A variant erythroferrone disrupts iron homeostasis in SF3B1-mutated myelodysplastic syndrome

See allHide authors and affiliations

Science Translational Medicine  10 Jul 2019:
Vol. 11, Issue 500, eaav5467
DOI: 10.1126/scitranslmed.aav5467

Loading up on iron

Myelodysplastic syndromes are clonal disorders of hematopoiesis with a variety of manifestations and potential complications. One category of myelodysplastic syndromes, those associated with ring sideroblasts, is known for causing iron overload even in the absence of transfusions, but the reasons for this are unclear. Bondu et al. discovered that a variant form of erythroferrone, a hormone that helps regulate bone marrow function, is produced in this disorder and results in deregulation of iron loading through suppression of the hormone hepcidin. Moreover, the variant erythroferrone is not present in healthy bone marrow cells, indicating its potential utility as a biomarker of clonal hematopoiesis.


Myelodysplastic syndromes (MDS) with ring sideroblasts are hematopoietic stem cell disorders with erythroid dysplasia and mutations in the SF3B1 splicing factor gene. Patients with MDS with SF3B1 mutations often accumulate excessive tissue iron, even in the absence of transfusions, but the mechanisms that are responsible for their parenchymal iron overload are unknown. Body iron content, tissue distribution, and the supply of iron for erythropoiesis are controlled by the hormone hepcidin, which is regulated by erythroblasts through secretion of the erythroid hormone erythroferrone (ERFE). Here, we identified an alternative ERFE transcript in patients with MDS with the SF3B1 mutation. Induction of this ERFE transcript in primary SF3B1-mutated bone marrow erythroblasts generated a variant protein that maintained the capacity to suppress hepcidin transcription. Plasma concentrations of ERFE were higher in patients with MDS with an SF3B1 gene mutation than in patients with SF3B1 wild-type MDS. Thus, hepcidin suppression by a variant ERFE is likely responsible for the increased iron loading in patients with SF3B1-mutated MDS, suggesting that ERFE could be targeted to prevent iron-mediated toxicity. The expression of the variant ERFE transcript that was restricted to SF3B1-mutated erythroblasts decreased in lenalidomide-responsive anemic patients, identifying variant ERFE as a specific biomarker of clonal erythropoiesis.


Myelodysplastic syndromes with ring sideroblasts (MDS-RS) are clonal hematopoietic stem cell (HSC) disorders with a prominent erythroid dysplasia of the bone marrow (BM) responsible for a macrocytic anemia. Mitochondrial iron accumulation and apoptosis of mature erythroblasts cause ineffective erythropoiesis (1, 2). In contrast to other MDS subtypes, patients with MDS-RS exhibit signs of systemic iron accumulation that is reflected by increased ferritin and nontransferrin bound iron concentrations even before they become transfusion dependent and subsequently develop parenchymal iron overload (3, 4).

Splicing factor gene SF3B1 is mutated in ~90% of MDS-RS, and the diagnosis is considered whenever the gene is mutated, although the percentage of RS is relatively low, between 5 and 15% (57). These mutations arise in the HSCs (810). Aberrant splicing events are reported in MDS and other SF3B1-driven cancers, including uveal melanoma and chronic lymphocytic leukemia (CLL) (1114). The selection of an alternative branch site resulting in the use of a cryptic 3′ splice site (ss) is the most frequently detected abnormality. Computational analysis revealed that most of the cryptic 3′ss are located upstream of canonical 3′ss at nucleotide distances that are not multiples of 3, suggesting that the aberrant transcripts would likely contain a premature termination codon (PTC) and be degraded by nonsense-mediated decay (NMD). It has been predicted that half of the aberrantly spliced transcripts in SF3B1-mutated cells are NMD sensitive and that the canonical isoforms are down-regulated (11). For instance, ABCB7 transcript encoding a mitochondrial transporter involved in the export of Fe-S clusters is aberrantly spliced and undergoes NMD. It is also possible that NMD-insensitive aberrant transcripts are translated into proteins with altered function (11). Globally, missplicing may contribute to defective messenger RNA (mRNA) production and deregulation of cellular pathways (15, 16). How these aberrant splicing events contribute to the disease phenotype and in particular to systemic iron overload is unclear.

In contrast with other MDS subtypes, MDS-RS are associated with lower expression of the iron homeostasis regulator, hepcidin, and as a consequence, the absorption of iron by duodenal enterocytes and its release from erythrophagocytic macrophages may be increased (1722). Inappropriately low hepcidin concentrations in MDS-RS could depend on the degree of ineffective erythropoiesis linked to impaired iron incorporation into heme because of mitochondrial iron trapping and/or to increased expression of hepcidin repressors (23). Growth differentiation factor 15 (GDF-15) and twisted gastrulation (TWSG1), two members of the transforming growth factor–β superfamily, have been proposed as pathological suppressors of hepcidin in ineffective erythropoiesis (24, 25). More recently, erythroferrone (ERFE), a C1q–tumor necrosis factor–related family of proteins (CTRP) member, has been described as a major erythroid regulator of hepcidin, involved in the pathological suppression of hepcidin in patients with β-thalassemia (26, 27).

In the present study, we identify a variant transcript of ERFE specific to SF3B1-mutated MDS. The expression of the variant ERFE is restricted to the erythroid lineage, and the variant contributes to increased concentration of ERFE protein, resulting in hepcidin suppression and systemic iron accumulation in patients. This variant appears to be a pertinent biomarker of clonal erythropoiesis with potential use for monitoring treatment response of anemic patients with SF3B1-mutated MDS.


Up-regulation of ERFE using an alternative 3′ss in SF3B1-mutated MDS

To investigate the mechanism of systemic iron overload, we established a cohort of 156 patients with lower-risk MDS, including 60 MDS-RS with single lineage dysplasia (SLD), 17 MDS-RS with multilineage dysplasia (MLD), 2 5q-syndrome, 17 MDS-SLD, 42 MDS-MLD, and 18 MDS with type 1 excess of blasts (EB1) (table S1). The Revised International Prognostic Scoring System score was very low in 15, low in 95, intermediate in 36, and high in 4 patients. Twenty-six genes commonly mutated in MDS were sequenced in the BM mononuclear cell (MNC) fraction. The SF3B1 gene was mutated in 94 patients with MDS, including 63 (67%) affected by an SF3B1K700E mutation. Among the 62 patients with no SF3B1 mutation, other splicing genes, SRSF2, U2AF1, or ZRSR2, were mutated in 27 cases, and no splicing gene mutation was observed in 35 cases. Some patients presented with two splicing mutations (fig. S1). We evaluated the consequences of SF3B1 gene mutations on gene expression and splicing by sequencing the transcriptome of the BM MNCs isolated from 27 patients in this cohort, including 21 with SF3B1MUT MDS, 6 with SF3B1WT MDS, and 5 healthy controls. Differential analyses of gene expression and splice junctions were conducted using DESeq2 (28). We detected 6343 genes as differentially expressed between SF3B1MUT MDS and SF3B1WT MDS with P < 0.05 (data file S1). Principal components analysis of gene expression profiles separated SF3B1MUT and SF3B1WT MDS (Fig. 1A). The differentially expressed genes were enriched in 73 specific gene ontology (GO) terms with an absolute log2 fold change (FC) > 1 and a Benjamini-Hochberg (BH)–adjusted P value < 0.05 (Fig. 1B). Genes with log2(FC) < −1 were involved in serine/threonine kinase signaling, apoptosis, myeloid differentiation, inflammation, and cell-cell adhesion, whereas those with a log2(FC) > 1 were involved in heme metabolism, erythrocyte differentiation, and iron homeostasis (Fig. 1C). We plotted the log2(FC) of 16 differentially expressed genes belonging to the IRON_ION_HOMEOSTASIS gene set (GO:0055072; and showed that the FAM132B/ERFE transcript encoding ERFE was increased (Fig. 1D). The FAM132B/ERFE transcript was similarly expressed in samples with an epigenetic TET2MUT/DNMT3AWT/SF3B1WT genotype compared to healthy controls or with a DNMT3AMUT/TET2WT/SF3B1MUT compared to DNMT3AWT/TET2WT/ SF3B1MUT genotype, suggesting that the deregulation of ERFE transcript expression was linked to SF3B1 mutation (table S2).

