Common Lymphoid Progenitors and B Cell Subpopulations in the Bone Marrow Are Predictive of Treatment-Free Remission in Chronic Myeloid Leukemia

Background: The DESTINY clinical trial (NCT01804985) recruited chronic phase-chronic myeloid leukemia (CP-CML) patients (n=174) in molecular remission (MR⁴ and MR³) taking imatinib, nilotinib or dasatinib for an attempt at treatment-free remission (TFR), consisting of a 12-month 50% tyrosine kinase inhibitor (TKI) reduction and subsequently stopping TKI if remained in MR³, with 24 months of further follow-up. TFR at the 36-month endpoint was 72% and 36% in the MR⁴ and MR³ groups, respectively. To date, there are no established biomarkers which can accurately predict TFR success, although numerous features of the CML immunological landscape have been proposed.

Objectives: Using bone marrow (BM) mononuclear cells (MNCs) from patients enrolled in the DESTINY trial immediately prior to TKI de-escalation, we aimed to quantitate a broad range of stem/progenitor and immune effector cell types, identify features of the immune landscape which can predict TFR success and facilitate the development of predictive biomarkers for CP-CML patients attempting TFR.

Methods: BM-MNCs from DESTINY patients were profiled using flow cytometry to quantify a range of immunophenotypic cell populations. Univariate logistic regression models were built using relapse/TFR as a binary outcome, and models with p<0.1 were used to build a multivariate logistic regression model through iteration. RStudio (v.2023.09.0) and GraphPad Prism (v.9.4.1) were used to conduct all statistical analyses. To assess differentiation capacity, CD34⁺ cells isolated from BM-MNCs were co-cultured in vitro with OP9 or OP9-DL1 stromal cells and growth factors selected to facilitate either T, B or NK cell differentiation.

Results: The fraction of common lymphoid progenitors (CLPs; Lin^-CD34⁺CD10⁺) within BM-MNCs was reduced in DESTINY patients who relapsed (n=16) compared with those who sustained TFR (n=17; p_adj=0.005). However, other progenitor populations, including granulocyte–macrophage (GMPs; Lin^- CD34⁺CD38⁺CD123⁺CD45RA⁺), common myeloid (CMPs; Lin^-CD34⁺CD38⁺CD123⁺CD45RA^-) and megakaryocyte–erythroid (MEPs; Lin^-CD34⁺CD38⁺CD123⁺CD45RA⁺) progenitors, were not altered in number within the Lin^- population. RNA sequencing of isolated CLPs (n=4 relapse, n=4 TFR) identified, through gene set enrichment analyses, that mTORC1 pathway target genes were downregulated in CLPs from patients who experienced molecular recurrence, as compared to TFR (FWER=0.04 [adjusted significance statistic]).

Furthermore, using an in vitro OP9/OP9-DL1 co-culture model, we found that CD34⁺ MNCs isolated from DESTINY patients’ BM had a higher capacity for differentiation towards T (CD3⁺CD19^-) cells (p=0.049), but not NK (CD3^-CD56⁺) or B cells (CD3^-CD19⁺), in patients who experienced molecular recurrence.

In contrast, we observed no difference in the proportions of total differentiated T (CD3⁺CD19^-), NK (CD3^-CD56⁺) or B (CD3^-CD19⁺) cells in DESTINY BM-MNCs between outcome groups (p=0.983, p=0.577 and p=0.827, respectively). However, results indicate the naïve B cells (CD3^-CD19⁺IgD⁺CD27^-CD38^low) were reduced in patients who went on to relapse (n=12 relapse; n=13 TFR; p_adj=0.033), while switched memory B cells (CD3^-CD19⁺IgD^-CD27⁺CD38^lowIgM^low) were increased in relapsing patients, compared to TFR (p_adj=0.031).

Univariate logistic regression analyses found that the BM CLP fraction size was positively predictive of TFR (OR=1.73; p=0.048). The size of the naïve B cell-containing fraction (CD3^-CD19⁺IgD⁺CD27^-) and switched memory B cell-containing fraction (CD3^-CD19⁺IgD^-CD27⁺) had predictive value also (OR=8.71, p=0.051; OR=0.87, p=0.088; respectively). Using these parameters, through iterative model comparison, a multivariate predictive model for TFR was developed using the naïve B cell-containing fraction (OR=4.90) and CLPs (OR=1.61); the ROC curve AUC=0.826 (p=0.008).

Conclusion: Our data suggest that whilst on TKI therapy and in MR³/MR⁴, increased numbers of CLPs and IgD⁺CD27^- B cells in the BM are both predictors of successful TFR. Furthermore, downregulated mTORC1 signalling in CLPs was associated with molecular recurrence and may underpin the biology of skewed CLP numbers. Further functional and transcriptomic analyses to better understand the immune microenvironment in CML patients attempting TFR will assist biomarker development for identifying patients most likely to maintain successful TFR.