MDS-associated SF3B1 Mutations Enhance Pro-Inflammatory Gene Expression in Patient Blast Cells

Abstract

Two factors known to contribute to the development of Myelodysplastic Syndrome (MDS) and other blood cancers are (1) somatically acquired mutations in components of the spliceosome and (2) increased inflammation. Spliceosome genes, including SF3B1, are mutated at high frequency in MDS and other blood cancers; these mutations are thought to be neomorphic or gain-of-function mutations that drive disease pathogenesis. Likewise, increased inflammation is thought to contribute to MDS pathogenesis; inflammatory cytokines are strongly elevated in these patients, with higher levels correlating with worsened patient outcome. In the current study, we used RNAseq to analyze pre-mRNA splicing and gene expression changes present in blast cells isolated from MDS patients with or without SF3B1 mutations. We determined that SF3B1 mutations lead to enhanced pro-inflammatory gene expression in these cells. Thus, these studies suggest that SF3B1 mutations could contribute to MDS pathogenesis by enhancing the pro-inflammatory milieu in these patients.

Graphical Abstract

Summary:

SF3B1 mutations enhance pro-inflammatory gene expression in blast cells, possibly contributing to an overall inflammatory milieu in MDS patients.

INTRODUCTION

Genes encoding components of the spliceosome are somatically mutated at high frequency in hematologic malignancies and at lower frequency in solid tumors [1–5]. In fact, spliceosome gene mutations are the most common class of mutation in patients with myelodysplastic syndrome (MDS) [6]. The spliceosome genes that are most frequently mutated in MDS are SF3B1, U2AF1, and SRSF2 [1–5]. The SF3B1 mutation frequency ranges from 5–7% in de novo MDS to ~80% in patients with MDS subtype Refractory Anemia with Ring Sideroblasts (MDS-RARS) [1–5]. In solid tumors, SF3B1 mutation frequencies range from ~2% in breast cancer to 15–20% of uveal and mucosal melanomas [1, 2, 7]. These somatically acquired SF3B1 mutations are heterozygous point mutations in the protein’s HEAT domains; the most common SF3B1 mutation found in MDS patients is SF3B1-K700E. SF3B1 mutations are thought to be gain of function or neomorphic driver mutations that promote neoplasia by altering pre-mRNA splicing, although the precise mechanism by which SF3B1 acts has remained elusive [1–5].

A hallmark of MDS is that hematopoietic stem cells fail to differentiate properly [8, 9], instead developing into immature blast cells that display aberrant lineage markers [10–13]. In MDS patients, blast cells represent a significant fraction of the marrow and peripheral blood cell pool, with blast cell percentages as high as 20% in MDS patients [8, 9]. In this study, we used RNAseq to interrogate gene expression and pre-mRNA splicing in blast cells isolated from MDS patients’ bone marrow, comparing patients with SF3B1-K700E mutations to patients without spliceosome mutations. As expected, we observed significant changes in pre-mRNA splicing and gene expression induced by the SF3B1-K700E mutation in patient blast cells. Of note, genes in pro-inflammatory signaling pathways were substantially upregulated by the SF3B1-K700E mutation in patient blast cells. This is consistent with the growing body of evidence that SF3B1 and other spliceosome gene mutations enhance inflammatory signaling [14–17].

MATERIALS AND METHODS

Ethics Statement and MDS patient samples

This study was approved by the National Jewish Health Institutional Review Board (IRB) and the Colorado Multiple Institutional Review Board. All subjects gave written informed consent. Bone marrow aspirate from patients with MDS was collected, total mononuclear cells were isolated by standard Ficoll procedures [18], and cells were cryopreserved in liquid nitrogen as part of an IRB-approved tissue banking protocol at the University of Colorado. Clinical parameters for MDS patients were collected through retrospective chart review as described previously [19]. Genotypes of spliceosome genes and other MDS-relevant genes were determined by capillary sequencing for patients diagnosed prior to 2014 and by using a next generation targeted 54 gene resequencing panel for patients diagnosed after 2014 as described [19].

RNAseq Analysis

The RNAseq procedure and analyses are described in the Supplementary Materials and Methods.

