TO THE EDITOR:
Congenital dyserythropoietic anemias (CDAs) are rare disorders defined by morphologic abnormalities of erythroid precursors that lead to ineffective erythropoiesis.1,2 CDA type III (CDA III), characterized by multinucleated erythroblasts in the bone marrow, represents the rarest form, with only ∼60 patients described in the literature. Studies in 2 independent families identified a causative dominant missense mutation in KIF23, which encodes MKLP1,3 the kinesin subunit of centralspindlin, a key regulator of cytokinesis.4 In addition to the familial cases, a few sporadic cases of CDA III have been reported indicating an autosomal recessive pattern of inheritance.1,2 However, no genetic cause has been identified in such cases. Here, we describe a sporadic CDA III case associated with compound heterozygous variants in RACGAP1, which encodes CYK4/MgcRacGAP, the other subunit of centralspindlin. We demonstrate these variants cause cytokinesis failure in human erythroid cells as a result of altered target specificity in Rho-family GTPase signaling.
The proband initially presented at age 3 years with unexplained macrocytic anemia that has persisted throughout his life, although never requiring a transfusion (supplemental Figure 1A). After ruling out common causes of macrocytic anemia, a bone marrow biopsy was performed. Marked dyserythropoiesis with megaloblastoid changes accompanied by multinucleated erythroid precursors and characteristic gigantoblasts were seen in ∼25% of erythroblasts (Figure 1A), establishing the diagnosis of CDA III. Clinical whole-exome sequencing did not find the previously reported pathogenic KIF23 mutation3 or other potential deleterious variants in KIF23. However, compound heterozygous variants in RACGAP1 were identified, namely, c.1187T>A; p.Leu396Gln (L396Q; maternally inherited) and c.1294C>T; p.Pro432Ser (P432S; paternally inherited; Figure 1B). The parents were not anemic, as assessed by 2 independent complete blood counts (supplemental Figure 1B), suggesting both alleles are recessive. Both variants are extremely rare; they have not been seen in the genome aggregation database,5 and both variants have a combined annotation-dependent depletion score of >25, with numerous other in silico algorithms predicting a deleterious effect. Indeed, they cause nonconservative amino acid substitutions of highly conserved residues in the GAP domain (Figure 1C). These results suggest the variants identified in this patient may represent the first known genetic cause of an autosomal recessive form of CDA III.
To determine the role of RACGAP1 in erythroid differentiation, we performed shRNA-mediated knockdown of RACGAP1 in primary human CD34+-derived erythroid cells.6 Efficient knockdown of RACGAP1 (supplemental Figure 2A) did not affect terminal differentiation, as measured by surface expression of GPA (supplemental Figure 2B) or band3/α4 integrin (supplemental Figure 2C). However, we found a striking defect in enucleation (supplemental Figure 2D), along with a significant increase in multinucleated erythroblasts (Figure 1D), the hallmark of CDA III. These results indicate that RACGAP1 has critical functions in human erythropoiesis consistent with its role in the pathogenesis of CDA III.
We next examined the effect of the 2 variants on RACGAP1 function by transgene-mediated rescue of shRNA-treated erythroid cells. Technical limitations did not allow for such experiments in the primary CD34+ culture system. We instead used an immortalized erythroid cell line derived from human cord blood CD34+ cells (ImCBE) (data supplement). Knockdown of RACGAP1 in ImCBE (supplemental Figure 3) resulted in an increase of multinucleated erythroblasts (Figure 1E-F), similar to what was observed in primary erythroid cells. ImCBE cells were transduced with an RNAi-resistant RACGAP1 or green fluorescent protein (GFP) transgene along with the RACGAP1-targeting shRNA. The WT transgene efficiently suppressed the multinucleation (Figure 1E-F). In contrast, L396Q failed to rescue the multinucleation as compared with WT (Figure 1E-F). The effect of P432S was even more drastic, with almost no rescue of the multinucleation (Figure 1E-F). These results demonstrate both the variants identified in the proband affect the function of RACGAP1 in human erythropoiesis.
