Splicing factor 3b subunit 1 (Sf3b1) haploinsufficient mice display features of low risk Myelodysplastic syndromes with ring sideroblasts

Background The presence of somatic mutations in splicing factor 3b subunit 1 (SF3B1) in patients with Myelodysplastic syndromes with ring sideroblasts (MDS-RS) highlights the importance of the RNA-splicing machinery in MDS. We previously reported the presence of bone marrow (BM) RS in Sf3b1 heterozygous (Sf3b1+/−) mice which are rarely found in mouse models of MDS. Sf3b1+/− mice were originally engineered to study the interaction between polycomb genes and other proteins. Methods We used routine blood tests and histopathologic analysis of BM, spleen, and liver to evaluate the hematologic and morphologic characteristics of Sf3b1+/− mice in the context of MDS by comparing the long term follow-up (15 months) of Sf3b1+/− and Sf3b1+/+ mice. We then performed a comprehensive RNA-sequencing analysis to evaluate the transcriptome of BM cells from Sf3b1+/− and Sf3b1+/+ mice. Results Sf3b1+/− exhibited macrocytic anemia (MCV: 49.5 ± 1.6 vs 47.2 ± 1.4; Hgb: 5.5 ± 1.7 vs 7.2 ± 1.0) and thrombocytosis (PLTs: 911.4 ± 212.1 vs 878.4 ± 240.9) compared to Sf3b1+/+ mice. BM analysis showed dyserythropoiesis and occasional RS in Sf3b1+/− mice. The splenic architecture showed increased megakaryocytes with hyperchromatic nuclei, and evidence of extramedullary hematopoiesis. RNA-sequencing showed higher expression of a gene set containing Jak2 in Sf3b1+/− compared to Sf3b1+/+. Conclusions Our study indicates that Sf3b1+/− mice manifest features of low risk MDS-RS and may be relevant for preclinical therapeutic studies. Electronic supplementary material The online version of this article (doi:10.1186/s13045-014-0089-x) contains supplementary material, which is available to authorized users.


Background
Myelodysplastic syndrome (MDS) is a heterogeneous group of hematopoietic stem cell disorders characterized by peripheral blood (PB) cytopenias, dysplastic bone marrow (BM), and increased risk of transformation to acute myeloid leukemia (AML). Within MDS, refractory anemia with ring sideroblasts (RARS) is a low-grade disease characterized by anemia, erythroid dysplasia, and the presence of 15% or more RS [1]. Some patients with RARS also present with marked thrombocytosis (RARS-T), a form of myelodysplastic/ myeloproliferative neoplasm (MDS/ MPN) associated with mutations in JAK2, TET2, and MPL genes [2][3][4][5]. The presence of RS is a key pathologic criterion for the diagnosis of both RARS and RARS-T. RS are erythroblasts with an abnormal localization of mitochondrial iron which appears in the shape of a blue ring by light microscopy. Studies investigating the mechanisms of RS formation in MDS implicated the mitochondrial genes ALAS2 and ABCB7 based on the gene expression differences detected in CD34-positive cells of RARS and RARS-T patients compared to healthy individuals [6,7]. The discovery of recurrent somatic mutations in splicing factor 3b, subunit 1 (SF3B1), a component of the RNA splicing machinery in approximately 60% of RARS and 82% of RARS-T patients opened a new area of study in MDS [8][9][10][11][12].
SF3B1 is a core component of the U2 small nuclear ribonucleoprotein (U2 snRNP). The function of SF3B1 is to recognize the 3' splice site at the intron-exon boundaries of pre-nascent RNAs. SF3B1 protein interacts with the 3′splice-site recognition of U2AF65 and other splicing factors such as SF3B14 to facilitate the successive steps of RNA splicing [13,14]. Although SF3B1 has been associated with MDS-RS, the biological role and the functional consequences of the genetic alterations in this gene on the pathogenesis of MDS-RS have not been fully elucidated. We previously reported that a mouse model characterized by Sf3b1 haploinsufficiency (Sf3b1 +/− ) have RS in the BM [15]. Isono et al. generated the SF3B1 +/− mice by replacing 4 exons of Sf3b1 (chromosome 1qC1.2) with a neo-cassette to investigate the interaction of Sf3b1 protein with the polycomb group of proteins. In 2005, they reported that Sf3b1 −/− mice were embryonic lethal whereas Sf3b1 +/− mice survived and exhibited several skeletal abnormalities [16]. However, the long-term dynamics of the hematologic phenotype of this mouse model was not analyzed. The diagnosis of human MDS is strictly based on blood counts, BM morphology, and cytogenetic criteria. Similarly, the criteria established by the Mouse Models of Human Cancers Consortium are also only weighted on PB counts and morphologic features. It is for this reason that we focused our investigation on the long term PB and BM morphologic characteristics of Sf3b1 +/− mice to help establish if this mouse model displays features of MDS and can therefore serve as a robust mouse model to study human RARS and RARS-T and a platform to test new therapies.

