MMS22L is a novel key actor of normal and pathological erythropoiesis

Abstract

The emergence of next‐generation sequencing techniques has led to the genetic characterization of numerous congenital erythroid disorders, emphasizing crucial pathways in both normal and pathological erythropoiesis. In this study, whole exome sequencing of a single patient with atypical congenital pure red cell aplasia revealed a mutation in the CDAN1 gene, typically associated with congenital dyserythropoietic anemia type 1 (CDAI), together with a previously unreported mutation in the MMS22L gene. Combined mms22l and cdan1 haploinsufficiency results in severe anemia in a zebrafish model. In human erythroid progenitors, loss of MMS22L leads to proliferation and differentiation arrest associated with activation of the p53 pathway and global epigenetic alterations, showing that MMS22L plays an indispensable role in erythropoiesis. Furthermore, MMS22L and CDAN1 are involved in the same protein complex whose nuclear import is mediated by the importin 4 (IPO4) protein, and MMS22L nuclear import is impaired in CDAI patients due to a defective interaction between CDAN1 and IPO4. Overall, through the genetic description of a single case characterized by digenic inheritance, we identified MMS22L as a novel key factor in erythropoiesis and brought new insights into normal erythropoiesis regulation and CDAI pathophysiology.

INTRODUCTION

Red blood cells (RBCs) are generated from hematopoietic stem and progenitor cells through the step‐wise process of differentiation called erythropoiesis. The production of approximately two million RBCs every second requires intricate coordination between intrinsic erythroid programs and extrinsic factors. ¹

Next‐generation sequencing (NGS) techniques have brought important insights into the molecular mechanisms of normal RBC production and homeostasis and how these processes may be perturbed in human diseases. ² , ³ , ⁴ In addition, these approaches might be of great interest to better characterize congenital erythroid disorders in which a subset of patients does not carry any mutations in the presumed causal gene and for which transgenic mouse models do not recapitulate the phenotype. ⁵

Congenital dyserythropoietic anemia of type I (CDAI) is an autosomal recessive disease characterized by ineffective erythropoiesis with distinct morphologic features in bone marrow late erythroblasts (presence of binuclear erythroblasts and chromatin bridges between nuclei). ⁶ Approximately 90% of CDAI cases are caused by bi‐allelic mutations in CDAN1 or CDIN1 genes, ⁷ which are crucial for DNA repair and/or chromatin assembly following DNA replication. ⁸ , ⁹ However, the exact mechanisms involved in CDAI pathophysiology remain unknown.

Here, we report the original case of an adult female with atypical congenital pure red cell aplasia (PRCA) in which whole exome sequencing identified heterozygous mutations in the CDAN1 and MMS22L genes. While monoallelic mutation or invalidation of the genes, respectively, did not result in an abnormal erythroid phenotype, the combined mutations led to red cell aplasia in vivo. Further functional analyses revealed that MMS22L and CDAN1 interact in a complex with importin 4 (IPO4), which mediates their nuclear translocation, and that nuclear import of this complex is impaired in CDAI patients. Finally, we found that loss of MMS22L resulted in proliferation and differentiation arrest associated with activation of the p53 pathway and global epigenetic alterations, showing that MMS22L plays an indispensable role in erythropoiesis.

RESULTS

Clinical presentation

The proband presented at the age of 48 with severe nonregenerative anemia associated with chronic leg ulcers and facial dysmorphia. The onset of her anemia was unknown since no laboratory record was available. She was the last‐born of five children from a consanguineous couple (first cousins). Her family history included a late miscarriage and the death of two siblings in their first year of life (Supporting Information S1: Figure 1A).

Bone marrow examination showed severe PRCA with neither associated dyserythropoiesis nor internuclear bridges, typical features of CDAI. Medullary conventional karyotyping did not show any cytogenetic abnormalities. No disease or condition associated with acquired PRCA, such as autoimmune disorders, lymphoproliferative disorders, solid tumors, viral infections, or drug exposure, was found. Bone marrow culture in methylcellulose resulted in a drastic decrease of erythroid colony formation (CFU‐E) (data not shown), with no inhibitory effect of autologous sera, confirming an intrinsic red cell defect and excluding immunological PRCA.

Blood transfusions, as well as various treatments, including steroids, immunoglobulins, cyclosporine, and cyclophosphamide, were administered without long‐term efficiency. The proband ultimately developed aplastic anemia and died at the age of 62 from septic shock in the context of febrile neutropenia.

The presence of extra hematological signs, unusual family history, and atypical presentation prompted us to suspect a late‐onset congenital disorder. Given the initial phenotype of PRCA, a diagnosis of Diamond‐Blackfan anemia (DBA) was first considered. Targeted sequencing did not identify any mutations in the coding region of the 21 genes reported to be involved in DBA. ¹⁰ In addition, the constitutional karyotype did not show any abnormalities, and Mitomycin C culture did not reveal chromatid breaks or quadriradial, ruling out the diagnosis of a chromosomal breakage syndrome such as Fanconi anemia. Whole‐exome sequencing (WES) was then performed on the proband, her healthy brother, and her son to identify genetic events responsible for her hematological phenotype.

