Hypomorphic ATP11c is a novel regulator of decreased efficacy of transfused red blood cells in humans and mice

Abstract

Chronic red blood cell (RBC) transfusion sustains patients with diverse hematologic disorders, but repeated transfusion leads to iron overload and alloimmunization. Reducing transfusion burden requires identifying donor units that circulate more effectively after storage, yet determinants of this variability remain incompletely defined. Here, we integrate forward genetics in mice, multi‐omics analyses of over 13,000 human donors, and studies of two families with hereditary ATP11c mutations to reveal a central role for this phospholipid flippase in transfusion efficacy. We show that common ATP11C variants, including the missense SNP V972M, and rare familial loss‐of‐function alleles impair RBC survival by disrupting membrane lipid remodeling and cytoskeletal stability—a mechanism distinct from oxidative damage pathways. Together, these findings establish ATP11c as a novel determinant of transfusion outcomes across species and genetic contexts, and highlight opportunities for donor stratification and improved storage technologies to advance precision transfusion medicine.

INTRODUCTION

Chronic transfusion of red blood cells (RBCs) is a necessary and life‐sustaining therapy for a variety of pathologies (e.g., sickle cell disease, thalassemia, aplastic anemia, myelodysplastic syndrome, certain neoplasias, etc.). However, two major sequelae occur in chronically transfused patients (CTPs) in a dose‐dependent fashion. First, chronic transfusion can result in iron overload and eventual toxicity that can cause permanent organ damage and/or death in some patients despite aggressive iron chelation therapy. ¹ , ² Second, ongoing alloimmunization to multiple blood group alloantigens can make compatible blood increasingly difficult to find, resulting in morbidity and/or mortality due to insufficient RBC units. ³ There has been a vociferous debate in the field as to whether older units of RBCs are harmful to patients. ⁴ However, the sequelae of iron overload and alloimmunization are uncontroversial, dose‐dependent, and affect numerous CTPs with different pathologies.

It is now understood that there is substantial donor‐to‐donor variability in how well stored RBCs circulate post‐transfusion. ⁵ However, there have historically been no methods to identify units that will circulate better post‐transfusion. Recently, we have reported that stored RBCs from X‐linked glucose 6‐phosphate dehydrogenase‐deficient (G6PDdef) donors have decreased post‐transfusion circulation in humans, ⁶ which has now been confirmed in real‐world clinical treatment of patients with sickle cell disease (SCD). ⁷ Because approximately 10% of RBCs transfused into SCD patients are G6PDdef, ⁸ this demonstrates an opportunity to improve therapy for CTPs by screening for and excluding G6PDdef units. However, considerable variability in post‐storage RBC circulation remains in 90% of RBC units that are not G6PDdef, and there are currently no approaches that can identify units that will circulate well versus those that will circulate poorly. Likewise, there are no methods to identify specific donors whose blood will store better than others. As such, a deeper mechanistic understanding of the determinants of RBC storage is highly important.

We have reported that oxidative damage in the form of lipid peroxidation correlates with poor posttransfusion recovery (PTR) in both humans and mice and that preventing oxidative damage during storage improves post‐transfusion circulation. ⁹ , ¹⁰ , ¹¹ , ¹² Along these lines, we have now identified a series of genes mechanistically involved in oxidative stress biology for which differential expression in RBC donors affects RBC storage, including G6PD, ⁶ STEAP3, ¹¹ , ¹³ FADS1/2, ¹⁴ , ¹⁵ LPCAT3, ¹⁶ SLC22A16, ¹⁶ and Gpx4. ¹⁷ Dysregulation of these gene products results in a ferroptosis‐like demise of the stored RBCs, in both humans and mice. ¹¹ Here, we identify hypomorphic genetic variants of ATP11c that affect RBC storage and PTR through cellular changes in membrane and cytoskeletal remodeling that are distinct from the oxidative damage observed in previously reported ferroptosis‐like biology. ¹¹

MATERIALS AND METHODS

Please refer to Supporting Information methods for an extended version of this section.

Mice

A/J (Cat#000646), KK/HIJ (Cat#002106), FVB (Cat#001800), and UBC‐GFP mice (Cat#004353) were purchased from Jackson Labs. All procedures ¹⁸ complied with UVA IACUC protocols.

Blood storage, transfusion, and post‐transfusion circulation

RBC storage and transfusion followed established protocols. ¹⁸ Briefly, blood was stored at 4°C and spiked with mCherry RBCs for tracking; PTR was assessed 24 h post‐transfusion by flow cytometry.

REDS RBC‐Omics study

The REDS‐III RBC‐Omics study enrolled 13,091 consented donors across four US centers; hemolysis and omics assays were performed on stored RBCs as described. ¹⁹ , ²⁰ A total of 643 donors scoring in the 5th and 95th percentile for hemolysis parameters at the index phase of the study were invited to donate a second unit of pRBCs, a cohort henceforth referred to as “recalled donors.” These units were assayed at storage Days 10, 23, and 42 for hemolytic parameters and mass spectrometry‐based high‐throughput metabolomics, ²¹ proteomics, ²² lipidomics, ²³ and ICP‐MS analyses. ²⁴ Under the aegis of the REDS‐IV‐P project, ²⁵ a total of 1929 samples (n = 643, storage Day 10, 23, and 42) were processed with this multi‐omics workflow.