Genomic analysis of Sf3b1 mice
Embryos of the Sf3b1 mice were purchased from RIKEN. Mating of Sf3b1 mice was conducted in-house at the Cleveland Clinic. All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of the Cleveland Clinic.
None of the Sf3b1 +/− or Sf3b1 +/+ mice died immediately after birth and no obvious skeletal abnormalities were noted. There were no reported early deaths in either cohort. A total of 78 mice were analyzed (Sf3b1 +/− /Sf3b1 +/+ = 33/45). There were no homozygous Sf3b1 −/− mice. Tissues from tail and toes were taken in the first 10 days of life and used as a source of genomic DNA. PCR analysis showed that Sf3b1 +/− mice carried 2 PCR products: a wild type (WT) band at 1.5 kb and knock-out (KO) band at 0.9 kb as shown for mice # 1, 2, 4, and 7 in Additional file 1: Figure S1.
Hematologic findings of Sf3b1 +/− mice Mouse models of MDS demonstrate specific features resembling human MDS disease albeit at variable time points [17]. This fact underlines the importance of long term follow-up of mouse models to accurately capture diseaserelated events. We examined the standard hematologic parameters of Sf3b1 +/− (n = 5) and Sf3b1 +/+ (n = 5) starting from 6 months of age every month. After 6 months of age, fertility of breeding pairs dropped dramatically. No progeny was produced by mice of this age or older. The mechanism for this decline is unclear at present and it is under active investigation.
In terms of Hgb levels, Sf3b1 +/− mice tend to have lower values compared to Sf3b1 +/+ at 6,8,9,11, and 12 months of age. Statistically significant difference was noted at month 11 (6.97 g/dL ± 1.60 vs 10.04 g/dL ± 0.73; P = 0.008) of age ( Figure 1B). As expected, the trend of the RBC values paralleled the trend of the Hgb levels with statistical significance being reached at 11 (5.96 M/uL ± 1.02 vs 8.28 M/uL ± 0.48; P = 0.008) and 12 (4.52 M/uL ± 1.30 vs 6.09 M/uL ± 0.82; P = 0.027) months of age ( Figure 1C). PLT counts increased at month 6 until month 12 of age with a significant difference at month 10 (731 K/ uL ± 105.36 vs 579 K/uL ± 92.66; P = 0.008) ( Figure 1D). We also observed that after 12 months of age, some of the mice (n = 3) started to show a decline in overall activity characterized by reduced movements and difficulty walking which culminated in death a few weeks later. Two of the deaths were in the Sf3B1 +/+ group while 1 occurred in the Sf3b1 +/− cohort.
Since somatic heterozygous mutations in SF3B1 were also identified in a specific cohort of chronic lymphocytic leukemia (CLL) patients [18] we also measured and analyzed the leukocyte counts of the mice. The leukocyte compartment was primarily enriched with lymphocytes. However, the values were variable over time in both mice groups (data not shown). Moreover, mast cells were also evaluated in BM cells derived from Sf3b1 +/− (n = 2) and Sf3b1 +/+ (n = 2) by performing immunohistochemistry for CD117 (c-Kit) (Additional file 2: Figure S2). Mast cells noted as CD117 positive cells were rare and scattered and no difference was detected between both groups.

