Red cell shape regulation by band 3–ankyrin–spectrin linkage: implications for clinical severity of bovine hereditary spherocytosis

To survive in the circulation while being subjected to continuous shear stress conditions for >100 days, the red blood cell (RBC) must be highly flexible and durable, undergoing marked reversible deformation without fragmentation or the loss of membrane surface area. The RBC's deformability and stability are maintained by the structural organization of the membrane skeletal proteins, transmembrane proteins, and lipid bilayer.

The membrane skeleton supports the structural integrity of human RBCs by providing the mechanical stability against shear deformation.^1-4 Spectrin, actin, and protein 4.1R are the principal components of the membrane skeletal network. Spectrin is a long, flexible rod-like protein consisting of multiple spectrin repeats. Spectrin α and β chains associate with each other to form the αβ heterodimer in an antiparallel fashion, and spectrin tetramers are formed by head-to-head self-association of 2 dimers. Distal ends of spectrin dimers are connected to short F-actin filaments and 4.1R within the junctional complex.^4-6 The spectrin skeletal network is linked to the lipid bilayer through interactions with transmembrane proteins, specifically spectrin–ankyrin–band 3 and spectrin–4.1R–glycophorin C associations.^7-9 Such vertical interactions allow the membrane skeleton network to stabilize the membrane lipid bilayer.^4,10-13 Various mammalian red cell membranes seem to share these principal structural and functional organizations.^14-20 

Defects in these interactions lead to membrane fragmentation or surface area loss associated with abnormal shape changes, such as hereditary elliptocytosis (HE) and hereditary spherocytosis (HS) and resulting hemolytic anemias. The principal lesion of HE involves horizontal interactions, primarily spectrin tetramer formation, due to mutations at the NH₂ terminus of α-spectrin or the COOH terminus of β-spectrin.^12,21,22 However, a weakening of the vertical interaction is a common feature of HS. The lipid bilayer with reduced support by the membrane skeleton is destabilized, leading to release of skeleton-free vesicles and resultant spherocytosis.¹³ 

We previously revealed that the total deficiency of band 3 due to a premature termination mutation in SLC4A1 in bovine RBCs led to loss of band 3–ankyrin–spectrin linkage, thereby markedly reducing the cohesion between the lipid bilayer and the spectrin-based skeleton resulting in severe spherocytosis.¹⁴ However, the RBC membrane pathology differed significantly in one respect from that of the band 3-null mouse models.^15,16 The spectrin content of bovine band 3–deficient RBCs was reduced by ∼50%, and vesicle fragments from these RBCs contained spectrin, suggesting that bovine band 3 deficiency is implicated not only in reduction of the membrane surface area but also in the lateral interactions of the membrane skeletal proteins. In contrast, the spectrin content and the structure of the membrane skeleton in band 3–deficient murine RBCs and band 3-null human RBCs were nearly normal.^15,16,23,24 

In this regard, an earlier study²⁵ suggested that spectrin–ankyrin interaction of high affinity (K_a = 10⁻⁷ M) constrains the spectrin to a narrow submembranous space resulting in high local concentration of spectrin (>10⁻⁴ M) enabling effective tetramer formation. Furthermore, the interactions between spectrin, ankyrin, and band 3 in vitro were coupled in a positively cooperative manner.²⁶ Moreover, disruption of the band 3–ankyrin–spectrin link led to dissociation of a large proportion of the spectrin tetramers into dimers.²⁷ Collectively, these findings have suggested that the band 3–ankyrin–spectrin linkage contributes not only to the lipid bilayer stabilization but also to the formation of spectrin tetramers. However, to date, there have been no clinical cases or biological models to clearly demonstrate this hypothesis in vivo.

In this study, we report that a novel substitution E91K in α-spectrin, found in several cases of bovine band 3 deficiency, induces structural disruption of the spectrin repeat α1 leading to destabilization of the spectrin tetramer. Although the E91K substitution exacerbates the primary phenotype of band 3 deficiency, it causes only mild spectrin reduction and no significant abnormalities in RBCs with normal band 3 content. These findings provide substantial in vivo evidence to support a pivotal role for band 3–ankyrin–spectrin linkage in promoting the mechanical properties of RBC membrane skeleton.

