Clonal trajectories and cellular dynamics of myeloid neoplasms with SF3B1 mutations

To the Editor:

Molecular lesions, either structural cytogenetic anomalies or gene mutations of known significance, occur in a step-wise fashion and are morphologic and prognostic markers in myeloid neoplasms (MNs). Current World Health Organization (WHO) classification categorizes patients with myelodysplastic syndrome (MDS) with ringed sideroblasts (RS) ≥15% or ≥5% RS with an SF3B1 mutation (SF3B1^MT) as MDS-RS [1], a subtype in which the presence of SF3B1 mutations confers a favorable prognosis [2–6]. Although most patients with SF3B1^MT have a classic phenotype, there is considerable heterogeneity within this subcategory including individual diversion of the originally favorable clinical phenotype. The disappearance of RS can be observed during the disease course of MNs, suggesting that new cellular shifts, due to the acquisition of additional lesions cooperating with/ or suppressing SF3B1^MT might go along with evolution to acute myeloid leukemia (AML). Such courses are likely a result of the type and configuration of SF3B1^MT, which potentially impact the disease course.

Inspired by a seminal work describing SF3B1^MT as a disease-defining molecular lesion in terms of survival and independent risk of clonal evolution [6], we looked closer at our preliminary observations suggesting that important biological clues can be extrapolated from the molecular associations of SF3B1^MT and their clonal architecture in the context of MNs [7]. To that end, we reviewed clinical and molecular annotations of patients with MNs with the intent of dissecting the SF3B1 mutatome and describing whether the clonal rank (ancestral/dominant vs. sub-clonal/secondary) might alter the phenotypic cell trajectories. We identified 209 SF3B1^MT in about 6% of MN patients (209/3673). Clinical and molecular data of the cohort were collected at The Cleveland Clinic Foundation and retrieved from publicly available datasets (refer to Supplementary Material) for comparisons’ purposes. The demographic, clinical and cytogenetic characteristics are summarized in Table S1. Using variant allele frequencies (VAFs) estimation that was previously confirmed by PyClone method [7, 8], SF3B1^MT were categorized as: (i) dominant (SF3B1^DOM, n = 96, 46%), (ii) secondary (SF3B1^SEC, n = 68, 33%) and as expected, due to the lack of resolution of the bioinformatics methods deployed, sub-categorization included transitional category of (iii) “co-dominant” (SF3B1^COD, n = 45, 21%). Schematic representations of these configurations are described in Fig. 1A. Despite the different hierarchical assignments with respect to other concurrent mutations, SF3B1^MT VAFs in the 3 groups did not statistically vary (P = 0.07; Fig. S1). In terms of survival, SF3B1^COD had a similar median OS to that of SF3B1^SEC (24.4 vs. 21.9 mo.) and a poorer outcome vs. SF3B1^DOM (41.1 mo.; P = 0.03; Fig. 1B). In terms of disease phenotypes, SF3B1^COD cases had no significant association with disease-type or cytogenetics vs. SF3B1^DOM and SF3B1^SEC (Table S2). The mutational profile of SF3B1^COD resembled that of SF3B1^SEC, while differences were found compared to SF3B1^DOM with a significant association with ASXL1, DNMT3A, RUNX1 and TET2 mutations (Fig. S2A, B). Although some mutations (DNMT3A, 22%; TET2, 20%; RUNX1, 11%) were preferentially associated with SF3B1^COD (Fig. S3A), they had no prognostic power (Fig. S3B–D). Therefore, we focused on non-ambivalent hierarchical SF3B1^MT (mutations with >5% difference in VAFs; ΔVAF > 5%) as these showed higher degrees of distinction and a more stringent representation of the chronology of mutations defined as SF3B1^DOM vs. SF3B1^SEC.

Fig. 1. Characteristics, mutational associations and prognostic implications of clonal events in SF3B1 mutant myeloid neoplasms.

