SOCS1 mutation subtypes predict divergent outcomes in diffuse large B-Cell lymphoma (DLBCL) patients.

Suppressor of cytokine signaling 1 (SOCS1) is frequently mutated in primary mediastinal and diffuse large B-cell lymphomas (DLBCL). Currently, the prognostic relevance of these mutations in DLBCL is unknown. To evaluate the value of the SOCS1 mutation status as a prognostic biomarker in DLBCL patients, we performed full-length SOCS1 sequencing in tumors of 154 comprehensively characterized DLBCL patients. We identified 90 SOCS1 mutations in 16% of lymphomas. With respect to molecular consequences of mutations, we defined two distinct subtypes: those with truncating (major) and those with non-truncating mutations (minor), respectively. The SOCS1 mutated subgroup or the minor/major subtypes cannot be predicted on clinical grounds; however, assignment of four established gene-expression profile-based classifiers revealed significant associations of SOCS1 major cases with germinal center and specific pathway activation pattern signatures. Above all, SOCS1 major cases have an excellent overall survival, even better than the GCB-like subgroup. SOCS1 minor cases had a dismal survival, even worse than the ABC gene signature group. The SOCS1 mutation subsets retained prognostic significance in uni- and multivariate analyses. Together our data indicate that assessment of the SOCS1 mutation status is a single gene prognostic biomarker in DLBCL.

. Notably, the ratio of G/C (n=64) to A/T (n=12) in these substitutions was shifted towards a predominance of G/C mutations. This overrepresentation suggests that mutations arise through somatic hypermutation (see below). As an additional approach the base pair substitutions were analyzed using the SIFT score (sorting intolerant from tolerant) 1 which predicts whether an amino acid substitution affects protein function (Supplemental Table 1).
Analysis of somatic hypermutation. In total, 40.8% (31/76) of all point mutations were found in somatic hypermutation hotspot motifs 2-4 , whereas 13.1% of the sequenced SOCS1 region includes such somatic hypermutation motifs. When the type of mutation is put in the context of hotspot motifs, 23 were replacement and 8 silent mutations (ratio 3.4). Analysis by somatic hypermutation motifs showed that 32.9% (25/76) of the somatic hypermutation mutations were present in RGYW motifs. The guanine (G) in the RGWY motif represents 34 residues of 636 nucleotides in the SOCS1 coding region (5.3%). Moreover, 5.3% (4/76) targeted G of the DGYW and 8.0% (6/76) affected adenine (A) of the WA motif. The frequency of G within DGYW was 14 residues of SOCS1 coding region (636 bp; 2.2%, on both strands) and A within WA hotspot is represented with 35 residues (5.5%; on both strands). In summary, the somatic hypermutation mutation pattern in the SOCS1 gene in DLBCL samples was skewed towards G substitutions in RGYW hotspot.
Differences were not significant (P=0.774; Fisher's) and we also excluded an association of the rs149311 status with the SOCS1 mutation type (P=0.12; Fisher's).

SOCS1 mutations in the COSMIC database. The COSMIC database contained 67
individual SOCS1 mutation entries. These were derived from 49 samples composed of 31 patient samples and 18 cell lines (6 entries/samples contained no positional information). Regional analysis of SOCS1 mutant cases in comparison with the DLBCL cohort is provided in Figure 2 as well as Supplemental Table 2. Almost all cases (91.7%) had mutations that affected the region encoding the JAK-kinase domain and therein the majority of mutations directly affected the region encoding the SH2 subdomain (83%). In contrast, mutations affecting the 3' regions, encoding C-terminal domains such as the SOCS box (25%) or the recently discovered nuclear localization signal (NLS, 12,5%) were relatively rare (Supplemental Table 2). Nomenclature follows Human Genome Variation Society (HGVS, http://www.hgvs.org/mutnomen/; last accessioned Oct 1 th , 2012) and positional information refers to NM_003745 and ENSP00000329418 for DNA and amino acid, respectively. The SIFT (sorting intolerant from tolerant) score predicts whether an amino acid substitution affects protein function 1 . Abbreviations: AA, amino acid; c., affected position coding DNA; p., AA position; >, single base substitutions; _ range of changed sequence; del, deletion; dup, duplication; *, stop codon; fs, frame shift; red indicates mutations at somatic hypermutation motifs.

Supplemental Table 2. SOCS1 Mutation Frequency by Protein Domains
Abbreviations: AA, amino acids; Ref, supplemental reference; n, number of SOCS1-mutated cases; N, number of mutations in the indicated domain; COSMIC, Catalogue of Somatic Mutations in Cancer; DLBCL, diffuse large B-cell lymphoma; SH3, Src Homology 3 (XPpXP); PRR, proline rich-repeats contain diproline motifs PxxPxR (type I) and RPxPXXP (type II) and represent the defining determinants of the SH3 domain; JAK, Janus-kinase; KIR, kinase inhibitory region; ESS, extended SH2 subdomain; SH2, Src Homology 2; TEC, tyrosine kinase expressed in hepatocellular carcinoma; NLS, nuclear localization signal; SC, STAT-induced STAT inhibitor COOH-terminal; pY1007, phosphorylated tyrosine at position 1007. Symbols: % (n/N) percent of all mutations; [range %] percent of cases (per domain) with mutations that are predicted to encode for a C-terminally foreshortened SOCS1 protein. The range takes the spectrum of 5' mutational consequences into account (details see main paper). Briefly, the left number indicates a 'conservative' weighing where only the complete lack of C-terminally encoded domains is considered a deleterious event whereas the right number is derived from a more 'aggressive' weighing which also accounts for alterations in domain position or partial disruptions of domains. Here, ranges are provided by domain and a plot over the entire coding region is provided in Figure 2C of the main paper. Case-based distributions of transition to transversion ratios were compared using the Fisher's exact test. Abbreviations: TS, transition; TV, transversion; SHM, somatic hypermutation