GATA1 erythroid-specific regulation of SEC23B expression and its implication in the pathogenesis of congenital dyserythropoietic anemia type II

Biallelic mutations in SEC23B gene cause congenital dyserythropoietic anemia type II (CDAII), an autosomal recessive disorder characterized by ineffective erythropoiesis, hemolytic anemia, and splenomegaly.1 Approximately 80 SEC23B mutations have been described so far.31 However, several cases show an incomplete inheritance pattern or no mutations within the gene. Mutations in cis-acting regulatory elements of SEC23B as well as digenic inheritance was hypothesized as additional genetic etiology of CDAII.2 CDA variants have been associated to mutations in GATA1, encoding for the homonymous DNA binding protein that plays an essential role in development and maintenance of both megakaryocytic and erythroid lineages.4 GATA1-related cytopenias are hallmarked by thrombocytopenia that can be associated to impaired erythropoiesis, dyserythropoietic anemia, congenital erythropoietic porphyria (CEP), thalassemia,5 or Diamond Blackfan anemia-like disease.6 The pathogenesis of these conditions is mostly related to altered GATA1 protein levels.87

In this study we described the functional interaction between GATA1 and SEC23B genes in two patients with suspected CDAII (Table 1). See methods in Online Supplementary Data.

Table 1.Clinical data.

Case1 (A-II.3) was a 22-year-old Italian male from healthy parents (Figure 1A). At birth, respiratory distress, haemorrhagic manifestations, and hypospadias were observed. Family history was not indicative for anemia, but both brothers were monorchid. At 4 months, anemia with thrombocytopenia was observed, and blood and platelets (PLT) transfusion regimen was started. Bone marrow (BM) analysis revealed dyserythropoiesis with bi- and multi-nucleated erythroblasts, while electron microscopy (EM) highlighted erythroblasts with double membranes, a typical feature of CDAII. Additionally, ultrastructural abnormalities in megakaryocytes, such as fairly normal granulation pattern, lack of demarcation membranes, and presence of cytoplasmic vacuoles were observed. The clinical data affecting both erythroid and megakaryocytic lines prompted us to perform sequencing analysis of both GATA1 and SEC23B. We identified the causative mutation of X-linked thrombocytopenia with dyserythropoietic anemia (XLTDA) GATA1-p.G208R,9 but no causative mutations in SEC23B gene.

Figure 1.Genetic study and ex vivo analysis of GATA1 and SEC23B expression. (A) The proband A-II.3 was hemizygous for the GATA1 mutation c.722G>A, p.G208R; according to X-linked inheritance pattern, the mutation was inherited from the mother. (B) Proband B-II.2 showed digenic inheritance pattern, with SEC23B intronic mutation, c.1905+3G>T, and GATA1 c.-183G>A polymorphic variant. DNA from relatives was not available. (C) In silico analysis of c.1905+3G>T mutation predicted a slight reduction of the score between WT and mutated donor site sequence (WT score = 95.15, mut score = 91.28, variation = −4.06%). Accordingly, amplification by PCR of the exon region 13–17 of SEC23B cDNA from PBL of the patient B-II.2 highlighted the presence of two bands on agarose gel: the top band (expected size fragment) of 539-bp, and the bottom band in lane 2 of 377-bp (exon-16 skipped transcript). Sequencing analysis of the 377-bp PCR product is shown. The chart shows the expression analysis of the exon-16-skipped transcript compared to the full length transcript by qRT-PCR analysis. In agreement with in silico prediction, the exon-16-skipped transcript is produced at very low level compared to full transcript (2.6%) HC(s), healthy control(s). (D) GATA1 and SEC23B expression in both patients compared to HC (n = 8) is shown. We observed a marked downregulation (80%) of GATA1-G208R mRNA expression in the proband A-II.3 compared to HC (P=0.01), and a slight reduction (50%) of GATA1-c.-183G>A gene expression in the proband B-II.2 compared to HC (P=0.03). Also SEC23B gene expression resulted down-regulated in both probands compared to HC (P=0.03). Data are presented as mean ± SE. P value by Student t-test. (E) WB analysis of GATA1 and SEC23B in PBL from both patients compared to a HC. Sizes (in kDa) are on the right. (F) The charts show the densitometric quantification of WB presented in the panel E. A marked reduction of GATA1 level in both probands A-II.3 (70%) and B-II.2 (80%) compared to HC was observed. SEC23B protein levels similarly decreased in both patients. OD, optical density. Data are presented as mean ± SD.