Determination of hemoglobin and bilirubin increment via the vein‐to‐vein database

Association of ATP11C SNPs with hemoglobin increments was performed by interrogating the vein‐to‐vein database, as described in Roubinian et al. ²⁶ , ²⁷

Blood samples from two families with novel ATP11C mutations

Blood was obtained from two families with hereditary hemolytic anemia (Policlinico di Milano, Italy), under IRB‐approved protocols. ²⁸ RBCs from HEM, HET, and HD individuals were processed for omics analysis.

RESULTS

Poor storage of KK/HIJ RBCs is not due to an inherent decrease in RBC circulatory lifespan

We have previously reported a wide range of post‐transfusion RBC circulation as a result of RBC storage in 13 distinct inbred strains of mice ²⁹ (Figure 1A). Decreased expression of STEAP3 was identified as a major determinant responsible for increased PTR in the three best storing strains. ¹³ However, there was a wide range of PTR in the remaining 10 strains, which was not explained by STEAP3 (Figure 1A, labeled as Steap3 High). Of these strains, A/J and KK/HIJ were chosen to study both because they have the highest and lowest PTRs, respectively, and also because KK/HIJ is from a phylogenetic arm not well represented in previous studies using outbred mice. ¹⁰ , ¹¹ The goal of the current study was to perform a genome‐wide association study (GWAS) on the F2 generation of breeding between A/J and KK/HIJ mice, similar to the approach used to isolate STEAP3 from an F2 cohort of B6 and FVB mice. ¹³

Figure 1.

Naturally occurring genetic variants of ATP11c differentially regulate murine RBC post‐transfusion circulatory capacity. We previously reported that stored RBCs from different mouse strains store differently, ¹⁰ and that these differences are in part explained by genetic heterogeneity in the region coding for Steap3. ¹³ , ¹⁴ However, significant heterogeneity in post‐transfusion recovery (PTR %) is observed across mouse strains coding for the same Steap3 variant (A). Crossing female A/J and male KK/HIJ mice (B, C)—two strains with high Steap3 levels but significantly different PTR—we observed a sex dimorphism in F2 progeny as a function of chromosome X recombination (C). Multiple breeding cycles contributed to isolating on mice on an A/J background a region on chromosome X derived from KK mice that is associated with poor PTR (D).

Only humans with normal peripheral blood parameters are accepted as blood donors. As such, only mice with normal hematological parameters are a reasonable model for RBC blood storage. Analysis of the MGI phenome database of inbred strains of mice indicated that KK/HIJ mice have normal hematological parameters and reticulocyte counts, consistent with a normal RBC circulatory lifespan (Supporting Information S1: Figure 1). However, to test RBC lifespan directly, KK/HIJ mice were subjected to a full‐body biotinylation pulse‐chase protocol. ³⁰ KK/HIJ RBCs had a circulatory life‐span consistent with the general value of 55‐65 days in mice and was the same as B6 RBCs (Supporting Information S1: Figure 2A). In contrast, the better‐storing A/J strain had a slightly shorter RBC lifespan. Together, these data demonstrate that the poor storage of KK/HIJ mice is not the result of an intrinsic RBC circulation defect, nor are KK/HIJ mice anemic. As a result, KK/HIJ mice meet the criteria as a model of poor storage of RBCs donated by volunteers with normal peripheral blood parameters.

Isolation of a locus on the X chromosome associated with poor RBC storage in mice

KK/HIJ sires were bred with A/J dames. RBCs from F1 mice had the same PTR as A/J mice (Figure 1B), suggesting that genes conferring poor storage quality in KK/HIJ mice were recessive in nature. We thus predicted that RBCs from approximately 25% of F2 mice would have poor storage. Testing of 117 F2 mice demonstrated that female mice had a unimodal population with a normal distribution (Shapiro–Wilk P = 0.437) and a mean 44.9% PTR, consistent with the parental A/J strain (Figure 1C). In contrast, the PTRs from male F2 mice had a non‐Gaussian distribution that fit a bimodal population based on a Bayesian Information Criterion (BIC = 780) or a trimodal population (BIC = 774). Treating the population as bimodal, 42% of mice were in the low PTR group (mean = 11.6%), which performed significantly worse than the female population (Mann–Whitney, P = 4.9 × 10⁻¹³). In contrast, there was no significant difference in PTRs between female F2 mice and male high PTR group (Mann–Whitney, P = 0.083). This result was consistent with an X‐linked inheritance pattern wherein the locus causing poor storage was recessive in female mice but manifested in 50% of male mice that were hemizygous due to having only one X chromosome.

