Oncogene-induced TIM-3 ligand expression dictates susceptibility to anti–TIM-3 therapy in mice

Leukemia relapse is a major cause of death after allogeneic hematopoietic cell transplantation (allo-HCT). We tested the potential of targeting T cell (Tc) immunoglobulin and mucin-containing molecule 3 (TIM-3) for improving graft-versus-leukemia (GVL) effects. We observed differential expression of TIM-3 ligands when hematopoietic stem cells overexpressed certain oncogenic-driver mutations. Anti–TIM-3 Ab treatment improved survival of mice bearing leukemia with oncogene-induced TIM-3 ligand expression. Conversely, leukemia cells with low ligand expression were anti–TIM-3 treatment resistant. In vitro, TIM-3 blockade or genetic deletion in CD8+ Tc enhanced Tc activation, proliferation, and IFN-γ production while enhancing GVL effects, preventing Tc exhaustion, and improving Tc cytotoxicity and glycolysis in vivo. Conversely, TIM-3 deletion in myeloid cells did not affect allogeneic Tc proliferation and activation in vitro, suggesting that anti–TIM-3 treatment–mediated GVL effects are Tc induced. In contrast to anti–programmed cell death protein 1 (anti–PD-1) and anti–cytotoxic T lymphocyte–associated protein 4 (anti–CTLA-4) treatment, anti–TIM-3-treatment did not enhance acute graft-versus-host disease (aGVHD). TIM-3 and its ligands were frequently expressed in acute myeloid leukemia (AML) cells of patients with post–allo-HCT relapse. We decipher the connections between oncogenic mutations found in AML and TIM-3 ligand expression and identify anti–TIM-3 treatment as a strategy for enhancing GVL effects via metabolic and transcriptional Tc reprogramming without exacerbation of aGVHD. Our findings support clinical testing of anti–TIM-3 Ab in patients with AML relapse after allo-HCT.

Results show mean ± SEM of 3 independent experiments.

Suppl. Figure 4: Metabolic changes in Tc isolated from anti-TIM-3 treated mice
(A, B) Gene-set-enrichment analysis (GSEA) using the R package "fgsea" with metabolic pathway definitions from KEGG and measured fold changes as input.First, metabolomics data was quantile normalized to compensate for cell size effects before calculating the logarithm of the fold change as a measure of difference between conditions.To compensate for method bias, only the sub-set of KEGG pathways that was covered by the employed metabolomics platform was considered.show mean ± SEM from 3 independent experiments.P-values were calculated using a two-way Anova followed by Tukey's multiple comparisons test.
Suppl.fresh and were isolated by FICOLL density centrifugation.CD3 + Tc depletion was performed by magnetic separation.On day 2, 5x10 4 human CD3 + Tc isolated from peripheral blood of healthy donors were injected i.v.

GVHD mouse model
Briefly, BALB/c recipient mice were irradiated (using a lethal dose divided into two doses of 5.08Gy within a 4-hour interval) prior to the injection (i.v.) of allogeneic 5x10 6 C57BL/6 BM cells and 4x10 5 Tc. From day 1 to day 5, mice were injected (i.p.) with 150µg of anti-TIM-3 / isotype, anti-PD-1 (RMP1-14) / isotype (anti-rat IgG2a) or anti-CTLA-4 (9D9) / isotype (mIgG2b) Ab.At d7, mice were sacrificed by cervical dislocation.Organs (liver, SI and colon) were collected and stained with hematoxylin and eosin.Histopathological GVHD severity was assessed on the basis of a published histopathology scoring system ( 6) by an experienced pathologist blinded to the experimental groups.

Generation of retroviral supernatants
For all transductions, we used the Murine Stem Cell Virus (MSCV) retroviral expression vector system.We used the MSCV backbone plasmid with different oncogenes and IRES-GFP (pMIG).
For generation of virus supernatant, we used Platinum-E (Plat-E) retroviral packaging cell line.
Plat-E cells were transfected with 10 µg MSCV vectors using Lipofectamin TM 2000 transfection reagent (Invitrogen™) according to manufacturer's protocol.Transfection was checked using fluorescence microscopy after 24h and transfection mix was replaced by 5 ml of fresh Dulbecco's Modified Eagle Medium (DMEM, Gibco) supplemented with 10% FCS and 1% P/S.Virus supernatant was collected three subsequent times with 12h between.Virus supernatant was filtered using a 0.45µm syringe filter and stored for less than one week at 4°C.
Retroviral supernatant was added to the medium before and 3 rounds of spin infection (each 12 hours) were performed (1200g, 90 min, 32°C).Transduction efficiency was measured using flow cytometry detecting GFP-expressing cells.