Fig. 1 Differential gene expression in SF3B1MUT MDSs.

(A) Biplot of the first two principal components (PCs) showing 45.4% of the variability within the data (PC1, x axis; PC2, y axis). (B) Volcano plot showing differentially expressed transcripts in SF3B1MUT BM MNCs compared to SF3B1WT BM MNCs. Fold change (FC) on the x axis and negative log10 of Benjamini-Hochberg (BH) corrected P values on the y axis. Dotted vertical and horizontal lines reflect the filtering criteria (FC < 0.5 or > 2.0; and BH-corrected P value < 0.05). The red dots represent differentially up-regulated transcripts, the blue dots represent differentially down-regulated transcripts, and the gray dots represent transcripts without differential expression. (C) GO enrichment analysis of differentially expressed genes with an absolute log2 (FC) >1 and a P value < 0.05 in SF3B1MUT MDS. The most down-regulated and up-regulated gene sets are represented. (D) Log2(FC) of the expression of 16 among the 96 genes of the IRON_ION_HOMEOSTASIS gene set (GO:005072) deregulated in SF3B1MUT MDS.

We then identified 1528 differentially expressed 5′ and 3′ junctions, including annotated 5′ donor and 3′ acceptor ss, ambiguous junctions, and canonical junctions with BH-adjusted P values ≤ 10−5 and absolute log2(FC) ≥ 1 (data file S2). These junctions allowed the hierarchical clustering of the 21 SF3B1MUT and 6 SF3B1WT MDS samples (Fig. 2A). After excluding differentially expressed canonical junctions, we then considered the 1147 alternative junctions, among which we identified 786 3′ acceptor junctions (68.5%), 176 5′ donor junctions (15.3%), and 185 ambiguous junctions (16.1%) attributed to either the alternative 5′ or 3′ss or both alternative 5′ and 3′ss (Fig. 2B). The analysis of distances between the alternative and canonical 3′ss showed that most of the alternative 3′ss (AG′) were located between −24 and −9 nucleotides (nt) preceding the canonical 3′ss (AG) (Fig. 2C). The proportion of alternative AG′ junctions in these novel splice variants generally increased in SF3B1 mutants and less frequently in SF3B1WT samples (Fig. 2D). To detect alternative transcripts that were likely to generate substantial amounts of variant proteins with a modified length, we applied a filter selecting transcripts with an additional stretch of nucleotides numbering in multiples of 3, a ratio of alternative junction coverage (AG′) to alternative and canonical junctions coverage (AG′ + AG) of more than 0.1, and an expression ratio of the alternative junction in SF3B1MUT versus SF3B1WT samples of more than 10 (fig. S2A). We obtained 66 alternative junctions in 63 genes (data file S3 and fig. S2B). These genes were involved in epigenetic regulation and transcription, mRNA processing and translation, intracellular transport, cell cycle and migration, signaling and apoptosis, mitochondrial metabolism, and iron homeostasis (Fig. 2E). Among the 66 alternative junctions, 29 were related to an in-frame insertion of 9 to 27 nt, and 26 of them were differentially expressed between SF3B1MUT and SF3B1WT MDS with an absolute log2(FC) > 0.3 (Fig. 2F). The FAM132B/ERFE gene was identified among the up-regulated genes with an alternative junction due to the use of a cryptic 3′ss located between exons 2 and 3 (chr2: 239,070,357 to 239,071,364) and no PTC. The aberrant transcript contained 12 additional nucleotides in the open reading frame and from here on is referred to as ERFE+12 (Fig. 2G). It was systematically detected in all SF3B1MUT MDS samples and represented a mean percentage of 24.8% of FAM132B/ERFE transcripts in SF3B1MUT MDS versus 0.2% in SF3B1WT MDS. The mutation pattern of samples expressing ERFE+12 is shown in fig. S2C. This indicates that the expression of ERFE+12 was strongly linked to the presence of a mutation in the SF3B1 gene.

Fig. 2 Differential splice junctions in SF3B1MUTMDSs.

(A) Hierarchical clustering and heat map of differential splice junctions between SF3B1MUT and SF3B1WT MDS samples. Values indicating percent usage of the differential splice junction versus all other junctions sharing the same ss are normalized as Z scores across patients and limited to a maximum of |Z| = 2. Rows are splice junctions with indicated types: acceptor (red), donor (blue), ambiguous (green), and differentially expressed canonical junction (gray). Columns are patients. (B) Venn diagram of the number of differential alternative 5′ donor and 3′ acceptor junctions in SF3B1MUT compared to SF3B1WT MDS. The overlapping area represents ambiguous junctions (n = 185). (C) Distances between the alternative (AG′) and canonical (AG) 3′ss within the 50 nt upstream of the AG plotted as a histogram. (D) Comparison of the expression of alternative junctions. The ratio AG′/AG′ + AG for each junction in SF3B1MUT (red) versus SF3B1WT (gray) MDS is shown. (E) Distribution of the biological functions of 63 genes affected by one or two aberrant 3′ss junctions located at <50 bases of the canonical 3′ss, an additional sequence multiple of 3 nt, a ratio AG′/(AG′ + AG) > 0.1 in SF3B1MUT samples, and an FC > 10. (F) Ratio of AG′/(AG′ + AG) in 26 genes whose expression was up- or down-regulated in patients with SF3B1MUT. (G) Sashimi plot of 3′ss canonical (*) and aberrant (**) junctions in the FAM132B/ERFE gene in three BM MNC samples, one SF3B1WT and two SF3B1K700E MDS. The ratio AG′/(AG′ + AG) is indicated in parentheses.