Due to the X‐linked pattern of inheritance, a modified congenic backcrossing approach was undertaken to isolate the KK/HIJ locus on the X chromosome responsible for poor storage (Figure 1C). Since males are hemizygous for the X chromosome, only females can have X chromosome‐based recombination during gametogenesis. Thus, heterozygous females were generated by breeding poorly storing males from the F2 generation with A/J females. The resulting F2.N1 dames were then crossed with A/J sires, and male offspring were both genotyped and phenotyped for PTR. Males with low PTRs that had the smallest amount of KK/HIJ DNA in the X chromosome were then chosen as sires, and the two‐generation breeding strategy was repeated. After four generations, a male was isolated that had PTR consistent with the parental KK/HIJ strain, with an isolated congenic region of KK/HIJ DNA in an otherwise A/J X chromosome. Two more rounds of backcrossing to A/J removed any detectable A/J derived DNA in other chromosomes. Storage of RBCs from the resulting male congenic mice had PTRs as low as the KK/HIJ parental strain (Figure 1D). Mice were then interbred to generate a new strain designated A.KK‐ChrX^KK.

Based upon flanking SNPs, the isolated region was between 1112 and 2510 kb in size. A conservative estimate defines the region as 61 485 356–58 975 265 bp on the X chromosome. This region contains 10 protein‐encoding genes, 22 pseudogenes, 2 snRNAs, 2 miRNAs, and 16 lncRNAs (Supporting Information S2: Table 1); 27 QTL have been reported in this region, but none of the mapped phenotypes were measured in RBCs. The only QTL with biology potentially related to RBCs mapped splenic weight and liver iron content, respectively. However, neither KK/HIJ nor A/J strains were used in the mapping studies that identified these QTLs.

A.KK‐ChrX^KK has decreased expression of ATP11c on RBCs

Of the 10 protein‐coding genes in the congenic region of A.KK‐ChrX^KK, only ATPase phospholipid transporting 11c (ATP11C) is detected in the RBC proteome, both in mice (Supporting Information S1: Figure 2B) and humans, as determined from published deep characterization of murine and human proteomes of ultra‐pure RBCs from our group ³¹ and others. ³² However, it remains formally possible that non‐coding elements within the congenic interval—including regulatory snRNAs, miRNAs, lncRNAs, or pseudogenes not captured by our proteomic analyses—may contribute to or modulate the observed phenotype, including through effects on ATP11c expression or other linked loci.

ATP11c flips phosphatidylserine (PS) and phosphatidylethanolamine (PE) to the inner leaflet of the phospholipid bilayer in an ATP‐dependent fashion, thereby maintaining phospholipid asymmetry. ³³ ATP11c is conserved throughout evolution with an 89% identity and 94% similarity between mouse and human orthologues. ³⁴ Because extracellular PS on the RBC surface promotes clearance by phagocytes through recognition by scavenger receptors, ³⁵ ATP11c activity is predictably essential for RBCs to continue circulating. The same phenotype, characterized by hemolysis and mild anemia, is observed in both humans with mutations resulting in very low expression and in ATP11c knockout mice. ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ Because there are no amino acid differences between ATP11c in KK/HIJ and A/J mice, we hypothesized that KK/HIJ and A.KK‐ChrX^KK RBCs have decreased ATP11c expression. Consistently, A/J RBCs had easily detectable membrane levels of ATP11c by mass‐spectrometry, while KK/HIJ and A.KK‐ChrX^KK RBCs had extremely low/undetectable levels of ATP11c.

Metabolomic and lipidomic analyses reveal oxidative and lipid remodeling signatures

Metabolomics and lipidomics profiles (Figure 2A,B) separated A/J and A.KK‐ChrXKK RBCs, particularly after 3 days of storage. Storage of ATP11c‐deficient RBCs was marked by increased lysophospholipids and reduced storage lipid pools and long‐chain fatty acids (Figures 2B and 3C,D). Together, these data suggest that A.KK‐ChrXKK RBCs have an impaired biology of lipid remodeling, especially after storage. Pathway enrichment analysis confirmed significant dysregulation in glycerophospholipid metabolism, carnitine shuttle activity, and ceramide biosynthesis (Figure 3B). In contrast to the strong association of lipid peroxidation with poor RBC storage, A.KK‐ChrX^KK RBCs had decreased levels of lipid peroxidation products, including lipid alcohols 9‐,13‐hydroxyoctadecadienoic acid (HODEs) and 9,15‐hydroxyeicosatetraenoic acid (HETEs) (Figure 3A). No significant differences were observed in major RBC antioxidant pathways, such as the pentose phosphate pathway and glutathione (Supporting Information S1: Figure 3).

Figure 2.

Omics perturbations in fresh and stored RBCs from A.KK‐ChrX ^KK mice (ATP11c mice). (A) Fresh and (3‐day‐)stored RBCs from A/J mice or hypomorphic ATP11c mice are characterized by significant metabolic, lipidomics (B), and proteomics alterations (C). Lipid and protein abbreviations follow standard LipidMaps and Official Gene Symbol annotations.

Figure 3.