MMS22L and CDAN1 germline heterozygous mutations cooperate to cause pure red cell aplasia

WES analysis confirmed the absence of mutations in genes known to be affected in inherited bone marrow failure syndromes and revealed a missense heterozygous (VAF = 46%) mutation (c.2173C>T; p.R725W) in the CDAN1 gene (NM_138 477.2) (Figure 1A). However, her healthy brother was also harboring this mutation, suggesting that another gene variant was involved in the phenotype of the proband. After filtering out variants shared with her healthy brother, we detected another heterozygous nonsense mutation (c.2194C>T; p.Q732X; VAF: 52%) in the exon 15 of the MMS22L gene (NM_198 468.2) (Figure 1A), which has never been reported in public variant databases. This variant was also detected in DNA from saliva in the heterozygous state, indicating its germline origin (Supporting Information S1: Figure 1B). MMS22L messenger RNA (mRNA) level in the patient's PBMCs was reduced by half compared to control PBMCs (Supporting Information S1: Figure 1C), likely due to nonsense‐mediated decay, suggesting that MMS22L acts in a haplo‐insufficient manner. Of note, MMS22L mutations have never been reported in published cohorts of inherited bone marrow failures, CDAs, or DBAs. ¹¹ , ¹²

Figure 1.

Combined MMS22L and CDAN1 haplo‐insufficiency alters in vivo erythropoiesis. (A) Electropherograms of the patient, her brother, and son showing Sanger sequence validation of the MMS22L Q732X and CDAN1 R725W mutations. Arrows indicate the position of the nucleotide's substitutions. (B) Quantification of the o‐dianisidine positive area in the yolk sac of each embryo injected with the indicated morpholinos. Analysis was performed using the ImageJ software. Data are representative of two independent experiments. (C) Representative images of 48 hpf o–dianisidine–stained embryos injected with the indicated morpholinos. MOs used in this experience result in a 50% downregulation of gene expression. P‐values are determined by a two‐tailed t‐test. ns: not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

A heterozygous frameshift MMS22L mutation (c.843delT; p.V393fs) was identified in our in‐house database (Polyweb database, Imagine Institute) in an unrelated individual with no hematological abnormalities. In line with this observation, heterozygous Mms22l ± mice generated by the International Mouse Phenotyping Consortium (http://www.mousephenotype.org) do not display any overt phenotype. These findings suggest that a single heterozygous loss‐of‐function mutation in MMS22L is likely insufficient on its own to cause the hematological phenotype observed in the patient.

Several lines of evidence suggest that the CDAN1 R725W mutation is a hypomorphic mutation. First, her healthy brother was also harboring this mutation. Second, this variant had already been reported in a CDAI patient with compound heterozygous CDAN1 mutations. ¹³ Third, CDAN1 R725W is located within the same domain as the CDAN1 R714W variant, and was shown to impair protein function by disturbing its ability to form a complex with the histone chaperone ASF1, therefore, preventing its sequestration in the cytoplasm during limiting steps of nucleosome assembly. ⁹ Also, this mutant has lost the ability to arrest cells in S phase and inhibit DNA synthesis. ⁹ Lastly, we introduced the CDAN1 R725W mutation using CRISPR‐Cas9 gene editing in both heterozygous and homozygous states in the human erythroblast cell line BEL‐A ¹⁴ (Supporting Information S1: Figure 2A) and found that only the homozygous CDAN1 R725W mutation results in dyserythropoiesis (Supporting Information S1: Figure 2B–D). Taken together, these findings suggest that the CDAN1 R725W mutation likely acts as a hypomorphic variant.

Thus, we tested the cooperation between deleterious MMS22L and CDAN1 mutations using a zebrafish knockdown model. One‐cell stage zebrafish embryos were injected with antisense morpholinos (MOs) targeting mms22l, cdan1, or co‐injected with both morpholinos. We selected MOs resulting in a 50% downregulation of gene expression to mimic the genetic condition of the patient (Supporting Information S1: Figure 3A,B). In accordance with murine models, quantification of hemoglobin (Hb) positive cells in the yolk sac of each embryo by o‐dianisidine staining revealed that knocking down mms22l and cdan1 separately does not significantly alter the percentage of Hb positive cells. However, double knockdown of mms22l and cdan1 led to severe anemia, as indicated by the significant decrease in the percentage of Hb‐positive cells compared to the controls (Figure 1B,C). Of note, co‐injected embryos presented morphological defects, despite no known developmental role for MMS22L or CDAN1 (Figure 1C).

These data suggest that MMS22L and CDAN1 heterozygous mutations cooperate to cause PRCA.

MMS22L is required for the proliferation and differentiation of normal erythroid progenitors

Considering that combined heterozygous MMS22L and CDAN1 mutations result in erythroid aplasia, we suggest that MMS22L plays an important role in human erythropoiesis. To verify this hypothesis, we targeted MMS22L mRNA in cord blood‐derived hematopoietic progenitors using an MMS22L shRNA (ShMMS22L), resulting in a knockdown efficiency of ~80% (Supporting Information S1: Figure 4A,B). We then cultured the cells in vitro under erythroid differentiation conditions. In the presence of EPO, MMS22L inactivation resulted in proliferation arrest (Figure 2A) due to a blockade of the cells in the G1 phase of the cell cycle (Figure 2B). A slight increase of apoptosis, albeit not statistically significant, was also observed (Figure 2C).

Figure 2.

