Cell-type–specific transcriptional regulation of PIGM underpins the divergent hematologic phenotype in inherited GPl deficiency

Hypomorphic coding mutations in genes of the glycosylphosphatidylinositol (GPI) biosynthetic pathway cause inherited GPI deficiency (IGD).^1-6  Unique to PIGM-associated IGD is a homozygous hypomorphic mutation (−270C>G) in the core promoter of this mannosyltransferase-encoding gene.^7,8 PIGM transcription in B cells is dependent on the generic transcription factor (TF) Sp1, and in IGD B cells, the pathogenic mutation abrogates binding of Sp1 to its cognate −270GC-rich motif.^8,9  This leads to histone hypoacetylation and polycomb-dependent PIGM transcriptional repression and thus to IGD^8,9  Patients with PIGM-associated IGD suffer from life-threatening splanchnic vein thrombosis and intractable epilepsy but not clinically significant intravascular hemolytic anemia.⁷  The latter, caused by lack of the GPI-linked complement regulators CD55 and CD59 from the surface of erythrocytes, is a defining feature of paroxysmal nocturnal hemoglobinuria (PNH),^10,11  a disorder caused by somatic mutations in PIGA in hematopoietic stem cells and their progeny.^12,13  Although the −270C>G mutation in PIGM-associated IGD is constitutional, the degree of IGD is variable and differs between different cell types, a feature shared with the other forms of IGD.^1-5,7  In particular, despite its marked deficiency in IGD B cells, expression of GPI in erythroid cells in PIGM-associated IGD and presumably in other forms of IGD is largely preserved.

Here, we investigate the transcriptional mechanisms that underpin the blood cell type-specific differences in GPI expression in PIGM-associated IGD.

The patient with IGD and the heterozygous parents were previously described.⁷  Blood samples were obtained and stored under approval by the local research ethics committees; verbal informed consent was provided in accordance with the Declaration of Helsinki.

Detailed methods are described in the supplemental Methods available on the Blood Web site.

Although surface expression of GPI in primary PIGM IGD red blood cells is near normal (>90% GPI positive), >70% of primary B cells (supplemental Figure 1) and >90% of patient-derived lymphoblastoid B cell lines (LBCLs) are GPI negative.⁹  To address whether differential GPI expression reflects differences in transcriptional regulation of PIGM, we first assessed GPI and PIGM messenger RNA (mRNA) levels in primary B and erythroid cells from a patient with IGD. In line with the near-normal GPI expression in mature erythrocytes, we found that cell-surface GPI levels (as assessed by anti-CD59 staining) and PIGM mRNA in erythroid precursors from a patient with IGD (P1) and his heterozygous parent (Het1) were actually higher than in normal control cells (Figure 1A-B). In contrast, PIGM mRNA levels in sorted GPI-negative primary B cells from P1 were lower than in their GPI-positive counterparts and Het1 B cells (Figure 1C). In line with intact PIGM transcription in erythroid precursors, in the erythroid K562 cell line, a 2-kb PIGM promoter fragment with the C>G mutation was nearly as active as the corresponding WT fragment (Figure 1D).

These findings suggest that the −270GC-rich motif is required for PIGM transcriptional activity in B cells but not in erythroid cells.

To investigate whether the differential role of the −270GC-rich motif in PIGM transcription in erythroid vs B cells entails differences in promoter accessibility, we mapped nucleosome occupancy by nucleosome protection assay in the core and proximal promoter of PIGM in K562 cells and LBCLs (Figure 1E). We found that in B cells, nucleosome occupancy at the promoter region spanning the −270 motif is higher compared with K652 and epithelial HeLa cells in which the core promoter is nearly nucleosome-free. These findings are specific to PIGM, because nucleosome occupancy at the core promoter of the housekeeping gene GAPDH was uniformly low in all 3 cell types. Next, we addressed whether these differences impact on transcriptional activity. We found that PIGM mRNA levels in LBCLs were lower than in K562 and HeLa cells (Figure 1F), in keeping with higher nucleosome occupancy at the PIGM promoter in B cells and suggesting that chromatin configuration at the promoter of PIGM is less permissive for transcription in B cells than other cell types. Given these differences in transcriptional potential, we propose that the pathogenic C>G mutation is sufficient to tip the promoter into a repressive state in B but not erythroid cells, thereby in part explaining the differential cellular phenotype in PIGM-associated IGD.