Histological examination of bone marrow, spleen, and liver
We first examined and compared the BM cellularity between the Sf3b1 +/− and Sf3b1 +/+ mice. H&E stain showed trilineage hematopoiesis with an adequate number of megakaryocytes in both groups of mice (Additional file 3: Figure S3, panel A). No difference in the number of BM cells was also observed at the end of the study between Sf3b1 +/− and Sf3b1 +/+ mice (57.9 ± 9.5 vs. 60.2 ± 8.4; P = 0.74) (Additional file 3: Figure S3, panel B). We next examined the morphology of BM cells derived from Sf3b1 +/− and Sf3b1 +/+ mice. BM (2-3 × 10 5 ) cells were spotted on cytospin slides and stained with Wright-Giemsa. BM cells from Sf3b1 +/− showed dyserythropoietic features including nuclear budding or nuclear irregularity ( Figure 2, red arrows) similar to what is observed in human MDS. Similar features were also noted in slides stained with Prussian blue (Additional file 4: Figure S4, black arrows). We originally reported the presence of rare RS in BM slides from Sf3b1 +/− mice [15]. We confirmed this observation by performing Prussian blue staining on fresh BM cytospin slides and finding occasional RS ( Figure 3, black arrows) in the BM of Sf3b1 +/− mice while BM cells from Sf3b1 +/+ only showed iron accumulation in histiocytes. RS were noted in several BM slides as shown in Additional file 5: Figure S5, black arrows.
The spleen and liver from both groups of mice were also dissected, measured, and histopathologically examined at the end of the study. Spleen and liver weights were compared between Sf3b1 +/− (n = 4) and Sf3b1 +/+ (n = 3) (0.10 ± 0.02 vs. 0.08 ± 0.01; P = 0.08; 1.13 ± 0.15 vs. 1.36 ± 0.21; P = 0.14). Microscopic examination of the spleen tissues showed significant expansion in the red pulp of Sf3b1 +/− mice with finding of extramedullary hematopoiesis (EMH) with hematopoietic elements, increased megakaryocytes with hyperchromatic nuclei, increased hemosiderin deposits and signs of fibrosis ( Figure 4) but no hepatomegaly or microscopic abnormalities in the liver were noted (Additional file 6: Figure S6).

