Comparison of Anti-PD1, Anti-CTLA4 and Anti-TIM3 in Treatment of AML Patients with Single-Cell Transcriptomics Displays Different Effects on Immune Subtypes

Introduction Although immune checkpoint inhibitors (ICI) in combination with hypomethylating agents (HMA) have had limited activity in acute myeloid leukemia (AML), a subset of patients has achieved durable responses. This emphasizes the need to define the immunomodulatory effects and immune phenotypes that associate with response to different ICIs. Further, understanding individual responses can point towards new immunotherapeutic strategies in AML. Methods We combined single-cell RNA and T cell receptor sequencing (scRNA+TCRαβ-seq) data from AML patients (N=37) treated on clinical trials with the combination of HMA with either anti-CTLA4 (responders CR/CRi/PR: 10/18; Penter et al., Blood 2023), anti-PD1 (CR/CRi/PR: 5/12; Abbas Hu et al., Nat Comm 2021, Goswami et al., JITC 2022) or anti-TIM3 (CR/CRi/PR: 3/7; Huuhtanen et al., ASH 2021). Overall, we pooled together 122 longitudinal bone marrow samples and obtained 605,877 quality-controlled and harmonized cells, making it the largest scRNA+TCRαβ-seq dataset of AML treated with immunotherapy thus far. Results First, we defined immune checkpoint expression across immune cells. Expression of PD1 (PDCD1) and CTLA4 was mostly restricted to T cells, where PD1 was mainly expressed in CD8+ T cells (Tem, Temra) and CTLA4 in CD4+ T cells (Treg). This contrasted with TIM3 (HAVCR2), that was highly expressed in different NK cells (CD56bright, CD56dim, adaptive NK cells) and different myeloid cells, with lower levels in CD8+ T cells, suggesting different treatment effects. As immune checkpoint expression profiles differed particularly in NK cells, we explored their role in ICI responses. Responders in all treatment groups had more mature (cytotoxic CD56dim and memory-like adaptive) NK cells prior to treatment than non-responders (p=0.016). Even though there was no overall significant difference in NK cell expansion, responders to anti-TIM3 or anti-PD1 had more transcriptional changes (number of DEGs) in NK cells, than those treated with anti-CTLA4. Finally, anti-PD1 upregulated IFN-γ and IFN-α pathways, while anti-TIM3 upregulated NF-κB pathway, suggesting different activation mechanisms. Given high PD1 expression in CD8+ T cells, we wondered whether this would translate into different ICI-related mechanisms of response. Indeed, anti-PD1 responders upregulated IFN-g and IFN-a pathways, unlike anti-TIM3 and anti-CTLA4 responders. Additionally, when analysing scTCRab data, more CD8+ T cell clones expanded (Fisher's exact test) in response to anti-PD1 (mean number of expanding clones 20.7, p=0.036) and to anti-TIM3 (mean 23.1, p=2.5x10-7) than to anti-CTLA4 (mean 6.6). However, expanding T cell clones did not correlate with treatment outcomes and were also detected in non-responders. Since T cell clones also expanded in non-responders, we asked whether their phenotypes or frequencies associated with response to ICI. In responders to anti-PD1, we found large expanded T cell clones (>1% of CD8+ TCR repertoire) whose transcriptional profiles shifted from an exhausted (PDCD1+, TOX+) to a cytotoxic CD8+ Temra phenotype, consistent with reversal of T cell exhaustion. In contrast, in anti-TIM3 responders, the size of expanding T cell clones was small (<0.1% of CD8+ TCR repertoire in pre-treatment samples), and we noted decreased CD8+ Temra populations while exhausted T cells increased. This suggests that anti-TIM3 might not reverse T cell exhaustion, but potentially increases priming of T cells. After anti-CTLA4, no clear differences in the phenotypes of the expanded CD8+ T cell clones were noted, consistent with the expression of CTLA4 on CD4+ T cells. Conclusions Our analysis suggests different immune activation mechanisms for anti-PD1, anti-CTLA4, and anti-TIM3 in AML. Both anti-PD1 and anti-CTLA4 exhibited more effects on T cells compared to anti-TIM3. In responders, anti-PD1 therapy upregulated IFN-γ and IFN-α pathways and reversed exhaustion in CD8+ T cell clones. Conversely, anti-TIM3 had more pronounced effects on NK cells via upregulation of NF-κB pathway in responders. None of the immune cell populations were able to predict the treatment response, highlighting that the choice of ICI should be personalized to each patient.