MMS22L inactivation impairs in vitro erythropoiesis. (A) Cell proliferation of shMMS22L and shCTRL‐transduced erythroid progenitors was monitored for 96 h by real‐time videomicroscopy using the Incucyte® system. Results are represented as the fold increase of cell confluence compared to Day 0 and show mean ± SD for three technical replicates. Data are representative of three independent experiments. (B) Cell cycle analysis of shMMS22L and shCTRL‐transduced erythroid progenitors. Percentage of cells in each phase of the cell cycle is determined by FACS analysis at Day 6 of differentiation. Results are represented as the mean ± SD of three independent experiments. (C) Percentage of apoptotic shMMS22L and shCTRL‐transduced cells determined by FACS analysis as the percentage of Annexin V positive cells amongst the PI negative population at Day 4 and Day 6 of differentiation. Results are represented as the mean ± SD of three independent experiments. (D) Erythroid differentiation of shMMS22L and shCTRL‐transduced erythroid progenitors was assessed by FACS as the percentage of GPA‐positive cells at Day 5, Day 7, Day 9, and Day 11 of differentiation. Results are represented as the mean percentage ± SD of three independent experiments. (E) Cell proliferation of shMMS22L and shCTRL‐transduced CD36‐positive and CD36‐negative cells was monitored for 120 h by real‐time videomicroscopy using the Incucyte® system. Results are represented as the fold increase of cell confluence compared to Day 0 and show mean ± SD for three technical replicates. (F) Percentage of apoptotic shMMS22L and shCTRL‐transduced cells differentiated toward the erythroid or the graulo‐monocytic lineage, determined by FACS analysis as the percentage of Annexin V positive cells amongst the PI negative population at day 6 of differentiation. Results are represented as the mean ± SD of three independent experiments. (G) Erythroid differentiation of shMMS22L and shCTRL‐transduced CD36 positive cells was assessed by FACS as the percentage of GPA‐ positive cells at day 12 of differentiation. Granulo‐monocytic differentiation of shMMS22L and shCTRL‐transduced CD36‐negative cells was assessed by FACS as the percentage of CD11B and CD16 positive cells at Day 12 of differentiation. (H) Representative images of MGG‐stained cells at Day 7 of erythroid (top panel) or granulo‐monocytic (bottom panel) differentiation. Basophilic and polychromatic erythroblasts can be observed in the shCTRL transduced CD36+ cells, whereas shMMS22L‐tranduced cells, CD36+ resemble granulocytic and monocytic progenitors. In both shCTRL and shMMS22L‐transduced CD36− cells, progenitors differentiate into mature macrophages and granulocytes. P‐values are determined by a two‐tailed t‐test. ns: not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

As previously reported, ¹⁵ , ¹⁶ , ¹⁷ MMS22L inactivation in nonerythroid HeLa cells resulted in accumulation of the cells in the G2/M phase (Supporting Inforation S1: Figure 4C), suggesting that MMS22L may act differently in the erythroid lineage.

To address the effect of MMS22L inactivation in erythroid differentiation, we monitored Glycophorin A (GPA) expression over time. EPO‐induced erythroid differentiation of hematopoietic progenitors was severely impaired by MMS22L downregulation, as the percentage of GPA+ cells did not increase over time (Figure 2D). Of note, we even observed a decrease in the percentage of GPA+ cells over time, suggesting that those cells underwent apoptosis. Together, these data demonstrate that MMS22L plays an important role in erythroid survival, proliferation, and differentiation.

To test whether this effect was specific to the erythroid lineage, we assessed the consequence of MMS22L knockdown in the granulo‐monocytic (GM) lineage. In this experiment, shMMS22L and shCTRL‐transduced cells were sorted on the basis of CD36 expression. CD36+ progenitors were cultured with EPO to undergo erythroid differentiation, and CD36− progenitors were cultured with GM‐CSF to undergo GM differentiation (Supporting Information S1: Figure 5A). Although MMS22L expression level was much lower (Supporting Information S1: Figure 5B) in CD36‐ cells compared to CD36+ cells, its inactivation also resulted in proliferation arrest (Figure 2E). In addition, MMS22L knockdown led to a significant increase in apoptosis only in the CD36+ subpopulation (Figure 2F). Moreover, while cell differentiation was blocked in the erythroid lineage in line with the above data, shMMS22L‐transduced CD36− cells successfully differentiated, as assessed by expression of the CD11B and CD16 GM markers (Figure 2G). This was also confirmed by May‐Grünwald‐Giemsa (MGG) staining, which revealed an absence of erythroblasts within the MMS22L‐depleted CD36+ compartment but the presence of mature granulocytic and monocytic cells within the MMS22L‐depleted CD36‐ compartment (Figure 2H).

