Multiomics-guided ex vivo drug sensitivity testing in Relapsed/Refractory Acute Myeloid Leukemia Free

R/R AML carries significant therapeutic challenges and has a 5-year OS below 10%. The integration of multiomics profiling with ex vivo drug sensitivity (DS) presents an opportunity to develop a more precise approach to the management of AML. Ex vivo drug screening with patient samples allows sensitivity and resistance analysis of anti-cancer drugs prior to translating to the bedside. In this study, we aim to integrate multiomics and clinical data to identify biological signatures that predict response to 64 approved/investigational anti-cancer ex vivo akin to BEAT AML study (Bottomly et al, 2022).

Methods The clinical and genomic characteristics of 16 patients with R/R AML from Karmanos Cancer Institute were reviewed, including NGS and therapeutic regimens. Mononuclear cells obtained from the bone marrow or peripheral blood of R/R AML patients with circulating blasts were sent viably for robotics guided exvivo high-throughput drug sensitivity with 64 anti-cancer drugs. These drugs included kinase inhibitors, JAK inhibitors, MEK inhibitors, epigenetic modulators, HDAC inhibitors, BET inhibitors, anti-apoptotic pathway drugs, and p53 activators. Other drugs were hypomethylating agents, immunomodulatory drugs, and drugs targeting oncogenic drivers or DNA damage repair. Gene expression was calculated using log₂TPM (transcripts per million). Cell viability was measured with CellTiter-Glo and expressed as fraction of DMSO control median. Per drug, 10 concentrations were used and DS defined in terms of IC₅₀ (half maximal inhibitory concentration). IC₅₀ thresholds were generated per drug. Patients were tagged as responders/non-responders based on having a DS score below/above threshold, respectively. All samples were also subjected to whole exome and whole trancriptome including micro RNA sequencing.

Results 10 of 16 samples had sufficient luminescence signal and viability after sampling and sequencing and were subsequently included for analysis. The median age was 58 (32-75 yrs), with 6 female patients. AML subtypes included 8 de novo AML and 2 secondary AML. Majority of patients (7/10) had abnormal cytogenetics. Patients received a median of 5 lines of treatment. The most common mutations in the cohort were DNMT3A (3), FLT3 (4), and IDH2 (3).

Among the 64 drugs, the median number of responders per drug was 5 (range 2-5). The median number of genes overexpressed in responders was 1916, and 1129 were underexpressed in non-responders. The median number of overlapping genes was 1009. To identify biological signatures of drug sensitivity, overlapping genes were identified and filtered for target genes relevant to each drug. 22 drugs were found to have target genes which were upregulated in responders and downregulated in non-responders.

Patient samples that was sensitive to navitoclax (3/10) had upregulation of BCL2L1 and was associated with post-transplant relapse clinically (OR 8.4, p=0.18). DNMT3A mutation suggested response to navitoclax and KPT-330 (OR 8.4, p=0.18; OR 4.2, p=0.5, respectively). Associations were also found between FLT3 mutation or short relapse latency among navitoclax responders (OR 4.2, p=0.5; each respectively).

FLT3 mutants were associated with ponatinib sensitivity which was linked to FGFR3/FGFR4 upregulation in responders (OR 4.2, p=0.5). Furthermore, samples of patients who relapsed quickly were sensitive to eprenetapopt, wherein responders had upregulation of target gene MDM2 (OR 2.7, p=0.6). All patient samples were highly sensitive to Panobinostat. 9/10 samples also were highly sensitive to CDK-IN-2.

Our cohort was enriched in hsa-miR-15a-5p, hsa-let-7a-5p (targets MYC/BCL-XL, enhance apoptosis resistance), miR-181 family (targets HOX/MEIS) and hsa-miR-182-5p, (targets MCL1 in apoptotic escape). To add, we found 5 novel miRNAs that target cell cycle arrest and promote therapy resistance.

Conclusion Our study demonstrates that integrating multiomics profiling with exvivo DS testing can potentially identify meaningful biological signatures predictive of therapeutic response in R/R AML patients. Crucially, this proof-of-concept results supports a precision-medicine framework combining genomic insights with functional drug screening, might offer actionable options for a patient group that is otherwise limited to supportive care alone. Further validation in larger cohorts of specific targets and translating and repurposing drugs post exvivo DS through potential clinical trials are ongoing.