Mechanisms of immune escape and altered tissue homing in leukemia cutis Free

Extramedullary acute myeloid leukemia (eAML) relapse is frequent after allogeneic hematopoietic cell transplantation (HCT), suggesting the emergence of immune escape from graft-versus-leukemia (GvL) responses. The previous observation of complete responses of post-HCT leukemia cutis (LC) to CTLA-4 blockade points towards opportunities to restore GvL through immunomodulation. However, mechanisms of extramedullary tropism and associated immune populations remain unknown, limiting rational therapy development. We therefore dissected the immune microenvironment and leukemia features of eAML using single-cell RNA sequencing (scRNA-seq) on prospectively collected biopsies from 7 individuals.

We performed scRNA-seq on 16 eAML biopsies and 6 peripheral blood (PB) samples prior to and following immune checkpoint blockade (ICB). For comparison, we included publicly available external scRNA-seq datasets of post-HCT AML relapse bone marrow (BM) (n=7) and healthy donor skin (n=6), yielding 150,475 cells. Targeted genotyping (nanoranger) revealed leukemic involvement in progenitor cells of recipient origin (souporcell). In contrast to BM, LC was enriched for regulatory T cells (Tregs) (1.8% vs 12.9%; p < 0.001), despite similar number of T cells. T cells in LC were donor-derived (median chimerism 99%, range 33-100%), confirming the potential for GvL to be active in extramedullary sites.

To explore programs driving extramedullary tropism, we performed differential gene expression analyses (DGEA) of leukemia cells from LC versus BM. 4 of 7 extramedullary and 1 of 7 medullary AML cases displayed reduced HLA class II expression, indicating immune escape. We found increased expression of the adhesion molecule ICAM1 (encoding CD54) and decreased PECAM1 (encoding CD31) expression in LC. Further, integrin genes were differentially expressed with higher levels in medullary (ITGA4, ITGA6, ITGAL, and ITGB4) or extramedullary AML (ITGA5 and ITGAV). These differences were directly detectable in two cases with concomitant circulating and extramedullary disease, supporting tissue-dependent expression of AML. Thus, extramedullary tropism associates with immune escape and altered homing receptor expression.

To define how T cell gene expression profiles differed between LC and BM, we performed DGEA and found CD4⁺ and CD8⁺ T cells to differentially express genes associated with T cell homing such as CD44 (higher in LC) or SELL (encoding CD62L), ITGB2, ITGB7, ITGA4 and ITGAL (higher in BM). Further, LC-associated T cells showed increased expression of immune checkpoint molecules (PDCD1, LAG-3 or CTLA-4), suggesting higher levels of T cell exhaustion. Comparison of T cell states across BM, PB, eAML and healthy skin tissues revealed the number of T cells with high expression of an exhaustion signature and reduced levels of a memory score was most frequent in LC. TCR repertoire analysis with matched BM showed that T cell clonotypes unique to LC had a higher expression of an exhaustion signature than those shared (10.3% vs 2.3%), consistent with local antigen recognition. Together, LC associates with an exhausted T cell infiltrate, suggesting a potential mechanistic basis for clinical activity of ICB.

Finally, we performed immune monitoring prior to and during consecutive PD-1 and CTLA-4 blockade in one post-HCT LC case. We tracked circulating WT1 (HLA-A*01:01)-specific CD8⁺ T cells (flow cytometry) in 11 PB samples over 32 months. Onset of LC associated with reduced WT1-specific T cells (1.7% vs 0.8%), while nivolumab led to re-expansion (1.2%). Following ipilimumab and until further LC progression, WT1-specific T cells contracted once again (0.8%), demonstrating associations with the clinical course. ICB also induced changes in tissue-resident T cells: CXCL13⁺ CD4⁺ T cells with high expression of PDCD1 contracted post PD-1 blockade, while Tregs expanded. Throughout, overlap of local and circulating TCRs remained high (~50%), suggesting a minority of T cells to be leukemia specific. Overall, ICB reshaped circulating and local T cell repertoires, illustrating that immunomodulation may overcome immune escape in eAML.

In sum, we provide a comprehensive single-cell characterization of LC, uncovering immune evasion, altered homing, and T cell exhaustion, highlighting fundamental differences between medullary and extramedullary AML. This sets the stage for future studies to dissect tissue-specific leukemic niches and informs rational design of immunotherapies.