To further dissect the lineage‐specific effects of MMS22L downregulation during hematopoietic differentiation, we performed single‐cell RNA sequencing (scRNA‐seq) on cord blood‐derived hematopoietic progenitors transduced with either shCTRL or shMMS22L lentivirus. Cells were analyzed at two key time points: the first immediately after sorting transduced cells (Day 0), to capture early transcriptomic changes during lineage commitment; and the second after 7 days of in vitro erythroid culture (Day 7), to assess the impact of MMS22L knockdown on terminal erythropoiesis. After quality control and filtering, we retained 70,657 single‐cell profiles (36,007 shCTRL and 34,650 shMMS22L), which were combined to construct and annotate a metacell manifold model ¹⁸ (Supporting Information S1: Figure 6A,B). UMAP visualization of the combined dataset revealed well‐defined trajectories of all myeloid lineages (Figure 3A). However, mapping the distribution of shMMS22L‐derived cells onto this landscape showed a marked depletion of erythroid progenitors and precursors in the MMS22L knockdown condition (Figure 3B). Quantification of lineage composition confirmed this pattern, with a reduction in the fraction of Megakaryoctyte‐Erythroid Progenitors (MEPs) and erythroid cells in the shMMS22L condition at Day 0, while other myeloid lineages were preserved (Figure 3C). Notably, the fraction of megakaryocyte (MK) cells was increased upon MMS22L depletion, suggesting that differentiation of MEPs was skewed toward the MK lineage at the expense of erythroid differentiation.

Figure 3.

scRNAseq dissects lineage‐specific consequences of MMS22L loss during hematopoietic differentiation. (A) Annotated two‐dimensional Uniform Manifold Approximation and Projection (UMAP) of our combined metacell (MCs) manifold comprising 70,657 single‐cell transcriptomes (36,007 shCTRL and 34,650 shMMS22L) collected at Day 0 and Day 7 of in vitro differentiation. MCs are shown as color‐coded points. The MC model was built based on the transcription profile of each MCs at the different time points, and annotation was performed based on known markers. BEMPS, basophils‐mast‐eosinophils progenitors; GMP, granulocyte‐monocyte progenitor; MEP, megakaryocyte erythroid progenitor; Mo/DCs, monocyte/dendritic cells; MK, megakaryocytes. (B) Combined MC model of in vitro differentiation with color intensity indicating the fraction of shMMS22L‐derived cells within each MC. (C) Stacked bar plots showing the cell type composition of shCTRL and shMMS22L transduced primary cells at Day 0 and Day 7 of differentiation, based on metacell models generated separately for each sample (see Supporting Information S1: Figure 6A for individual models).

At Day 7, erythroid cells (including erythroid progenitors, basophilic, and orthochromatic erythroblasts) represented approximately 90% of the shCTRL population. In contrast, erythroid cells accounted for only ~1% of shMMS22L cells at the same time point (Figure 3C and Supporting Information S1: Figure 6C). Instead, MMS22L‐depleted cells exhibited a skewed differentiation toward other myeloid lineages, with a marked enrichment in basophils. Notably, macrophage differentiation was observed exclusively in the shMMS22L condition and not in shCTRL, suggesting that MMS22L depletion not only disrupts erythropoiesis but also induces monocytic differentiation toward macrophages.

MMS22L inactivation causes selective activation of the p53 pathway in the erythroid lineage

To further investigate the molecular basis of the erythropoiesis defect, we performed differential gene expression analysis between MEPs (the only erythroid primed population with sufficient cells for robust comparison) from shCTRL and shMMS22L samples. We found significant enrichment of the TP53 pathway in MMS22L‐depleted cells, with key p53 targets such as BTG2, CDKN1a, and GDF15 being amongst the most upregulated genes (Figure 4A,B and Supporting Information S1: Table 1). In line with the scRNA‐seq data, we confirmed p53 pathway activation at the protein level in sorted primary CD36+ cells following MMS22L depletion, as evidenced by increased TP53 and P21 expression (Figure 4C). Notably, p53 transcript level remained unchanged in MMS22L‐depleted MEPs (Supporting Information S1: Figure 7A), which we confirmed by RT‐qPCR in sorted primary CD36+ cells (Supporting Information S1: Figure 7B), suggesting that TP53 expression is regulated by MMS22L through post‐transcriptional mechanisms.

Figure 4.

MMS22L depletion results in DNA double‐strand breaks‐independent p53 pathway activation, decreased level of chromatin‐bound H3.1, and increased genome‐wide H3K4me3 occupancy. (A) Volcano plot showing the differentially expressed genes (P < 0.05, log‐fold change >0.8) between transcriptionally defined shCTRL and shMMS22L MEPs. Genes associated with the TP53 pathway are shown in red. (B) Boxplots showing expression of the indicated genes in shCTRL and shMMS22L MEPs. Log₂‐transformed normalized expression values of selected genes were extracted from single‐cell RNA‐seq data. Boxplot centers, hinges, and whiskers represent median, first and third quartiles, and 1.5× interquartile range, respectively. P‐values from the Kolmogorov–Smirnov test are shown on the plot. (C) Western blot analysis showing the expression of p53, p21, and phosphorylated RB in shCTRL and shMMS22L‐transduced CD36‐positive cells. Protein levels were compared to actin. Data are representative of two independent experiments. (D) Comparative average metaprofiles of H3K4me3 Cut&Run enrichments from shCTRL and shMMS22L‐transduced UT‐7 cells. Dashed lines indicate standard deviation. (E) Violin plots showing that the H3K4me3 levels are equally increased at erythroid‐specific (“upregulated erythroid” and “stable erythroid”) and nonerythroid genes (“non erythroid”). Erythroid‐specific genes were defined based on their up‐regulation or constant expression throughout terminal differentiation (“upregulated erythroid” and “stable erythroid,” respectively). The “non‐erythroid” gene set was defined as genes being suppressed during differentiation. P‐values are determined by a two‐tailed t‐test. ns, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