A Fish plots and pie charts depicting different examples of SF3B1 mutations clonal status and its secondary mutations in the case of SF3B1 dominant (SF3B1^DOM), SF3B1 co-dominant (SF3B1^COD) and SF3B1 secondary mutations (SF3B1^SEC). B Kaplan–Meier curves showing the overall survival of cases with SF3B1 mutations per clonal status (dominant, secondary, co-dominant). The curves represent the percent survival of patients with dominant, co-dominant and secondary SF3B1 mutations. Comparisons of survival in dominant versus secondary SF3B1 and co-dominant versus dominant and secondary were done. Levels of statistical significance was calculated by P values. C Scatter plot describing the percentage of ringed sideroblasts in patients with SF3B1 dominant (blue) vs. secondary (green) mutations. Levels of statistical significance was assessed by P values using Mann–Whitney U test. D Bar graph showing the frequency (in percentage) of a panel of myeloid gene mutations co-occurring with SF3B1 mutations in dominant vs. secondary status. E Univariate analysis showing the odds ratio representing the strength of the association of SF3B1 mutations with other gene mutations. Levels of statistical significance is indicated in purple and orange colors (P < 0.0001 and P≤ 0.05) using Fisher’s Exact test. F Multivariate analysis showing the odds ratio representing the strength of the association of SF3B1 mutations with other gene mutations. Levels of statistical significance is indicated in purple and orange colors (P < 0.0001 and P≤ 0.05) using Fisher’s Exact test. G Kaplan–Meier curves showing the impact of TET2 mutations (TET2^MT) on SF3B1 mutant cases per hierarchical SF3B1 mutational configuration: SF3B1 dominant (SF3B1^DOM) vs. secondary (SF3B1^SEC). Levels of statistical significance was calculated by P values. H Kaplan–Meier curves showing the impact of the presence of FLT3 mutations (FLT3^MT) on SF3B1 dominant (SF3B1^DOM) cases. Levels of statistical significance was calculated by P values. Single cell-DNA sequencing revealing the clonal trajectories of one sample with (J) ancestral/founder SF3B1 mutation and another sample with secondary/subclonal SF3B1 mutation (I). Different colors represent different mutations.

As compared to SF3B1^SEC, SF3B1^DOM were enriched in MDS (33 vs. 18%; P = 0.007), particularly with RS (27 vs. 9%; P = 0.006) while less present in sAML (12 vs. 23%; P = 0.03) (Table S3, Fig. S4). Eighty-percent of SF3B1^DOM and 85% of SF3B1^SEC occurred in the elderly (age ≥ 60 years). No differences in sex distribution and hematological parameters were reported except for bi-cytopenia (52 vs. 36%; P = 0.01), more dysplastic myeloid cells (53 vs. 29%; P = 0.01) and bilineage dysplasia (47 vs. 26%; P = 0.02, Table S3) in patients with SF3B1^SEC vs. SF3B1^DOM. In contrast, the bone marrow of SF3B1^DOM patients was more often normocellular (46 vs. 29%; P = 0.02) and enriched with RS (Table S3); the latter also correlated with higher SF3B1^MT VAFs (Fig. 1C, Fig. S5). Interestingly, although patients with SF3B1^SEC had shorter OS than those with SF3B1^DOM (21.9 vs. 41.1 mo.; P < 0.03; Fig. 1B), when dichotomized according to different SF3B1^MT VAF ranges (>40%, 20–40%, <20%), no significant OS differences were observed (Fig. S6A) even when the SF3B1^MT clonal ranks were considered (Fig. S6B–D).

Therefore, we investigated the prognostic importance of distinct molecular associations in SF3B1^DOM and SF3B1^SEC.The analysis of the mutational burden showed that patients with SF3B1^DOM had fewer numbers of mutations per individual compared to those harboring SF3B1^SEC [1.0 (95/96) vs. 2.6 (179/68)]. The frequencies of mutations in SF3B1^DOM vs. SF3B1^SEC are depicted in Fig. 1D. In total, 274 concurrent somatic mutations were detected, of which 4% included co-hits in other RNA-splicing factor genes (PRPF8, SRSF2, U2AF1, ZRSR2; Fig. S7A, B and S8A–E) as recently described [9]; and of which 75% had a normal karyotype (Fig. S7C). Targeted deep sequencing of the most recurrently mutated genes in MNs revealed that concurrent myeloid mutations were more associated with SF3B1^SEC than SF3B1^DOM, suggesting that this increased accumulation of concomitant mutations might reflect a more pronounced susceptibility to clonal expansion. Indeed, univariate analyses showed significantly higher odds of mutations in TET2 (41 vs. 11%; P < 0.0001), RUNX1 (31 vs. 7%; P < 0.0001), ASXL1 (18 vs. 2%; P = 0.0005), BCOR/BCORL1 (15 vs. 5%; P = 0.05), DNMT3A (18 vs. 6%; P = 0.02) and STAG2 (13 vs. 3%; P = 0.01) in individuals with SF3B1^SEC vs. SF3B1^DOM (Fig. 1E). We then applied multivariate analysis to confirm the independent relationships between SF3B1 clonal status and other distinct genomic associations and found that ASXL1, DNMT3A, IDH2, NF1, STAG2, TET2 and RUNX1 significantly shaped the molecular profile of SF3B1^SEC (Fig. 1F). Moreover, mutations preceding SF3B1^SEC mainly affected lineage-restricted genes associated with repression of erythroid programs (RUNX1, 12%), cohesin complex (STAG2, 9%), transcriptional corepressors (BCOR/ BCORL1, 7%), terminal monocytic differentiation (TET2, 7%), chromatin remodeling (ASXL1, 6%) and leukemogenesis (DNMT3A and FLT3, 4% each) (Fig. S9). Thereby, we investigated whether specific concurrent mutations can impact the survival outcomes of SF3B1^MT MNs and further distinguish SF3B1^DOM vs. SF3B1^SEC subsets. Our analysis revealed markedly unfavorable outcomes in SF3B1^SEC compared to SF3B1^DOM when particular genes, e.g., TET2, co-existed (11.7 vs. 96.1 mo., P = 0.01; Fig. 1G) while the presence of FLT3^MT contributed to dismal prognosis only in SF3B1^DOM (8.4 vs. 47.6 mo., P = 0.01; Fig. 1H). The extent of prognostic variations between both SF3B1^MT configurations suggests that new lesions might represent key contributors to the distinct clinical course of SF3B1^MT MNs. The association between TET2 and SF3B1 mutations has been previously seen in MNs and in vivo functional studies described an exacerbated anemic phenotype when present [10]. The prognostic implication of FLT3 mutation concurrence with SF3B1 mutation might highlight a distinct subgroup of patients with more proliferative features.

