Mass cytometry identifies expansion of double positive and exhausted T cell subsets in the tumor microenvironment of patients with POEMS syndrome.

Abstract

We sought to dissect the tumor microenvironment in a small cohort (N=10) of patients with POEMS at diagnosis and after therapy using mass cytometry. We included 10 MGUS patients as controls. We identified 29 immune cell subsets in the CD45+ and CD3+ compartments. Double positive T cells and PD-1 positive CD4 T cells were expanded and naïve CD4 T cells were decreased in the bone marrow of patients with newly diagnosed /progressing POEMS. These findings provide evidence for possible antigenic driven selection as a driver of disease pathogenesis in POEMS.

Introduction

POEMS syndrome is a paraneoplastic disorder characterized by a low clonal plasma cell (PC) burden and a constellation of signs/symptoms predominant among which is the presence of peripheral neuropathy. Most patients derive clinical responses with PC-directed therapies including radiation of solitary plasmacytomas which suggests that the clonal PC are implicated in the pathogenesis of the syndrome. However, other patients demonstrate clinical responses even in the absence of hematologic responses or have no clinical relapse in the presence of hematologic relapse ¹, which suggests that PC-extrinsic factors are implicated in disease pathogenesis. One of these factors that have not been adequately explored is the immune tumor microenvironment (iTME). In this study, we sought to dissect the diversity of the bone marrow iTME in a small cohort of patients with POEMS at diagnosis and after therapy using mass cytometry (CyTOF).

Methods

After institutional IRB approval, we included cryopreserved mononuclear cells from 10 bone marrow aspirate samples from patients with POEMS syndrome at various stages of their disease (diagnosis, post-transplant, post- immunomodulatory therapy) and 10 newly diagnosed MGUS patients as controls. Our antibody panel, sample preparation, staining and acquisition protocols and data analysis methods have been previously described ² and are included in the supplemental methods.

Statistical analyses were performed using JMP (SAS) v 14. CyTOF analyses were performed in Cytobank. To compare continuous and categorical variables the Wilcoxon rank-sum and Fischer’s exact tests were used, respectively. Given the multiple comparisons performed, P values of <0.001 were considered statistically significant whereas p values of <0.05 were considered indicative of potential trends.

Results

The baseline demographics and clinical characteristics of patients are shown in the supplemental table. POEMS patients were younger (p=0.01) and, as expected, more likely to be lambda restricted (p=0.03). Of the 10 patients with POEMS, 3 had samples at diagnosis-ND, 1 at progression-PROG (clinical and hematologic progression), 3 at day+100 post-autologous stem cell transplant (ASCT) and 3 after 3, 9 and 12 cycles of ixazomib, lenalidomide and dexamethasone (post-IRD), respectively, administered as part of a clinical trial (NCT02921893).

Mass cytometry identifies a diverse T cell compartment in patients with POEMS syndrome.

Excluding T cells and lineage negative cells, FlowSOM identified 11 well-characterized immune clusters in the CD45+ compartment (supplemental figure 1). Separate FlowSOM clustering was performed on manually gated CD8 and CD4 T cell subsets to better dissect their heterogeneity. A total of 18 distinct subsets were identified (figure 1). We observed some similar phenotypes between CD8 and CD4 cells such as the terminally differentiated (CD57+) T5 and T15 subsets and the HLA-DR+/CD-16+ T6 and T10 populations. The latter are novel for POEMS syndrome and dysproteinemias in general and are thought to represent “innate-like” subsets³ that can be activated in the absence of T cell receptor stimulation. The CD38+/PD-1+ T3 subset is also novel for POEMS and along with the T16 subset, it was characterized by CCR5 and PD-1 positivity. These common phenotypes suggest that T cells are exposed to a similar milieu of stimuli and undergo similar polarization patterns. Of note, the immunosenescent (CD27-/CD28-) T4 subset was the most common CD8 T cell subset. Other subsets of interest included the CXCR3/CCR4+ T12 subset with a Th1-like phenotype and the ICOS/CD38+ positive Treg-1 T17 subset. The former is thought to have an immunosurveilling role in multiple myeloma ⁴. The latter is thought to be a highly immunosuppressive T regulatory subset associated with poor outcomes in several malignancies⁵.

Figure 1.

CyTOF analyses of CD8 and CD4 positive T Cells in the tumor microenvironment of patients with POEMS syndrome. viSNE maps and respective heatmaps from patients with POEMS syndrome. Markers expected to be uniformly negative across all subsets are not shown for clarity. Relative median frequencies of total CD3+ cells are shown as a bar graph on the right.

Double positive (DP) T cells and exhausted CD4 T cells are increased whereas naïve CD4 T cells are decreased in the iTME of patients with POEMS compared to patients with MGUS.