SF3B1MUT-restricted expression of alternative ERFE+12 transcript

To ascertain that expression of the ERFE+12 transcript is dependent on the presence of an SF3B1 mutation, we transfected the erythromegakaryocytic cell line UT7/erythropoietin (EPO) with a pLVX plasmid encoding a synthetic full-length SF3B1WT or mutant SF3B1K700E complementary DNA (cDNA). We then designed a sensitive fluorescent polymerase chain reaction (PCR) allowing the detection of ERFE+12 and ERFEWT transcripts as 162- and 150-nt fragments by capillary electrophoresis (table S3). Twenty-four hours after transfection, ERFE+12 transcript was detected in SF3B1K700E-transfected cells but not in SF3B1WT cells (Fig. 3A). To validate the splice pattern induced in the context of SF3B1 mutation in ERFE pre-mRNA, we performed a minigene splicing assay. An ERFE sequence of about 200 nt located on both sides of the cryptic 3′ss (AG′) was cloned in an Exontrap vector. The alternative junction in ENOSF1 gene (chr18: 683,395 to 685,920) cloned in the same vector was used as a control (13). These minigenes were transfected into the murine G1E-ER4 cell line in which the SF3B1K700E mutation was edited by CRISPR-Cas9 technology. As expected due to the lack of intron homology between species, sequencing the transcriptome of G1E-ER4 SF3B1K700E cell line or isogenic SF3B1WT cell line demonstrated that endogenous murine ERFE and ENOSF1 did not undergo alternative splicing (12). After transfection, the alternative 3′ss AG′ ERFE+12 was detected by capillary electrophoresis in G1E-ER4 SF3B1K700E cells but not in the isogenic G1E-ER4 SF3B1WT or parental cells (Fig. 3B). The usage of alternative 3′ss AG′ was detectable for ENOSF1 gene in the SF3B1WT cell line and became predominant in the SF3B1K700E cell line, suggesting that the mutation favored the usage of alternative AG′. Then, we performed a rescue experiment by adding a destabilization domain (DD) tag to the mutant SF3B1R625G allele in the Mel202 cell line, as previously described (29). The DD-tagged protein is stabilized by interaction with the DD ligand Shield-1 and is degraded upon Shield-1 withdrawal. In Mel202 clone 26 containing one DD-tagged SF3B1R625G allele, Shield-1 removal abrogated ERFE+12 expression. This supports a causal relationship between SF3B1 mutation and the aberrant splicing isoform of ERFE (fig. S3).

Fig. 3 SF3B1-dependent expression of 3′ss aberrant ERFE+12.

(A) Induction of ERFE+12 by expressing SF3B1K700E in human SF3B1WT UT-7/EPO cell line. Cells were transfected with a pLVX plasmid encoding a synthetic SF3B1WT or SF3B1K700E cDNA. Nontransfected (NT) UT-7/EPO cells are shown as a control. The canonical ERFE and aberrant ERFE+12 transcripts were detected by capillary electrophoresis of fluorescent PCR products. The x axis represents molecular size (nt for nucleotides) of PCR amplicons, and the y axis represents relative fluorescent units (RFU). The peak at 150 nt corresponds to the canonical transcript, whereas the peak at 162 nt refers to the alternative transcript due to cryptic AG′ usage. (B) Analysis of alternative AG′ and canonical AG usage of ERFE and ENOSF1 minigenes transfected into murine CRISPR-Cas9 SF3B1WT (clone 9.2) and SF3B1K700E (clone 5.13H) G1E-ER4 cells by fluorescent PCR. Transfected and NT parental G1E-ER4 cells are shown as controls. The peak at 135 nt corresponds to the transcript generated by a canonical AG usage, whereas the peak at 147 nt refers to the alternative transcript due to cryptic AG′ usage. (C) Detection of ERFE+12 depends on the presence of an SF3B1 mutant in MDS. BM MNC RNAs from three patients with SF3B1 mutations (SF3B1K700E, SF3B1H622Q, and SF3B1N626D), three with mutations in other splice genes (U2AF1Q157P, SRSF2P95H, and ZRSR2H191Y), and three with IDH1R132C, ASXL1dupG646, or no mutations were analyzed. ERFE+12/ERFE+12 + ERFEWT ratios are indicated. (D) Quantification of ERFEWT and ERFE+12 transcripts by RT-qPCR in BM samples depicted in (C). Results are expressed as normalized ratio quantities (NRQ) to ACTB and B2M housekeeping genes. (E) Detection of ERFEWT and ERFE+12 transcripts in SF3B1MUT or SF3B1WT erythroid progenitors or basophilic/polychromatic erythroblasts (Baso/polyE) in culture in comparison with BM MNC. (F) Analysis of ERFE transcripts in SF3B1WT diseases with ineffective erythropoiesis. The image shows peripheral blood (PB) MNC from one patient with MDS-RS with TET2 and SRSF2 and no SF3B1 mutations, samples from three patients with congenital sideroblastic anemias (two BM samples with ALAS2 mutation and one PB sample with GLRX5 mutation), and one PB sample from a patient with severe β-thalassemia (β-Thal). Samples were analyzed by capillary electrophoresis of ERFE PCR products. One PB sample from a patient with SF3B1K700E MDS was used as a positive control.

We then investigated the expression of ERFEWT and ERFE+12 in primary BM samples of patients with MDS using fluorescent PCR and real-time quantitative PCR (RT-qPCR) (table S3). Among 46 patients with lower-risk MDS, SF3B1MUT was present in 25 patients, including 20 with MDS-SLD-RS/MDS-MLD-RS, 3 with MDS-MLD, and 2 with MDS-EB1. By fluorescent PCR, ERFE+12 was detected in the BM MNC fractions of all SF3B1MUT MDS. ERFE+12 was not detected in any other cases of MDS with mutations in SRSF2 (n = 10), U2AF1 (n = 1), or ZRSR2 (n = 1), or in 9 patients with MDS with mutations in other genes, as shown in Fig. 3C and table S4. ERFEWT and ERFE+12 mRNA amounts measured by RT-qPCR were up-regulated in SF3B1MUT samples compared to any SF3B1WT samples with other splicing or epigenetic mutations or without recurrent mutation (Fig. 3D and table S4). In one patient with an SF3B1 monoallelic deletion and a G742D substitution on the remaining allele, the expression of alternative ERFE+12 exceeded that of the canonical ERFEWT (fig. S4). We then amplified primary erythroblasts from the BM CD34+ cells of two patients with SF3B1MUT and one patient with SF3B1WT MDS and showed that ERFE+12 was expressed in SF3B1MUT but not in SF3B1WT erythroid cells (Fig. 3E). This further supports the idea that ERFE+12 is related to SF3B1 mutation. SF3B1MUT MDS are characterized by the enrichment of the BM with erythroid cells. To investigate whether ERFE+12 could be detected in cells derived from erythroblastic BM with another genetic background, we collected samples from one patient with MDS-RS with 38% of BM erythroblasts, more than 15% of RS, a normal karyotype, and no SF3B1 mutation but one SRSF2 and one TET2 mutation; three patients with a congenital sideroblastic anemia due to an ALAS2 mutation/deletion or a GLRX5 mutation; one patient with a severe β-thalassemia; and one patient with SF3B1K700E MDS. By fluorescent PCR, ERFEWT was present, whereas ERFE+12 was not detectable in any sample except the SF3B1K700E MDS (Fig. 3F). This confirms that the onset of an aberrant ERFE+12 is dependent not on the amplification of the erythroid compartment or the presence of RS but on the presence of a mutant SF3B1.

