Single cell T cell landscape and T cell receptor repertoire profiling of AML in context of PD-1 blockade therapy

In contrast to the curative effect of allogenic stem cell transplantation in acute myeloid leukemia via T cell activity, only modest responses are achieved with checkpoint-blockade therapy, which might be explained by T cell phenotypes and T cell receptor (TCR) repertoires. Here, we show by paired single-cell RNA analysis and TCR repertoire profiling of bone marrow cells in relapsed/refractory acute myeloid leukemia patients pre/post azacytidine+nivolumab treatment that the disease-related T cell subsets are highly heterogeneous, and their abundance changes following PD-1 blockade-based treatment. TCR repertoires expand and primarily emerge from CD8+ cells in patients responding to treatment or having a stable disease, while TCR repertoires contract in therapy-resistant patients. Trajectory analysis reveals a continuum of CD8+ T cell phenotypes, characterized by differential expression of granzyme B and a bone marrow-residing memory CD8+ T cell subset, in which a population with stem-like properties expressing granzyme K is enriched in responders. Chromosome 7/7q loss, on the other hand, is a cancer-intrinsic genomic marker of PD-1 blockade resistance in AML. In summary, our study reveals that adaptive T cell plasticity and genomic alterations determine responses to PD-1 blockade in acute myeloid leukemia.


REVIEWER COMMENTS
Reviewer #1 (Remarks to the Author): Abbas et al. use a state-of-the-art combination approach of single cell gene expression profiling and single cell TCR sequencing to assess the T cell phenotype and clonotype in BM cells from AML patients treated with anti-PD-1 (nivolumab) plus azacitidine combination therapy at a single cell level. BM samples were derived from 8 patients enrolled in the NCT02397720 study of nivolumab+azacitidine in r/r AML and 2 healthy donors. Paired scRNA and scTCR profiling revealed changes in cellular subset composition and TCR clonotypes pre-and post nivolumab+azacitidine treatment. They found that the most abundant clonotypes expanded primarily in patients that responded or had stable disease while the most abundant clonotypes retracted in non-responders. Furthermore, the authors suggest GZMK expression in CD8 T cells and MAIT cells as a potential marker for improved outcome in AML. Finally, they identified that chromosome 7/7q loss was associated with resistance to anti-PD-1/azacitidine treatment in a larger cohort of 57 patients. While this study provides an in-depth look at T cell phenotypes and TCR repertoires in patients before and after anti-PD-1 based therapy, one major concern is that it remains at a merely descriptive level and I am not not convinced that it will be practice changing in the clinic. In addition, a major part of the results is based on quantitative assessment of cell population proportions, which has to be viewed with caution because of the limited number of patients assessed. While differences in gene signatures can be quantitatively assessed using scRNA-seq, mere quantification of cell populations should preferably be analyzed with other methods that allow for larger patient and cell numbers, e.g. flow cytometry.
Major Comments -The authors should be cautious with claiming quantitative changes in certain cell types (e.g. CD4 vs. CD8 cell subset) due to the very low patient numbers per group. It cannot be ruled out that single patients with high cell numbers drive specific changes. Single cell sequencing should be rather utilized to compare changes in gene expression in specific cell subsets between the patient groups, e.g. did T cell effector signatures change in the CD8 subset in response to treatment. Quantitative changes in cell subset proportions should rather be confirmed by flow with a) larger patient numbers and b) larger cell numbers, to be able to statistically assess these changes. -Why is timepoint C not presented in Figure 2C? Since this is a merely descriptive depiction of the clonality, all timepoints should be considered and reported. Does the time point "post" in Figure 2B and 3B refers to time point B or C -in particular in responders? -Did decrease in clonality at timepoint C predict subsequent relapse? Did patients with persistent increase in T cell clonality stay in remission longer? -In Figure 3G the text implies that only responders are depicted. How did the exhaustion scores change in non-responders? -Pseudotime analysis revealed that GZMB expression was higher at a later pseudotime. Did GZMB expression correlate with exhaustion signature? -The findings of co-expression of TCF7 and EOMES in the GZMK+ CD8 T cells, as well as their stem cell like properties are interesting, but rather descriptive. The manuscript would be improved by confirming these findings by flow and in functional assays.
-The MAIT cell data are interesting but have to be viewed with caution, since 90% of the cells were contributed by a single patient.
-The authors identify chr7/7q deletion as a resistance mechanism to checkpoint blockade and attribute this to expansion of regulatory T cells and reduced IFNg signaling. What were the percentages/phenotype of Tregs in the patients in this study at different time points? Were these correlated with the duration of response/survival? Also here, functional assays to explore the role of IFNg signaling in response to azacitidine/nivolumab would be highly beneficial.
Minor Comments - Figure 2B and 3B would be clearer if the response of the individual patients was indicated as in Figure 6A.