In vitro Tc proliferation assays
αCD3/CD28-mediated Tc stimulation experiments were performed using splenic mouse Tc purified using the pan Tc isolation kit (Miltenyi Biotec) cocultured with Dynabeads™ Mouse T-Activator CD3/CD28 (ratio 1:1, Thermofisher Scientific) and mIL-2 (30U/mL).When indicated, anti-TIM-3 or isotype Ab was added to the culture (10 µg/mL, every two days).For in vitro exhaustion analysis, cells were continuously exposed to CD3/CD28 Abs in the presence of mouse IL-2 for 14 days before analysis by FC.

In vitro killing assay
Splenic Havcr2 cko mouse Tc were isolated using the pan Tc isolation kit (Miltenyi Biotec) and cultured with Dynabeads™ Mouse T-Activator CD3/CD28 (ratio 1:1, Thermofisher Scientific) and mIL-2 (30U/mL).Cells were split when needed and mIL-2 (30U/mL) was added to the culture every two days.On day 6, αCD3/CD28 beads were removed and Tc were cocultured with WEHI-3B GFP+ cells for 6 hours at the indicated ratios.Specific killing of leukemia cells was analyzed by FC.
The luciferase signal was quantified in photons per second per mouse.Acquisition, analysis, and visualization of BLI were performed using Living Image Software (PerkinElmer).

Single-cell energetic metabolism by profiling translation inhibition (SCENITH)
To assess the metabolic activity of Tc subsets, we used the method named single cell energetic metabolism by profiling translation inhibition (SCENITH) (8).Briefly, cell suspension was prepared from the spleens and isolated cells were divided into four different conditions. 1 × 10

Spectral Flow cytometry
For high dimensional analysis of spectral flow cytometry data, cells were stained using the antibodies specified in Suppl.Table 3. obtain expression values between 0 and 1.This was followed by unifold manifold approximation and projection (UMAP) using the umap package.Automated clustering and metaclustering of the percentile normalized data was performed with the FlowSOM package.This was followed by expert-guided merging of some clusters based on their median marker expression profile.

Cells
Differential abundance analysis was performed with a Wilcoxon test, differential expression was calculated using the limma test implemented in the diffCyt package.

Analysis of human scRNA sequencing datasets
To quantify HAVCR2, LGALS9, and HMGB1 expression in healthy bone marrow and AML cells at diagnosis, we used provided cell type annotations (12).We scaled data with the Seurat version 4.9.9.9050 function ScaleData and scale.max= 4. Next, we selected the relevant cell types and visualized gene expression using the ComplexHeatmap package version 2.16.0 (13)

Survival analysis
Reanalyses of survival differences in AML was based on the expression level of HAVCR2 across published datasets of AML cohorts (19).RNAseq data was collected from the GenomicDataCommons (GDC) library.A query was used to download public accessible STAR-Counts: TARGET-AML.Additionally, the data were filtered by tissue type: Primary Blood Derived Cancer -Peripheral Blood, Primary Blood Derived Cancer -Bone Marrow, Blood Derived Normal, Bone Marrow Normal, Recurrent Blood Derived Cancer -Peripheral Blood and Recurrent Blood Derived Cancer -Bone Marrow.High HAVCR2 expression was determined by calculating the mean.Gene expression more than one standard deviation above the mean was considered as high expression.

A
P<0.0001 were analyzed on day 23.(A) Representative FC plots displaying the proportions of exhausted TIM-3 + PD-1 + CD4 + Tc and exhausted TIM-3 + PD-1 + CD8 + Tc in the spleen of leukemia-bearing mice.(B-C) Relative TIM-3 MFI of TIM-3 + PD-1 + (B) in CD4 + Tc or (C) in CD8 + Tc in the indicated organ was assessed by FC. Results represent mean ± SEM of 2 independent experiments and Pvalues were calculated using an unpaired Student's t-test.(D-E) The potential interference of anti-TIM-3 blockade with TIM-3 detection via FC was assessed on day 10 of in vitro Tc stimulation.Tc were then incubated with 10 μg/mL of anti-TIM-3 (clone 5D12) or isotype Ab (mIgG1) for 2 hours.(D) Overlaid histograms showing the expression of TIM-3 after the cells had been exposed to anti-TIM-3 Ab (red line) or the isotype Ab (dark dotted line).Exposure to anti-TIM-3 does not reduce TIM-3 detection compared to exposure to isotype Ab.Representative data from one of three biological replicates.(E) The bar diagram shows the relative TIM-3 MFI among viable cells.