Translation of ERFE+12 into a variant protein that represses hepcidin

Human ERFE encodes a 354–amino acid polypeptide. The addition of 12 nt in ERFE mRNA generates an ERFE variant (further referred to as ERFEVPFQ) containing a valine-proline-phenylalanine-glutamine (VPFQ) insertion immediately upstream of the collagen domain (Fig. 4A and fig. S5A). To investigate whether the mutant ERFEVPFQ protein was produced in vivo, we prepared cell lysates of erythroblasts derived in culture from the BM CD34+ cells of patients with SF3B1MUT MDS. Through liquid chromatography–tandem mass spectrometry (LC MS/MS) protein identification, we obtained several peptide-matching propositions including the ALHELGVYYLPDAEGAFR peptide (fig. S5B) already reported in public databases (, and we identified a cryptic peptide VPFQFGLPGPPGPPGPQGPP GPIIPPEALLK corresponding to the VPFQ insertion at position 108 of ERFE (Fig. 4B). This confirms that the alternative ERFE+12 transcript is translated into ERFEVPFQ in SF3B1MUT erythroblasts.

Fig. 4 Identification of ERFEVPFQ peptide by mass spectrometry and hepcidin repression by recombinant ERFEVPFQ protein.

(A) Amino acid sequence of ERFEWT and ERFEVPFQ peptides. VPFQ (red), collagen domain (blue). (B) Identification of a specific ERFEVPFQ peptide in erythroblast cell lysates by mass spectrometry using nano liquid chromatography coupled with a Q Exactive Plus mass spectrometer. The calculated peptide mass was 3227.650782 from 3+ ion observed at mass/charge ratio (m/z) 1076.890870 with measured Δ = 3.6 ppm, Mascot score = 17, expectation value = 0.069. HY, hydroxylated proline residues. (C) SDS–polyacrylamide gel electrophoresis and Coomassie blue staining of purified ERFEWT and ERFEVPFQ in reducing and nonreducing conditions. (D) Hep3B and HepG2 hepatocellular carcinoma cells were treated with purified recombinant human ERFEWT or ERFEVPFQ (2 μg/ml) for 16 hours. HAMP was quantified by RT-qPCR and normalized to HPRT. Data shown are means ± SEM of three independent experiments and represent an FC of hepcidin mRNA expression in ERFE-treated compared to untreated (CTRL) cells. Two-tailed Student t test for P values; ***P < 0.0001; **P < 0.001.

ERFE represses hepcidin in mice and contributes to pathological hepcidin suppression in patients with transfused and nontransfused β-thalassemia (26, 27). Whether ERFEVPFQ also repressed hepcidin was tested. For this purpose, recombinant proteins ERFEWT and ERFEVPFQ were produced in human embryonic kidney (HEK) 293F cells. SDS–polyacrylamide gel electrophoresis analysis of ERFEWT and ERFEVPFQ in reducing and nonreducing conditions demonstrated that the insertion of four amino acids does not change the molecular weight of the protein or its multimerization pattern, such that both proteins were predominantly found in high–molecular mass forms (>250 kDa) and in a trimeric form of ~200 kDa (Fig. 4C). Then, Hep3B and HepG2 hepatocellular carcinoma cells were treated for 16 hours with purified ERFEVPFQ or ERFEWT. Both proteins similarly caused a fivefold reduction in the expression of HAMP mRNA encoding hepcidin compared to controls (Fig. 4D). These data indicate that insertion of four amino acids upstream of the collagen domain did not affect the bioactivity of the protein.

Prediction of hyperferritinemia by overproduction of ERFE in SF3B1MUT MDS

We then measured the plasma concentration of ERFE in the training cohort of 156 patients with MDS and 20 healthy non–blood donor controls using a validated immunoassay (27). We first verified that the ERFE immunoassay was able to detect both ERFEWT and ERFEVPFQ. Human ERFE enzyme-linked immunosorbent assay (ELISA) detected similar amounts (1 μg/ml) of each recombinant protein in the supernatants of HEK293F cells transiently transfected with ERFEWT and ERFEVPFQ expression vectors. This established that both isoforms were detectable by ELISA. The mean concentration of ERFE in SF3B1MUT or SF3B1WT MDS was higher than that in 20 non–blood donor healthy controls (P < 0.0001; Fig. 5A). Among the MDS samples, the ERFE concentration was higher in SF3B1MUT (135.0 ± 72.5 ng/ml) compared to SF3B1WT (62.1 ± 36.7 ng/ml) MDS (P < 0.0001; Fig. 5A). High concentrations of circulating ERFE correlated with high expression of ERFE+12 transcript (Pearson test; P < 0.0001; fig. S6A). Consistently, the ERFE concentration was also higher in MDS-RS compared to all other World Health Organization (WHO) MDS subtypes (fig. S6B). Ferritin concentrations were significantly higher in patients with SF3B1MUT compared to patients with SF3B1WT (P < 0.0001; Fig. 5B). We also measured plasma hepcidin and confirmed that the concentration of hepcidin in SF3B1MUT MDS was significantly lower compared to SF3B1WT MDS (P = 0.031; Fig. 5C). We found no difference between SF3B1MUT MDS and healthy controls. The hepcidin/ferritin ratio was lower in SF3B1MUT than in SF3B1WT MDS (P < 0.0001; Fig. 5D) and also in MDS-RS compared to other WHO subtypes (fig. S6C), as a result of both a lower concentration of hepcidin and a higher concentration of ferritin in patients with SF3B1MUT. The plasma concentration of ERFE was inversely correlated to the hepcidin/ferritin ratio (Pearson test; P < 0.0001; r = 0.600; fig. S6D). Our analysis also highlights that an ERFE concentration above a threshold of 100 ng/ml repressed hepcidin more efficiently (fig. S6D). Serum EPO concentration was equally increased in patients with SF3B1MUT and SF3B1WT compared to normal values (5 to 36 U/liter; table S1), and although ERFE is regulated by erythropoietin in mice, we did not find any correlation between serum EPO and ERFE concentration (fig. S6E) (26). The increased concentration of plasma ERFE was associated with a more pronounced degree of ineffective erythropoiesis, as assessed by a significant increase in plasma concentration of soluble transferrin receptor (sTfR) in SF3B1MUT compared to patients with SF3B1WT MDS (P < 0.0001; fig. S6F).

Fig. 5 Increased plasma concentration of ERFE in patients with SF3B1MUT MDS.

Quantitative analysis was performed in plasma samples collected from 20 non–blood donor healthy volunteers (black), 156 patients with MDS including 94 SF3B1MUT (red) and 62 SF3B1WT (gray) representing the training cohort (A to D), and 55 patients with MDS including 42 SF3B1MUT (red) and 13 SF3B1WT (gray) representing the validation cohort (E to H). The graphs show quantification of erythroferrone (A and E), ferritin (B and F), hepcidin (C and G), and hepcidin/ferritin ratio (D and H). Results are expressed as medians and interquartile ranges (IQRs). The boxplots represent the median and the first and third quartiles, and the whiskers represent the lowest and the highest values still within the 1.5 IQR of the lower and upper quartiles. Mann-Whitney for P values.

To validate these findings, we prospectively enrolled an external cohort of patients with lower-risk MDS in our study until the proportion of patients with SF3B1MUT MDS was comparable to that in the training cohort (table S5). This validation cohort consisted of 55 patients with MDS, 42 (76.3%) of whom had MDS with SF3B1 mutation. Patients with SF3B1MUT and SF3B1WT in this cohort had a similar transfusion burden with a mean number of 4 red blood cell (RBC) units per 8 weeks (table S5). The mean concentrations of ERFE and ferritin significantly increased (P = 0.0005 and P = 0.0488, respectively; Fig. 5, E and F), whereas hepcidin and hepcidin/ferritin ratio significantly decreased in patients with SF3B1MUT (P = 0.0038 and P = 0.0002, respectively; Fig. 5, G and H). This confirms the results of the training cohort and suggests that the increase of ERFE concentration in patients with SF3B1MUT MDS is independent of RBC transfusions.