Stored RBCs from hypomorphic ATP11c mice have lower lipid peroxidation and higher levels of lysophosphatidylserines (LPS) and lysophosphatidylethanolamines (LPE). Lipid alcohols hydroxyeicosatetraenoic acid (HETEs) and hydoxyoctadecadienoic acid (HODEs) accumulate during storage in A/J mice and—significantly less so—in A.KK‐ChrX^KK mice (A). Lipid pathway analysis via BioPan ⁵⁰ revealed defects in storage lipids (triacyl‐ ad diacyl‐glycerols—TG and DG), LPEs and LPSs in hypomorphic ATP11c mice (B). Compared to parent A/J mice, these mice showed lower end of storage levels of poly‐ and highly unsaturated free fatty acids (PUFA and HUFA–C), with rewiring of elongation and desaturation pathways (D). In (E) pathway analysis of the most significant (two‐way ANOVA by storage duration and genotype) proteomics changes in hypomorphic ATP11c mice compared to A/J mice. ANOVA, analysis of variance.

Proteomic profiles of ATP11c‐deficient RBCs highlight membrane remodeling and stress responses

Widespread proteomics changes were observed in numerous gene products not contained in the congenic KK/HIJ region in A.KK‐ChrX^KK mice (Figure 2C,D). The top discriminatory features included elevation in proteins involved in vesicle‐mediated transport and oxidative stress response. Pathway analysis demonstrated an enrichment in proteins involved in chromatin remodeling, ribosomal assembly, and protein translation (Figure 3E).

ATP11c variants in humans associate with altered RBC storage phenotypes

To investigate if the identification of ATP11c as a factor in RBC storage translates to human populations, we queried the REDS‐III RBC‐Omics cohort (n = 13,091 donors). Multiple single‐nucleotide polymorphisms (SNPs) in the region coding for ATP11c, including several missense mutations (i.e., rs55724992 (V972M), rs17281983 (Y522C), and rs140504622 (R194C)), were identified with appreciable minor allele frequencies (up to 20%) across the donor population (Figure 4A–C). SNPs rs12687833, rs6654371, and rs2485724 were significantly associated with RBC osmotic fragility (Figure 4D). An additional SNP of potentially high importance emerged (rs55724992), which is a missense variant coding for V972M in close proximity to the ATP11C substrate‐binding site (Figure 4B). rs55724992 (V972M) was most prevalent in hemizygous male donors of African descent, tendentially younger and with higher BMIs (Figure 4E). RBCs from donors carrying this SNP had the lowest levels of N6‐methyl‐lysine, nicotinamide, and ornithine, and the highest levels of highly unsaturated very‐long chain fatty acids (22:5 and 22:6) (Figure 4F). Additional multi‐omics correlates to other ATP11c SNPs are presented in Supporting Information S1: Figure 4. Of note, the association between ATP11c genotypes or protein levels was lost when focusing on the freshest available units from REDS Recalled donors (i.e., storage Day 10), with overlapping omics phenotypes between ATP11c high and low donors (n = 308 per group), with no significant lipidome changes noted and limited effects on the proteome (Supporting Information S1: Figure 5).

Figure 4.

ATP11c polymorphism is associated with osmotic fragility and is prevalent among donors of African descent in REDS RBC Omics Index volunteers. Of the 13,091 blood donor volunteers enrolled in the index phase of the REDS RBC Omics study (A), multiple missenses mutations were observed (B), along with several intronic SNPs with minor allele frequencies ~20% (C). These SNPs were significantly linked to osmotic fragility (D) and were prevalent in male donors of African descent (E). Donors carrying one or two alleles of the missense rs55724992 V972M SNP had higher RBC levels of PUFAs (FA 22:5; FA 22:6) and pentose phosphate isomers (F). In this figure, 0, 1, 2 indicate alleles (2 merges homozygous females and hemizygous males).

ATP11c protein levels are modulated by age and storage duration in human RBCs

Donors ranking below the 5th or above the 95th percentile for hemolysis parameters in the 13,091 index cohort were invited to donate a second unit of blood (“recalled donor cohort”), which was stored for a total of 42 days and sampled at storage Days 10, 23, and 42 for multi‐omics analyses (Figure 5A). Specifically, proteomics analyses allowed us to directly quantify ATP11c in 643 recalled donor RBC units, which revealed significant inter‐donor variability (Figure 5B). ATP11c levels were not significantly associated with donor age (Figure 5B) but were significantly higher in blood units from females after three weeks of storage (Figure 5C), or at donation if the unit came from donors of Hispanic descent (Figure 5C).

Figure 5.

RBC levels of ATP11c proteins in REDS Recalled donors are linked to hemolysis. Of the 13,091 index donors, 643 REDS RBC omics donors were invited to donate a second independent unit for multi‐omics analyses (A). In these units, RBC protein levels of ATP11c were measured through mass spectrometry and linked to donor age (B) or storage duration (C) as a function of donor sex and ethnicity. RBC levels of ATP11c were inversely correlated with complement and coagulation component deposition on stored RBCs (D), markers of hemolytic propensity. In (E), heat map of the top 50 significant multi‐omics signatures in REDS Recalled donors as a function of storage duration or ATP11c protein levels (n = 100 donors with the highest or lowest ATP11c protein level out of the full Recalled population). In (F, G) elevated levels of ATP11c proteins are associated with lower osmotic and oxidative hemolysis in REDS Recalled donors (n = 100 per group).