Because both variants lie in the GAP domain of CYK4 (Figure 1C), we sought to understand the enzymatic effects of both variants. GAP domains stimulate the GTP hydrolysis by a GTPase, such as RhoA, Cdc42, or Rac1, using a conserved arginine residue (R385 in CYK4) and thus suppress the signaling mediated by the GTP form of these GTPases.7-9 The GAP domain of CYK4 stimulates the GTPase activity of Cdc42 and Rac1 with much greater efficiency than RhoA.10-13 The structural basis for this specificity is unclear, because most residues on the GAP-binding surface of the Rho-family GTPases are conserved (Figure 2A; supplemental Figure 4A). We noticed that the region on CYK4 encompassing R385, L396, and P432 is located close to the GTPase residues that are distinct among RhoA, Cdc42, and Rac113,14 (Figure 2A; supplemental Figure 4B). Therefore, we hypothesized that the RACGAP1 variants might affect the GAP activity of CYK4. This was tested by examining the increase of the GTPase rate by the CYK4 GAP domain14,15 (Figure 2B; supplemental Figure 5). As expected, WT GAP showed a significantly stronger ability to stimulate the GTPase activity of Cdc42 and Rac1, as compared with that of RhoA, whereas the enzymatically inactive R385A control failed to stimulate the activity of any of the 3 GTPases (Figure 2B). The L396Q variant showed a significant decrease in the ability to stimulate Cdc42 and Rac1 GTPases but no change in the ability to stimulate RhoA (Figure 2B). In contrast, P432S showed no effect on Cdc42 or Rac1 but had a significant increase in the ability to stimulate RhoA GTPase (Figure 2B). To our knowledge, this is the first mutation that significantly promotes the activity of a Rho-family GAP. These results suggest that both variants differentially affect CYK4 GAP activity to undermine its target specificity.
To further probe the mechanistic consequences of both variants, we established multiple independent HeLa cell lines that expressed an RNAi-resistant WT, L396Q, or P432S16 CYK4 GFP, along with a GFP-only control (Figure 2C-D). Immunoblotting confirmed the transgenes expressed at comparable levels to the endogenous CYK4 with some variation (supplemental Figure 6A-B). Similar to what we see in erythroid cells (Figure 1D-E), depletion of endogenous CYK4 with RNAi treatment in the GFP-only control resulted in frequent cytokinesis failure, whereas this was strongly suppressed with WT CYK4 GFP (Figure 2C). GFP fluorescence allowed us to establish a threshold of WT CYK4 GFP levels that was necessary to significantly suppress the cytokinesis failure (supplemental Figure 6C). Using this threshold, 1 stable cell line (L15) with high expression of the L396Q transgene showed modest defects in rescuing the phenotype, whereas the other line (L6) showed much stronger defects in rescuing the cytokinesis failure (Figure 2C). Strikingly, as observed in erythroid cells (Figure 1E-F), the P432S transgene nearly completely failed to rescue the cytokinesis failure, regardless of the expression level (Figure 2C; supplemental Figure 6). These results demonstrate both variants show similar defects in cytokinesis in HeLa and erythroid cells (Figures 1E-F and 2C).