Given the fact that inferring clonal architecture from bulk sequencing might not accurately describe that SF3B1^MT can exist in a secondary configuration and acquired during clonal evolution, we applied single-cell DNA sequencing to an MDS case with complex molecular architecture and dissected its clonal trajectory. We found a subclonal SF3B1^MT (p.K700E) to an ancestral TET2^MT (p.Y1618X), that coexisted in clones carrying TET2 and JAK2 (p.V617F) and TET2, JAK2 and TP53 (c.276+1G>A) (Fig. 1J). Additionally, to confidently assign the mutations we performed serial DNA sequencing on 21 patients’ specimens to deeply define the clonal trajectories of SF3B1^MT in MNs. The study of the evolution of SF3B1^MT VAFs (Fig. 2A–J) demonstrated that a founder SF3B1 clone (p.E622D) remained as such and increased in VAF (case #1, 2; the latter confirmed also by single-cell DNA sequencing; Fig. 1I), became secondary due to acquisition of other hits (cases #3) or codominant (case #11). Of note, we found that in one case, SF3B1^DOM disappeared at the 2nd time point (case #4) possibly suggesting that the composition of the bone marrow at that time-point contained only a minimal fraction of cells carrying SF3B1^MT. In addition, SF3B1^COD clones remained co-dominant (cases #5, 6), became secondary (case #7) or dominant (case #8). The SF3B1^SEC clone remained secondary (case #12), disappeared (case #13, 16) or became dominant due to the disappearance of other founder mutations (case #14, 15). The clonal evolution elucidated by serial samples as well as bone marrow reviews, unveiled that an evolving AML phenotype was accompanied by the acquisition of novel lesions (FLT3, PTPN11, RUNX1, TET2, WT1 in cases #17–19) or emerging cytogenetic abnormalities [inv(3) and del(5q) in cases #20, 21] driving transformation.

Fig. 2. The clonal trajectories of SF3B1 mutations.

Serial analysis of patients with myeloid neoplasia and SF3B1 mutations. DNA specimens from 21 samples with SF3B1 mutations were analyzed at 2 or 3 time points. Targeted sequencing panel for myeloid genes was applied to all samples. Line graphs indicate variant allele frequency in percentage for mutations considered somatic. Red, blue and green colors indicate mutations at first time point, second and third time point, respectively. Dx diagnosis, VAF variant allele frequency, mo months.

The mutational rank within clonal hierarchy is different from just assigning clonal burden and correlates with the outcomes of the mutations including SF3B1. Our study supports that specific disease phenotypes and clinical outcomes of patients with MNs are inferred by the clonal hierarchy of SF3B1^MT, not only its VAF%, with respect to other founder or subclonal myeloid events. This distinction allows for an even higher level of precision in predicting clinical outcomes of patients with SF3B1 mutations. Both, methodological feasibility of VAF based clonal ranking and predictive value of SF3B1 clonal hierarchy assessment in MN will need further confirmation. Additionally, we deciphered the clonal trajectories of SF3B1^MT and uncovered key events driving leukemic evolution in SF3B1^MT low-risk MDS. Hence, SF3B1^MT hierarchical configuration can potentially delineate the fate of SF3B1^MT MNs and further define this newly proposed disease-entity.