Band 3–deficient Japanese black cattle homozygous for the R664X mutation of SLC4A1¹⁴ and healthy control Japanese black cattle were kept at the animal experimentation facility of the Veterinary School of Hokkaido University. All experimental procedures were approved by the Laboratory Animal Experimentation Committee, Graduate School of Veterinary Medicine, Hokkaido University, with an approval number 18040. All other cattle were housed at several different locations in Japan, and their blood samples anticoagulated with EDTA were transported to Hokkaido University for analysis.

Routine hematological parameters were determined using the hematological analyzer ProCyte (IDEXX Laboratories). Microscopic examination of red cells and reticulocytes was carried out as described previously.¹⁴ 

Scanning electron micrography was performed as described previously.¹⁴ RBC vesiculation/fragmentation was examined under phase-contrast light microscopy after incubation of washed RBCs in phosphate-buffered saline (PBS, 10 mM sodium phosphate, pH 7.4, 154 mM NaCl) and at ambient temperature for 6 hours. RBCs bearing extruded vesicles were counted for 200 RBCs.

Preparation of RBC membrane ghosts and Triton X-100–extracted membrane skeletons (Triton shell), sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), and immunoblotting to analyze membrane proteins, including band 3, spectrin, 4.1R, and ankyrin, were performed as described previously.^{14,17,18,20,28,29} The abundance of different polypeptides was determined by densitometric scanning of SDS gels stained with Coomassie brilliant blue using a ChemiDoc imager (Bio-Rad) and Image Lab software (Bio-Rad). Detection and imaging of immunoblots were similarly performed.

RBCs were washed in PBS and suspended at hematocrit of 10%. Then, 1 mL of RBC suspension was applied to a 5-mL syringe attached to a Millex-HA membrane filter (Millipore, SLHA025BS) and pressurized manually at a rate of 1 drop per 5 seconds until the filtration was stopped. The vesicles in the filtrate (∼0.6 mL) were collected by centrifugation at 120 000g for 30 minutes at 4°C through a cushion of 20% sucrose in PBS and washed once in PBS. Vesicles were directly dissolved in 20 μL sample buffer for SDS-PAGE and stored frozen at −20°C until analysis by SDS-PAGE.

For decreasing the interactions between membrane skeleton and lipids, RBCs suspended in PBS were incubated for 30 minutes at 37°C in the presence or absence of 50 μM 4,4′-diisothiocyanostilbene-2,2′-disulfonate (DIDS; Cayman Chemical), washed, and resuspended in PBS.³⁰ 

Deformability and mechanical stability of the RBC membranes were examined as described previously.^31,32 

All data are expressed as the mean ± standard deviation (SD). Statistical significance was determined by using unpaired Student t tests, 1-way analysis of variance, or the Mann-Whitney test as indicated in the legends, and differences with a value of P < .05 were considered statistically significant. All statistical analyses were performed using GraphPad Prism 9.0 (GraphPad).

Analysis of complementary DNA and genomic DNA of bovine α- and β-spectrin, genotyping, preparation of plasmids, production of recombinant proteins, and their characterization are described in the supplemental Methods.

We first demonstrated that band 3–deficient cattle homozygous for the R664X mutation of SLC4A1 can be clearly divided into 2 groups, type 1 and type 2 (Figure 1A-C). Both type 1 and type 2 RBCs had spherocytosis and anisocytosis, with little noticeable difference in morphology (Figure 1A). However, owing to incubation at ambient temperature for several hours, ∼70% of type 1 RBCs had prominent blebs and protrusions under phase-contrast microscopy, whereas no such obvious vesicle extrusion was found in the RBCs from type 2 and control animals (Figure 1B-C).

