Identification of predictive protein signatures of treatment-free remission in patients with chronic myeloid leukaemia

Abstract Introduction: In the DESTINY trial (NCT01804985) bone marrow (BM) samples were taken from 160 patients with chronic myeloid leukaemia (CML) who discontinued tyrosine kinase inhibitors (TKIs), with treatment-free remission (TFR) as an outcome. Patients were recruited in major molecular response (MR3; BCR: : ABL10. 1%) and MR4 (BCR: : ABL10. 01%). Patients had an initial TKI de-escalation period for 12 months of half-dose before TKI cessation. After a further 24 months, 72% of MR4 and 36% of MR3 patients maintained TFR. In order to understand the changes occurring in the BM and how they may contribute to relapse, we looked at the proteomic profile and supportive stroma of the BM microenvironment (BMM). Methods: DESTINY BM plasma samples were evaluated from patients with successful TFR at both trial entry (TFRE) (n=20) and end of trial (TFRX) (n=10), and those with molecular recurrence, at both trial entry (RE) (n=20) and at point of recurrence (RX) (n=17). Samples from 3 healthy donors (N) were also assessed. Over 5000 proteins were identified and profiled from BM plasma using label-free mass spectrometry and analysed with Spectronaut software. Differentially expressed proteins (DEPs) were calculated at p. adj0. 05 and log fold change of 0. 5. The predictive ability of the proteins for TFR was analysed using a ROC curve and area under the curve (AUC). Mesenchymal stem cells (MSCs) were extracted from a subset of BM patient samples in RE (n=5) and TFRE (n=5) groups for RNA-sequencing (RNA-seq). All statistical and enrichment analyses were carried out using Rstudio (v 4. 5. 0). Results: Gene ontology (GO) analysis of DEPs identified downregulation of 'humoral immune response’ (GO: 0006959), 'antibacterial humoral response’ (GO: 0019731) and 'antimicrobial humoral response’ (GO: 0019730) pathways when comparing RE to TFRE patients. The DEPs within each pathway were used to calculate an AUC for their prediction of outcome, which was 0. 8, 0. 71 and 0. 8 respectively. Gene set enrichment analysis (GSEA), based on the identified proteins, also showed downregulation of 'humoral immune response’ (NES= -2. 1, p. adj= 0. 0008) and 'complement activation’ (NES= -2. 2, p. adj= 0. 001) in the same comparison, suggesting a compromised immune response in patients with subsequent molecular recurrence. Furthermore, GSEA pathway analysis of RE vs TFRE, based on the proteomics data, showed upregulation of a range of metabolic processes including 'purine-containing compound metabolic process’ (NES= 1. 7, p. adj= 0. 0008) and 'purine nucleotide metabolic process’ (NES= 1. 8, p. adj= 0. 001) indicating not only a dysregulation of metabolic processes within patients with molecular recurrence but also potential changes to immune cells and inflammation within the BMM. When comparing RX to RE, we identified an increase in 'myeloid leukocyte activation’ (GO: 0002274), 'granulocyte chemotaxis’ (GO: 0071621) and 'leukocyte chemotaxis’ (GO: 0030595) pathways, suggesting an increased pro-inflammatory profile. We identified DEPs: RHAG, CA1, RHCE, COL1A2 and SLC2A1 (p0. 05) that were upregulated when comparing RE vs TFRE and decreased in RX samples, which could be potential predictive biomarkers of molecular recurrence. COL1A2 is important for bone development, the decreased expression could indicate potential BM remodelling in patients with molecular recurrence. Additionally, RNA-seq of BM MSCs identified increased expression of FMO3 and CXCL12 in RE vs TFRE. This increase in FMO3 was also observed in the BM plasma (p0. 05) with FMO3 previously reported to have a role in metabolism of drugs including TKIs in CML. When comparing RE vs TFRE HLA-B was downregulated in RE vs TFRE, potentially demonstrating a weaker immune response in patients who experience molecular recurrence. The comparison of RX vs RE showed upregulation of MOK and AQP1, previously found to regulate MSC migration, immune responses, and cytokine release. Conclusions: Our findings indicate the supportive stroma of the BM may be involved in the change in protein expression presaging TFR. Additionally, patients who experience molecular recurrence after TKI de-escalation/discontinuation have altered inflammatory profile and a dysregulated immune response at trial entry which appears to continue at point of relapse whereas patients who maintain TFR have a more robust humoral immune response. Further proteomic studies are ongoing to validate these results and identify predictive biomarkers of TFR.