Iron homeostasis changes after RBC transfusions, which ameliorate the anemia and increase the concentrations of circulating iron, with both effects expected to increase hepcidin. To assess the influence of RBC transfusion in our analysis, we delineated a subset of 61 patients with MDS in the training cohort, including 25 patients with SF3B1MUT and 36 SF3B1WT with a low transfusion burden before inclusion (<4 RBC units per 8 weeks). In this subset, patients with SF3B1MUT or SF3B1WT MDS were equally transfused (mean, 0.5 RBC unit/8 weeks), but ferritin and plasma iron concentrations remained higher in patients with SF3B1MUT (fig. S7, A and B). The hepcidin/ferritin and hepcidin/plasma iron ratios were significantly lower in patients with SF3B1MUT (P < 0.0001 and P = 0.019, respectively; fig. S7, C and D), and the circulating ERFE concentration remained significantly increased in low transfusion burden patients with SF3B1MUT compared to patients with SF3B1WT (P < 0.0001; fig. S7E). Erythropoiesis-stimulating agents (ESAs) are used as first-line treatment of anemia, with low serum EPO and low transfusion burden as predictors of response (23). We investigated the impact of iron parameters on the response to epoetin ζ for 12 weeks in 59 patients with low-risk MDS included in a clinical trial of the Groupe Francophone des Myélodysplasies (GFM), GFM-Retacrit-2013 (NCT 03598582; table S6) (30). Plasma ERFE, ferritin, and hepcidin concentrations at enrollment were similar in responding and nonresponding patients, suggesting that these parameters were not predictive of the erythropoietic response in this cohort (fig. S8). We then explored the determinants of hyperferritinemia >300 μg/ml in patients with low transfusion burden. By univariate analysis, ERFE (P = 0.005), hepcidin (P = 0.013), and SF3B1 mutation (P = 0.006), but not sTfR concentration or the number of transfused RBC units, were significantly linked to the concentration of serum ferritin (Table 1). By multivariate analysis, ERFE, hepcidin, and SF3B1 mutation remained independent predictors of high ferritin concentration (Table 1). Together, these results indicate that before patients reach a critical threshold of transfusion dependence, hyperferritinemia can be caused by SF3B1MUT-induced expression of ERFE, which in turn lowers hepcidin.

Table 1 Erythroferrone, hepcidin, and SF3B1 mutation as independent predictors of hyperferritinemia in low transfusion burden patients with MDS.

Erythroferrone, hepcidin, sTfR, number of RBC units per 8 weeks, and SF3B1 mutation were evaluated in the group of 60 patients receiving less than 4 RBC units per 8 weeks on the basis of ferritin with a cutoff value of 300 ng/ml. Parameters are indicated as means and 95% confidence intervals (95% CI) or ranges. For univariate analysis, Mann-Whitney and χ2 tests were used to compare the variables between SF3B1MUT and SF3B1WT MDS. Multivariate logistic regression analysis was performed for variables with P value < 0.1 in univariate analysis.

View this table:

Erythroid lineage–restricted expression of ERFE+12

In mice, ERFE mRNA expression in the BM is regulated by EPO and is predominant in basophilic and polychromatic erythroblasts (26). To investigate whether ERFE and ERFE+12 expression is restricted to the erythroid lineage, we amplified in parallel the erythroid and granulocytic precursors derived from the BM CD34+ cells of one SF3B1MUT and one SF3B1WT sample. The purity of each lineage was assessed by the cytological examination of May-Grünwald-Giemsa–stained cytospins, and the quantification of lineage-restricted markers was assessed by RT-qPCR (Fig. 6A and fig. S9A). To compare the amount of each transcript isoform at the different stages of differentiation, we quantified the canonical ERFEWT and the aberrant ERFE+12 by RT-qPCR. The expression of the canonical transcript ERFEWT expressed as normalized ratio quantities (NRQ) increased in both SF3B1MUT and SF3B1WT MDS erythroblasts. In granulocytes, the expression of ERFEWT was close to the limit of detection (Fig. 6B). The expression of ERFE+12 was restricted to the erythroid lineage, increased with the differentiation of SF3B1MUT erythroblasts, and was higher in SF3B1MUT compared to SF3B1WT erythroblasts (Fig. 6B). These results indicate that ERFE+12 expression is specific to the erythroid lineage.

Fig. 6 Erythroid cell-restricted expression of ERFE+12.

(A) May-Grünwald-Giemsa–stained cytospins of erythroid and granulocytic precursors obtained from liquid culture of BM SF3B1WT or SF3B1K700E CD34+ progenitors. (B) Quantification of ERFEWT and ERFE+12 in erythroid and granulocytic precursors by RT-qPCR. Results are expressed as NRQ ± SEM to ACTB and B2M housekeeping genes. (C) ERFE+12 is absent in SF3B1MUT CLL. BM MNCs from a patient with SF3B1K700E CLL + MDS or PB MNCs from patients with SF3B1T663I CLL, SF3B1WT CLL, and SF3B1K700E MDS were collected for analysis by capillary electrophoresis of ERFE and MAP3K7 fluorescent PCR products. ERFE+12 was detected as a 162-nt fragment and MAP3K7+20 was detected as a 170-nt fragment. PB samples from one patient with an SF3B1WT CLL and one patient with an SF3B1MUT MDS are shown as controls. (D) ERFE+12 expression is restricted to SF3B1MUT myeloid lineage. CD19+CD5 B cells, CD19+CD5+ B CLL cells, CD3+ T cells, and myeloid cells were sorted from the BM MNC fraction of an SF3B1K700E CLL + MDS, an SF3B1K700E MDS, and an SF3B1WT MDS and from the PB MNCs of an SF3B1T663I CLL. ERFE and MAP3K7 transcripts were analyzed by fluorescent PCR. (E) The sequencing of SF3B1 was performed on cDNA of each cell fraction. +, mutated; WT, wild type; NA, not available.