The membranes from type 1 RBCs had a marked reduction in the major membrane skeletal proteins, spectrin, 4.1R, and actin, along with the total lack of band 3 and protein 4.2, as reported previously.¹⁴ Spectrin content was markedly reduced to 49.7% ± 3.8% (mean ± SD, n = 3; P < .01) of control membranes (mean ± SD, 100.0% ± 4.4%, n = 3). The Triton shells from type 1 RBCs also had a marked reduction in the content of membrane skeletal proteins. However, type 2 band 3–deficient animals had no apparent reduction in membrane skeletal proteins in both RBCs and Triton shells: the spectrin content in RBC membranes was similar to that in control membranes (mean ± SD, 106.7% ± 4.7%, n = 3, P = .135; Figure 1D). Interestingly, both types exhibited inclusion of albumin in the membrane fraction, indicating invagination of the RBC membrane.¹⁴ 

We tested whether vesicles released from type 2 band 3–deficient RBCs contained spectrin, as we previously reported for type 1 RBCs.¹⁴ Because type 2 RBCs had no spontaneous fragmentation, we mechanically induced RBC vesiculation. Compared with control RBCs, suspensions of the 2 different types of band 3–deficient RBCs had lesser resistance to filtration. Noticeably, however, the volume of vesicles obtained from type 2 RBCs was almost the same as that obtained from control RBCs and approximately one-fifth obtained from type 1 RBCs. Whereas vesicles from type 1 RBCs contained readily detectable amounts of α- and β-spectrin, vesicles from type 2 RBCs contained no appreciable amounts of spectrin and had a protein profile very similar to that of vesicles from control RBCs, except for the absence of band 3 (Figure 1E).

These findings demonstrate that type 2 RBCs represent the predominant phenotype of band 3 deficiency, membrane surface area loss without significant reduction in membrane skeletal components, as reported in mouse models and human band 3 deficiency.^15,16,23,24 Thus, marked fragility of type 1 RBCs is a composite phenotype caused by band 3 deficiency and an additional defect.

We analyzed the exon sequences of erythroid SPTA1 in a control and 1 for each type 1 and type 2 band 3–deficient cattle, and we found 3 independent alleles with different nucleotide sequences, SpαA, SpαB, and SpαBK91 (supplemental Figure 1). The SpαA and SpαB alleles produce α-spectrin isoforms A and B, respectively, that are distinct from each other at 14 amino acid residues (Figure 2A; supplemental Figure 1). The SpαBK91 allele contains an additional substitution in the SpαB backbone, generating the E91K variant of isoform B. The SpαBK91 was originally found in a type 1 band 3–deficient animal, and the following survey revealed that all band 3–deficient individuals with type 1 RBCs (n = 3) were heterozygous for the SpαBK91 allele, whereas no SpαBK91 allele was detected in individuals with type 2 RBCs (n = 3). Importantly, those with type 1 band 3 deficiency with heterozygous SpαBK91 were more anemic with larger decreases in hematocrit and hemoglobin values than those with type 2 band 3 deficiency which did not harbor the spectrin E91K substitution (Table 1), implying that the E91K substitution increased the clinical severity of spherocytosis due to band 3 deficiency.

To assess the impact of the E91K substitution independent of band 3 deficiency, we analyzed spectrin content and membrane mechanical properties of RBCs from healthy cattle with various SPTA1 genotypes. Genotyping of 167 individuals for SPTA1 (Figure 2A; supplemental Figure 2) revealed that SpαA and SpαB were the major alleles in both Japanese black and Holstein Friesian cattle. The SpαBK91 allele was only found in Japanese black cattle with an allele frequency of ∼15%. Quantification of the relative abundance of spectrin revealed that spectrin content was reduced by ∼15%, in RBC membranes homozygous and heterozygous for K91 α-spectrin (K/K and E/K, respectively; Figure 2B; supplemental Figure 3), compared with the membranes containing only E91 α-spectrin (E/E; Figure 2B). Comparison among different SPTA1 genotypes further revealed that spectrin was also decreased in RBCs from individuals with SpαB/SpαBK91 and SpαBK91/SpαBK91 genotypes (B/BK91 or B91K/B91K), compared with SpαB/SpαB RBCs (B/B; Figure 2C). The E91K substitution is thus the dominant factor in reducing spectrin content.