Flow cytometry analysis of patient blast cells

MDS bone marrow mononuclear cells were thawed, washed with ice-cold staining buffer (HBSS with 2% FBS), and Near IR Live-Dead staining was performed (Invitrogen, L10119) to exclude dead cells, according to the manufacturer’s instructions prior to staining. Cells were blocked with 2.5 μg Fc-block (BD Biosciences, 564219) for 15 minutes, then surface stained for 30 min at 4°C in staining buffer containing antibodies against human CD34 (BD Biosciences, 562577), CD123 (BD Biosciences, 564197), CD45 (BD Biosciences, 560566), and CD14 (BD Biosciences, 563420). Following surface staining, cells were washed in stain buffer, fixed for 20 min at room temperature using Cytofix/Cytoperm (BD Biosciences), washed in PermWash buffer (BD Biosciences) and stained with anti-human S100A8 (BD Biosciences, 566010) diluted in PermWash buffer at room temperature for 60 minutes. After washing in stain buffer, samples were analyzed on a FACSCelesta (BD Biosciences). Flow data were analyzed using FlowJo 10 (FlowJo, LLC). Statistical significance was assessed with unpaired t-tests using Prism software (GraphPad); significant differences were considered p<0.05.

Analysis of K562 cells

The source of wild type and SF3B1-K700E mutant K562 cells and description of the qPCR and ELISA analyses are described in the Supplementary Materials and Methods.

RESULTS AND DISCUSSION

The SF3B1-K700E mutation alters gene expression and pre-mRNA splicing in blast cells from MDS patients

We used flow cytometric sorting (gating strategy in Supplementary Fig. S1) to purify blast cells from seven untreated MDS patients. These included four patients with SF3B1-K700E mutations and three patients without mutations in SF3B1, U2AF1, or SRSF2 (demographics in Supplementary Table S1). RNA was isolated from sorted cells, and RNAseq analysis was performed. We used two independent pipelines to analyze differential gene expression (one using HISAT2 for mapping, one using STAR for mapping; see Supplementary Materials and Methods) and two independent pipelines to analyze differential splicing (one using DRIM-seq, one using rMATS; see Supplementary Materials and Methods).

As expected, SF3B1 mutation led to substantial changes in gene expression (Fig. 1A, Supplementary Table S2 Tabs 1 and 2) and isoform usage (Fig. 1B, Supplementary Table S2 Tabs 3–8). As observed in other cell types, SF3B1 mutation induced changes in gene expression and pre-mRNA splicing in many genes in patient blast cells. Despite the strong effect on gene expression, and perhaps because MDS is a heterogeneous disease, changes in splicing correlated better with SF3B1 mutation status than did gene expression changes (compare Fig. 1B to Fig. 1A). SF3B1-K700E altered multiple classes of pre-mRNA splicing events (Fig. 1C, Supplementary Table S2 Tabs 3–8). Even though rMATS reports many more alternative splicing events than DRIM-seq, the two pipelines report a similar frequency of alternative splicing events with skipped exons being the most common (Fig. 1C; note that DRIM-seq does not report mutually exclusive exon usage).

Fig. 1. The SF3B1-K700E mutation alters gene expression and pre-mRNA splicing in blast cells from MDS patients.

Panels A and B depict unsupervised hierarchical clustering of genes based on differential gene expression using HISAT2 mapping (A) or altered pre-mRNA splicing using DRIM-sq (B). WT 1–3 indicates three MDS patients without mutations in SF3B1, U2AF1, or SRSF2; K700E 4–7 indicates four MDS patients with SF3B1-K700E mutations. The schematics on the left in panel C depict different types of alternative pre-mRNA splicing, including exon skipping (ES), intron retention (IR), alternate 5’ splice site usage (A5SS), alternate 3’ splice site usage (A3SS), and mutually exclusive exons (MXE). The black lines indicate the canonical splicing event; the green lines indicate the alternative event. The pie charts depict the relative frequency of each type of alternative pre-mRNA splicing event identified using DRIM-seq (n=344 events total) or rMATS (14,397 events total). DRIM-seq does not report the mutually exclusive exon usage (MXE) class; rMATS does report this class of events.

SF3B1-K700E mutation alters splicing of largely distinct sets of genes in blast cells and CD34⁺ cells from MDS patients

To determine how the effect of SF3B1 mutation in blast cells compares to that in other cells, we compared SF3B1-K700E-induced splicing changes in blast cells to two prior studies examining the effect of SF3B1 mutations in CD34⁺ hematopoietic stem and progenitor cells [20, 21]. For a fair comparison, we re-analyzed these prior RNAseq studies using our DRIM-seq pipeline to identify pre-mRNA splicing changes induced by SF3B1 mutations, comparing MDS patients with SF3B1 mutations to MDS patients without any spliceosome mutation (Supplementary Table S2 Tabs 9 and 10). Interestingly, individual pre-mRNA splicing events that were altered by SF3B1 mutation in patient blast cells were largely distinct from those identified in the two prior CD34⁺ studies (Fig. 2A), with only 6% of the changes in isoform usage identified in blast cells also present in either of the CD34⁺ studies, although we note that overlap between the two CD34⁺ studies themselves also was fairly moderate. This is consistent with previous observations that U2AF1 mutations alter splicing of largely non-overlapping sets of genes in different cell types [22, 23] and that SF3B1 mutations exhibit only partial overlap of pre-mRNA splicing changes in different cell types [24].

