen, an increased frequency of perforin-expressing donor CD4⁺ T cells was observed, together with previously described increases in Tregs²⁶ (Figure 3G). These data demonstrate that the prevention of IL-6R signaling in CD4⁺ T cells promotes cytolytic CD4⁺ T-cell expansion, an effect restricted to the Eomes⁺ fractions (Figure 3H-I), consistent with the recently described role of Eomes in this subset.⁸ We also noted a reduction in the proportion, median fluorescence intensity, and/or number of Tbet⁺ Th1 and, particularly, Th17 subsets in recipients of Cd4-Cre⁺ × Il6ra^f/f grafts (IL-6R^−/−; Figure 3J-K). These findings were similar in recipients of grafts from B6.Cd4-Cre⁺ or B6.Cd4-Cre^neg × Il6ra^f/f control donors, excluding Cre recombinase effects that were independent of the floxed allele (supplemental Figure 3A-B). Taken together, these data demonstrate the dependence of IL-6 signaling on cytolytic CD4⁺ T-cell effector differentiation pathways after experimental BMT.

Figure 3.

Genetic deletion of IL-6R signaling promotes the expansion of donor cytotoxic effector memory CD4⁺ T cells. Lethally irradiated B6D2F1 recipients received BM and T-cell grafts from either IL-6R^−/− (B6.Cd4-Cre⁺ × Il6ra^f/f) or littermate control (B6. Cd4-Cre^neg × Il6ra^f/f) donors. Spleen and blood were harvested on day 14 after BMT. For scRNA-seq analysis, samples (n = 6 per group, with pooled samples from every 2 mice) were prepared as described elsewhere.⁸ (A) UMAP representing all mouse CD4⁺ T-cell transcriptomes in spleen and blood from both groups. (B) UMAPs depicting human scRNA-seq data set reclustered from Figure 2 (left) and projection of the mouse scRNA-seq data set onto the human data set (right). (C) Proportion analysis of cellular transcriptomes across clusters in blood from recipients of B6.Cd4-Cre^neg × Il6ra^f/f or B6.Cd4-Cre⁺ × Il6ra^f/f donor grafts. (D-E) UMAP displaying (D) cytotoxic gene signatures and (E) Eomes within all mouse CD4⁺ T-cell transcriptomes in spleen and blood from both groups. For flow cytometry analysis, proportions of CD4⁺ T-cell subsets in (F) blood and (G) spleen. (H) Representative FACS plots and (I) frequency of Eomes⁺ CD4⁺ T cells in the spleen. (J) Representative FACS plots (left), Tbet MFI, and proportion and number of Th1 cells (right) in the spleen. (K) Proportion and number of Th17 cells in the spleen. Data combined from 2 replicate experiments (n = 10-12 per group from 2 experiments) are shown. Data were analyzed using the Mann-Whitney U test and presented as mean ± standard error of the mean (SEM). ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. Eff., effector; Mem., memory MFI, mean fluorescence intensity; NS, not significant.

A combination of a calcineurin inhibitor (CNI) and methotrexate was the standard GVHD prophylaxis strategy when clinical studies of TCZ were undertaken. Surprisingly, we were unable to document a modulation of Th17 (or Th1) differentiation after effective IL-6R inhibition with TCZ in these studies.^3,4 This is despite clear data demonstrating the critical role of IL-6R signaling in these T-cell differentiation paradigms after experimental BMT.⁶ We consequently postulated that the impact of IL-6 signaling on T-cell differentiation may be mitigated by the presence of calcineurin inhibition. We confirmed experimentally that this was indeed the case, with Th17 and Th1 differentiation being independent of IL-6 signaling in the presence of systemic tacrolimus administration that results in therapeutic levels.¹⁶ As expected, Th1 and Treg differentiation was attenuated in the presence of tacrolimus, as was Eomes⁺IL-10⁺ type 1 T-cell (Tr1) differentiation (supplemental Figure 4A-C). Together, these data help explain the apparent lack of effect of IL-6 inhibition on classical T-cell differentiation in patients receiving additional pharmacological immune suppression with a CNI.