MMS22L forms a complex with the NFKBIL2/TONSL protein, a homolog of the plant DNA repair protein Tonsoku/Brushy1/Mgoun3. ¹⁵ , ¹⁶ , ¹⁷ , ¹⁹ This MMS22l/TONSL complex is involved in homologous recombination (HR) dependent repair of stalled or collapsed replication forks. Specifically, it mediates the loading of the strand exchange protein RAD51 on ssDNA, which is necessary for the strand invasion step of HR. ²⁰ Accordingly, MMS22L downregulation in nonerythroid cell lines has been reported to cause increased DNA double‐strand breaks (DSBs). ¹⁵ , ¹⁶ , ¹⁷ , ²⁰ As an increase of DSBs could explain p53 pathway activation, we assessed the rate of DSBs in MMS22L‐depleted primary erythroblasts. Confocal imaging analysis showed that MMS22L inactivation did not result in an increased number of yH2AX foci per nucleus as compared to control cells (Supporting Information S1: Figure 7C). These findings suggest that p53 activation upon MMS22L downregulation was not due to genomic instability and suggest a specific and unexplored role of MMS22L in erythropoiesis. Of note, we observed an upregulation of genes associated with endoplasmic reticulum (ER) stress in MMS22L‐depleted MEPs, including DDIT3, TXNIP, PPP1R15A, and SQSTM1 (Supporting Information S1: Table 1), suggesting that unfolded protein response (UPR) may contribute to p53 pathway activation upon MMS22L loss.

MMS22L downregulation induced global epigenetic changes

MMSS22‐TONSL has been reported to be a histone reader as its recruitment to sites of DNA replication is mediated by the interaction between TONSL and newly synthesized histones that are unmethylated on histone H4 lysine 20 (H4K20). ²¹ Additionally, TONSL‐MMS22L also exhibits histone chaperone activity. ²² Therefore, we sought to explore whether MMS22L loss may induce global and/or specific changes in epigenetic landscape using a Cleavage Under Targets and Release Using Nuclease (Cut&Run) assay. ²³ We compared genome‐wide mapping of H3K4me3 and H3K27ac, two well‐recognized epigenetic markers of active promoters and enhancers, ²⁴ , ²⁵ respectively, in shCTRL and shMMS22L‐transduced UT‐7 EPO‐dependant cell line. Although we did not observe significant changes in specific regulatory regions of erythroid genes, MMS22L downregulation was associated with a global increase in H3K4me3 levels (Figure 4D,E and Supporting Information S1: Figures 7D and 8). These preliminary observations suggest that the observed blockade in erythroid differentiation following shMMS22L knockdown is not directly related to epigenetic changes on erythroid‐specific genes, but rather appears to be associated with global changes in chromatin modifications. Given the limitations of the UT‐7 cell model, further studies using primary cells and more refined epigenomic approaches will be necessary to determine the relevance of these changes in the context of erythroid differentiation.

CDAN1 inactivation prevents MMS22L trafficking to the nucleus through importin 4

In addition to H3 and H4, MMS22L‐TONSL also interacts with ASF1, a histone chaperone that promotes their loading onto ssDNA during homologous recombination. ²⁶ Besides, CDAN1 has been reported to directly bind to ASF1 within a cytosolic ASF1–H3‐H4–importin‐4 (IPO4) complex. Specifically, CDAN1 acts as a negative regulator of ASF1 by mediating its sequestration in the cytoplasm, resulting in the blocking of histone delivery and S‐phase arrest. ⁹

The above‐described cooperation of MMS22L and CDAN1 mutations and the known interaction of MMS22L‐TONSL and CDAN1 with ASF1 and H3, suggest that MMS22L and CDAN1 are involved in a common pathway. Therefore, we attempted to study the cross‐talk between both proteins. First, we tested whether they interact in HeLa cells using Proximity Ligation Assay (PLA). As shown in Supporting Information S1: Figure 9, we found that MMS22L and CDAN1 showed positive PLA signals, mainly in the nucleus, demonstrating that they are in close proximity (Figure 5A). These results suggest that MMS22L is also part of the CDAN1‐ASF1‐H3‐H4‐IPO4 complex.

Figure 5.