To better understand how proteomic alterations caused by the ATP11C V972M variant affected metabolic flux, we clustered proteomic data obtained from REDS Recall donors ³¹ within HD and HET phenotype groups (Supporting Information S1: Figure 6A). Out of 1877 samples with genotyping for ATP11C V972M, 1836 samples from HD donors and the 35 samples from HET donors were selected into 18 and 8 representative samples, respectively. The cluster centroids were treated as representative samples of the HD phenotype and HET phenotypes, while the 6 HEM samples in their original state were treated as representative samples. We formulated proteome‐constrained models and performed flux variability analysis to obtain the maximum flux, effective flux range, and associated enzyme abundances. Using the Spearman rank correlation coefficients calculated between maximum flux and enzyme abundance for 905 flux‐carrying enzyme‐catalyzed reactions, we classified 441 reactions as abundance‐dependent (ρ ≥ 0.8), 78 reactions as abundance‐correlated (0.5 ≤ ρ < 0.8), and 386 reactions as abundance‐independent (ρ < 0.5) (Supporting Information S1: Figure 6B). Analysis of the top 150 altered flux ranges with statistical significance (P < 0.002) revealed increased flux capacity within membrane remodeling involving ether‐lipids as well as protein dephosphorylation activity (Supporting Information S1: Figure 6C,D). Of note, PS‐flippase ⁴¹ activity was classified as completely abundance‐dependent (Supporting Information S1: Figure 6D). These results suggest that blood donors who are hemizygous for ATP11C V972M will exhibit a collapse in membrane asymmetry earlier than RBCs without this phenotype, likely leading to an altered hemolytic propensity.

Correlation of multi‐omics results to ATP11C protein levels identified a negative correlation with complement and coagulation components (Figure 5D,E), previously linked to hemolytic propensity. ⁴² SNPs rs55724992 and rs17281983, which encode missense changes, showed strong associations with shifts in membrane lipid metabolites and redox biology, consistent with findings in RBCs from ATP11c hypomorphic KK/HIJ and A.KK‐ChrX^KK mice (Figure 5E). Comparing the top and bottom quartiles of donors ranked by ATP11c protein abundance identified low ATP11c levels as associating with increased osmotic and oxidative hemolysis (Figure 5F,G), which would predictably result in lower hemoglobin increments after transfusion. ⁴²

ATP11c SNPs correlate with decreased transfusion efficacy in humans

The REDS III RBC Omics vein‐to‐vein database tracks hemoglobin increments post‐transfusion in patients receiving units that also have multi‐omics data. ⁴³ Recipients of units from homozygous females or hemizygous males encoding the rs55724992 (V972M) SNP had significantly lower hemoglobin increments following transfusion (n = 4460 transfusion events; Figure 6A,B). This effect was gene dose‐dependent, as no significant difference was observed in females who are heterozygous for rs55724992. Since similar effects on lower hemoglobin increments had been previously reported for relatively prevalent G6PD SNPs (e.g., V68M—~6% of the total REDS donor population ²⁶ , ⁴⁴ ), and both ATP11c and G6PD are coded by genes on chromosome X, we formally tested whether any of the ATP11c reported in the present study—especially the rs55724992 one—were in linkage disequilibrium (LD) with any of the G6PD SNPs previously linked to hemolytic propensity of stored RBCs in REDS. Sex‐aware LD analyses confirmed that ATP11C V972M is not in meaningful linkage disequilibrium with the common G6PD variants, including functional and most prevalent (MAF > 6%) A+ N126D and A− V68M variants (Supporting Information S1: Figure 7A–C).

Figure 6.

Missense ATP11c mutation V972M is linked to significantly lower hemoglobin increments. Interrogating the vein‐to‐vein database, we evaluated the effect on hemoglobin increment of a single‐unit transfusion event involving blood units from donors carrying 0, 1, or 2 alleles of the missense ATP11c mutation V972M (rs55724992—A). Results indicate significantly (FDR‐adjusted P < 0.05) lower hemoglobin increments (40%–60% decrease) in recipients of blood units from donors carrying two alleles of this SNP (n = 4460 transfusion events—B). This observation is consistent with a likely functional role of the V to M amino acid substitution at residue 972 in proximity to the substrate‐binding pocket of the ATP11c enzyme (mapped against its published structure—6LKN.pdb in C).