The SF3B1 gene is mutated in 15% of patients suffering from CLL, and the SF3B1MUT allele is present in CD19+ lymphocytes (14, 31). To investigate whether ERFE+12 was detectable in SF3B1MUT CLL, we collected one sample from a patient with CLL followed by MDS, whose BM MNC expressed a clonal SF3B1K700E mutation, and three peripheral blood (PB) MNC samples from two patients with CLL, one of whom harbored a clonal SF3B1T663I mutation with a variant allele frequency of more than 40% and the second had no mutation in the SF3B1 gene, and from a patient with SF3B1K700E MDS. We analyzed the MAP3K7 transcript, which is alternatively spliced in SF3B1MUT CLL or MDS using a cryptic 3′ss (14, 32). A 170-nt fragment corresponding to the alternative MAP3K7 transcript was enriched in all SF3B1MUT samples compared to the SF3B1WT CLL (Fig. 6C). ERFE+12 was detected in the PB MNC of the SF3B1K700E MDS but not in the SF3B1T663I or SF3B1WT CLL. ERFE+12 was not detected in the BM MNC of the SF3B1K700E CLL + MDS sample. To get further insights into the cell types expressing ERFE+12, we sorted CD19+CD5 B cells, CD19+CD5+ pathological B cells, and CD3+ T cells and myeloid cells containing erythroblasts from the BM MNC fraction of the patients with SF3B1K700E CLL + MDS, SF3B1K700E MDS, and SF3B1WT MDS and from the PB MNC of the patient with SF3B1T663I CLL (fig. S9B). The number of cells in the myeloid fraction of SF3B1T663I CLL was too small for further studies. The 170-nt fragment of MAP3K7 was detected in CD19+CD5+ pathological B cells of SF3B1K700E CLL + MDS and SF3B1T663I CLL and also in the myeloid fraction but not in the CD19+CD5 B cells of SF3B1K700E MDS (Fig. 6D). By sequencing SF3B1 RNA, we demonstrated that the mutation was present in the cell populations in which the alternative MAP3K7 transcript was detected (Fig. 6E). The alternative ERFE+12 transcript was not detected in the CD19+CD5+ pathological B cells of SF3B1K700E MDS + CLL or SF3B1T663I CLL, and its expression was restricted to SF3B1K700E myeloid MDS cells. Together, these results indicate that ERFE is expressed in erythroid cells and ERFE+12 is restricted to SF3B1MUT MDS myeloid cells.

Correlation between changes in ERFE+12 expression and the response to lenalidomide

Fifty percent of patients with lower-risk MDS, including patients with MDS-RS, experience primary resistance or secondary failure to treatments they receive to cure their anemia. Whether the mechanism of resistance involves the persistence of clonal erythropoiesis is always unknown. We have previously shown that lenalidomide administered to patients with ESA-resistant non-del(5q) MDS targets the malignant clone and, in some cases, eliminates the dominant SF3B1MUT clone for the duration of response (33). However, the frequency of the SF3B1 variant allele, which is expressed in erythroid and myeloid cells, does not only reflect the abundance of clonal erythroblasts. Here, we retrospectively monitored the expression of erythroid-specific ERFE+12 transcript for the follow-up of patients included in clinical trials of the GFM: GFM-Retacrit-2013 and GFM-LenEpo-2008 (30, 34). For this purpose, we performed a fluorescent PCR and integrated ERFE+12 and ERFEWT peak heights as a ratio ERFE+12/ERFE+12 + ERFEWT in patients with SF3B1MUT MDS. Then, the ratio was measured in paired RNA samples at enrollment and first evaluation. In GFM-Retacrit-2013, four nonresponders and six responders were compared, and no significant variation was observed (Fig. 7A and fig. S10A). By contrast, in GFM-LenEpo-2008, the ratio ERFE+12/ERFE+12 + ERFEWT decreased in six responding patients but remained stable in eight nonresponding patients (Fig. 7B and fig. S10B). The percent variation of ratios between samples obtained during screening and after four cycles of treatment was significantly different between responding and nonresponding patients (Mann-Whitney test, P = 0.0013; Fig. 7B, right). The percent variation of SF3B1 variant allele frequency in BM MNCs between screening and evaluation after four cycles was not significantly different between nonresponding and responding patients (fig. S10C). This confirms that lenalidomide may target clonal erythropoiesis. By contrast, ERFE protein quantities did not vary with the response in both cohorts (fig. S11). Last, we addressed the prognostic value of the ERFE+12/ERFE+12 + ERFEWT ratio for overall survival (OS) in a cohort of 90 patients with low-risk MDS, including 24 patients with SF3B1MUT and 66 with SF3B1WT enrolled at diagnosis with a median follow-up of 36.9 months (table S7). In this cohort, a receiver operating characteristic (ROC) analysis established that a value of 0.008 was the threshold of positivity of the ERFE+12/ERFE+12 + ERFEWT ratio with a specificity and sensitivity of 100%. As shown in Fig. 7C, an ERFE+12/ERFE+12 + ERFEWT ratio > 0.008 was predictive of better OS (log-rank test; P = 0.019) and was correlated with the presence of an SF3B1 mutation. Among patients expressing the variant ERFE+12 transcript, there was no significant correlation between ERFE+12/ERFE+12 + ERFEWT ratio and OS (log-rank test; P = 0.064; fig. S12). Our results suggest that ERFE+12 expression may correlate with OS for patients with SF3B1MUT MDS.

Fig. 7 ERFE+12 expression as a marker of clonal erythropoiesis and survival.

Fluorescent PCR was performed at screening and evaluation in (A) 10 paired samples from patients with SF3B1MUT MDS enrolled in the GFM-Retacrit-2013 clinical trial (four nonresponding and six responding patients) and (B) 14 paired samples from patients with SF3B1MUT MDS enrolled in the GFM-LenEpo-2008 clinical trial (eight nonresponding and six responding patients). Peak heights of ERFE+12 and ERFEWT signals were integrated as ERFE+12/ERFE+12 + ERFEWT ratios. Percent variations of ratios are indicated (right) as medians and IQR (25 to 75%). Mann-Whitney test for P values. (C) Overall survival according to ERFE+12/ERFE+12 + ERFEWT ratio shown as a Kaplan-Meier curve. A threshold of positivity of 0.008 was determined by ROC analysis. Log-rank test for P value. ns, not significant.


In this study, we show that in patients with SF3B1MUT MDS, an alternative FAM132B/ERFE+12 transcript is translated into an ERFEVPFQ protein and, together with the canonical transcript, contributes to the overexpression of ERFE. Similarly to ERFE, the ERFEVPFQ protein efficiently represses hepcidin. ERFE+12 is specifically induced by the clonal erythropoiesis.

In SF3B1MUT cancers, alternative 3′ss usages are the most frequent splicing aberrancies (11, 13, 14, 35). The recently resolved crystal structure of the SF3b complex helps to understand this feature (36). In the spliceosome, the SF3b complex interacts with the pre-mRNA and is involved in the branch site selection during splicing. Mutations in SF3B1 affect the structure of the SF3b RNA-binding platform, resulting in the selection of alternative branch site and alternative transcripts, of which more than 50% are subjected to NMD (11, 13, 36). On the basis of these findings, we focused our attention on the transcripts generated by the use of an alternative AG′. The analysis of the distance separating AG from AG′ revealed strong peaks at 15, 18, and 21 nt, showing that inserts could be multiples of 3 nt. Among alternative in-frame junctions, we identified two regulators of iron homeostasis, ABCB7 and ERFE. As already reported, the alternative ABCB7 mRNA results from the addition of 21 nt between exons 8 and 9 and, in our sample set, was slightly down-regulated in SF3B1MUT compared to SF3B1WT MDS (11, 32, 37). By contrast, the aberrant ERFE transcript modified by the addition of 12 nt was up-regulated 2.3-fold in SF3B1MUT MDS. ERFE+12 has not been reported in other SF3B1MUT cancers despite some overlap between the sets of alternative 3′ss transcripts in SF3B1MUT uveal melanoma, CLL, and MDS (11, 13, 14, 38, 39). We showed the existence of a variant protein containing VPFQ sequence immediately upstream of the collagen domain, which is putatively involved in protein-protein interactions in other CTRP family members (40). VPFQ may not influence protein conformation, because the apparent molecular weight was similar for recombinant ERFEWT and ERFEVPFQ proteins, which both repressed HAMP gene expression. However, we could not exclude an effect of the VPFQ insertion on protein stability and/or still unknown functions.

