Timing Antigenic Escape in Multiple Myeloma Treated with T-Cell Redirecting Immunotherapies

Introduction T-cell redirecting therapies, such as chimeric antigen receptor T-cells (CART) and T-cell Engagers (TCE), have advanced the treatment of relapsed/refractory multiple myeloma (RRMM), however, most patients experience relapse. Relapse is often driven by antigen escape mechanisms, whereby tumor cells may either lose or modify the cell surface target (i.e., BCMA or GPRC5D); (Lee et al. Nat Med 2023). However, it is not known if antigen escape variants are selected (e.g., present at baseline) or acquired (e.g. induced by therapeutic pressure). Methods We developed a novel temporal workflow to time the acquisition of antigenic escape mechanisms as follows: DPClust was applied to sequential pre- and post- TCE/CART 80x whole genome sequencing (WGS) samples to reconstruct phylogenetic trees, assigning mutations into trunk (clonal and shared) or branch (sample-level unique). We then leveraged chemotherapy-associated mutational signatures (e.g., melphalan, SBS99) as temporal barcodes linked to a discrete clinical exposure. Since presence of SBS99 in bulk WGS is predicated by the requisite clonal expansion of a single exposed cell bearing a unique melphalan mutational barcode, detection of SBS99 in a higher-level branch than a GPRC5D or TNFRSF17 somatic escape event implies that the event was acquired following, and not before, HDM-ASCT. We applied this workflow to 24 pre-/post- CART/TCE samples from 11 patients. 10/11 patients received high-dose melphalan and autologous stem cell transplant (HDM-ASCT), 7 had anti-BCMA, 1 had anti-GPRC5D, and 3 had both T-cell redirecting therapies. Results 7/11 patients exhibited genomic alterations in TNFRSF17 or GPRC5D post-therapy that were undetectable at baseline using bulk WGS. In total, we could demonstrate that 4/7 patients had antigen escape events that were not present before HDM-ASCT, indicating that these mutations were acquired post-therapy. Each case is described below. Patient MM-17 was treated with both anti-BCMA CART and TCE, with samples investigated pre-CART, post-CART and post-TCE. The latter sample showed presence of a monoallelic deletion 16p in a 1st-level branch, which was undetectable in the pre-CART/TCE, but became clonal post-TCE. From this cluster, two 2nd-level branches emerged with two independent mutations in TNFRSF17 (cancer cell fraction; CCF 90% and CCF 10% respectively). Because SBS99 was only detected in the trunk, these two were acquired, and not present before HDM-ASCT (and therefore TCE). Patient MM-15, treated with anti-BCMA TCE, exhibited a clonal TNFRSF17 mutation coupled with a monoallelic loss of TNFRSF17 (1st-level branch), which were not detectable before treatment but clonal after TCE. SBS99 in the trunk reveals that both BCMA events were not present at the time of HDM-ASCT. Patient MM-19 was treated with both anti-GPRC5D and anti-BCMA TCE. This patient showed two 1st-level branches, with one of them containing a biallelic GPRC5D loss and presence of SBS99 within the trunk again suggesting that loss of the antigen occurred following melphalan exposure. A final patient was exposed to tandem HDM-ASCT followed by bendamustine (which causes a similar mutational profile to SBS99) and later anti-GPRC5D TCE treatment. At relapse, a deletion encompassing GPRC5D was identified as part of 1st-level branch cluster. Profound convergent evolution occurred with 4 distinct clusters arising from the cluster carrying the deletion, each with its own second hit on GPRC5D. Here, the SBS99-like signature was truncal, such that all GPRC5D antigen escape subclones were not originally present at initial diagnosis. Looking at these alterations across the CoMMpass data set we did not find any of these events. Longitudinal tracking of these mutations using droplet digital PCR in 4 patients consistently showed that genomic events promoting antigen escape were not detectable during the initial months of therapy but began to emerge nearly 1 year post therapy initiation. Conclusion Our findings reveal that antigen escape mechanisms in RRMM are often acquired post-therapy rather than pre-existing at baseline. This diminishes the rationale for baseline diagnostic testing for predicting treatment outcome and highlights the importance of monitoring for on-treatment emergence of resistance, as well as highlighting the need to investigate which patients are at a higher risk of developing these events under therapeutic pressure.