Given that CNI- and antimetabolite-based immunosuppression has now been replaced by PTCy, we next examined the effect of IL-6 signaling in CD4⁺ T-cell differentiation in this context. B6D2F1 recipients received T-cell–replete grafts from B6.Cd4-Cre⁺ × Il6ra^f/f (IL-6R^−/−) or B6.Cd4-Cre⁺ (littermate controls) donors with standard PTCy administered on days +3 and +4 after transplant. In both blood and spleen, a reduction in the proportion of Th17, but not Th1, subsets was seen in recipients of IL-6R^−/− grafts compared to recipients of IL-6R–intact grafts. The Th17 effect was not abrogated by PTCy on a per-cell basis, whereas marginal reductions in Th1 differentiation were noted in the spleen (Figure 4A-B). Consequently, unlike with a CNI, IL-6–dependent CD4⁺ T-cell differentiation was not perturbed by PTCy. We also examined the effect of PTCy on the Tr1 and Treg subsets. Surprisingly, PTCy and the absence of IL-6 signaling synergized to significantly increase the proportion of Tr1 and Treg subsets, whereas no effects were seen in isolation (Figure 4C-D). PTCy was effective in preventing GVHD lethality in these systems (Figure 4E) regardless of IL-6 signaling in donor T cells, and consistent with the augmentation of Treg differentiation, the combination of PTCy and the absence of IL-6 signaling further reduced clinical GVHD scores relative to recipients of PTCy, where IL-6 signaling remained intact (Figure 4F). Consequently, the protective effects of PTCy are not dependent on IL-6 signaling and, indeed, can be augmented by IL-6R inhibition.

Figure 4.

Benefit of IL-6R inhibition is retained in the PTCy setting. Lethally irradiated B6D2F1 recipients received BM and T-cell grafts from either B6.Cd4-Cre (−) Il6ra^f/f, B6.Cd4-Cre (+), or B6.Cd4-Cre (+) Il6ra^f/f donors. PTCy was delivered on days 3 and 4 after transplant at 50 mg/kg. Spleen and blood were harvested on day 14 after BMT. Proportions of CD4⁺ T-cell cytokine-producing subsets in blood (A) and spleen (B). (C) Frequency of FoxP3⁺ regulatory CD4⁺ T cells in the blood and spleen. (D) Representative flow cytometry plots and enumeration of Eomes and IL-10 expression in the FoxP3⁻ CD4⁺ T-cell fraction in the spleen. (E) Overall survival of recipients, pooled from 2 independent experiments (n = 10-20 mice per group). Data were analyzed using the Kaplan-Meier method, and comparisons were conducted using the log-rank test. (F) Clinical scores of recipients at D35 after transplant. Data were pooled from 2 independent experiments (n = 3-20 mice per group). Data in panels A-D were pooled from 2 independent experiments (n = 5-10 mice per group). Data are presented as mean ± SEM and were analyzed using 1-way analysis of variance (for normally distributed data) or the Kruskal-Wallis test. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. IFN-γ, interferon gamma; MFI, mean fluorescence intensity; NS, not significant; TCD, T-cell depleted.