Ektacytometry analysis revealed that the deformability index (DI) values increased with increasing applied shear stress for red cells with normal band 3 of all 3 genotypes E/E, E/K, and K/K with E/E RBCs having a difference in DI curve at low shear stress possibly due to tumbling of RBCs.^33,34 Interestingly, the maximal DI values for K/K and E/K RBCs were similar and higher than E/E RBCs (Figure 2D). In terms of membrane mechanical stability, K/K and E/K RBCs readily fragmented under high sheer stress conditions (750 dynes per cm²), reaching minimum DI values within 20 seconds, whereas E/E RBCs fragmented gradually >100 seconds (Figure 2E), implying the E91K substitution is the dominant determinant of membrane mechanical stability. Taken together with the lack of significant differences in messenger RNA levels among different SPTA1 genotypes (supplemental Figure 4), the allele-specific reduction in spectrin is attributable to impaired assembly of the membrane skeleton.

Despite marked instability of RBC membranes in vitro under nonphysiologically high shear forces, the individuals possessing K91 α-spectrin had no noticeable abnormalities in RBC morphology and red cell indices (supplemental Figure 5), implying that fragmentation and loss of membrane components due to E91K substitution proceed slowly under physiological conditions in the circulation.

To investigate how the E91K substitution affects the mechanical stability of the membranes, we assessed the impact of the E91K substitution on the structure of recombinant bovine α[0-1] and α–β-fused mini-spectrins (Figure 3A). Structure prediction of bovine α[0-1] by the Phyre2 server³⁵ depicted a domain structure similar to its human counterpart in both the unbound³⁶ and bound²² states with no notable difference in the overall structure between α[0-1]E91 and α[0-1]K91 (Figure 3B-C), although several changes due to the E91K substitution were suggested in the vicinity of the E91K substitution site in α[0-1]K91, including the disappearance of interhelical hydrogen bonds found for the bound state of α[0-1]E91 (Figure 3D).

The circular dichroism (CD) of α[0-1]E91 had typical α-helical secondary structure with an α-helix content of 56.3%, whereas α[0-1]K91 had decreased α-helix content of 46.7%, indicating some unfolding due to E91K substitution. Reduction in α-helix content was also apparent in mini-SpK91 (Figure 4A). The thermal denaturation profile of α[0-1]E91 revealed clear cooperative transition from an α-helical to a disordered structure between 46 and 60°C with a transition midpoint (T_m) of ≃54°C, consistent with the previously reported T_m for α1^37,38 or α[0-1]¹⁴ of human α-spectrin. In contrast, α[0-1]K91 had a reduced ellipticity even at 25°C and an almost linear decrease with no discriminate transition in ellipticity from 25 to 70°C, reflecting progressive unfolding (Figure 4B). The thermal denaturation of mini-SpE91 exhibited at least 2 transitions, first T_m at ≃43°C followed by the second at ≃51°C suggesting cooperative multistate unfolding. Notably, mini-SpK91 had markedly reduced ellipticities from 25 to 60°C with a clear helical-to-random transition with T_m ≃43°C which corresponded to the first transition for mini-SpE91, implying that the E91K substitution destabilizes local structure of α[0-1] without significant structural alterations in distal repeats.

In the ¹H-¹⁵N HSQC spectra at 25°C, numerous signal peaks were distributed evenly in the spectra for both ¹⁵N-α[0-1]E91 and ¹⁵N-α[0-1]K91 (Figure 4C). However, less dispersion of chemical shifts was apparent for ¹⁵N-α[0-1]K91. At 40°C, most of the signal peaks were lost in the spectrum for ¹⁵N-α[0-1]K91, consistent with the reduced thermal stability of α[0-1]K91 in CD analysis. Notably, Hε-Nε signals presumably derived from 2 conserved tryptophan residues (W59 and W131) observed for ¹⁵N-α[0-1]E91 disappeared in the ¹⁵N-α[0-1]K91 spectrum (¹H, 10.2–10.5 ppm; ¹⁵N, 125–126 ppm), indicating less ordered structure in the α1 region of α[0-1]K91 in the vicinity of K91 and spatially adjacent conserved tryptophan residues likely involved in hydrophobic interhelical interactions^39,40 (Figure 3D).