Figure 2
Bone marrow morphology in Sf3b1 +/− compared to Sf3b1 +/+ mice. Bone marrow (BM) cells were extracted by flushing femurs of Sf3b1 +/− (n = 5) and Sf3b1 +/+ (n = 5) in media supplemented with 10% fetal bovine serum. Cells (2-3 x 10 5 ) were washed and spotted on cytospin slides prior immersion in buffered Wright-Giemsa staining solution. Budding and irregular nuclei are indicated in red arrows and are also magnified in the right quadrants. This feature was also observed in slides of BM cells from Sf3b1 +/− mice subjected to iron staining (Additional file 4: Figure S4). The image is presented for 1 Sf3b1+/-mouse.
down-regulation [fold change (FC) = 0.75; P = 0.075, rank = 309] in Sf3b1 +/− vs Sf3b1 +/+ mice. Since this Sf3b1 +/− mouse model was originally developed to study the interaction of Sf3b1 protein and proteins of the polycomb (PcG) complex, we also evaluated the status of known PcG genes, finding a trend towards lower mRNA levels of Ezh2 (FC = 0.02; P = 0.185, rank = 1387). We also found higher mRNA levels of Bmi1 (FC = 1.69; P = 0.138, rank = 471), a component of the PcG repressive complex, which is involved in axial skeletal development. This is likely a consequence of repression of the Hox genes. Bmi1 has been associated with progressive loss of proliferative capacity of hematopoietic stem cells and anemia. In addition, gene expression analysis of genes important in MDS pathogenesis showed weak evidence for lower mRNA expression levels of Npm1 (FC = 0.01; P = 0.184, rank = 1363) and no evidence of changes for Asxl1 and Runx1 (FC = 1.25; P = 0.296 and FC = 1.21; P = 0.471) in Sf3b1 +/− vs Sf3b1 +/+ . Because in human MDS, SF3B1 clones are found in early hematopoietic stem cells, [19] we interrogated gene sets and genes related to hematopoietic stem cell function and signaling. In total 39 gene sets were selected from the MSigDB c2 collection (gene set results in Additional file 8: Table S2; gene results for members of the gene sets in Additional file 9: Table S3) and showed that hematopoietic receptors mainly expressed in myeloid cells like Trem1 and transcriptional factors involved in hematopoietic development like Ptsg2 (Cox2) were over-expressed in Sf3b1 +/− (FC = 2.80, P = 0.011 and 2.43, P = 0.028). Thrombospondin-1 (Thbs1), a glycoprotein involved in the in-vitro proliferation of megakaryocytes was also one of the highest ranked genes and was found to be overexpressed (FC = 2.67, P = 0.008). Haploinsufficiency of Nr4a1 and Nr4a3, two nuclear receptors expressed in hematopoietic stem and myeloid cells, has been shown to cause MDS/MPN and leukemic evolution in mice [20]. In patients with MDS carrying SF3B1 mutations, the risk of AML transformation is less compared to those with WT SF3B1. In this study, Nr4a1 was found to be over-expressed in Sf3b1 +/− mice compared to Sf3b1 +/+ mice (FC = 2.29, P = 0.038) but Nr4a3 was not detected. We also observed some evidence of down-regulation in Sh2b3 (Lnk) and Calr in Sf3b1 +/− (FC = 0.22, P = 0.261 and FC = 0.33, P = 0.193) compared to Sf3b1 +/+ mice. Mutations in both genes have been found in human MPNs. In addition, a Stat5 target gene set showed some evidence of increased expression in Sf3b1 +/− mice (gene set P = 0.064).
From our global gene set analysis of collections c1 through c7, we found 1 significant gene set in the human Figure 4 Histology of splenic tissues of Sf3b1 +/− compared to Sf3b1 +/+ mice. Spleens from Sf3B1 +/− and Sf3b1 +/+ mice were fixed in 4% formaldehyde/ PBS and embedded in paraffin. Sections were stained with Haematoxylin-Eosin and showed extramedullary hematopoiesis with all 3 hematopoietic elements, increased megakaryocytes with hyperchromatic nuclei (black arrows), increased hemosiderin deposits (blue arrows) and evidence of fibrosis. chromosomal location collection (c1), chr9p24 (gene set P = 0.00032). This gene set contained Jak2 and showed higher expression in Sf3b1 +/− compared to Sf3b1 +/+ mice. We also evaluated genes associated with mitochondrial function (Abcb7, Alas2, and Sod2), which may possibly explain the anemia phenotype in Sf3b1 +/− mice. Only Abcb7 showed a weak evidence of increased expression in Sf3b1 +/− mice (FC = 1.58, P = 0.096).

