Mutation Burden in Multiple Myeloma Is Captured By Gene Expression Profiles

In the current study we sought to determine whether mutation burden (MB) is reflected in gene expression patterns. We generated 389 gene expression profiles and 311 mutation profiles from 182 cases, split into a training set of 97 and test set of 84. Copy number variants (CNV), rearrangements (TX) and short variant mutations (SV) were identified with FoundationOne Heme (F1) and U133Plus2.0 (u133) gene expression data, GEP70 and GEP80 risk scores, subgroup and CNV calls provided by Signal Genetics. 145 Kegg pathways, transcription factor binding sites (TFbs) for 195 TF mapped to u133 genes and 1161 F1 features and u133 GEP signatures/scores were evaluated for enrichment of defined genes. u133 data for normal plasma cells (n=22), MGUS (n=44), smoldering MM (n=12), relapsed MM (n=76), and MM cell lines (n=42) were derived from GSE31162. Newly diagnosed MM samples came from GSE31162 (n=584), GSE19784 (n=321), GSE15695 (n=247) and E-MTAB-317 (n=233). Of 593 genes assayed by F1, 293 were mutated at least twice. There were a total of 3454 mutations (average = 11, minimum = 3, and maximum = 37). KRAS was mutated in 40 tumors, while TP53 was mutated 45 times in 31 tumors. TP53 and 62 other genes had 2 or more unique mutations in a single tumor. A linear curve of MB exhibited a sharp upward inflection at 19 mutations. We sought to determine if GEP could identify characteristic features of MM flanking the inflection point. A training set of 97 (86 <= 15 MB and 11 >= 21 MB) and a test set of 84 (49 <= 15 MB, 28 >15 but < 21 MB and 7 >= 21MB) was produced. A mean ratio identified 576 genes exhibiting 2-fold higher expression in MM with high MB (hiMB) and 1617 genes from low MB (loMB). Notably, forty-four of the 293 mutated genes were in this list of genes. A geometric mean ratio of the two gene sets was then calculated for all samples. The mean of the resulting score (MB.2) was higher in MM with hiMB (1.66) than MM with loMB (-0.235) in the training set. MM with hiMB (0.329) had a higher MB.2 score than the group with intermediate MB.2 (0.158) and both higher than MM with loMB (-0.122). MB.2 was lowest in normal PC (-0.518) and progressively increased with disease progression: MGUS (-0.341), SMM (-0.308), MM (0.199), relapsed MM (0.334), MM cell lines (1.168).[SP1] 57% of the 182 cases harbored only SV mutations, 32% had SV and CNV, 32% SV and TX and 7% had SV, CNV and TX mutations. SV only mutations were present in 76% of MB.2 quartile 1 (MBq1) and 30% of MBq4. SV, CNV and TX mutations were present in 4% of MBq1 and 17% of MBq4. MB.2 was positively correlated with GEP70, GEP80, proliferation index, and TP53 target genes in MM and genes modulated by thalidomide and dexamethasone in PGx studies, in at least 6 of the 7 cohorts studied. The CD2 subtype, a myeloid classification and GEP70 low risk were significantly overrepresented in both MBq1 and GEP70q1 in all 7 cohorts. Conversely, the MF, MS, and PR subtypes, GEP70 and GEP80 high risk, as well as +1q, amp1q21, and del13q were significantly overrepresented in MBq4 and GEP70q4 in all 7 cohorts. MB.2 [SP2] genes derived from MM with loMB where enriched in 45 of 148 Kegg pathways. Notable were Hedgehog, Prostaglandin, Tx factors in cancer, HOX, MYB signaling, ephrin-B reverse signaling and embryonic stem cells. Five pathways related to B-cell biology were enriched. Mitotic cell cycle, integrin signaling, chromatin acetylation, ubiquitin ligation, and G1 to G1/S were underrepresented. MB.2 genes from MM with hiMB were enriched in TP53, lipid lysis, complement cascade, adherens junctions, Wnt regulation of CYR61, cyclins, prostaglandins, cell cycle, and MYC target pathways. Interferon signaling, TNF-NF-kB, EGFR, NOD, endoplasmic reticulum, ubiquitin ligation, Wnt-Hedgehog-NOTCH and BMP-SMAD modules were underrepresented. An enrichment of Rel and NF-kB TFbs was observed for genes negatively correlated with MB.2. and genes positively correlated with MB.2 and GEP70 were enriched for E2F and TP53 binding site. In conclusion, we show that MB can be captured by GEP in MM, that MB increases with disease progression, and pathways enriched by hiMB and loMB are different and may imply differences in pathogenesis as well as treatment.

[SP1]This is an important finding - suggest emphasizing more

[SP2]Starting with this para suggest referring to groups are low vs high mutation burden for improved readability.