Fig. 2. SF3B1 mutations alter splicing of different genes in patient blast cells and CD34⁺ stem cells.

In panel A, the Venn diagram depicts the number of alternative pre-mRNA splicing events induced by SF3B1 mutations in either patient blast cells (the current study), CD34+ cells in study #1 [20], or CD34+ cells in study #2 [21]. In all studies, the analyses compared cells with SF3B1 mutations to those without any spliceosome mutation using DRIM-seq. In panel B, the Venn diagram depicts the number of genes with altered isoform usage induced by SF3B1 mutations in these same three groups.

In contrast to the limited overlap in splicing changes induced by SF3B1-K700E in blast cells and CD34⁺ cells when considered at the individual isoform level, there was substantial overlap in splicing changes when analyzed at the gene level (Fig. 2B), with 53% (93 of 177) of genes altered in patient blast cells also undergoing alternative splicing in CD34⁺ cells. This suggests that specific genes may be particularly susceptible to SF3B1-K700E induced alternative splicing, even though individual splicing events may vary.

SF3B1-K700E mutation affects splicing of genes with “weaker” intronic splicing regulatory sequences

Several conserved sequences in introns regulate pre-mRNA processing. One key sequence that defines the 3’ end of introns is the polypyrimidine (pY) tract, which lies just upstream of the conserved AG dinucleotide. Upstream of the pY tract is the branch point sequence, which is recognized by the U2 snRNP and associated proteins including SF3B1 [25]. Introns with pY tract or branch point sequences that differ from the canonical sequences are more likely to be alternatively spliced.

In other cell types and cancers, genes that exhibit altered splicing in the presence of SF3B1 mutations have “weak” intronic regulatory sequences [26, 27]. To investigate if alternative splicing induced by SF3B1-K700E in blast cells from MDS patients likewise involves altered intronic sequences, we examined the pY tracts in genes in which SF3B1-K700E induced alternative 3’ splice site usage or exon skipping of the downstream exon. Compared to control introns in which SF3B1-K700E did not induce splicing changes, introns that led to altered splicing in the presence of SF3B1-K700E exhibited a shorter or weaker pY tract (Fig. 3A,B). This is evidenced by more adenosine and fewer thymidine residues present in the pY tracts of introns that undergo alternative splicing, particularly in the −20 to −30 positions, which is consistent with these introns containing “weaker” pY tracts.

Fig. 3. Identification of intron features that are altered in genes that undergo SF3B1-K700E-induced alternative pre-mRNA splicing.

Panels (A) and (B) depict sequence logo plots for genes with altered 3’ splice site usage (A) or exon skipping (B) induced by the SF3B1-K700E mutation compared to control introns in genes whose splicing is not affected by SF3B1 mutation (comparisons using DRIM-seq). Panel (C) depicts the predicted strength of the branch point in genes that undergo SF3B1-K700E alternate 3’ splice site usage compared to genes that are not affected by SF3B1 mutation. Branch point strength was calculated as outlined in [28]. A higher score indicates a stronger predicted branch point.

We also used a computational method [28] to identify and assess the strength of branch points in introns that underwent alternative 3’ splice site usage compared to control genes and found a moderate but statistically significant decrease in the predicted strength of those branch points (Fig. 3C). This also could contribute to SF3B1-K700E-induced alternative splicing in patient blast cells.