Prophylactic IL-6R inhibition has been shown to delay engraftment in clinical studies and does not prevent severe GVHD,^4,27 such that it is not routinely used in the HSCT setting unless CRS is present, which is only seen in the setting of PTCy-based immune suppression. In contrast, IL-6R inhibition is the US Food and Drug Administration–approved mainstay of treatment for significant CRS that occurs predictably after CAR T-cell therapy. We consequently sought to understand the effect of IL-6 signaling on CAR T-cell differentiation, recognizing that, in the clinic, patients with grade ≥2 CRS would typically receive IL-6R inhibition with TCZ. We thus accessed a publicly available scRNA-seq data set from the study by Li et al,¹¹ in which patients received the CD19-targeted CAR T-cell product axicabtagene ciloleucel for the treatment of high-grade B-cell malignancies. We used the scComp differential composition model to compare T-cell subsets in low-grade (CRS <2) vs high-grade (CRS ≥ 2) CRS, whereby the latter would routinely receive TCZ. This revealed a compositional shift between CD4⁺ T-cell subsets, with an increase in cytotoxic CD4⁺ T cells and impaired Th1 differentiation in patients with high-grade (and putatively TCZ treated) CRS (Figure 5A). Given that we are unable to deconvolute the effects of IL-6R inhibition from high-grade CRS in these data sets, we next investigated the effects of IL-6 on CD19-targeted CD4⁺ CAR T-cell expansion and the subsequent control of B-cell acute lymphoblastic leukemia (ALL) in our established preclinical systems,^8,20 in which the effects could be effectively disentangled. Lymphodepleted B6 recipients received eGFP⁺ ALL on day 0 with wild-type (WT) and IL-6R^−/− (CD4 and CD8 in a 1:1 ratio) CD19-directed CAR T cells (EGFR⁺) on day +3, as depicted in Figure 5B. As expected, the expansion of CD4⁺ CAR T cells was impaired in the absence of IL-6 signaling early after transfer (Figure 5C). However, the proportion of Eomes⁺ and FoxP3⁺ CD4⁺ CAR T cells (Figure 5D-E) was increased in the absence of IL-6 signaling, whereas interferon gamma and granulocyte-macrophage colony-stimulating factor (GM-CSF) expression was reduced (Figure 5F). We also confirmed a reduction in Tbet and an increase in perforin expression characteristic of Tr1 cells (Figure 5G).^8,28 To assess the functional capacity of CD19-targeted CAR T cells to control tumor in the absence of IL-6 signaling, we repeated these experiments with higher leukemia burdens and quantified tumor control 1 week after CAR T-cell transfer. Despite the quantitative reduction in CD4⁺ CAR T-cell expansion seen in blood and BM, control of B-cell ALL was enhanced in the absence of IL-6 signaling (Figure 5H), consistent with qualitative improvements in CAR T-cell function.

Figure 5.

IL-6 impairs Eomes⁺ CD4⁺ CAR T-cell expansion and tumor control. (A) Results of the scComp differential composition model comparing T-cell subsets (from the study by Li et al¹¹) between low-grade CRS (CRS <2) and high-grade CRS (CRS ≥2) that typically receive IL-6R inhibition, demonstrating increased cytotoxicity and impaired Th1 differentiation in the latter. The points represent posterior estimates of the compositional effect size for each cell type. The horizontal lines indicate the corresponding 95% credible intervals. Effect sizes are on the log-relative abundance (log ratio) scale. The vertical dashed line at 0 represents no compositional shift. (B) Schematic showing the experimental design: B6 recipients were lethally irradiated and received syngeneic BM and 2.5 × 10⁵ (C-G) eGFP⁺ B-ALL on day 0 and 1 × 10⁶ WT and IL-6R^−/− CD19-directed CAR T cells (EGFR⁺) at a 1:1 CD4-to-CD8 T-cell ratio on day +3, as outlined. (C) Relative expansion of CD4⁺ and CD8⁺ CAR T cells in blood and BM at day 10 after transfer. Concatenated flow cytometry plots with relative differentiation of Eomes⁺ CD4⁺ CAR T cells (D) and Treg expansion (E), cytokine expression (F), and Tbet and perforin expression (G). Data were pooled from 2 independent experiments (n = 10 per group). (H) B6 recipients were lethally irradiated and received syngeneic BM and 5 × 10⁵ eGFP⁺ B-ALL on day 0 and 1 × 10⁶ CD4^cre or IL-6R^−/− CD19-directed CAR T cells (EGFR⁺) at a 1:1 CD4-to-CD8 T-cell ratio on day +3, as outlined. Concatenated flow cytometry plots in blood of eGFP⁺ B-ALL and EGFR⁺ CAR T cells and absolute numbers in blood and BM a week after CAR T cell transfer. Data were pooled from 2 independent experiments (n = 12-13 per group). Data are presented as mean ± SEM. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. EGFP, enhanced green fluorescent protein; EGFR, epidermal growth factor receptor; FDR, false discovery rate; GM-CSF, granulocyte-macrophage colony-stimulating factor; IFN-γ, interferon gamma; NS, not significant; pfp, preforin/pore forming protein; TBI, total body irradiation; WT, wild-type.

