The study was approved by the Medical Ethics Committee of UCLouvain and University Medical Center Utrecht (Study 17–450). After informed consent, blood from the patient and 4 healthy volunteers was collected by venipuncture into K⁺/EDTA-coated tubes at the University Medical Center Utrecht. Whenever possible, healthy donors were selected to be age- and gender-matched. After collection, the tubes were transferred to the research laboratory at UCLouvain. Two splenectomised healthy donors were included in the study. Before experiments, RBCs were collected through 10-fold blood dilution in a glucose- and HEPES-containing medium [Dulbecco’s modified eagle medium (DMEM), Invitrogen]. Diluted blood was centrifugated at 200 g for 2 min, the supernatant removed and RBCs suspended in medium. RBCs were washed a second time by centrifugation at 200 g for 2 min and resuspended, as in Cloos et al. (2020). The number of RBCs used for experiments were as follows: (i) Hb release, 12.5 × 10⁶; (ii) intracellular ATP content, 5 × 10⁶; (iii) calcium and ROS contents, 15 × 10⁶; (iv) lipidomic and FA analyses, 2.5–6 × 10⁸; (v) PS surface exposure, 0.5 × 10⁶; (vi) cholesterol assay, 37.5 × 10⁶; and (vii) fluorescence imaging, 12.5 × 10⁶.

Chemical treatments were performed on washed RBCs. To activate calcium entry through PIEZO1, RBCs were incubated with 0.1 μM Yoda1 (Bio-Techne) for 30 s at RT. To deplete the intracellular calcium content, RBCs were preincubated in a calcium-free medium containing 1mM EGTA (Sigma-Aldrich) for 10 min at RT and maintained during the experiment. To induce oxidative stress, RBCs were submitted to 100 mM H₂O₂ for 60 min at 37°C.

RBC morphology was determined (i) on blood smears, (ii) upon suspension of living RBCs in μ-dish IBIDI chambers, (iii) by SEM on fixed RBCs, and (iv) after immobilization of living RBCs on poly-L-lysine (PLL)-coated coverslips. For the RBCs in suspension, washed RBCs were diluted 24-fold in DMEM, deposed in μ-dish IBIDI chambers and observed with a wide-field fluorescence microscope (Observer.Z1; plan-Apochromat 100× 1.4 oil Ph3 objective). For SEM microscopy experiments, RBCs were prepared and analyzed exactly as in Cloos et al. (2020). For RBC immobilization, PLL was deposed on coverslips for 30 min at 37°C and then washed. RBCs were then spread on the PLL-coated coverslip during 4 min and the coverslip was then placed upside down in a medium-filled LabTek chamber (Fisher Scientific). RBCs were observed with the fluorescence microscope Observer.Z1 as above.

Washed RBCs were incubated for 10 min at RT in isotonic (320 mOsm, DMEM) or increasingly hypotonic media (264–0 mOsm, DMEM mixed with water to obtain the desired osmolarity) and then pelleted by centrifugation at 200 g for 2 min. Supernatants and pellets broken with 0.2% (w/v) Triton X-100 were both assessed for Hb at 450 nm in 96-well plates (SpectraCount^TM, Packard BioScience Co.). Hb release in the supernatant was expressed as percentage of the total Hb present in the sample.

Deformability of RBCs was measured with the Laser Optical Rotational Red Cell Analyzer (Lorrca, RR Mechatronics, Zwaag, Netherlands). In this ektacytometer, RBCs are exposed to shear stress in a viscous solution (Elon-Iso, RR Mechatronics), forcing the cells to elongate in an elliptical shape. The diffraction pattern, that is generated by a laser beam, is measured by a camera. The vertical axis (A) and the horizontal axis (B) of the ellipse are used to calculate the elongation index (EI) by the formula (A − B)/(A + B). The EI reflects the deformability of the total population of RBCs. Two different forms of ektacytometry were used: osmotic gradient ektacytometry and cell membrane stability test (CMST). Osmotic gradient ektacytometry measurements of RBCs were obtained using the osmoscan module, according to the manufacturer’s instructions and as described elsewhere (Da Costa et al., 2016; Lazarova et al., 2017). Briefly, 250 μl whole blood was standardized to a fixed RBC count of 1000^∗10⁶ and mixed with 5 ml of Elon-Iso. RBCs in polyvinylpyrrolidone (PVP) were exposed to an osmolarity gradient from approximately 60–600 mOsmol/L, whereas shear stress was kept constant (30 Pa). The CMST was performed using the CMST module on the ektacytometer. To perform a CMST, 50 μl of whole blood was standardized to a fixed RBC count of 200^∗10⁶ and mixed with 5 ml of Elon-Iso. In the CMST, RBCs are exposed to a shear stress of 100 Pa for 3,600 s (1 h) while the EI is continuously measured. The change in the elongation index (ΔEI) was calculated by determining the median of the first and the last 100 s of the CMST and subsequently calculating the difference between the medians. The ΔEI depicts the capacity of the RBCs to shed membrane and resist shear stress.

The intracellular ATP level was determined by a chemiluminescence assay kit (Abcam) as described in Conrard et al. (2018) and Cloos et al. (2020). ATP levels were then reported to the corresponding Hb content measured by spectrophotometry.