Discussion
MDS is a heterogeneous disease with a variety of clinical, morphologic, and biological features. Mouse models may provide helpful insight into the mechanisms whereby specific genetic alterations can contribute to disease pathogenesis and can serve as platforms to study therapies that may be useful to treat these diseases. In MDS, the Human Cancer Consortium Mouse Model Group established the set of criteria that defines a MDS mouse model. There are 3 main criteria including the presence of at least one PB cytopenia (anemia, neutropenia or thrombocytopenia), the presence of a maturation arrest in a non-lymphoid hematopoietic component demonstrated in the form of dysplasia, and the absence of criteria of a non-lymphoid leukemia [17,21]. Here we report the hematologic and some of the biologic characteristics of a mouse model with Sf3b1 haploinsufficiency. The Sf3b1 +/− mice demonstrated macrocytic anemia, thrombocytosis, dyserythropoiesis, RS and EMH in the spleen. These are findings clinically and pathologically observed in human RARS and RARS-T. We also observed that PLT levels were increased and the spleen was enlarged as demonstrated by EMH in Sf3b1 +/− which are important clinical features of RARS-T patients. In humans, somatic mutations in JAK2 have been associated with disorders characterized by increased number of PLTs like RARS-T and related MPNs [5]. Patients with JAK2 mutations are also frequently found to have an enlarged spleen and an evidence of EMH. Definitive evidence of RS in erythroid precursors were once again consistently identified although in small numbers in the Sf3b1 +/− and not in the Sf3b1 +/+ mice supporting our initial report that demonstrated rare RS in this mouse model. Somatic mutations in SF3B1 have also been found in 7-15% of CLL patients and associated with aggressive phases of the disease, relapsed and chemorefractory CLL [18]. The link between SF3B1 mutations and CLL pathogenesis remain unclear. In MDS, mutations have been associated with a better survival outcome and a lower rate of AML transformation. Interestingly, we noticed an enrichment of the lymphocyte compartment in our mouse model although the increase was variable over time. Studies of Sf3b1 haploinsufficiency identified a reduction of hematopoietic stem cell pool confined in the myeloid compartment. Our data differ from a recent paper where Sf3b1 +/− haploinsufficiency appears to only lead to an impairment in the stem cell function but does not lead to MDS features in the same mouse model [22]. Matsunawa et al. investigated the functional role of Sf3b1 in normal hematopoiesis in this mouse model describing that besides a decrease in the number of hematopoietic cells and a reduced capability of hematopoietic reconstitution, no features of MDS were observed. Based on their results there was no change in the number of WBC and PLTs and in the content of Hgb up to 44 weeks (Additional file 1: Figure S1, panel A) [22]. Morphologically, Matsunawa et al. did not detect any RS and any change in spleen size by weight estimation. Although, the same mouse model was used, there are key differences in the methodology that significantly affected the outcomes of both studies. Our current study aimed to study specifically the morphologic features of this mouse model using conventional routine techniques used in the assessment of clinico-pathologic features of human MDS and MPN and during long term follow-up. This is an important difference since some mouse models exemplified by Sall4 (14.5 months), Evi1/Evi1t (12 months), NPM-1 (6-18 months) and Arid4a (12-22 months) did not show their respective phenotypes until the mouse models were much older and had longer follow-up [17]. This is in keeping with human MDS, where the vast majority of patients are diagnosed at an elderly age with a median age of diagnosis of 71 years old [23]. Next Matsunawa et al. did not analyze specifically the dysplastic morphologies and no images of cellular morphology of the BM aspirates have been shown. The tabulated hematologic results presented in their study showed a lower percentage of erythroid cells in Sf3b1 +/− mice compared to Sf3b1 +/+ further supporting our findings (P = 0.07). The histomorphologic features of the spleen, a frequently affected organ in human RARS-T were also not studied in the prior study. Our study showed that the spleen of the Sf3b1 +/− was not just enlarged but displayed architectural changes consistent with EMH akin to patients with human RARS-T. The RNA-sequencing results also support the fact that Sf3b1 +/− mice have a pattern more close to low rather than to high-risk MDS. In human MDS, ASXL1 mutations have been found enriched in patients with high-risk rather than in low-risk MDS and are correlated with unfavorable outcomes and AML transformation. In addition patients with ASXL1 mutations carry concomitant RUNX1 mutations and lower incidence of SF3B1 mutations [24]. Studies in mice showed that Asxl1 haploinsufficiency leads to a reduced hematopoietic stem cell pool, decreased hematopoietic repopulating capacity, and mild features of MDS [25]. On the same line, mice expressing the RUNX1 frameshift mutation (S291fs) develop signs of MDS including excess of blasts and dysplasia of the erythroid compartment [26]. In our mouse model, we observed minimal changes in the expression levels of both Asxl1 and Runx1, factors traditionally associated with more inferior outcomes in patients with MDS further supporting the natural history of human RARS-T probably due to the fact that Sf3b1 +/− mice do not manifest a late stage higher risk MDS disease. Indeed we did not observe any increased in blasts percentage and any sign of AML development despite the long term follow-up.
Our clinicopathologic results are further supported by RNA sequencing analysis where we found an overexpression of Jak2 and a down-regulation of Sh2b3 and Calr mRNA levels consistent with what is observed in human RARS-T. In regards to RS we consistently identified RS in the BM of these mice by using two blinded independent hematopathologists and this is unlikely to be simply a matter of chance. Lastly, using the guidelines established by the hematopathology subcommittee of the Mouse Models of Human Cancers Consortium, [17,21], it clearly shows that this mouse model fulfills the criteria for an MDS mouse model (Additional file 10: Figure S7).