MMS22L and ASF1B trafficking to the nucleus is mediated by IPO4 and is impaired in CDAI patients because of a defective interaction between IPO4 and CDAN1. (A) Quantification of PLA signals between MMS22L and CDAN1 in HeLa cells displayed as the mean number ±SD of dots per nucleus and cytoplasm. Analysis was performed using the ImageJ software. (B) Western blot analysis showing the expression of MMS22L and ASF1B in the nuclear (NE) extracts of shCTRL and shCDAN1‐transduced UT‐7 cells. Protein levels were compared to TBP expression. Data are representative of three independent experiments. (C) Western blot analysis showing the expression of MMS22L, CDAN1, and ASF1B in the nuclear (NE) extracts of shCTRL and shIPO4‐transduced UT‐7 cells. Protein levels were compared to TBP expression. Data are representative of two independent experiments. (D) Upper panel: western blot analysis showing the expression of MMS22L and ASF1B in the nuclear extract (NE) of B‐LCLs established from healthy controls (CTRL 1, 2, 3) and CDAI patients (CDAI 1, 2, 3). Protein levels were compared to TBP expression. Data are representative of two independent experiments. Lower panel: Quantification of relative protein expression level of MMS22L (left panel) and ASF1B (right panel) in CDAI patients. Results are represented as the mean ± SD of the three controls versus the mean ± SD of the three CDAI patients, all normalized to the controls. (E) Quantification of PLA signals between CDAN1 and MMS22L performed in B‐LCLs established from healthy controls (CTRL 1, 2) and CDAI patients (CDAI 1, 2, 3). Results are displayed as the mean percentage ±SD of red dots located in the nucleus and in the cytoplasm. Data from the 2 CTRL B‐LCLs and from the 3 CDAI B‐LCLs were pooled together. Analysis was performed using the ImageJ software. (F) Quantification of PLA signals between CDAN1 and IPO4 performed in B‐LCLs established from healthy controls (CTRL 1, 2) and CDAI patients (CDAI 1, 2, 3). Results are displayed as the mean percentage ±SD of red dots locating in the nucleus and in the cytoplasm. Data from the 2 CTRL B‐LCLs and from the 3 CDAI B‐LCLs were pooled together. Analysis was performed using the ImageJ software. P‐values are determined by a two‐tailed t‐test. ns, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Next, we assessed in the UT‐7 cell line whether shRNA‐mediated silencing of one protein would impair the expression or the localization of the other. While MMS22L inactivation did not impact CDAN1 expression or localization (Supporting Information S1: Figure 10A), we found that CDAN1 knockdown using shCDAN1 lentiviral transduction (knock down efficiency of 70% in whole cell extract, Supporting Information S1: Figure 10B) resulted in a severe decrease of MMS22L nuclear abundance (Figure 5B), which was not restored by treating the cells with the proteasome inhibitor MG132 (Supporting Information S1: Figure 10C). Moreover, nuclear level of ASF1B, an ASF1 paralog particularly involved in cell proliferation, ²⁷ was also decreased upon CDAN1 knockdown (Figure 5B). Consequently, we suggest that CDAN1 knockdown leads to a defect in MMS22L and ASF1B trafficking from the cytoplasm to the nucleus. As it has already been reported that IPO4 protein mediates the import of H3 and H4 to the nucleus, ²⁸ we hypothesized that the whole MMS22L‐CDAN1‐ASF1‐H3‐H4‐IPO4 complex was translocated to the nucleus through IPO4. According to this hypothesis, we found that IPO4 knockdown (Supporting Information S1: Figure 10D) severely reduced MMS22L, ASF1B, and CDAN1 nuclear expression (Figure 5C). Altogether, these data suggest that CDAN1 acts as a scaffold protein recruiting MMS22L and CDAN1 to IPO4 to allow their appropriate import into the nucleus.

MMS22L trafficking to the nucleus is impaired in CDAI patients because of a defect in CDAN1 and IPO4 interaction

As CDAN1 seems to be essential for MMS22L and ASF1B trafficking to the nucleus, we evaluated whether it was impaired in CDAI patients. We examined nuclear expression of both proteins in B‐LCLs established from three CDAI patients. Two of these patients were brothers, carrying the same CDAN1 compound heterozygous mutations (c.3128A>T; p.C1043V and c.451_457del; p.R151fs), and the third one carried a unique heterozygous CDAN1 mutation (c.1864G>A; p.E622K) (Supporting Information S1: Figure 11). In accordance with our CDAN1‐silencing experiments, we found that MMS22L and ASF1B trafficking to the nucleus was impaired in all CDAI B‐LCLs, as compared to three control B‐LCLs (Figure 5D). We next examined by PLA whether the impairment of MMS22L trafficking to the nucleus in CDAI patients was due to a defect in its interaction with CDAN1. We found that CDAN1 binding to MMS22L was not impaired in CDAI patients but that their interaction shifted from the nucleus to the cytoplasm with a significant increase of dots being in the cytoplasm in CDAI B‐LCLs, as compared to control B‐LCLs (Figure 5E and Supporting Information S1: Figure 12A,B). Moreover, we found that CDAN1 interaction with IPO4 was altered in CDAI patients with a significant decrease in the number of PLA dots in CDAI B‐LCLs compared to the control B‐LCLs (Figure 5F and Supporting Information S1: Figure 12C,D). Altogether, those results demonstrated that reduced MMS22L‐CDAN1 import into the nucleus observed in CDAI patients was related to a defective interaction between CDAN1 and IPO4. We therefore suggest that CDAI physiopathology involves a defect in the nuclear import of MMS22L and most likely of the whole MMS22L‐CDAN1‐ASF1‐H3‐H4 complex.