The SF3B1-K700E mutation increases expression of genes in pro-inflammatory signaling pathways

To determine what the global effect of SF3B1-K700E may be on cell function, we determined if changes in pre-mRNA splicing or gene expression affected particular signaling pathways. Changes in pre-mRNA splicing affected signaling pathways involved in various aspects of RNA processing including pre-mRNA splicing itself and transcriptional regulation (Supplementary Table S2 Tabs 11 and 12), which is consistent with prior studies in other cell types [20, 21]. While SF3B1 mutation decreased expression of many genes, pathway analyses of this decrease in gene expression identified only one pathway that reached statistical significance using either analysis pipeline: DNA repair (Supplementary Table S2 Tabs 11 and 12). Thus, SF3B1-K700E may induce a largely non-specific overall decrease in gene expression in patient blast cells; this may reflect the undifferentiated nature of these cells. In contrast, genes whose expression was increased by SF3B1-K700E mutation were overrepresented in signaling pathways that involved an enhanced inflammatory/host defense response (Table 1 and Supplementary Table S2 Tabs 11 and 12). These pathways included “innate immune response,” “immune response,” and “inflammatory response” as well as many other pro-inflammatory/pro-immune gene ontology (GO) categories. This increased gene expression in pro-inflammatory signaling pathways was identified using both RNAseq analysis pipelines (compare Supplementary Table S2, Tabs 11 and 12). Thus, we conclude that SF3B1-K700E strongly enhances pro-inflammatory gene expression in patient blast cells. Similar effects on pro-inflammatory signaling pathways also have been observed in CD34⁺ cells; for example, Dolatshad et al. [20] determined that several innate immune signaling pathways (including Lymphotoxin β receptor signaling and Toll-like receptor signaling) exhibited altered gene expression in CD34⁺ cells from patients with SF3B1 mutations.

Table 1.

Gene Ontology categories with genes overexpressed in SF3B1-K700E Blast cells with Benjamini corrected p<0.1

Pathways	# of Genes	P-Value
GO:0045087~innate immune response	25	7.70E-16
GO:0006955~immune response	20	4.24E-11
GO:0006954~inflammatory response	18	5.45E-10
GO:0042742~defense response to bacterium	10	5.37E-07
GO:0050729~positive regulation of inflammatory response	8	5.51E-07
GO:0006935~chemotaxis	9	1.56E-06
GO:0006898~receptor-mediated endocytosis	10	4.26E-06
GO:0030593ñeutrophil chemotaxis	7	5.05E-06
GO:0050776~regulation of immune response	9	2.54E-05
GO:0050832~defense response to fungus	5	2.92E-05
GO:0006968~cellular defense response	6	5.77E-05
GO:0007166~cell surface receptor signaling pathway	10	9.12E-05
GO:0050830~defense response to Gram-positive bacterium	6	2.59E-04
GO:0050715~positive regulation of cytokine secretion	4	5.97E-04
GO:0060333~interferon-gamma-mediated signaling pathway	5	0.001292
GO:0002430~complement receptor mediated signaling pathway	3	0.001916

Complete data including gene list in each GO category in Supplementary Table S2 Tab 11. GO analysis of data from HISAT2 gene expression mapping.

To confirm the enhanced inflammatory gene expression observed in SF3B1-K700E blast cells, we used flow cytometry to validate at the protein level expression of S100A8, which was upregulated in SF3B1-K700E blast cells at the mRNA level (Fig. 4A and Supplementary Table S2 Tabs 1 and 2). S100A8 is a key pro-inflammatory mediator that is elevated in the sera of MDS patients and may contribute to disease pathogenesis [29, 30]. We monitored S100A8 production in blast cells from 6 MDS patients with SF3B1-K700E mutations and 4 MDS patients without spliceosome mutations (demographic data in Supplementary Table S1). We found that S100A8 protein levels were increased in patient blast cells (Fig. 4B), consistent with the mRNA data. S100A8 levels also were increased in the CD34⁺ sub-population (Fig. 4C), although this did not quite reach statistical significance.

Fig. 4. The SF3B1-K700E mutation increases production of the pro-inflammatory mediator protein S100A8.

(A) depicts gene expression data from the RNAseq analysis (transcripts per million, TPM) for S100A8 in patient blast cells. The figure also depicts the results of flow cytometry analysis to monitor S100A8 protein levels in either patient blast cells (B) or in the CD34⁺ sub-population (C). These plots display S100A8 levels as gMFI. WT = samples from patients without a spliceosome gene mutation. K700E = samples from patients with a SF3B1-K700E mutation. (D,E) S100A8 was monitored at the mRNA level by qPCR or at the protein level by ELISA in K562 cells expressing either WT SF3B1 or SF3B1-K700E. The qPCR results were normalized relative to βactin with expression in WT cells defined as 1. Data represent mean, SEM.