Fluorescence spectra from W59 and W131 in α1 and the susceptibility of the fluorescence intensity to a hydrophilic collisional quencher acrylamide were analyzed. The α[0-1]E91 and α[0-1]K91, including their W59F mutants, had fluorescence spectra with the λ_max (maximum emission intensity wavelength) of 330 to 340 nm and reduction in the fluorescence intensities by the addition of acrylamide in a concentration-dependent manner (Figure 4D). The K_SV values for α[0-1]K91 and α[0-1]K91/W59F were significantly higher than those for α[0-1]E91 and α[0-1]E91/W59F, respectively (Figure 4E-F). In contrast, α[0-1]E91/W131F and α[0-1]K91/W131F had blue shifts of λ_max to ∼330 nm, and their K_SV values were nearly equal to each other (Figure 4D-F). Therefore, tryptophan residues in α[0-1]K91 and α[0-1]K91/W59F are more susceptible to acrylamide compared with those in their counterparts containing E91,^41-43 indicating higher exposure of the W131 residue presumably due to destabilization of α1.

Collectively, these data indicate that the E91K substitution causes structural destabilization greater than predicted (Figure 3B-D), including unfolding and unwinding in the triple-helical bundle of the α1 domain.

To unravel how the disordered α1 leads to the impaired membrane skeletal organization, we assessed the effect of E91K substitution on oligomer states of α–β-fused mini-spectrin “dimers” as previously described.⁴⁴ Purified “tetramer (dimers of α–β-fused mini-spectrin)” fractions, with a Stokes radius of ∼80 Å, of SpE91 and SpK91 were collected and concentrated. Following incubation at 37°C for 1 hour, mini-SpE91 and mini-SpK91 eluted in 2 peaks corresponding to the tetramer (80 Å) and dimer (54 Å) fractions (Figure 5A). The relative abundance of the tetramer in mini-SpK91 was markedly lower than that in mini-SpE91 (means ± SD, 38.9% ± 18.6% vs 86.6% ± 8.0%, n = 3, P < .05; Figure 5B). The 1:1 mixture of mini-SpE91 and min-SpK91 also demonstrated intermediate values between that of either species (68.9% ± 11.5%, n = 3), indicating that the tetramer containing mini-SpK91 is less stable and more readily dissociated to dimer than that formed by mini-SpE91.

Finally, we evaluated the impact of the loss of membrane bilayer–skeleton linkage, characteristic of band 3 deficiency, on the stability of RBC membranes with reduced spectrin tetramerization by treatment of RBCs with DIDS which reduces band 3-ankyrin association³⁰ (Figure 5C-H; supplemental Figure 6). Vesicles obtained from E/E and E/K RBCs by filtration-induced hemolysis contained band 3 and negligible amount of spectrin (Figure 5C-D). Similarly, vesicles generated by E/E RBCs treated with DIDS before filtration had negligible content of spectrin. Importantly, the vesicles obtained from DIDS-treated E/K RBCs contained remarkable amount of spectrin (Figure 5C-G) and the junctional complex constituent 4.1R (Figure 5H), suggesting local fragmentation of the membrane skeleton and release of its components into the vesicles.

These data demonstrate that the E91K substitution impairs the mechanical stability of the membrane skeleton through reduced spectrin tetramerization and that disturbing the membrane–membrane skeletal association exacerbates the membrane loss.

This study revealed that weakened spectrin tetramerization due to the E91K substitution in α-spectrin markedly exacerbates spherocytic phenotypes due to band 3 deficiency. Generation of spectrin-free vesicles in filtration-induced hemolysis from type 2 band 3–deficient RBCs without E91K substitution, as well as those from control RBCs (E/E RBCs), indicated that the membrane skeleton in type 2 spherocytes was nearly intact. In marked contrast, in type 1 band 3–deficient RBCs possessing E91K substitution, increased dissociation of spectrin tetramers into dimers causes local accumulation of disrupted spectrin skeleton and consequent release of the membrane vesicles containing fragmented membrane skeleton, including spectrin and 4.1R. Thus, HS in type 1 RBCs with extremely fragile membranes, originally reported for bovine band 3 deficiency,¹⁴ is caused by combined phenotypes involving the surface area loss due to complete lack of band 3 and membrane fragmentation due to impaired spectrin self-association, features of HS and HE, respectively.^12,13,21 Thus, the E91K substitution in bovine α-spectrin is a novel SPTA1 allele-specific modulator of membrane defects, including band 3 deficiency.

