We focused on a family in which two male siblings were diagnosed with DBA (Figure A). While DBA generally demonstrates autosomal dominant inheritance (1
), both parents in the family had entirely normal hematological analyses; assuming that the disease has full penetrance, this suggests X-linked or autosomal recessive inheritance. The affected siblings (II-1 and II-3) remain alive and most recently required chronic red blood cell transfusions for treatment (Table ). Both siblings showed robust clinical responses to corticosteroid therapy for a period of 4 (II-1) and 6 (II-3) years, as is typical for DBA (Table ). The siblings have macrocytic anemia, consistently low reticulocyte counts, and a modest elevation of fetal hemoglobin levels — features characteristic of DBA (Table and Supplemental Table 1; supplemental material available online with this article; doi:
). Sibling II-1 did not have an elevation of erythrocyte adenosine deaminase levels (Supplemental Table 1), which is seen in some DBA patients (1
). Interestingly, sibling II-1 has also had a mildly low platelet count beginning at the age of 17 years, and both siblings were noted to have had occasional mild reductions in the neutrophil count, with those in sibling II-1 being consistently lower (Table ). However, these siblings showed neither clinical signs of abnormal bleeding nor an increased propensity for infections. To confirm the clinical diagnosis of DBA, we reevaluated the original bone marrow aspirates and biopsies from these patients. Independent evaluation by hematopathologists agreed with the diagnosis of DBA. The bone marrow was noted to have erythroid hypoplasia without abnormalities of the other hematopoietic lineages (Figure , D–F).
Identification of GATA1 mutations in DBA.
Hematologic parameters for index DBA cases
We performed whole-exome sequencing on the 2 siblings with DBA, obtaining at least 10-fold coverage for more than 93% of the target bases (Supplemental Table 2 and refs. 8
). We reasoned that pathogenic mutations leading to DBA were likely to be rare in unaffected populations and therefore filtered out all variants identified from the latest draft of the 1000 Genomes Project, dbSNP build 132, and 95 exomes sequenced for the National Institute of Environmental Health Sciences Environmental Genome Project (Supplemental Table 2, Methods, and ref. 9
). After filtering, a total of 74 variants were identified as being shared by the 2 affected siblings. We genotyped these mutations in the other family members (Figure A). Of the 74 mutations, 31 (42%) were found in the 2 affected siblings but not in an unaffected sibling (Supplemental Table 3). No variants were identified that would fit an autosomal recessive model of inheritance. Only a single variant on the X chromosome, within the GATA1
gene, showed appropriate segregation for an X-linked disorder with full penetrance (Figure , A and B, and Supplemental Table 3).
encodes a transcription factor necessary for erythroid differentiation (10
), and therefore it is biologically plausible that this gene is involved in DBA. The mutation in the GATA1
gene is a G→C transversion at position 48,649,736 on the X chromosome (hg19 coordinates) (Figure B) and results in the substitution of leucine for valine at amino acid 74 of the GATA1 protein. The mutation occurs at the last nucleotide of the exon 2 donor splice site and therefore would also be predicted to affect splicing of GATA1
RT-PCR on peripheral blood–derived RNA samples from control individuals confirmed prior findings (11
) that there are normally 2 splice variants of GATA1
produced: a full-length form involving splicing of exons 1, 2, and 3 with subsequent exons and a shorter GATA1s
form involving splicing of exons 1 to 3 directly, with skipping of exon 2 (Figure C). By contrast, RT-PCR analysis of samples from the patients showed that the GATA1 mutation greatly favors the production of GATA1s
mRNA, which lacks exon 2 (Figure C). Quantitative RT-PCR analysis of GATA1
exon 2 demonstrated that individuals II-1 and II-3 had only trace amounts of mRNA containing this exon (3%–5% of control levels), while their mother (I-2), who carries the GATA1
mutation, had a level at 53% of controls (Figure ). This suggests that trace amounts of properly spliced full-length GATA1
mRNA may possibly be produced with this mutation. A lack of exon 2 would only allow translation to initiate at codon 84, resulting in the formation of the GATA1s protein that lacks the first 83 amino acids, which contain the transactivation domain of this transcription factor (Figure and refs. 12
GATA1 exon 2 mutation results in trace amounts of GATA1 mRNA containing exon 2.
A model of how GATA1 mutations in DBA favor production of GATA1s alone.
We then sought to identify additional DBA patients carrying mutations in GATA1 by screening 62 additional male DBA patients without known pathogenic mutations. We identified one patient with a deletion of one of 2 adjacent G nucleotides (X chromosome positions 48,649,736–48,649,737) at the same genomic position as the GATA1 mutation found in the 2 brothers above (Figure G). This mutation would also be predicted to favor production of GATA1s, as a result of impaired splicing and frameshift of the full-length GATA1 open reading frame (Figure G). This patient has anemia that has responded to treatment with corticosteroids and has not had other hematologic abnormalities (Supplemental Table 4).
Interestingly, a mutation identical to the G→C transversion in exon 2 has been reported to result in dyserythropoietic anemia in humans, and other GATA1
germline mutations are associated with variable types of anemias and thrombocytopenias (11
). These latter cases are due to missense mutations in the zinc fingers of GATA1 and are distinct from the mutations affecting the production of different isoforms (Figure ). The variability among phenotypes seen in the different mutations favoring GATA1s
production may be attributable to differences in the levels of GATA1
expressed. This phenomenon has been seen with mouse hypomorphic mutations, where even slight differences in Gata1
levels can lead to variable phenotypes involving survival defects, unrestrained proliferation, or impaired differentiation of erythroid progenitors (16
). We speculate that alterations in GATA1
expression may also underlie the phenotypic variability seen over time in the DBA patients. In addition, similar mutations that lead to the production of GATA1s alone are acquired somatically in all cases of Down syndrome–associated acute megakaryoblastic leukemia and transient myeloproliferative disease (12
). Mice with a mutation resulting in expression of only GATA1s have apparently normal erythropoiesis (18
), which emphasizes the species-divergent functions of GATA1. Specifically, these findings demonstrate that the full-length form of GATA1 is required for normal erythropoiesis in humans, but not in mice.
Systematic sequencing of GATA1
mutations in other cases of DBA will likely unveil similar mutations and reveal the extent to which such mutations contribute to this disease. All 3 of the patients with GATA1
mutations meet all of the current clinical diagnostic criteria for DBA and many of the supportive criteria, with the exception of elevations in erythrocyte adenosine deaminase levels (19
). However, discovery and further phenotypic analysis of DBA patients with GATA1
mutations may uncover unique differences between this set of patients and cases due to ribosome protein gene mutations, which may lead to revision of the current diagnostic criteria for DBA (19
). While the majority of studies on DBA pathogenesis have been focused on the role of ribosomal biogenesis, the finding of GATA1
mutations in DBA opens new avenues for studying the underlying basis of this disorder. These findings may provide insight into the erythroid specificity of this disease, which remains an enigma. Additionally, these DBA cases, coupled with the phenotypes described for other human GATA1
), increase our understanding of how this transcription factor plays a role in specifying human erythropoiesis.