In MDS, serum EPO is increased but ineffective in stimulating erythropoiesis because maturing erythroid cells undergo apoptosis in terminal phases of differentiation (1, 2). Increased EPO and sTfR concentrations have been inversely correlated with hepcidin concentrations in MDS (17, 18, 41, 42). Our study indicates that plasma hepcidin concentration in SF3B1MUT MDS is similar to healthy non–blood donor controls and inappropriately low compared to SF3B1WT MDS, suggesting that its production could be dysregulated by severe ineffective erythropoiesis and/or a more specific mechanism (17). We showed the cell-autonomous SF3B1MUT-dependent expression of ERFE+12 and a role for increased concentrations of circulating ERFE in hepcidin reduction. This confirms the involvement of ERFE in human pathologies (27).

ERFE is the major erythroid regulator of hepcidin (26). Other candidate regulators of hepcidin linked to erythropoiesis have been proposed, including EPO and GDF-15. Although EPO administration decreases hepcidin concentration (43), this effect is predominantly indirect. The other candidate, GDF-15, is poorly induced by EPO in human volunteers. Although serum GDF-15 concentration was increased in β-thalassemia or congenital dyserythropoietic anemia, it was not inversely correlated with hepcidin in MDS (17, 24, 44). Low hepcidin preserves ferroportin on enterocytes and macrophages, causing increased intestinal absorption and macrophage release of iron (45, 46). Compared to SF3B1WT MDS, the lower concentrations of hepcidin in SF3B1MUT MDS may explain early iron overload before patients receive erythrocyte transfusions (17, 18). In patients with low transfusion burden, who had received a mean of 0.5 RBC units, hepcidin, ERFE, and SF3B1 mutation were independent predictors of hyperferritinemia. This demonstrates that iron overload is strongly related to the control of hepcidin through the SF3B1-regulated production of ERFE. Hepcidin increases with transfusion intensity in patients with MDS or β-thalassemia because of exogenous iron loading and transient suppression of ineffective erythropoiesis (27, 47, 48). Here, in regularly transfused patients with SF3B1MUT MDS, hepcidin remained lower and ERFE remained higher than in patients with SF3B1WT MDS. This establishes a driver role for SF3B1 mutation in the stimulation of ERFE expression and systemic iron overload that characterized patients.

The SF3B1 mutation confers a good prognosis in MDS (6, 49). However, when patients are transfusion dependent, iron overload in cardiac and liver tissues becomes clinically evident and may impair life expectancy (50). Chelation efficiently reduces iron burden in regularly transfused patients with MDS and serves as an adjuvant therapy for anemia, because deferasirox may improve hematopoiesis by protecting against oxidative stress (5153). Promising substitutes for ESAs in resistant patients include lenalidomide, which may transiently reduce SF3B1 mutant allele burden, and activin receptor ligand traps such as luspatercept, which promotes late-stage erythropoiesis in mouse models of β-thalassemia and MDS and diminishes iron overload in β-thalassemia mice (33, 5456). In lenalidomide-treated patients, but not in epoetin ζ–treated patients, the kinetics of ERFE+12 expression correlated with the response to treatment, indicating that in responding patients, lenalidomide targeted clonal erythroid precursors either directly or indirectly. Changes in plasma ERFE protein after lenalidomide were not correlated with the response to these treatments. The impact of transfusions on the regulation of ERFE expression at the post-transcriptional level should be examined in the future. ERFE+12 expression could also be useful for disease surveillance of patients with MDS-RS treated with other drugs, such as luspatercept. Further clinical studies will be required to validate ERFE+12 expression as a therapy-responsive biomarker of ineffective erythropoiesis in patients with SF3B1MUT MDS.

Last, our findings open therapeutic avenues for preventing iron accumulation in patients with MDS-RS. Increasing hepcidin could produce therapeutic benefits in SF3B1MUT MDS, because a moderate increase in hepcidin decreased iron and improved anemia in β-thalassemia mice (57). Therefore, either the administration of a hepcidin agonist or the targeting of ERFE overexpression may provide a potential strategy for preventing iron overload and improving erythropoiesis in patients with SF3B1MUT MDS.


Study design

The study involved identifying a splicing variant of ERFE by RNA-sequencing human primary BM samples and the variant peptide by mass spectrometry. Each in vitro experiment (transfection, minigene assay, Degron-KI, transcriptional repression of hepcidin) was repeated three times. Each experiment using primary cells was performed with at least three samples in each category. A ratio of ERFE variant transcript to total transcript was validated as a marker of clonal erythropoiesis on the basis of the response to lenalidomide or ESA in two cohorts of patients with MDS enrolled in clinical trials and as a prognostic marker of OS in a prospective multicenter cohort of patients with MDS. Predictive value of plasma ERFE and hepcidin concentrations for hyperferritinemia was measured in training and validation cohorts of patients with MDS. For these studies, randomization or blinding was not applicable.


For the training cohort, patients with MDS (n = 156) were enrolled between 2008 and 2017 ( GFM-LenEpo-2008, NCT01718379; GFM-Retacrit-2013, NCT 03598582). For the GFM-Retacrit-2013 cohort, the status of response to epoetin ζ was recorded. The prospective cohort of 90 patients with MDS with survival data was enrolled between 2010 and 2018. BM aspirates and PB plasma samples were collected after each patient gave informed consent for biological investigations according to the recommendations of the institutional review boards (IRBs): IdF X GFM-LenEpo-08 EudraCT 2008-008262-12, IdFII 2010-A00033-36, IdFIII 2010-2753, IdFV 212-A01395-38 EudraCT 2012-002990-7338, and OncoCCH 2015-08-11-DC. For the validation cohort, patients with MDS (n = 55) were enrolled prospectively for plasma collection in France (five centers; IRB Onco-CCH 2015-08-11 DC) and Germany (one center; IRB Ethikkommission an der TU Dresden; EK 115032015) between 2016 and 2018. BM samples from five age-matched controls and PB plasma samples from 20 healthy controls were collected. Patient characteristics are indicated in tables S1, S5, and S7 and according to the response to epoetin ζ in table S6. Low transfusion burden was defined as <4 RBC units per 8 weeks.

DNA and RNA sequencing

Mutations in a panel of 26 genes were screened by next-generation sequencing (NGS) (fig. S1). RNA was sequenced on an Illumina HiSeq 2500 platform using a 100–base pair paired-end sequencing strategy. TopHat (v2.0.6) was used to align the reads against the human reference genome Hg19 RefSeq (RNA sequences, GRCh37) downloaded from the University of California, Santa Cruz Genome Browser ( Analyses of differential gene expression and differential junction read counts were performed using DESeq2 (13, 28).

CRISPR-Cas9 generation of isogenic Sf3b1K700E and Sf3b1WT cell lines

The murine erythroid cell line G1E-ER4 (58) was used to generate isogenic Sf3b1K700E and Sf3b1WT cell lines using CRISPR-Cas9–stimulated homology-mediated repair.

ERFE and ENOSF1 minigenes

The ERFE minigene was synthesized by insertion of the ERFE alternative junction in pET01 Exontrap vector (MoBiTec). G1E-ER4 9.2 (SF3B1WT) and G1E-ER4 5.13H (SF3B1K700E) cells were transfected and processed for fluorescent PCR.