Because centralspindlin functions as a key signaling hub for cytokinesis,17,18 we tested whether the variants affected CYK4 localization. As previously reported, WT CYK4 showed diffuse cytoplasmic localization during metaphase but began to accumulate at the spindle midzone at the onset of anaphase and strongly concentrated to the midbody4,12,16,19-21 (Figure 2D; supplemental Figure 7 for unmerged images). The L396Q variant showed a similar localization pattern, although a subpopulation of the L396Q cells showed late furrow regression (Figure 2D). In contrast, the P432S variant localized to the equatorial cell cortex with a ring-like pattern instead of accumulating at the spindle midzone (Figure 2D magenta arrows). This abnormal localization was often accompanied by a failure to form a cleavage furrow or by formation of a transient shallow furrow that failed to complete (Figure 2D). These results demonstrate that the severe cytokinesis defect by the P432S variant is likely due to the elevated GAP activity toward RhoA, which would interfere with the contractile ring formation, and the mislocalization of centralspindlin, whereas the modest cytokinesis defect by the L396Q variant seems to be primarily due to the diminished GAP activity toward Rac and Cdc42.12
Here, we report the first known genetic cause of an autosomal recessive form of CDA III. The variants identified provide novel insights into the substrate specificity of the CYK4 GAP domain and showcase the importance of the precise balance between the activities of Rho-family GTPases during cytokinesis.12,22-24 We propose a model that tries to explain the recessive nature of the identified variants based on the altered target specificity of the CYK4 GAP domain (Figure 2E). Interestingly, a point mutation at the corresponding residue of P432 in the Drosophila melanogaster ortholog was also shown to affect its in vivo function.25 The cortical mislocalization of the P432S variant (Figure 2D), which exhibits elevated GAP activity against RhoA but not against Rac1 or Cdc42 (Figure 2B), indicates the importance of the proper target specificity of the GAP domain for the subcellular localization of centralspindlin. This implies a novel mechanism for the cortical recruitment of centralspindlin via the RhoA-GAP interaction, which is hidden during normal cell division but might contribute to cleavage signaling when the central spindle is disrupted.26,27
A perplexing feature of most CDAs is that variants in widely expressed genes lead mostly to erythroid-restricted phentoypes.1,2 Presumably, this reflects the erythroid-specific vulnerabilities of these genes. Indeed, it is starting to be better appreciated that distinct cell types are variably susceptible to alterations in cytokinesis machinery.28 More work is needed to address why erythroid cells are particularly vulnerable to centralspindlin dysfunction. Our findings establish that the cause of CDA III is not limited to KIF23 but could also include other defects in centralspindlin function. Mutations in other proteins that interact and function with centralspindlin29 may also result in CDA III–like conditions. Indeed, viewing diseases as aberrances of genetic pathways vs defects in individual genes may provide the proper framework to better understand and classify diseases.
The authors thank Sergey Lekomtsev and Mark Petronczki for plasmids and Vimla Aggarwal, Julie Canman, Kartik Ganapathi, Kazutaka Murayama, Mikako Shirouzu, Steve Spitalnik, and Jenny Yang for helpful discussions and/or critical comments on the manuscript. The authors gratefully acknowledge CAMDU (Computing and Advanced Microscopy Unit) at Warwick Medical School for its support and assistance with live imaging. The Columbia University Irving Medical Center institutional review board (IRB) deemed this study exempt from IRB approval because it was considered a case study.The authors apologize to those authors whom we could not cite because of reference limitations.
This work was supported by grant 1K08NS119567 from the NINDS, National Institutes of Health (NIH) (S.N.W.), program grant C19769/A11985 from Cancer Research UK (M.M.), and grants HL140625 and HL149626 from the NHLBI, NIH (X.A.).
Contribution: S.N.W. designed research studies, collected and analyzed clinical data, and wrote the manuscript; M.B. designed research studies, performed and analyzed enzymatic assays and HeLa cell studies, and edited the manuscript; H.Z., C.G., and Y.H. designed research studies and performed and analyzed studies in erythroid cells; S.T., B.H.D., and M.T.L. collected and analyzed clinical data; X.A. designed research studies, analyzed data, and edited the manuscript; and M.M. designed research studies, analyzed data, and wrote the manuscript.
Conflict-of-interest disclosure: The authors declare no competing financial interests.
Correspondence: Sandeep N. Wontakal, Department of Pathology and Cell Biology, Columbia University Irving Medical Center, New York, NY 10032; e-mail: email@example.com; Masanori Mishima, Division of Biomedical Sciences, Warwick Medical School, University of Warwick, Coventry CV4 7AL, United Kingdom; e-mail: firstname.lastname@example.org; and Xiuli An, Laboratory of Membrane Biology, New York Blood Center, New York, NY 10065; e-mail: email@example.com.
The online version of this article contains a data supplement.
S.N.W., M.B., and H.Z. contributed equally as first authors.