Extreme ATP11C hypomorphism in humans is associated with distinct rbc metabolic and proteomic remodeling

Although anemic individuals are deferred from donation, assessing the effects of extreme ATP11c deficiency on RBC biology is both informative in understanding ATP11c effects on RBCs and also important to understand the molecular pathology of individuals with hemolysis and mild anemia due to ATP11c deficiency. We analyzed RBCs from two unrelated families with rare pathogenic ATP11c variants, recently reported in a clinical case study, ²⁸ but not yet characterized through omics approaches. In both families, hemizygous males (HEM) presented with non‐transfusion‐dependent mild hemolytic anemia. ²⁸ In Family 1, the affected patient (second child, with splenomegaly) harbored the L448Yfs*9 mutation, while in Family 2 the proband carried an insertion of FQ residues between K145 and V146 (c.436‐7A>G) (Figure 7A). Of note, the patient in family 1 also had a concomitant de novo ANK1 mutation, which aggravated the phenotype and contributed to a diagnosis of hereditary spherocytosis. In both cases, heterozygous female carriers (HET) were clinically asymptomatic, while hemizygous males (HEM) had macrocytosis, elevated reticulocyte counts, and chronic compensated anemia. Notably, in one case (Family 1), hemolytic symptoms were exacerbated following a transient parvovirus infection, consistent with increased susceptibility to bone marrow stress (Figure 7B).

Figure 7.

Extreme ATP11C hypomorphism in humans is associated with distinct RBC metabolic and proteomic remodeling. Pedigrees of two unrelated families with pathogenic ATP11C variants. Hemizygous (HEM) males presented with hemolytic anemia and splenomegaly, while heterozygous (HET) female carriers were clinically asymptomatic (A). Family 1 carried a frameshift mutation (L448Yfs*9); Family 2 carried an insertion of F and Q residues between amino acids K145 and V146 (c.436‐7A>G—A). Structural model of ATP11C (PDB: 6LKN) indicating variant sites L448 and V145‐K146 region, relative to the substrate binding pocket (B). Heatmap of untargeted metabolomics data from RBCs of healthy donors (HD), HET, and HEM individuals, showing distinct clustering by genotype. Highlighted metabolites include increases in lysophospholipids (LPE, LPC), bilirubin, and spermine in HEM RBCs (C). Box plots quantifying genotype‐dependent differences in representative metabolites: lysophosphatidylethanolamine (LPE 20:3—D), bilirubin (a marker of hemolysis—E), and intracellular ATP (F). HEM samples exhibited significant increases compared to HD and HET. Statistical significance determined by one‐way ANOVA with multiple comparisons (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; ns, not significant). Heatmap of RBC proteomics profiles from HD, HET, and HEM individuals. HEM samples showed decreased abundance of cytoskeletal and membrane‐organizing proteins (e.g., ankyrin, spectrin, band 3 complexes), consistent with compromised structural integrity and enhanced vesiculation (G). HET samples displayed intermediate patterns, suggesting partial dosage effects and potential mosaicism due to X‐inactivation.

Untargeted metabolomics and proteomics analyses were performed on RBCs from HEM, HET, and HD (canonical variant hemizygous) individuals and revealed distinct separation between genotypes (Figure 7C). RBCs from ATP11C‐deficient individuals (HEM) showed increased levels of lysophospholipids (LPE 20:3; Figure 7D) and hemolysis (bilirubin; Figure 7E). Of note, dysfunctional ATP11C was accompanied by increases in ATP pools (Figure 7F), consistent with a significant role for ATP11C‐dependent flippase activity in the economy of intracellular ATP content.

Proteomic analysis showed decreased abundance of cytoskeletal and membrane‐organizing proteins (e.g., ankyrin, β‐spectrin, and band 3 complexes) in HEM RBCs compared to HD or HET (Figure 7G). These changes suggest compromised structural integrity and increased susceptibility to vesiculation or clearance. Notably, the HET samples displayed intermediate profiles, indicating partial dosage effects even in the absence of overt clinical anemia (Figure 7G). However, this may represent a bimodal distribution with two distinct RBC phenotypes due to Barr body inactivation of either the canonical or variant ATP11c gene.

DISCUSSION

This study identifies levels of ATP11c expression and/or activity as a determinant of RBC storage quality and post‐transfusion efficacy in both mice and humans. Importantly, these findings represent a qualitatively distinct mechanism of poor RBC storage. Other identified genetic determinants of poor RBC storage (e.g., G6PD, ⁶ STEAP3, ¹¹ , ¹³ GPX4, ¹¹ , ¹⁷ FADS1/2, ¹⁴ LPCAT3, ¹⁴ SLC22A16, ¹⁶ PFKP ⁴⁵ ) are all regulators of oxidative stress biology and have a common endpoint of lipid peroxidation, or its mitigation (e.g., l‐carnitine and glutathione), which results in ferroptosis of the RBC. In contrast, ATP11c hypomorphism results in the opposite effect—decreased oxylipin formation. Rather, ATP11c hypomorphism is associated with alterations in RBC lysophospholipid metabolism and membrane remodeling. Presumably, ATP11c has not been detected in other murine GWAS studies since the ATP variants in the KK/HIJ genome are absent from other strains utilized for tested populations. ¹¹ , ¹³

The discovery of effects of ATP11c hypomorphism using an unbiased forward genetics approach in mice, followed by a gene‐focused association study that uses a hypothesis‐driven approach using human data, demonstrates the power of combining translatable models with human studies that avoid decreased statistical power from the multiple‐testing corrections present in genome‐wide association studies. Of course, the use of murine models as an instrument of discovery always has the potential vulnerability of murine biology not translating into human biology. However, the currently used model of RBC storage and transfusion has now been successfully translated for at least 4 genes, including the new ATP11c finding.

