AFFILIATIONS
1 Dana Farber Cancer Institute, Boston, Massachusetts, United States
2 Institut Universitaire du Cancer de Toulouse - Oncopole, Université de Toulouse, Toulouse, France
3 University of Washington, Seattle, Washington, United States
4 Dana Farber Cancer Institute, Boston, Massachusetts, United States
5 Dana Farber Cancer Institute, Boston, Massachusetts, United States
6 Institut Universitaire du Cancer de Toulouse - Oncopole, Université de Toulouse, Toulouse, France
7 Institut Universitaire du Cancer de Toulouse - Oncopole, Université de Toulouse, Toulouse, France
8 Dartmouth Hitchcock Medical Center, Lebanon, New Hampshire, United States
9 Dana Farber Cancer Institute, Boston, Massachusetts, United States
10 Dana Farber Cancer Institute, Boston, Massachusetts, United States
11 Institut Universitaire du Cancer de Toulouse - Oncopole, Université de Toulouse, Toulouse, France
12 Dana Farber Cancer Institute, Boston, Massachusetts, United States
13 Dana Farber Cancer Institute, Boston, Massachusetts, United States
ABSTRACT
15% of newly diagnosed MM patients relapse within 36 months despite being classified as standard risk. Although the IMS/IMWG CGS identifies the majority of genomically high-risk(HR) cases, a subset of functional HR cases remains undetected. To identify biological correlates of high-risk beyond CGS, we performed integrated WGS and RNAseq in a novel NDMM cohort enriched for early relapse and shorter survival.
We assembled a WGS and RNAseq dataset comprising 626 NDMM patients deliberately enriched for extremes of outcomes. This cohort included three groups: Group 1, PFS <18 months without del(17p) or t(4;14)(n=249); Group 2,PFS >36 months without these lesions(n=208); Group 3, PFS<18 months with del(17p) or t(4;14)(n=169). Findings are independently validated in 1108 additional patients with matched WGS and RNAseq data from multiple studies after excluding CGS HR cases.
Application of the CGS to the Group1 (MM with PFS<18 months lacking del17p and t(4;14)), demonstrated that only 20% were classified as HR, leaving the majority of early relapses genomically unexplained. Despite the absence of canonical HR lesion, these patients exhibited increased genomic instability(GI) identified by higher Genomic scar score(GSS), total mutations, and shorter telomere compared with low-risk patients (Kruskal-Wallis p<2.2e-16, 6.4e-06, and 3.7e-07). APOBEC and focal LOH/chromosomal instability were increased in both HR groups (adj. p<0.001). Rather than harboring recurrent hotspot alterations, Group 1 accumulated non-recurrent genome-wide abnormalities, explaining why lesion-based classifiers failed to identify these biologically aggressive cases.
We examined DNA repair pathways as a potential determinant of GI. LR patients(Group2) preferentially utilized high-fidelity repair through classical non-homologous end joining (c-NHEJ; p<1e-05), whereas both HR groups relied predominantly on error-prone alternative end joining and microhomology-mediated repair. Across individual patients, utilization of error-prone repair pathways correlated directly with GSS(R=0.34-0.43, p<1e-06) while high-fidelity repair showed an inverse relationship. These findings identify altered DNA damage repair as a reproducible biological hallmark of HR MM rather than simply a consequence of GI.
We asked whether additional DNA-based features could improve identification of these CGS-negative high-risk patients. A composite classifier including of gain1q, del8p, del16p, del16q, along with GSS, and genome integrity pathway alterations (≥2 events = high risk) significantly stratified OS (p<0.0001). However, performance was substantially decreased in the independent dataset (p=0.0058), suggesting that DNA alterations alone incompletely capture HR.
We therefore investigated RNA features unique to CGS-negative early-relapsing patients. Compared with LR patients, patients in Group 1 with no CGS-defining lesions, showed significantly diverged expression profiles. Their transcriptome was enriched in genome integrity programs including DNA repair and sister chromatid separation pathways (adj p=2.1e-04). From these analyses, we derived a MM-specific chromosome instability index comprising of 45 genes (CIN45) differentially expressed between Groups 1 and 2 after accounting for genomic background. Amongst the CGS-negative patients, CIN45 alone identified patients with significantly inferior survival (HR=3.59,OS p<0.0001) and was independently validated in 390 CGS-negative patients for both PFS and OS (HR=3.44,p<0.0001). In multivariate analysis RNA-based classifier remained an independent predictor of outcome.
Finally, integration of both DNA and RNA-based classifier provided the greatest prognostic discrimination. Combining the DNA composite with CIN45 identified a distinct DNA-high/RNA-high subgroup with ultra high-risk disease with worst outcome. Importantly, CIN45 provided independent information within each DNA-defined group (OS and PFS p<0.0001 in our and the independent dataset).
High-risk myeloma lacking CGS-defined lesion is characterized by genomic instability and defective DNA repair rather than recurrent genomic alterations. While DNA-based metrics improve risk prediction, they incompletely capture this biology and misclassify cases. We here demonstrate role for transcriptomic changes in influencing disease behavior and CIN45 provides novel independent, validated prognostic information beyond CGS, supporting integrated DNA and RNA profiling for improved risk assessment
DISCLOSURES
Mehmet Kemal Samur, PhD: BMS - Research Funding; K36 Therapeutics, CodingBio, Abbvie, Neuberg Center for Genomic Medicine - Consultancy (Includes expert testimony)