Next, we considered whether lineage-specific differences in the binding of the generic TF Sp1 at the promoter of PIGM are responsible for, or contribute to, active transcription of PIGM and intact GPI biosynthesis in IGD erythroid cells. As before, in B cells,⁹  we found high levels of Sp1 binding not only at the −270GC-rich box but also in upstream promoter regions where 3 GC-rich boxes are predicted to bind Sp1 (Figure 2A-B). By contrast, in K562 cells, we observed high levels of Sp1 binding corresponding to the upstream GC-rich boxes but not at the −270GC-rich box (Figure 2B). We further confirmed lack of binding of Sp1 to the −270GC-rich box in primary erythroid cells (Figure 2C). To further establish the functional role of Sp1 in PIGM transcription, we introduced a dominant-negative form of Sp1¹⁴  into K562 cells. This led to a >50% reduction in PIGM mRNA levels (Figure 2D), suggesting that Sp1 is required for active PIGM transcription in erythroid cells. Therefore, Sp1 regulates PIGM transcription in erythroid cells by binding to upstream, but not to the core promoter GC-rich boxes, providing a mechanism for the intact PIGM transcription and near-normal GPI expression in IGD erythroid cells.

We also tested the role of GATA1, a megakaryocyte-erythroid TF,¹⁵  and of KLF1,^16,17  an erythroid-specific TF that shares the same GC-rich binding motifs with the Sp TF family.^18,19  We found high-level binding of GATA-1 at the PIGM promoter in K562 cells (supplemental Figure 2A) but not in primary erythroid cells (supplemental Figure 2B). KLF-1, in primary erythroid precursors (KLF1 is not expressed at protein level in K562 cells²⁰ ), binds highly to the upstream PIGM promoter regions but not to the −270GC-rich box (supplemental Figure 2C). These data show that whereas GATA1 and KLF1 are likely to be important for PIGM transcription, their function is independent of the −270GC motif and thus the C>G mutation would not be expected to impact on their ability to regulate PIGM transcription in IGD erythroid cells.

Taken together, our data provide insights in 3 areas. First, they provide a molecular basis for the genotype-phenotype divergence between erythroid and B cells in patients with PIGM-associated IGD. However, the presence of a small fraction of GPI-negative red blood cells and GPI-positive B cells (supplemental Figure 1) in PIGM-associated IGD suggests that the differential binding of Sp1 to the −270GC-rich box in these cells might not be absolute. Perhaps in a stochastic manner, Sp1 occasionally fails to sufficiently activate PIGM transcription in IGD erythroid cells, whereas in a minority of primary B cells, either effective binding of Sp1 to the mutated motif or a yet to be defined compensatory transcriptional process activates PIGM transcription. Our findings also explain how defective GPI biosynthesis, a defining feature for both PNH and PIGM-associated IGD, leads to clinically significant intravascular hemolysis in the former but not the latter. Finally, they show that lineage-specific transcriptional regulation of an active gene might be determined by differential chromatin accessibility and binding of a generic TF to different areas of the promoter, thus highlighting a novel mechanism of gene transcription.

This work was supported by Fundação para a Ciência e Tecnologia, Portugal (J.R.C.), the Medical Research Council (V.S.C.), the Sir Halley Stewart Trust (K.M.), and the National Institute for Health Research Biomedical Research Centre at Imperial College Healthcare NHS Trust.

Contribution: J.R.C., K.M., and V.S.C. performed experiments and interpreted data, and A.K., I.A.G.R., A.M.A., and D.M.L. conceived and supervised the work and wrote the paper with input from all authors.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Anastasios Karadimitris, Centre for Haematology, Department of Medicine, Imperial College London, Hammersmith Hospital, Du Cane Rd, London W12 0NN, UK; e-mail: a.karadimitris@imperial.ac.uk.