DISCUSSION

Our results brought new insight into CDAI pathogenesis by identifying MMS22L as a component of a MMS22L‐CDAN1‐ASF1B‐H3‐H4‐IPO4 (MCAHI) complex, and by demonstrating a defect in its nuclear import in CDAI patients. However, the mechanisms by which this defect leads to hallmark features of CDAI, that is, internuclear bridges and spongy heterochromatin, remain unclear. Proteins of the MCAHI complex are involved in DNA repair and chromatin assembly. Huang et al. ²⁶ reported that both processes are related, as ASF1 and CAF‐1 dependent delivery of H3 and H4 histones onto ssDNA serves as a platform to specially recruit DNA repair proteins, such as MMS22L/TONSL and Rad51. Moreover, HP1 protein, a key player in heterochromatin structure and nucleolar organization, has been shown to interact with CAF1 ²⁹ and ASF1, ³⁰ and to be recruited by CAF1 to promote laser‐induced Rad51 nucleofilament formation. In CDAI erythroblasts, HP1 is sequestered in the Golgi, which has been suggested to account for the abnormal pattern of heterochromatin and altered cytokinesis. ³¹ , ³² One can, therefore, hypothesize that reduced shuttling of the MCAHI complex and more specifically of ASF1 could be responsible for the abnormal localization of HP1 and the defect in chromatin structure in CDAI.

While reduced nuclear translocation of MMS22L is detected in CDAI patients, a combination of a heterozygous inactivating MMS22L mutation together with CDAN1 mutation, as observed in our proband, results in a distinct clinical picture of PRCA. We propose a model in which varying nuclear quantity of proteins of the MCAHI complex could lead to different hematological phenotypes (Figure 4D and Supporting Information S1: Figure 13). In our proband, we suggest that the inactivation of one MMS22L copy combined with a CDAN1 mutation results in the absence of nuclear MMS22L. Functional experiments show that MMS22L silencing in primary erythroblasts leads to p53 activation, cell cycle arrest at the G0/G1 phase, differentiation blockade, and increased apoptosis, namely cellular patterns also observed in DBA, a congenital erythroid disorder characterized by PRCA. ³³ We, therefore, suggest that complete MMS22L nuclear loss disrupts erythropoiesis and underlies PRCA in our patient.

On the other hand, we showed that the MMS22L nuclear level is downregulated by about 50% in CDAI patients. We hypothesize that this milder defect would rather trigger replicative stress due to the accumulation of DNA DSBs and increased genomic instability, which might result in ineffective erythropoiesis, as already reported in acquired dyserythropoiesis. ³⁴ Hence, defective nuclear import of the MCAHI complex may represent a pathogenic mechanism in CDAI.

As MMS22L is a ubiquitous protein, it is striking that the consequences of its inactivation appear largely restricted to the erythroid lineage. Likewise, the erythroid‐specific nature of other diseases, such as CDAI or DBA, which involve ubiquitously expressed genes, is still not fully understood. Several mechanisms likely contribute to this specificity. First, erythroid differentiation is tightly coupled to rapid and sequential cell divisions, ³⁵ , ³⁶ unlike other myeloid lineages, which can differentiate under G1 arrest or in quiescent states. Our scRNAseq and functional analyses indicate that MMS22L depletion leads to TP53 pathway activation in MEPs, coinciding with cell cycle arrest and differentiation failure. While TP53 is a ubiquitous stress sensor, erythroid progenitors appear to have a particularly low tolerance for TP53‐induced cell cycle arrest, an observation consistent with several erythroid disorders, such as DBA and 5q– syndrome. ³⁶

Importantly, TP53 activation in our model was not associated with an increase in γH2AX foci. This suggests that MMS22L loss activates TP53 via nongenotoxic mechanisms. A relevant parallel can be drawn with DBA, in which ribosomal stress, rather than DNA damage, leads to TP53 stabilization through inhibition of MDM2 by free ribosomal proteins. ³³ Similarly, MMS22L loss may disrupt chromatin or nucleolar homeostasis, triggering TP53 activation through replication stress or altered nuclear proteostasis. In line with this, our preliminary observation of upregulated ER stress and UPR genes, such as DDIT3 and PPP1R15A, raises the possibility that ER stress contributes to TP53 activation in MMS22L‐depleted MEPs. Given the known role of MMS22L and ASF1B in chromatin assembly at replication forks, this pathway may be particularly essential during the rapid S‐phase transitions that characterize erythropoiesis.

Additionally, we observed that MMS22L depletion in erythroid progenitors leads to a global increase in H3K4me3 levels, suggesting epigenetic dysregulation. This alteration may further compromise erythroid gene programs and contribute to lineage‐specific sensitivity, as erythropoiesis depends on precisely orchestrated epigenetic remodeling. Future studies will be needed to define how MMS22L coordinates chromatin dynamics during erythroid lineage commitment.

In summary, this work emphasizes the value of NGS in diagnostically challenging cases for the identification of disease etiology and, more globally, for the discovery of novel players in erythropoiesis. We demonstrate here the crucial role of MMS22L in normal erythropoiesis and provide important novel insight into the pathophysiology of CDAI.

METHODS

Patients

Written informed consent was obtained from participants or their legal guardians, and the study was approved by the Comité de Protection des Personnes “Ile‐De‐France II.”

CDA patients described in this study were identified through the French national registry of DBA and CDA (Pr Lydie Da Costa).

Genomic DNA samples were isolated from peripheral blood mononuclear cells or saliva using standard procedures.

Whole‐exome sequencing, scRNA seq, and Cut&Run assay

Sequencing approaches, bioinformatics analyses, and data availability are detailed in the Supporting Information S1: Methods.

Zebrafish experiments

Morpholino injection, O‐Dianisidine staining, and quantification procedures are detailed in the Supporting Information S1: Methods.