Conclusions
In conclusion, our current data show that Sf3b1 haploinsufficiency in mice causes biological and morphological features resembling low risk MDS patients with RS specifically RARS and RARS-T opening the possibility that this mouse model can be helpful in testing therapeutic approaches in low risk MDS.

Mice
All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of the Cleveland Clinic. Sf3b1 +/− mice were originally developed by Isono et al. [16] Cryopreserved embryos of Sf3b1 +/− mice were purchased from Dr. H. Koseki and Dr. K. Isono from the Center for Integrative Medical Sciences (IMS) RIKEN (Japan) in early 2012. Embryos were successfully implanted in foster mothers and rederived mice were genotyped.
Long-term evaluation of Sf3b1 +/− mice A total of 5 Sf3b1 +/− (3 females/ 2 males) and 5 Sf3b1 +/+ (3 females/ 2 males) were maintained on a regular diet. Blood was collected by retro-orbital puncture in heparinized tubes every month. Blood was diluted 1:1 with PBS containing 2.7 mM EDTA and standard blood parameters [leucocyte counts, mean corpuscular volume (MCV), red blood cells (RBC), hemoglobin (Hgb), and platelets (PLTs)] were measured using a Hemavet 950 FS analyzer (Drew Scientific Incorporation, Dallas, TX). Mice were sacrificed at the endpoint of the study and tissues were collected as following: femurs from 2 mice per genotype were submitted for hematoxylin and eosin (H&E) stain, BM from all mice was flushed with Iscove's Modified Dulbecco's media plus 10% fetal bovine serum using a 25-gauge needle syringe from femurs and evaluated for cell count using a Vi-Cell™ XR cell viability analyzer (Beckman Coulter, Brea, CA). The spleen and liver were also fixed in 4% formaldehyde/PBS and stained with H&E.

Histomorphological analysis and Prussian blue staining
BM cells from femurs of Sf3b1 +/− and Sf3b1 +/+ mice were flushed with Iscove's Modified Dulbecco's medium supplemented with 10% fetal bovine serum (FBS). Cells (2×10 5 ) were washed once with PBS supplemented with 2% FBS and spotted on cytospin slides before Wright-Giemsa and Prussian blue stains were performed using standard histopathology staining procedures. Spleen and liver from Sf3b1 +/− and Sf3b1 +/+ mice were fixed in 4% formaldehyde/PBS and embedded in paraffin before H&E staining.