Case2 (B-II.2) was a 48-year-old Punjabi male from healthy parents (Figure 1B). Family history was not indicative of anemia. Overall clinical features mostly resembled those of CDAII biallelic patients, except for a slight reduction of PLT count (Table 1 and Online Supplementary S1). He showed jaundice, gallstones and liver iron overload with high transferrin saturation, low hepcidin/ferritin ratio and increased level of erythroferrone.10 Consequently to iron loading, he developed diabetes and complained of impotence. Mutations neither in HFE nor in HBA and HBB genes were detected. Deficit of pyruvate kinase and G6PD, porphyria, sideroblastic anemia and hepatitis were also excluded. BM examination highlighted hypercellularity with severe dyserythropoiesis, multinucleated erythroblasts, and internuclear bridges. Accordingly, mutations in CDAN1 and C15ORF41 genes were ruled out. Additionally, no mutations in CDAIV causative gene, KLF1, were found. Pathognomonic features of CDAII were observed: EM analysis showed binucleated erythroblasts with double membrane, and SDS-PAGE displayed hypoglycosylated band 3. Mutational analysis in this proband revealed an intronic SEC23B mutation, c.1905+3G>T, in heterozygous state. This variant was neither annotated in 1000 Genomes nor in ExAC browser. Both in silico and ex vivo analyses demonstrated that it impairs splicing, leading to the skipping of exon 16 in SEC23B mRNA with subsequent mRNA decay of the mutated allele (Figure 1C). We excluded additional mutations in both coding and splice site regions of SEC23B as well as large gene deletions or mutations in the promoter. Thus, the mutational pattern in this patient resulted incomplete. Sequence analysis of GATA1 showed the nucleotide transition c.-183G>A in the 5′ upstream region. This is a common variant (rs113966884), annotated in 1000 Genomes project with a minor allele frequency 0.02 (Online Supplementary Figure S1A), affecting the expression of both GATA1 isoforms (full length, fl and short, s), leading to a reduced expression of both transcripts in healthy controls (HCs) carrying rs113966884 A allele compared to G allele (Online Supplementary Figure S1B). Interestingly, we also observed a reduced SEC23B expression in control subjects with GATA1-rs113966884 A allele compared to G allele (Online Supplementary Figure S1C).

Expression analysis on peripheral blood leukocytes (PBL) from the 2 probands showed that both GATA1 mutations, p.G208R and c.-183G>A, account for a reduced expression of GATA1 itself at both mRNA and protein levels (Figure 1 D–E–F). Of note, in the GATA1-c.-183G>A patient we observed a slight reduction of GATA1 gene expression, despite the marked downregulation of GATA1 protein. This apparent discrepancy can be explained by the GATA1 positive feedback loop. Indeed, the upstream variant leads to a quantitative defect of GATA1 itself but not to a qualitative one, as in the case of missense p.G208R. Similarly to those observed in HCs carrying GATA1-rs113966884 A allele, in case B-II.2 we did not detect unbalanced expression of the two GATA1 isoforms; moreover, no abnormal megakaryopoiesis was observed, supporting the observation that GATA1s expression promotes mostly megakaryopoiesis, but not erythropoiesis.11

Although sequencing analysis revealed no mutations in SEC23B coding sequence in case A-II.3, expression analysis of this gene showed the same reduced trend of GATA1 (Figure 1D), with comparable results at protein evaluation (Figure 1 E–F). According to the ex vivo data, we observed a reduced GATA1 and SEC23B expression in HEK-293 cells transfected with GATA1-G208R mutant compared to cells transfected with GATA1 wild type (WT) (Figure 2 A–B–C). As expected, the reduced expression of GATA1-G208R in the proband A-II.3 also affected several other GATA1-regulated genes, similarly to SEC23B (Online Supplementary Figure S2). Thus, the phenotype of this patient could result from altered expression of many erythroid genes, although the specific abnormality of double membranes in erythroblasts could arise form altered SEC23B expression.

Figure 2.Functional interaction between GATA1 and SEC23B genes. (A) GATA1 and SEC23B expression of HEK-293 cells overexpressing pCDNA3.1-GATA1 and pCDNA3.1-GATA1-G208R compared to those transfected with empty vector (EV) were shown. Data are presented as mean ± SD. P value by Student t-test. (B) WB analysis of GATA1 and SEC23B in HEK-293 cells transfected with pCDNA3.1-EV, pCDNA3.1-GATA1 and pCDNA3.1-GATA1-G208R. Sizes (in kDa) are on the right. (C) Densitometric quantification of WB presented in the panel B; OD, optical density. Data are presented as mean ± SD. (D) ChIP-quantitative PCR analysis on GATA1 immunoprecipitated DNA in HEL at 6 days of differentiation by hemin. GATA1bs/ETO is the positive control (CTR+). HuSEC23B is the negative control (CTR−), a sequence flanking GATA1bs within HuSEC23B upstream region. Data are presented as mean ± SD. (E) Luciferase normalized activity (LUC activity) of HEK-293 cells co-transfected with PGL3-GATA1bs/HuSEC23B-2475 and −463 mutants and GATA1-WT showed a reduction of approximately 30 and 40% compared to GATA1bs-WT sequence. Data are presented as mean ± SD. (F) LUC activity of HEK-293 cells co-transfected with GATA1 mutants (GATA1-G208R, −R216W, −D218G, −V205M) and PGL3-HuSEC23B/3.44 in HEK-293 cell line. Data are presented as mean ± SD.