Notably, however, E/K and K/K RBCs with reduced spectrin tetramerization but normal band 3 contents did not demonstrate significant hematological lesions, although these cells exhibited marked mechanical instability under nonphysiologically high shear force in ektacytometry, suggesting that these RBCs possess some mechanism(s) that limits the damage due to altered spectrin tetramerization. Spectrin tetramerization occurs through head-to-head binding of adjacent spectrin dimers connected to the junctional complex. Since the binding affinity between dimers is extremely low,^33,45,46 the probability of dimers encountering head-to-head is a major determinant for tetramer formation. If adjacent dimers connected to the junctional complex are bound to the lipid bilayer through band 3–ankyrin association, the spatial movement of the dimer heads could be significantly restricted, promoting the self-association of spectrin dimers. Such a contribution of the band 3–ankyrin–spectrin linkage in spectrin tetramer formation was originally suggested by Morrow and Marchesi.²⁵ This hypothesis has been strengthened by the findings on the positively cooperative interaction between spectrin, ankyrin, and band 3²⁶ and the dissociation of a large proportion of spectrin tetramers into dimers by disrupting the spectrin–ankyrin–band 3 link.²⁷ 

Our analysis also revealed that membrane fractions prepared from type 2 band 3–deficient RBCs, but not from E/K and K/K RBCs, contained the plasma protein albumin indicating that the surface area loss in band 3 deficiency involves endocytic invagination followed by exocytic extrusion of microvesicles, as we previously reported for type 1 RBCs.¹⁴ This process is very similar to the release of exosomes in membrane remodeling during reticulocyte maturation.^47,48 Because band 3 is not contained in the exosomes released and remains totally associated with the RBC membrane during reticulocyte maturation,⁴⁹ exosome formation and thus exosome-like microvesicle formation in band 3–deficient RBCs seem to occur in membrane compartments lacking band 3. It will be interesting to determine whether such invagination is the consequence of an inward membrane curvature imposed by altered transmembrane protein–lipid interactions¹¹ or occurs through the interaction of such membranes with the spectrin skeleton as a scaffolding machinery.⁵⁰ 

A significant finding of this study is that the E91K substitution in the α1 domain affects the function of the adjacent dimer-dimer self-association site. Although many HE-associated α-spectrin mutations have been mapped to the partial repeat α0,¹² some HE mutations are located distant from the critical tetramerization site, mainly the linkers joining helices C and A or the C-terminal region of helix C adjacent to the linker.^12,22,38 The only exception reported to date is the common L207P mutation in the α2 domain,^51,52 which is located in the middle of helix B as is the case for E91K. The L207P mutation, as well as another common HE-associated mutation L260P, has been found to cause no extensive unfolding but shift the dimer-tetramer equilibrium toward closed dimers due to alterations in the triple-helical bundle of α2 to a more compact and stable structure, resulting in reduced and destabilized formation of spectrin tetramers.^6,53 In contrast, the E91K substitution seems to cause unfolding of α-helices and destabilization accompanied by unwinding of the triple-helical bundle in α1, suggesting a mechanism different from that for L207P mutation to reduce self-association of dimers. Intriguingly, in human α-spectrin, α0 is connected to α1 with a flexible linker in the unbound state and these regions are stabilized by forming a continuous α-helical structure consisting of α0, helix A in α1, and the linker region between α0 and α1 on binding with β-spectrin.^22,36 The tertiary state of the α1 domain with E91K substitution implies disturbed conformational change of α[0-1] or perturbation of proper relative positioning of β16, β17/α0, and α1 repeats required in tetramer formation.²² Moreover, the disordered structure of α1 may lead to the impairment of the supposed interstrand interactions of α1 (at K70 and E125/E126 in the vicinity of the E91K substitution site) with α2 to α3 repeats in the parallel strand of α-spectrin within the tetramer.⁶ The prediction by the MutationExplorer software (http://proteinformatics.org/mutation_explorer/)⁵⁴ reveals that the relevant mutation E100K in human α[0-1] causes structural destabilization with ΔΔG values of −2.78 and −6.47 in unbound and bound states, respectively, indicating that similar destabilization of spectrin tetramerization is expected in human RBCs. However, we still need to consider its potential effect on the allosteric effect of spectrin-ankyrin association to spectrin dimer self-association,²⁶ which could not be evaluated in our tetramer formation experiment using mini-spectrin proteins. Further kinetic and molecular dynamic studies for the interactions between band 3, ankyrin, and intact spectrin variants in red cells and in solutions would be required to precisely determine the mechanism for destabilized tetramer formation due to E91K substitution, in turn the physiology of spectrin tetramer stabilization by the band 3–membrane skeleton link.