-Please carefully check for typos: o p.7 "… in clinical trial settings" o p.2 "… stem-like properties likely." This sentence ends openly. o p.10 "…expression levels of the cytotoxic gene GNLY, which delivers…" Reviewer #2 (Remarks to the Author): In this manuscript, the authors performed single cell RNA-sequencing on a large number of bone marrow cells from patients with relapsed/refractory AML both pre-and post-treatment alongside healthy donors. They also profiled T-cell receptors in those same cells. In summary, the authors discovered drug response and cell type proportion heterogeneity within the T-cells, along with variable TCR repertoires between responders and non-responders. Their analysis found a developmental continuum of T-cells with differing levels of the GZMK gene as well as a potential mechanism for resistance to PD-1 blockade therapy. Overall, the authors effectively use biological and computational techniques in this manuscript to evaluate their hypothesis. The manuscript is clearly written, and the results novel and of value to the scientific community.
1. The author's state: "We identified 5 (2 conventional and 3 unconventional)48 T cell phenotypes in 25,798 T cells from 22 BM aspirates before and after treatment in AML patients (Fig. 3A). The 2 conventional phenotypes were CD4+ and CD8+ cells, constituting 53% and 35% of BM T cells at pretreatment, and 30.9% and 37.4% of BM T cells at posttreatment, respectively." This section goes further to highlight cell type proportion differences between highly similar cell types, such as CD8 and MAIT cells. I do not believe that bioinformatics tools in their current state are accurate enough in cell type identification to allow for definitively stating cell type proportion differences that are only moderately altered between the two patient groups. This section could be further supported using experimental methods to verify these proportion differences in some samples. 2. In a couple of places in the manuscript the authors acknowledge the heterogeneity in the patient population. One case is the following: "Of note, 89.9% of MAIT cells in our analysis were contributed by PT1 (responder) who had a unique clinical course." Given such a dramatic patientspecific effect, how are the authors assessing the significance of their results? 3. In the Chr7/7q section, the authors found 13 significantly downregulated genes on chr7q in AML cells and used these genes to perform gene set enrichment analysis. This section could be improved by relaxing the significance criteria and performing the analysis with a larger set of genes in addition to the original analysis, as 13 is a very small gene set. 4. The author's state that "All datasets generated during and/or analyzed during the current study will be available from the corresponding authors." Will these data become publicly available upon publication? 5. Doublets were identified using robust criteria, which should remove doublets effectively, but new computational tools for doublet detection, such as Scrublet, DoubletDecon, and Solo, have shown very high accuracy in biologically validated datasets. Given the high transcriptional similarity between the T-cell populations of interest and the difficulty in identifying doublets because of this with the given techniques, I would suggest using a computational approach for further validation. 6. The legend for Figure 1C refers to a UMAP-based pseudotemporal trajectory analysis, but figure shows only the UMAP and cell type proportion predictions with no trajectory analysis. 7. In Figure 1D, my immediate concern with the green and red UMAP was potential batch effects, given the separation between the PT1C and Normal BM cells in non-malignant cell types. Upon further reflection, this could simply be an overplotting issue. Using side-by-side UMAPs in the same dimensionality reduction space would better highlight the integration and differences and alleviate batch effect concerns. 8. For Figure 3A, within each cell type do the distinct subclusters align with the mutation breakdown? If not, what transcriptional change causes the small subsets of cells to distinctly separate from the larger cell population?
Reviewer #3 (Remarks to the Author): The authors analyze single-cell transcriptomics data from serial bone marrow samples of 8 patients treated for relapsed/refractory AML with azacytidine and the checkpoint inhibitor nivolumab. With respect to immune cells, transcriptomic data are analyzed in conjunction with TCR sequences. Main findings are: -The T-cell repertoire is much smaller in AML patients compared to healthy bone marrow samples.
-a relative expansion of unconventional T cell types after treatment -Presence of T cells from the same clonotype in several different maturation/activation subsets as defined by gene expression profiles.
-A relatively prominent GZMK+ T-cell population in AML patient which presumably reflects an intermediate differentiation stage.
-several findings from comparisons of serial samples are related to the treatment response, e.g. relative contraction versus expansion of T-cell clones; a weakly significant higher fraction of GZMK+ CD8 cells in responders; highly significant lower exhaustion scores in T cells of responders -del1/q7 as a possible negative predictor for treatment response The data and their analysis appear technically sound, but are mostly descriptive and exploratory. Data interpretation is hampered by the lack of appropriate control analyses, e.g. patients treated with azacytidine only. Few statistically significant findings are reported. In many instances where novel observations appear to be made, their validity cannot be ascertained by analysis of a validation cohort.
In line with the descriptive character of the manuscript, no specific hypotheses are being formulated, and many findings are described in comparison to immune responses to checkpoint inhibition in solid tumors. Since the mutanome in AML is much smaller than in tumors responding to checkpoint inhibition, and since the cited graft-versus-leukemia effect after allogeneic stem cell transplantation is biologically predominated by a graft-versus-hematopoiesis effect directed against minor histocompatibility antigens expressed in hematopoietic cells including AML, it is speculative in how far findings in solid tumors can be extrapolated and applied to AML. 3) The scientific style is frequently misleading. E.g., the term significant should be used exclusively in the context of an appropriate statistical context. 4) There numerous typographical mistakes throughout the manuscript.
5) The sentence "The clonotype size in healthy BMs ranged from 1 to 16 TCR clonotypes, compared to 1 to 1200 TCR clonotypes in AML" does not make sense semantically