Prior studies in other cell types have implicated an alternate isoform of the kinase MAP3K7 as a potential driver of enhanced inflammation induced by SF3B1 mutation [15, 16]. However, we did not detect an increase in this alternate isoform of MAP3K7 in SF3B1 mutated blast cells as assessed by RNAseq (Supplementary Table S2 Tabs 1 and 2) or qPCR (not shown). To determine if splicing changes in other genes might drive the increased gene expression in inflammatory signaling pathways, we used Ingenuity Pathway Analysis (IPA) to identify pathways with substantial changes in mRNA splicing induced by SF3B1 mutation (Supplemental Table 2, Tab 13–16). By comparing over-represented IPA pathways in our blast cell analyses and the two prior analyzed CD34⁺ studies, we identified many innate immune signaling pathways that exhibited altered splicing (Supplemental Table 2, Tab 17). For example, many genes in the Toll-like receptor signaling pathway exhibited altered splicing in both blast cells and CD34⁺ cells (Supplemental Table 2, Tabs 13–16), suggesting that multiple alternative splicing events may drive this increased inflammation. Consistent with this possibility are our prior studies that demonstrated that many genes in innate immune signaling pathways including TLR signaling pathways are unusually sensitive to perturbation of the spliceosome [31–33].

SF3B1-K700E drives pro-inflammatory gene expression in leukemia cell lines

MDS patients with SF3B1-K700E mutations are reported to have a lower IPSS-R score than MDS patients without these mutations [34], and we observed similar trends in our cohort (Supplementary Table S1). Therefore, to confirm that SF3B1-K700E was driving these pro-inflammatory gene expression changes, we examined the effect of SF3B1-K700E in a heterologous otherwise isogenic system. We quantified expression of a subset of these pro-inflammatory genes in K562 cells that either did or did not contain the SF3B1-K700E mutation. Cells expressing SF3B1-K700E had increased expression of six of seven pro-inflammatory genes tested including S100A8 (Figs. 4D, 5A–F). S100A8 protein levels also were increased in K562 cells with the SF3B1-K700E mutation (Fig. 4E). As a control, expression of a second housekeeping gene was not increased by SF3B1-K700E (Fig. 5G). Thus, SF3B1-K700E increases the expression of pro-inflammatory genes both in patient blast cells and in K562 cells. This suggests that the SF3B1 mutation is driving this pro-inflammatory gene expression change in patient blast cells.

Fig. 5. SF3B1-K700E mutations increase expression of pro-inflammatory genes in K562 cells.

Expression of the indicated genes was monitored by qPCR in K562 cells either carrying a SF3B1-K700E mutation (K700E) or wild type for SF3B1 (WT). All data were normalized so that expression in WT cells was defined as 1. qPCR data were analyzed relative to βactin. As a control, the data also were compared to a second control housekeeping gene, Gapdh (panel G). Data represent mean, SEM.

Conclusions

In conclusion, MDS-associated SF3B1 mutations induce substantial alternative pre-mRNA splicing and gene expression changes in patient blast cells. Strikingly, we observed that blast cells from MDS patients with SF3B1-K700E mutations exhibit increased expression of genes in pro-inflammatory signaling pathways including S100A8, which has been implicated in MDS pathogenesis [29]. Our prior studies demonstrated that inhibition of SF3B1 weakens inflammatory readouts [16, 31]. Conversely, we and others have shown that MDS-associated spliceosome mutations including those in SF3B1 lead to enhanced NFκB activity and pro-inflammatory gene expression [14–17]. These prior studies suggest that alternative splicing of many genes in innate immune signaling pathways may drive these differences, although the precise details remain to be elucidated. Our current study extends these findings to SF3B1-K700E patient blast cells. Blast cells represent a significant fraction of the cell pool in marrow and peripheral blood in these patients, and thus this enhanced cellular inflammatory response could contribute to an inflamed milieu in these patients. MDS patients are inflamed [35–38], and increased levels of some inflammatory markers such as IL-6 and IP-10 correlate with shortened survival [39]. Of note, other MDS-associated mutations lead to enhanced inflammation [35–38], and the current study suggests that SF3B1 mutations should be added to the growing list of MDS-associated mutations that enhance inflammation in these patients.

Supplementary Material

fS1

Fig. S1. Gating Strategy for Fluorescence-activated cell sorting of blast cells from MDS patients. The figure depicts the gating strategy for the isolation of patient blast cells. Cells were selected, singlets were chosen, live cells were selected by Near IR live dead staining, and blast cells were defined as SSC-H low and CD45 intermediate.

ABBREVIATIONS

AML

Acute Myeloid Leukemia

FACS

Fluorescence-activated cell sorting

MDS

Myelodysplastic Syndrome

pY Tract

Polypyrimidine tract