Also in case B-II.2, a reduced expression of SEC23B at both mRNA and protein levels was observed when compared to HC (Figure 1 D–E–F). Additionally, a correlation analysis on PBL from 27 HC revealed a significant direct correlation between GATA1 and SEC23B expression (P=0.00001) (Online Supplementary Figure S1D). Thus, we postulated a functional interaction between these two genes as a novel genetic basis of CDAII.

In order to corroborate our hypothesis, we characterized 4292-bp 5′ upstream sequence of human SEC23B (HuSEC23B) using a twofold approach: both in silico and in vitro data demonstrated that this genomic region is highly accessible to transcription in K562 cell line. In silico evaluation by ENCODE web-tool identified a 500-bp conserved region, named HuSEC23B/3.29-2.78 (3300-bp upstream ATG), which exhibited an enrichment of the H3K27Ac histone mark in K562 cells, suggesting a high accessibility of the chromatin to transcription in this locus. This genomic sequence is predicted to contain a 720-bp CpG island, partially unmethylated in K562 cells (Online Supplementary Figure S3). We tested 10 deletion mutants of HuSEC23B upstream region for the ability to drive luciferase expression by transient transfection of K562 cells. According to in silico data, only the larger fragment, containing the enhancer region HuSEC23B/3.29-2.78, showed a marked luciferase activity compared to the empty vector (Online supplementary Figure S4).

We demonstrated the direct binding of GATA1 to 5 predicted cis-elements (GATA1bs) within HuSEC23B upstream region by chromatin immunoprecipitation (ChIP) assay in both K562 and HEL cells at 6 days of erythroid differentiation (Figures 2D and Online Supplementary Figure S5), confirming the ChIPseq data showing GATA1 occupancy at SEC23B in primary human erythroid cells.12 Particularly, two sites, GATA1bs/HuSEC23B-2475 and GATA1bs/HuSEC23B-463, showed high relative occupancy of the immunoprecipitated factor (Figure 2D). Of note, when we transiently co-transfected GATA1-WT with both GATA1 binding sites mutated in the core GATA1 binding sequence, we observed a reduction of approximately 30–40% of luciferase activity compared to those observed in co-transfected cells with GATA1bs-WT sequence (Figure 2E). Similarly, when we transiently co-transfected GATA1 mutants associated to X-linked cytopenias (GATA1/G208R, GATA1/R216W, GATA1/D218G, and GATA1/V205M)5 and the HuSEC23B/3.44 in HEK-293 cells, we observed a reduction of HuSEC23B/3.44 luciferase activity induced by two specific mutants, GATA1-G208R (50%) and GATA1-R216W (40%), which are the causative variants of XLTDA and CEP, respectively, compared to that observed in co-transfected cells with GATA1-WT (Figure 2F). These results are in agreement with a CEP case recently reported, in which the severity of the anemia has been explained by the co-inheritance of the CEP-related GATA1-R216W with a rare SEC23B missense substitution.13

The identification of cis-acting regulatory elements in SEC23B upstream region could allow the definitive diagnosis of CDAII patients with both peculiar clinical phenotypes and incomplete mutation pattern. Although it has been classically conceived that phenotypes segregating in a Mendelian fashion are caused by alterations in a single gene or a unique locus, exceptions to this notion have been identified.14 Indeed, it now seems more likely that a lot of Mendelian disorders could be explained with the combinations of multiple disease-causing alleles, or their combination with polymorphic variants. It is likely that common genetic variants associated with small changes in erythrocyte production act by slightly altering the binding and activity of GATA1 and co-factors, resulting in mild to moderate changes in the expression of target genes.15 In the study herein we demonstrated that the clinical features of the case B-II.2 are the result of a complex genotype SEC23B-GATA1, suggesting a novel genetic etiology underlying CDAII. We described a GATA1 polymorphism as a genetic modifier, which exacerbates the phenotype induced by mutations in GATA1-dependent genes. This is a novel paradigm that paves the way to new investigations for evaluating the role of this GATA1 common variant as gene modifier in many other pathological conditions.