( A )
The significance scores (P values) of the top 10 enriched pathways are displayed.(B) The enrichment plot for pathway "Glycoysis/Gluconeogenesis".Suppl.Figure 5: Anti-TIM-3 Ab treatment after allo-HCT induces changes in myeloid populations (A) Heatmap showing the median expression of the 27 indicated markers used to identify the different Tc subsets.(B) Scaled expression of 9 phenotypic or functional antigens with the use of the FlowSOM algorithm among 8 myeloid phenotypic subsets.Log fold change of isotype Ab treatment compared to anti-TIM-3 Ab treatment is shown (blue color depicts higher expression in isotype, red color higher expression in anti-TIM-3).Differentially expressed proteins with P-value < 0.05 tested with moderated t-test of limma are presented.All antibodies used for spectral FC are summarized in Suppl.Table 3. histopathological aGVHD severity in the liver (A), SI (B) and the colon (C).(D-H) The percentage of (D) AML cells (H2-Kd + CD3 -CD45 + cells) (E) TOX + among CD8 + Tc (F) TCF-7 + among CD8 + Tc were assessed in the spleen.(G) Representative FC plots displaying the proportions of TOX + and TCF-7 + among CD8 + Tc. (H) Percentage of INF-γ + TNF-α + among TCF-7 + CD8 + Tc (TPEX subset) after pan-Tc stimulation for 5 hours (Cell Stimulation Cocktail, eBiosciences).Results

Figure 9 :
Sub-clustering of scRNAseq analysis (A-C) BALB/c recipient mice were injected i.v. with WEHI-3B cell line (BALB/c background) and 5x10 6 allogeneic Havcr2 fl/fl or Havcr2 fl/fl ;E8i cre/+ BM and Tc as indicated for each group.Tc were isolated at day 23 and stained with an oligo-tagged H-2Kb (donor) Ab allowing analysis of scRNAseq libraries from 4 donor CD3 + Tc samples (n=2 Havcr2 fl/fl and n=2 Havcr2 fl/fl ;E8i cre/+ ).(A) UMAP visualization revealed 21 distinct clusters before exclusion of non CD3 + Tc. (B) Overall expression of the CD3e gene in each cluster (C).The analysis was then divided into CD4 + and CD8 + Tc subclustering.Suppl.Figure 10: Genetic TIM-3 deletion in CD8 + Tc leads to remodeling of CD8 + Tc subsets (A-C) BALB/c recipient mice were injected i.v. with WEHI-3B cell line (BALB/c background) and 5x10 6 allogeneic Havcr2 fl/fl (n=2) and Havcr2 fl/fl ;E8i cre/+ (n=2) BM and Tc.Tc were isolated at day 23 and stained with an oligo-tagged H-2Kb (donor) Ab allowing scRNA-seq analysis of selected donor Tc. (A) Heatmap showing the Top 6 marker genes expressed for each clusters of CD8 + Tc. (B) UMAP showing metaclustering of CD8 + Tc in AML-bearing mice transferred with Havcr2 fl/fl (left panel) and Havcr2 fl/fl ;E8i cre/+ (right panel) BM/Tc.(Tem: effector memory Tc; Teff: effector Tc).(C) Dot plot showing the Top 3 marker genes expressed for each cluster of CD8 + Tc. (D) The geneset analysis represents the Top 20 most significantly regulated Hallmark gene sets in each individual CD8 + Tc cluster.Suppl.Figure 11: Deletion of TIM-3 in CD8 + Tc leads to changes in CD4 + Tc responses (A-E) BALB/c recipient mice were injected i.v. with WEHI-3B cell line (BALB/c background) and 5x10 6 allogeneic Havcr2 fl/fl and Havcr2 fl/fl ;E8i cre/+ BM and Tc.Tc were isolated at day 23 and stained with H-2Kb (donor) Ab allowing analysis of scRNA-seq libraries from 4 donor CD3 + Tc samples (n=2 Havcr2 fl/fl and n=2 Havcr2 fl/fl ;E8i cre/+ ).(A) The UMAP visualization shows 6 clusters of CD4 + Tc isolated from AML-bearing mice.(B) Feature plots showing the expression level of different marker genes expression relevant for the characterization of the clusters.(C) Bar diagram were incubated with along with True Stain FcX (BioLegend) and True Stain Monocyte Blocker (BioLegend) to prevent unspecific binding.For dead cell exclusion, cells were stained with Zombie NIR fixable viability dye (BioLegend, 1:500).Surface antigen staining was performed in PBS for 20 minutes at 4°C.The cells were then washed in PBS and fixed with the Foxp3 / Transcription Factor Staining Buffer Set (ThermoFisher) according to the manufacture's instruction.For intracellular cytokine staining, cells were incubated with the respective intracellular antibodies diluted in Permeabilization buffer (ThermoFisher).After washing, cells were resuspended in PBS and acquired on Cytek Aurora (Cytek Biosciences).Data pre-processing was carried out in FlowJo (TreeStar) for singlets and dead cell exclusion and CD45 + cell selection.Compensated and pre-gated cells were then imported into R studio using R version 4.0.2 using the read.flowSet()function of the flowCore package.The flow cytometry data was then transformed using a hyperbolic arcsine (arcsinh) transformation and percentile normalized to
For intracellular cytokine detection, fresh single-cell suspensions were incubated in a 96-well, round-bottom plate for 5 hours with the eBioscience Cell Stimulation Cocktail plus protein transport inhibitors (Fisher Scientific, catalog # 00-4975-93).Cells were stained for surface markers, before fixation, permeabilization, and intracellular staining.Multiparametric flow cytometry analyses were performed using a Celesta cytometer (BD Biosciences).