Proteogenomic Analysis of SF3B1mut CLL Reveals Altered Protein Signaling Cascades and Metabolomic Pathways

Although mutation in the U2 snRNP spliceosome component SF3B1 has been appreciated as a common driver of chronic lymphocytic leukemia (CLL) for over a decade, we still know little of its mechanism of action. Prior transcriptome-based characterizations have demonstrated widespread changes in alternative splicing in SF3B1^mut CLL. We hypothesized that protein level characterization could potentially yield fresh insights on alternative expression of proteins and on the impact of altered signaling pathways in patients bearing this mutation. To this end, we assembled CLL specimens from a cohort of 105 patients enrolled on the German CLL Study Group (GCLLSG) trials CLL8 and CLL11, collected prior to therapy. Of these, 52 patients expressed SF3B1^wt and 53 patients harbored clonal SF3B1^mut (CCF>0.8); 25 of these were the K700E mutation. Additionally, we characterized these samples by WES (available for 97 patients) and RNA-Seq (available for 96 patients).

Proteomes and phosphoproteomes from all 105 samples were analyzed using a multiplexed isobaric-tagged (TMT) high resolution mass spectrometry (MS) workflow. Briefly, cells were lysed and proteins were digested into peptides with LysC and trypsin. Protein yield per sample ranged from 20 μg to 300 μg. Peptides were labeled with TMT-11 reagent and combined into 11 multiplexes while accounting for input material availability and randomizing for SF3B1 mutation, IGHV status and ATM/TP53 mutation. To enable cross TMT-plex comparison, we constructed a common reference consisting of 40 samples with highest protein amount and an equal representation between SF3B1^wt and SF3B1^mut samples. Each plex was fractionated into 24 fractions by high pH reversed phase separation and 500 ng protein per fraction were analyzed by MS (Orbitrap Exploris, Thermo Fisher Scientific). For phosphoproteome analysis, peptides were pooled into 12 fractions and enriched by IMAC prior to MS analysis. Spectra were analyzed by Spectrum Mill v7.1 (Broad Institute) and searched against a human reference proteome database supplemented with non-canonical ORFs supported by Riboseq evidence of translation (nuORFdb v1.0). Patient specific proteins containing single amino variants, indels, and splice isoforms, derived from WES and RNA-Seq, were included in the search database.

We obtained a total of 10,419 proteins, with a median of 8,882 proteins per TMT-plex and 6,969 proteins that were quantifiable across all samples. The phosphoproteome analysis identified 32,410 sites; while each TMT-plex yielded a median of 19,380 phosphosites per plex, 5,810 were quantifiable across all patients. We integrated proteome data with patient specific mutation calls, RNA-Seq and copy number alteration (CNA) data using Panoply, a cloud-based platform for proteogenomic data analysis (Broad Institute). We confirmed the impact of chr11 and chr13 deletion as well as chr12 amplification on protein expression.

We readily identified peptides spanning the SF3B1^mut amino acid regions in corresponding samples. By unsupervised clustering, samples were grouped by SF3B1 status on the CNA and proteome level, a trend not observed based on RNA or phosphoproteome analysis. Gene set enrichment analysis revealed enrichment of TNFA and PI3K signaling in the SF3B1^wt clusters, while the SF3B1^mut associated clusters showed a strong association with metabolism changes. This analysis was highly concordant with a parallel interrogation comparing proteins expressed in SF3B1^wt CLL versus SF3B1^mut CLL. Here 1,955 proteins and 1,737 phosphosites were significantly altered (5% FDR). Proteins involved in B cell receptor and MAP kinase signaling were highly expressed in SF3B1^wt samples, while again amino acid metabolism pathways as well as oxidative phosphorylation were upregulated in SF3B1^mut samples.

To more deeply investigate the role of metabolism pathways in SF3B1^mut CLL, we subjected 6 independent CLL samples to MS-based metabolomics analysis. Inosine (P=0.029) and guanosine (P=0.047) were lower in the SF3B1^mut extracts (P=0.001), while acylcarnitines (P=0.0004) were increased, indicating fatty acid oxidation as a preferred energy source in SF3B1^mut CLL.

Altogether, our study reveals fresh evidence of altered energy metabolism in SF3B1^mut CLL. Ongoing studies focus on defining the mechanism by which SF3B1^mut exerts these metabolic changes and novel therapeutic opportunities afforded by these findings.