Discussion

IL-6 represents a dominant pathogenic cytokine after cellular immunotherapy and mediates immune toxicity after both BMT and CAR T-cell therapy. In the CAR T-cell setting in particular, IL-6R inhibition with TCZ represents the mainstay of US Food and Drug Administration–approved therapy for CRS. Here, we demonstrate that IL-6 acts as a suppressive cytokine, blocking the emergence of cytolytic CD4⁺ T cells and type I interferon responses after HSCT. Although the role of IL-6 in promoting Th1 and Th17 responses is well described, its ability to suppress Eomes⁺ Tr1 and cytolytic differentiation has not been previously described and was seen in both clinical and preclinical systems here.

Our data highlight that the previously unappreciated effects of IL-6 on T cell biology and the unexpected effects of concurrent clinically relevant pharmacological immune suppression provide an explanation for the apparently disparate effects of IL-6 signaling on CD4⁺ T-cell differentiation in preclinical and clinical studies. Interestingly, the clear effects of IL-6 in promoting Th17 and maintaining Th1 differentiation in preclinical systems⁷ are highly modulated by CNIs that represent the pharmacological mainstay of clinical immune suppression, helping explain this apparent discrepancy. In contrast, the new gold standard for GVHD prevention after HSCT is PTCy, and the protective effects of this strategy are IL-6 independent. Indeed, the combination of inhibition of IL-6 signaling and PTCy appears to have synergistic activity to promote Treg responses.

The induction of type I interferon signaling in monocytes is known to impart resistance to viral infection,²⁹ which is intriguing, given that a major consequence of IL-6R inhibition is the attenuation of cytomegalovirus reactivation.⁷ Cytolytic CD4⁺ T-cell lineages have been demonstrated in cancer and viral infection,³⁰ mediating the lysis of tumor cells or viral clearance in a major histocompatibility complex-II–dependent manner.^31,32 We have recently shown that Eomes controls the differentiation of CD4⁺ T cells along a cytolytic and anti-inflammatory trajectory, characterized by perforin/granzyme and IL-10 secretion, respectively.⁸ In those studies, this population attenuated GVHD by deleting antigen-presenting cells expressing alloantigen and mediating direct graft-versus-leukemia effects after allogeneic BMT.^8,28 In the CAR T-cell setting, we noted that the population attenuated CRS via IL-10 secretion and mediated direct antitumor effects and prevented antigen-low immune escape, primarily via perforin.⁸ Importantly, we confirmed that Eomes⁺ CD4⁺ CAR T-cell differentiation represents a dominant pathway in patients and is associated with long-term persistence and cure of the underlying malignancy.⁸

We have previously shown that Eomes⁺ CD4⁺ T-cell differentiation after allogeneic BMT is highly IL-27 dependent but that IL-27 is largely redundant in Eomes⁺ CAR T-cell differentiation.^8,28 In contrast, here we demonstrate that IL-6 suppresses this cytolytic CD4⁺ T-cell differentiation paradigm after both allogeneic BMT and CAR T-cell therapy and across both preclinical and clinical settings. Given the primacy of IL-6 in immune toxicity after CAR T-cell therapy, our data provide a logical framework for testing the prophylactic administration of IL-6R inhibition with TCZ before T-cell infusion, aiming to both prevent CRS and promote the generation of cytotoxic CD4⁺ T cells that mediate potent antitumor immunity. Whether indirect cytokine inhibition via STAT3 inhibition with agents such as ruxolitinib would have similar effects to TCZ on cytolytic CD4⁺ T-cell differentiation will need to be studied because these agents provide broader but less complete inhibition of any individual cytokine signaling. In summary, we demonstrate that IL-6 inhibition appears to confer a hitherto-unappreciated CD4⁺ T-cell differentiation paradigm associated with favorable immunotherapy outcomes.

Conflict-of-interest disclosure: G.R.H. has consulted for Generon Corporation, NapaJen Pharma, iTeos Therapeutics, CSL Limited, Cynata Therapeutics, Neoleukin Therapeutics, and Incyte Pharma and has received research funding from Compass Therapeutics, Syndax Pharmaceuticals, Applied Molecular Transport, Serplus Technology, Heat Biologics, Laevoroc Oncology, iTeos Therapeutics, Genentech, Incyte Pharma, and CSL Limited. C.G.W. owns shares in 10x Genomics, whose single-cell RNA sequencing reagents were used in this study. The remaining authors declare no competing financial interests.