In summary, our findings on naturally occurring cases of a novel substitution E91K in bovine SPTA1, in combination with the band 3 deficiency-causative SLC4A1 mutation, substantiate the longstanding hypothesis on an important role for band 3–ankyrin–spectrin linkage in maintaining the mechanical properties of the RBC membrane, in addition to its well-established function in stabilizing the membrane lipid bilayer. The findings also imply that combination of different membrane protein mutations working in conjunction can account for variable clinical severity of red cell membrane disorders, including HS.

The authors thank Akira Ban, Mikio Konno, Kenji Wada, Jun-ichi Sakai, and Yoshimi Ogata (Yamagata Prefectural Federation of Agricultural Mutual Relief Association) and Kazutaka Iwata (Iwata Animal Clinic) for collection of blood samples; Wataru Otsu, Daisuke Ito, Kota Sato, Keitaro Morishita, Takashi Kikukawa, and Momoko Fujimoto (Hokkaido University) for technical support; Kimiko Ito (Tokyo Women’s Medical University) for ektacytometry; Mitsuhito Matsumoto and Satoshi Tamahara (University of Tokyo) for technical support; Ken-ichiro Ono (University of Tokyo) for intellectual support; and Makoto Nakao and the late Sumie Manno (Tokyo Women’s Medical University) and the late Wataru Nunomura (Tokyo Women’s Medical University and Akita University) for constructive discussion.

This study was supported, in part, by Grants-in-Aid for Scientific Research 15658096, 16H05031, 20H03138, and 23K27065 from the Japan Society for Promotion of Science (M.I.) and DK32094 from the National Institute of Diabetes and Digestive and Kidney Diseases (N.M.).

Contribution: M.I. designed the research; K.M., M.T., T.T., O.I., and M.I. designed the experiments, performed the experiments, and analyzed the data; N.A., I.K., and Y.O.-Y. performed the experiments and analyzed the data; M.D. analyzed the data; K.K., A.D., and E.K. performed the experiments; K.M., M.T., and M.I. wrote the original manuscript; and N.M., Y.T., and M.I. analyzed the data, made critical intellectual contributions throughout the research, and edited the manuscript.

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

The current affiliation for E.K. is Division of Regenerative Medicine, Center for Molecular Medicine, Jichi Medical University, Tochigi, Japan.

The current affiliation for K.M. is Research Center, Mochida Pharmaceutical Co, Ltd, Gotemba, Japan.

The current affiliation for M.T. is Graduate School of Veterinary Science, Osaka Metropolitan University, Osaka, Japan.

The current affiliation for Y.O.-Y. is Division of Biomedical Information Analysis, Iwate Tohoku Medical Megabank Organization, Iwate Medical University, Iwate, Japan.

Correspondence: Mutsumi Inaba, Laboratory of Molecular Medicine, Graduate School of Veterinary Medicine, Faculty of Veterinary Medicine, Hokkaido University, Sapporo 060-0818, Japan; email: inazo@vetmed.hokudai.ac.jp.