Ex vivo erythropoiesis protocol

CD34+ cells were isolated from umbilical cord blood obtained immediately following delivery of a full‐term infant after informed consent and approval. CD34+ cells were then isolated by a magnetic sorting system (MACS Miltenyi Biotec) according to the manufacturer's recommendations. CD34+ cells were placed in an in vitro two‐phase liquid culture system as previously described by Fibach et al. ³⁷ (See supplemental Methods for more details).

CRISPR‐Cas9 gene editing of BEL‐A

BEL‐A cells (1 × 10⁵) were resuspended in CD34⁺ nucleofection kit buffer (Lonza Biosciences) containing 18 pmol Cas9, 45 pmol gRNA, and 100 pmol of ssODN template. The guide RNA CTCAGCCTTGTTCCACTCAC was used to create the CDAN1^+/R725W and CDAN1^R725W/R725W cell lines. A sense ssODN template was used for the edits with the desired mutation (GCACTGGTGA) flanked by 91 and 36 base pair homology arms, in accordance with the design principles described previously. ³⁸ Forty‐eight hours later, DRAQ5‐ cells were sorted into 96‐well plates using a BDInflux Cell Sorter.

For clone screening, genomic DNA was isolated, and the gene region of interest was amplified using GoTaq DNA polymerase (Promega) (CDAN1 Fw: 5′‐AGGCAGGTTAAAGGGCATCT‐3′, CDAN1 Rv: 5′‐AATCTGGACCCTCACTCTGC‐3′). PCR products were cleaned up (QIAquick PCR purification Kit; Qiagen) and sequenced by Eurofins Scientific.

Lentivirus production and transduction of primary erythroblasts

Lentiviruses were produced in 293T as previously described ³⁹ by the VVTG (Vecteurs Viraux et Transfert de Gène) platform of Necker‐Enfants Malades Institute (INEM). Sequences of ShRNAs are detailed in the Supplemental Materials and Methods. Primary cells from cord blood were transduced at Day 5 of the first phase of culture at an MOI of 15, except for the shCTRL, which was transduced at an MOI of 8 in the presence of 5 μg/mL of polybrene (Merck). Cells were cultured for two more days, and GFP‐positive cells were sorted using the Sony SH800 FACS cell sorter. Sorted cells were then placed in culture with the second phase medium to undergo erythroid differentiation.

Protein extraction, immunoblotting, flow cytometry, and immunofluorescence microscopy

Protein extraction, western blotting, flow cytometry, and immunofluorescence microscopy procedures, as well as a list of corresponding antibodies, are detailed in the Supporting Information S1: Methods.

Statistical analyses

Statistics procedures are detailed in the Supporting Information S1: Methods.

AUTHOR CONTRIBUTIONS

Elia Colin: Conceptualization; writing—original draft; investigation; funding acquisition; methodology; formal analysis; writing—review and editing; data curation; validation; visualization. Ivan Ferrer‐Vicens: Investigation; methodology; visualization; formal analysis; data curation. Dror Brook: Formal analysis; data curation; visualization; investigation; software; methodology. Mohammad Salma: Investigation; visualization; formal analysis; data curation; software; validation. Charlotte Andrieu‐Soler: Methodology; validation. Elisa Bayard: Methodology. Alicia Fernandes: Methodology. Chantal Brouzes: Validation; formal analysis. Carine Lefèvre: Methodology; formal analysis. Romain Duval: Methodology; formal analysis. Michaël Dussiot: Methodology; formal analysis. Thiago Trovati: Validation. Geneviève Courtois: Validation. Slim Azouzi: Validation. Mohammed Zarhrate: Methodology; formal analysis; software. Anne Lambilliotte: Resources. Sophie Park: Resources. Benjamin Carpentier: Resources. Martin Colard: Resources. Sandra Manceau: Resources; project administration. Despina Moshous: Resources. Patrick Mayeux: Validation; supervision. Emilie‐Fleur Gautier: Validation. Annarita Miccio: Formal analysis. Jean Soulier: Validation. William Vainchenker: Validation. Liran Shlush: Validation. Lydie Da Costa: Resources; validation; formal analysis. Jan Frayne: Validation; supervision. Eric Soler: Validation; methodology. Olivier Hermine: Supervision; resources; validation; conceptualization; writing—review and editing. Lucile Couronné: Conceptualization; investigation; funding acquisition; writing—review and editing; validation; supervision; writing—original draft.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

FUNDING

This work was supported by grants from Laboratory of Excellence (labex) GR‐Ex, reference ANR‐18‐IDEX‐0001. E.C was supported by a Ph.D. grant of the French Ministry of Higher Education, Research and Innovation and the Fondation pour la Recherche Médicale (FRM) and by a labex GR‐Ex fellowship. The labex GR‐Ex is funded by the IdEx program “Investissements d'avenir” of the French National Research Agency, reference ANR‐18‐IDEX‐0001. I.F‐V was supported by MRC grant MR/R009341/1.

Supporting information

Revised Supplementary data HemaSphere.

Supplementary Table1.

DATA AVAILABILITY STATEMENT

Data that support the findings of this study are openly available in the National Center for Biotechnology Information BioProject at https://www.ncbi.nlm.nih.gov/bioproject/, reference numbers PRJNA802558 and PRJNA803334.