Mass spectrometry analysis

Erythroblasts were lysed, and peptides were obtained by trypsin digestion (Promega) and analyzed by nano liquid chromatography coupled with a Q Exactive Plus mass spectrometer (Thermo Fisher Scientific). Data were analyzed using Mascot 2.5.1 (

Human ERFE quantification

Plasma ERFE concentration was determined as previously described (27).

Statistical analysis

For quantitative variables, values were compared using the Mann-Whitney test. Categorical variables were compared using χ2 or Fisher’s exact tests. P values < 0.05 were considered significant. ROC analysis was used to calculate the thresholds of positivity. Multivariate logistic regression analysis was adjusted for selected variables chosen with P < 0.1 in univariate analysis (JMP version 10.0.2, SAS Institute Inc.).


Materials and Methods

Fig. S1. Genomic characteristics of the training cohort of 156 patients with MDS.

Fig. S2. Identification of in-frame alternative transcripts in SF3B1MUT MDS samples by RNA sequencing.

Fig. S3. Allele-specific investigation of mutant SF3B1 causal relationship to aberrantly spliced ERFE+12 transcript using Degron-KI strategy.

Fig. S4. Massive use of FAM132B/ERFE cryptic junction in one MDS-RS with a bi-allelic alteration of the SF3B1 gene.

Fig. S5. Identification of ERFE peptide by LC MS/MS.

Fig. S6. Correlation of plasma ERFE concentrations with ERFE+12 transcript, WHO classification, and iron homeostasis parameters.

Fig. S7. Biological parameters of low transfusion burden patients with MDS in the training cohort.

Fig. S8. Plasma ERFE, ferritin, and hepcidin concentrations at enrollment in 59 patients with MDS according to the response to EPO.

Fig. S9. Erythroid-specific expression of ERFE variant transcript.

Fig. S10. Variations of ERFE+12 transcript expression SF3B1MUT low-risk MDS after therapy.

Fig. S11. Variations in plasma concentration of ERFE protein in SF3B1MUT low-risk MDS after therapy.

Fig. S12. Kaplan-Meier curve for the analysis of OS according to ERFE+12/ERFEWT + ERFE+12 ratio.

Table S1. Clinical and biological characteristics of the training cohort of 156 patients with MDS according to SF3B1 status.

Table S2. Differential expression of ERFE transcripts according to the mutational status of SF3B1, TET2, and DNMT3A genes by RNA sequencing.

Table S3. Primer sequences used for fluorescent PCR and RT-qPCR.

Table S4. Expression of total ERFE, ERFEWT, and ERFE+12 transcripts by fluorescent PCR and RT-qPCR in nine MDS BM samples.

Table S5. Clinical and biological characteristics of the validation cohort of 55 patients with MDS according to SF3B1 status.

Table S6. Clinical and biological characteristics of the cohort of 59 patients with MDS in the GFM-Retacrit-2013 cohort according to response status.

Table S7. Clinical and biological characteristics of the cohort of 90 patients with MDS with survival data according to SF3B1 and ERFE+12 status.

Data file S1. Differentially expressed transcripts between patients with SF3B1MUT and SF3B1WT MDS by RNA sequencing and custom code.

Data file S2. Differentially expressed 5′ and 3′ junctions between patients with SF3B1MUT and SF3B1wT MDS by RNA sequencing.

Data file S3. In-frame and differentially expressed 3′ cryptic ss junctions in SF3B1MUT MDS.

References (59, 60)


Acknowledgments: We thank F. Pflumio (INSERM UMR967, CEA/DSV/iRCM, Fontenay-aux-Roses), N. Taylor (Institut de Génétique Moléculaire, Montpellier), S. Vaulont (Institut Cochin), E.-F. Gautier (3P5 proteomic platform), J. Vinh (Ecole Supérieure de Physique et de Chimie Industrielles, Paris), B. Saint-Pierre (Institut Cochin), and D. Drubay (Institut Gustave Roussy) for fruitful discussions; S. Raynaud (Laboratoire d’oncohématologie, Hôpital Pasteur, Nice), J.-B. Arlet (Service de Médecine Interne, Hôpital Européen Pompidou Paris), and C. Kannengiesser (Service de Génétique, Hôpital Bichat Paris) for providing samples; F. Letourneur (Genom’IC, Institut Cochin), S. Baulande (sequencing platform, Institut Curie), L. Zaroili, M. Dejean, A. Marcon, V. Verjus-Lisfranc (Cochin Laboratory of Hematology), E. Benana (3P5 platform), and B. Billoré (INSERM U1220) for technical assistance; and M. Weiss (St Jude Children’s Research Hospital, Memphis, TN) and V. Paralkar (University of Pennsylvania, Philadelphia, PA) for providing the G1E-ER4 cell line and expert comments. Funding: This study was funded by INSERM, by the Institut National du Cancer INCa PLBio 2015 (INCa_9290), by INCa and the Direction Générale de l’Offre de Soins (DGOS) of the French Ministry of Social Affairs and Health through the Programme Hospitalier de Recherche Clinique (PHRC MDS-04; INCa-DGOS_5480), and by the Site de Recherche Intégrée sur le Cancer (SIRIC) CAncer Research for PErsonalized Medicine (CARPEM). C.L. and D.R. are the recipients of a salary funded by the Laboratory of Excellence on red cells GR-Ex. The Orbitrap Fusion MS was acquired with funds from FEDER and “Canceropole Ile de France.” L.K. received support from ANR-16-ACHN-0002-01 and from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (grant agreement no. 715491). T.G. received funding from the NIH by R01 DK 065029 (and the UCLA Center for Accelerated Innovation, under NIH grant U54HL119893, Palazzolo). Author contributions: S.B., A-S.A., C.L., G.J., F.G., T.L., P.P., S.K., M.W., A.R., and L.K. performed the experiments and analyzed the data; A.R., S.W., and A.-S.K. recorded patient clinical data; A.H., D.R., A.B., N.C., and I.B. performed bioinformatics analyses; A.T., S.P., D.B., H.P., L.A., U.P., T.C., M.H., A.S., V.S., and E.G. provided samples; S.P. performed statistical analyses and reviewed the manuscript; F.G., N.D., S.A., Z.K., H.P., M.W., R.M., M-H.S., P.M., E.N., T.G., and L.K. analyzed the data and reviewed the manuscript; M.F. and O.K. designed the study, analyzed the data, and supervised the work; M.F. wrote the manuscript. Competing interests: T.G., L.K., and E.N. are inventors on a patent application on ERFE. T.G. and E.N. are scientific founders of Intrinsic LifeSciences and Silarus Pharma, companies that have interests related to ERFE. M.F., O.K., L.K., F.G., M.-H.S., S.A., and A.H. are inventors on a patent application on variant ERFE. The other authors declare that they have no competing interests. Data and materials availability: RNA-seq data are available in the Gene Expression Omnibus (GEO) repository (accession number GSE132836). Material transfer agreement with H3 Biomedicine Inc. (S. Buonamici) is required to obtain synthetic full-length SF3B1WT or mutant SF3B1K700E cDNAs. The human erythroferrone assay is available from Intrinsic LifeSciences, La Jolla, CA. All other data associated with this study are present in the paper or the Supplementary Materials.

Stay Connected to Science Translational Medicine

Navigate This Article