We then sought to evaluate if there are any differences between the iTME of patients with newly diagnosed (n=3)/progressing (n=1) POEMS patients (ND/PROG) and that of patients with ND MGUS (figure 2). We though the latter is a more appropriate control group since it is a low plasma cell burden precursor unlike most other dysproteinemias. DP T cells and the T16 (CD4+/PD-1+) subset were increased whereas the naïve, T9 CD4 subset was decreased in patients with ND/PROG POEMS. The median mass intensity (MMI) of PD-1 on CD4+ T cells was also increased overall in patients with POEMS syndrome. This remained true (p<0.05) even when removing an outlier patient with ND POEMS who had the highest frequency of DP and T16 T cells as well as the highest PD-1 MMI in CD4+ T cells. We did not find any differences in PD-1 MMI for CD8+ T cells. We then evaluated if PD-1+ cells were increased by immunohistochemistry in the iTME of an expanded cohort of ND POEMS patients that included the 4 POEMS patients and 10 MGUS patients included in the CyTOF analyses and additional 10 ND POEMS patients. After blinded review and scoring, no differences in the frequency of PD-1+ cells were found, which suggests the differences noted could be limited to the specific subsets identified by CyTOF. Finally, a correlation matrix within the ND/PROG POEMS patients (supplemental figure 2) identified some highly correlated cell subsets. The T16 subset positively correlated with the terminally differentiated (CD57+) T5 and T7 subsets and negatively with the CD163 (an M2 polarization marker) negative monocyte-2 subset. DP T cells correlated significantly with the exhausted T3 and CD57+ T7 subsets. These correlations remained significant when considering all POEMS patients.

Figure 2.

Frequencies of the T9 and T16 T cell subsets, double positive T cells and PD-1 median mass intensity on CD4+ T cells in patients with newly diagnosed (ND) or progressing (PROG) POEMS syndrome.

Triplet induction therapy increases M2 polarized monocytes but decreases immunosenescent and effector memory T cell subsets

We evaluated 3 POEMS patients at day+100 post-ASCT and 3 patients after 3, 9 and 12 cycles of therapy with IRD, respectively (supplemental figure 2). In patients treated with IRD, CD163+ M2 polarized monocytes were increased whereas the CD27-/CD28- T5 subset and the CCR5+ effector memory T2 subset were decreased. The PD-1+ T16 CD4 T cell subset, which was increased in ND/PROG patients, was decreased in patients post-ASCT.

Discussion

In this study, our key finding is that DP T cells and PD-1 positive CD4 T cell subsets are increased whereas naïve CD4 T cells are decreased in patients with POEMS compared to MGUS. These findings are supportive of antigenic selection as a possible driver of disease pathogenesis in POEMS. The exhausted and DP T cells highly correlated with other exhausted and immunosenescent subsets, that also develop in response to chronic antigenic stimulation. Clonal PC in POEMS syndrome are highly restricted in terms of their immunoglobulin variable region gene usage ⁸, which is seen in normal B cells responding to antigen. DP T cells have been shown to expand in multiple malignancies and infections in a highly antigen specific manner and to produce a mostly Th2-skewed cytokine repertoire⁹ similar to the dominant cytokine profile encountered in the iTME of POEMS ¹⁰. The DP T cells in POEMS had an effector memory phenotype, seen in later stages of T cell differentiation and after exposure to antigen. In addition, PD-1 expression appeared to be increased across all CD4+ T cells in POEMS patients, and especially among CCR5+ subsets, which are recruited in response to antigen. It is unclear if the decrease of these CD4+/PD-1+ with ASCT explains its clinical benefits. It is notable though that most other PD-1+ subsets increase early post-ASCT². These results suggest that PD-1+ CD4 T cells of POEMS patients are antigen-experienced cells that are actively recruited in the iTME.

Lenalidomide and ixazomib increased M2 polarized monocytes but decreased immunosenescent CD8 T cells, which were the most abundant CD8 T cell subset at diagnosis. M2 macrophages are classically thought to promote tumor growth and have been associated with worse outcomes in various tumors including multiple myeloma. Lenalidomide has been shown to increase their numbers ^{2, 11} but can alter their function to a tumor controlling one¹². CD27-/CD28-/CD57+ T cells are terminally differentiated subsets in a state of replicative senescence that have been shown to have suppressor functions ¹³. Immunomodulating drugs, such as lenalidomide can decrease the frequency of these cells in the iTME¹⁴.

Our study has strengths but also significant limitations. It is the first of its kind using a multiparametric approach such as CyTOF. The use of MGUS as controls is ideal since these patients have a low burden of plasma cells, with similar transcriptional profiles to POEMS patients ¹⁵. We included patients treated with standard therapies (ASCT) but also novel triplet therapies (IRD) which are reasonable options for patients not eligible for ASCT. One major weakness is the small number of patients included, which was dictated by sample availability for this rare disease. Additionally, we did not evaluate myeloid immune subsets in detail because these cells are lost in a non-stochastic manner in cryopreserved samples, which could influence the results. For this reason, these results should be considered as preliminary and hypothesis-generating. Finally POEMS patients were younger, however, DP ⁶ and exhausted ⁷ T cells increase and naïve T cells decrease with age. Future research should validate these findings in larger cohorts, focus on identifying the functional attributes of these cells including their cytokine production profile and characterize the composition of the iTME in patients with relapsed disease.