RNA Sequencing (RNA-Seq) analysis Mapping
Total RNA was extracted from whole BM of 6-month-old female Sf3b1 +/− (n = 2) and Sf3b1 +/+ (n = 2) mice using NucleoSpin RNA II (Clontech Laboratories). PolyA cDNA was prepared from 3 μg of RNA and mouse RNAsequencing was run on Illumina HiSeq2000 by Otogenetics (Norcross, GA). 100 basepair paired-end RNA-sequencing reads were mapped to the mm10 RefSeq mouse transcriptome and spliceome by DNAnexus (http://dnanexus.com) using a Bayesian method where a read was mapped when its posterior probability of mapping exceeded 0.9. These filtered posterior probabilities were summed to generate fractional read counts per gene and per exon, with probabilities from splice-junction spanning reads counted for each relevant exon. We used rounded gene and exon read counts as inputs for our differential expression analyses.

Differential gene expression analysis
We used TMM [27] normalization and the voom-limma approach [28] from the R package limma version 3.17 with R version 3.0.1 in order to perform differential gene expression analysis for Sf3b1 +/− versus Sf3b1 +/+ samples. Before testing, we dropped all genes with read counts per million reads less than or equal to 1 in at least 2 samples to improve testing power while maintaining type I error rates. We used the lmFit function with empirical Bayes shrinkage to estimate fold changes, p-values and adjusted p-values obtained using the Benjamini-Hochberg method [29] for each filtered gene under the null hypothesis of common expression intensity across groups. Genes with adjusted p-values less than 0.10 were declared significant.

Differential exon usage analysis
We used the R package DEXSeq, version 1.6 (http://www. bioconductor.org/packages/release/bioc/html/DEXSeq.html), to perform differential exon usage analysis of Sf3b1 +/− versus Sf3b1 +/+ samples. DEXSeq uses a negative binomial (NB) distribution to model the exon read counts and shrinkage estimators to estimate the per-exon NB dispersion parameters. We defined a testable exon as one that had a total sum of at least 8 mapped reads across samples and was in a gene with no more than 70 exons. Before exon usage testing, we dropped any exons that were not testable or were in genes with less than 2 testable exons to improve testing power while maintaining type I error rates. We used the testForDEU function, which compares deviances from generalized linear model fits (assuming NB likelihood) to a chi-squared reference distribution, to estimate p-values and adjusted p-values obtained using the Benjamini-Hochberg method for each exon under the null hypothesis of common usage across groups. Exons with adjusted p-values less than 0.10 were considered significant. Logarithm base 2 fold changes (Sf3b1 +/− / Sf3b1 +/+ ) for each exon were estimated using the function estimatelog2FoldChanges.

Gene set differential expression analysis
We used CAMERA [30], Competitive Gene Set Test Accounting for Inter-Gene Correlation approach, as implemented in the camera function from the R package limma version 3.17.17, on TMM normalized and voom weighted expression data to test whether a set of genes was highly ranked relative to other genes in terms of differential expression, accounting for inter-gene correlation. We mapped mouse gene symbols to human gene symbols using the MGI mouse to human homology mappings (http://www. informatics.jax.org/homology.shtml). We used the MSigDB database [31] version 4.0 (http://www.broadinstitute.org/ gsea/msigdb/index.jsp), gene set collections c1 through c7 in our gene set analyses. For each MSigDB collection of gene sets, we ran the camera function to estimate p-values for the competitive null hypothesis that the genes in the tested gene set didn't show stronger average differential expression relative to all tested genes not in the gene set. Adjusted p-values were calculated on the gene-set p-values per collection using the Benjamini-Hochberg method to control for the number of gene sets tested within a collection. Any gene sets with adjusted p-values less than 0.10 were declared significant.

Statistical analysis
Comparison of hematologic parameters between Sf3b1 +/− and Sf3b1 +/+ mice were analyzed using two-sample Wilcoxon signed rank test and presented as mean ± standard deviations. Statistical analyses were performed using R (www.r-project.org). Data were considered statistically significant if the P value was ≤ 0.05.