It is important to acknowledge that the studies described herein observe powerful associations of ATP11c variation with RBC storage but do not formally establish causality. One major utility to model systems is the ability to test causality through approaches with the aim of modifying one (and only one) variable and observing alterations in effects. Indeed, this has been accomplished with regards to hereditary anemia caused by ATP11c deficiency through the reported anemia in ATP11c knockout mice. ⁴⁰ However, in the case of ATP11c in KK/HIJ and human blood donors, as with much modern biology research, the putative causal genetics is the effect of hypomorphic variants. Thus, to test causality, one must selectively increase or decrease the expression of a gene (in the correct tissue and potentially the correct stage of development or environmental induction) rather than simply deleting it.

KK/HIJ mice have multiple polymorphisms in ATP11c regulatory elements spanning large distances of DNA. This limits the feasibility of molecularly complicated gene editing experiments (i.e., knocking the KK/HIJ promoter and/or enhancer into A/J or B6 mice either by traditional ES approaches or CRISPR‐Cas9, not to mention any regulatory elements in introns or exons). Congenic backcrossing over multiple generations addresses this technical challenge; however, (as in this report), recombination around the congenic genetic elements is never perfect, and effects of other elements in linkage disequilibrium cannot be ruled out. As such, while it is our view that the current data provide sufficient evidence from multiple orthologous approaches to conclude a causal role for ATP11c in poor RBC storage, it remains formally possible that the association is not causal. However, the combined investigation of murine models of storage, over 13,000 REDS III RBC Omics donors, and two independent families carrying novel ATP11c mutations are, in our opinion, sufficient to establish a causal mechanistic role for ATP11c in RBC susceptibility to hemolysis in vivo and in vitro.

Given the known role of ATP11c as a flippase, it seems likely that the mechanism of decreased PTR In ATP11c hypomorphs is increased exofacial PS that is then recognized by scavenger receptors on phagocytes, a hallmark of RBC eryptosis, ⁴⁶ consistent with observations in the two Italian families with ATP11c mutations. ²⁸ However, owing to logistical, ethical approval or technical limitations, scramblase activities were not measured in the hypomorphic ATP11c mice or in the REDS RBC Omics samples. These results confirm and expand predictions on the deleterious role of storage‐associated depletion of ATP pools, which fuel ATP11c activity, in regulating PTR. ³¹ Our findings also raise the possibility of general membrane instability and lipid remodeling in hypomorphic ATP11c RBCs. This notion is further supported by the downregulation of cytoskeletal and structural proteins, suggestive of increased vesiculation and reduced mechanical stability, processes that favor RBC splenic sequestration and extravascular hemolysis after storage and transfusion. ⁴⁷ The current findings of increased ATP11c in stored RBCs from female donors may partially explain the observation that blood from female donors stores better (in general) than male donors, in both humans and mice. ⁴⁸

Of note, these findings may have direct implications for transfusion practice. While individuals with overt anemia are deferred from blood donation, carriers of hypomorphic ATP11C variants often present with compensated hemolysis and only mild or absent anemia. Current donor screening relies on complete blood count analyses, but does not assess hemolysis‐related markers, such as reticulocyte count or unconjugated bilirubin. As a result, in the absence of a clinical history of hemolysis or acute hemolytic events, donors harboring ATP11C variants could remain undetected and proceed to donation. The growing recognition that genetic heterogeneity among donors influences RBC storage quality and transfusion outcomes underscores a paradigm shift toward precision transfusion medicine. ⁴⁹ In this framework, genetic and molecular donor profiling can be used to match stored RBC units to recipients based on expected in vivo survival and hemolytic propensity. In the case of ATP11C, carriers of hypomorphic or loss‐of‐function alleles might not require permanent deferral from donation, but could be preferentially assigned to supply fresher units or used in contexts where storage duration is minimized. Conversely, donors with higher ATP11C expression or wild‐type alleles could be prioritized for chronic transfusion recipients—such as patients with sickle cell disease, thalassemia, or myelodysplastic syndromes—where prolonged RBC survival translates into fewer transfusions, reduced iron overload, and a lower risk of alloimmunization. These findings also align with ongoing efforts within the REDS‐IV‐P consortium to evaluate genomic typing arrays that integrate transfusion‐relevant SNPs, including those in ATP11C, G6PD, ⁶ STEAP3, ¹¹ , ¹³ PKFP, SLC7A5, ⁴² FADS1/2, ¹⁴ , ¹⁵ LPCAT3, ¹⁶ SLC22A16, ¹⁶ and GPX4, ¹⁷ into standard donor testing pipelines. As sequencing and array costs continue to fall, genotyping is expected to become routine at the time of first donation, effectively replacing current serologic blood typing at comparable cost while providing a more comprehensive view of donor biology. From a cost–benefit perspective, even modest improvements in transfusion efficacy for CTPs could translate into substantial savings and reduced morbidity at the population level.

In summary, the practical ramification of the growing number of genetic markers associated with poor RBC storage raises the possibility of genetic screening of donors to identify specific individuals whose RBCs are predicted to circulate for extended periods of time compared to the average RBC unit. Providing such units to CTPs has the potential to decrease the number of transfusions required, thereby decreasing iron burden (and potential toxicity) as well as decreasing the probability of alloimmunization that may complicate ongoing transfusion.

Beyond transfusion medicine, ATP11C variants may have broader evolutionary and physiological relevance. The high prevalence of the common missense variant V972M among donors of African ancestry raises the intriguing hypothesis that ATP11C hypomorphism may have been retained under selective pressure, possibly conferring partial protection against malaria, as has been demonstrated for G6PD deficiency and certain hemoglobin variants. Although our current data do not permit a definitive test of this hypothesis, it can be noted that RBC‐Omics samples were obtained from US African Americans, virtually all of whom descend from ancestors from locations where malaria is endemic. Future population‐genetic analyses of ATP11C variant frequencies in malaria‐endemic regions may shed light on this possibility and its evolutionary trade‐offs between pathogen resistance and RBC lifespan.

Mechanistically, our results suggest that decreased ATP11C expression impairs the maintenance of phospholipid asymmetry and membrane integrity rather than inducing the oxidative stress‐driven ferroptotic phenotype observed in other genetic determinants of poor RBC storage. Direct measurement of ATP11C‐dependent flippase activity in fresh human RBCs from variant carriers, as well as targeted CRISPR‐based modulation of ATP11C expression in murine or ex vivo erythroid models, will be important next steps to establish causality and further dissect the interplay between membrane lipid remodeling, vesiculation, and RBC survival.

Taken together, despite the limitations noted above, this study identifies ATP11C as a novel and evolutionarily conserved regulator of RBC membrane homeostasis, storage quality, and post‐transfusion efficacy. By combining murine forward genetics, large‐scale human donor cohorts, and rare familial cases, we delineate a cross‐species mechanism linking phospholipid flippase function to transfusion outcomes. These findings extend the molecular taxonomy of the red cell storage lesion beyond oxidative stress pathways and provide a roadmap for integrating genetic, metabolic, and proteomic donor profiling into clinical decision‐making, ultimately advancing the goal of personalized, genome‐informed transfusion medicine.

AUTHOR CONTRIBUTIONS

James C. Zimring: Conceptualization; investigation; funding acquisition; methodology; supervision; writing—original draft; project administration. Ariel M. Hay: Investigation; methodology. Monika Dzieciatkowska: Investigation; methodology. Daniel Stephenson: Investigation; methodology. Zachary B. Haiman: Investigation; software. Steven Kleinman: Supervision; conceptualization; funding acquisition. Philip J. Norris: Conceptualization; funding acquisition; supervision. Michael P. Busch: Conceptualization; investigation; funding acquisition; resources; supervision. Nareg Roubinian: Investigation; formal analysis; funding acquisition. Elisa Fermo: Investigation; conceptualization; resources. Paola Bianchi: Conceptualization; investigation; resources. Gregory R. Keele: Formal analysis; software; investigation; data curation. Grier P. Page: Supervision; conceptualization; investigation; software; formal analysis. Angelo D'Alessandro: Conceptualization; investigation; funding acquisition; methodology; supervision; writing—original draft; project administration.

CONFLICT OF INTEREST STATEMENT

Though unrelated to the present study, AD declares that he is a founder of Omix Technologies Inc., and an advisory board member for Hemanext Inc., Macopharma Inc., and SynthMed Bio. JCZ is a co‐founder of Svalinn Therapeutics, a company that develops therapies unrelated to RBC storage. None of the resources for the current studies were obtained from any companies or corporate interests.

FUNDING

A.D. and J.C.Z. were supported by funds from the National Heart, Lung, and Blood Institute (R21HL150032, R01HL146442, R01HL149714, R01HL148151). The REDS RBC Omics and REDS‐IV‐P CTLS programs are sponsored by the NHLBI contract 75N2019D00033, and from the NHLBI Recipient Epidemiology and Donor Evaluation Study‐III (REDS‐III) RBC Omics project, which was supported by NHLBI contracts HHSN2682011‐00001I, −00002I, −00003I, −00004I, −00005I, −00006I, −00007I, −00008I, and −00009I. N.R. received funding from NHLBI (R01HL126130). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. P.B. and E.F. received funding from the Italian MoH, FOMP‐RC2024 175/01.

Supporting information

Supplementary Methods and Figures.

Supplementary Table 1.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available in the supporting material of this article

All raw data and elaborations are included in Table S1.