Intracellular calcium content was determined by RBC labeling with Fluo4-AM (Invitrogen) as in Cloos et al. (2020). Labeled RBCs were then either analyzed by flow cytometry (FACSVerse, BD Biosciences) or fluorimetry (GloMax; Promega; λ_exc of 490 nm and λ_em of 520 nm). Flow cytometry data, obtained by medium flow rate on 10.000 events, was analyzed with the software FlowJo to determine the median fluorescence intensity (MFI) of analyzed RBCs. Fluorimetry data were normalized to the global Hb content determined spectrophotometrically as above.

RBC FA composition was assessed exactly as described in Cloos et al. (2020).

The cholesterol content of washed RBCs was determined with a fluorescent assay kit (Invitrogen) as in Cloos et al. (2020) and reported to the global Hb content as described above. The membrane composition in phospholipids, lysophospholipids, sphingolipids and oxysterols was analyzed by lipidomics according to Guillemot-Legris et al. (2016) and Mutemberezi et al. (2016), as described in Cloos et al. (2020).

Intracellular free ROS were assessed by RBC labeling with 2,7-dichlorodihydrofluoresceindiacetate (H₂DCFDA) in Krebs–Ringer–Hepes (KRH) solution for 30 min at 37°C. MFI of the whole RBC population was acquired as explained above for Fluo4-AM. Lipid peroxidation was evaluated by measurement of malondialdehyde (MDA) with a lipid peroxidation assay kit (Abcam) according to the high sensitivity protocol as in Cloos et al. (2020). Methemoglobin (metHb) was analyzed in RBC lysates, produced through repeated freeze–thaw cycles, with a sandwich Elisa kit (Lifespan Biosciences) as described in Cloos et al. (2020). Both MDA levels and metHb contents were reported to the global Hb content.

Endogenous cholesterol was evidenced through living RBC labeling with mCherry-Theta toxin fragment followed by immobilization on PLL-coated coverslips (Cloos et al., 2020). To visualize GM1 ganglioside, sphingomyelin and ceramide, living RBCs were spread onto PLL-coated coverslips and then labeled by trace insertion in the plasma membrane of BODIPY fluorescent analogs of those lipids (Invitrogen) as described in Conrard et al. (2018). Coverslips with living immobilized RBCs were placed in medium-filled LabTek chambers and observed with the wide-field fluorescence microscope (Observer.Z1; plan-Apochromat 100× 1.4 oil Ph3 objective).

To determine the presence of mitochondria fragments in RBCs, washed RBCs were labeled with 100 nM mitoTracker (Invitrogen) at RT for 30 min, washed, placed on PLL-coated coverslips and observed with the wide-field fluorescence microscope Observer.Z1.

RBCs were placed into IBIDI chambers as described above for RBC morphology in suspension. Microscopy images were then analyzed with the last version of Shape Analysis by Fourier Descriptors computation plugging for ImageJ as in Leonard et al. (2017b) and Pollet et al. (2020).

Exposure of PS was assessed by RBC labeling with Annexin-V coupled to FITC (Invitrogen) and analyzed by flow cytometry as in Cloos et al. (2020). The proportion of PS-exposing RBCs was then determined with the FlowJo software by positioning the cursor at the edge of the healthy RBC population.

AFM experiments were performed with a Bioscope Resolve AFM (Bruker) at ∼25–30°C in DMEM. PeakForce QNM Live Cell probes (Bruker) with spring constants of 0.10 ± 0.02 N m^–1 and tip radius of curvature of 65 nm were used. The spring constant of the cantilevers was calibrated with a vibrometer (OFV-551, Polytec, Waldbronn) by the manufacturer. The pre-calibrated spring constant was used to determine the deflection sensitivity (Schillers et al., 2017) using the thermal noise method (Hutter and Bechhoefer, 1993) before each experiment. In fast indentation experiments (Dumitru et al., 2018), the AFM was operated in PeakForce QNM mode and Force-distance (FD)-based multiparametric maps were acquired using a force setpoint of 300 pN. The AFM cantilever was oscillated vertically at 0.25 kHz with a peak-to-peak oscillation amplitude of 500 nm. Height and Young’s modulus maps were recorded using a scan rate of 0.2 Hz and 256 pixels per line. Slow indentation experiments were performed in Force-Volume (FV) mode. Individual FD curves were recorded in contact mode on the RBC surface with a force setpoint of 300 pN, using ramp speeds of 2 μm s^–1 for a 2 μm ramp size. Measurements were performed in the central region of RBCs to avoid substrate effects. Indentations ranged between 300 nm and 1 μm depending on cell type.

Immunolabeling of spectrin was performed as in Cloos et al. (2020) and Pollet et al. (2020). All coverslips were mounted with Dako and examined with a Zeiss Cell Observer Spinning Disk (COSD) confocal microscope using a plan-Apochromat 100× NA 1.4 oil immersion objective and the same settings for illumination.

RBC membrane area, RBC curvature and spectrin intensity/occupation were determined using the Fiji software. Lipid domain abundance was determined by manual counting and expressed by reference to the hemi-RBC projected area. The proportion of spiculated RBCs and of RBCs presenting lipid- or mitoTracker-enriched vesicles, patches or spicules was also assessed by manual counting on fluorescence images. For AFM, at least 64 FD curves were recorded per cell and analyzed using the Nanoscope Analysis v9.1 software (Bruker). Hertz model was used to extract Young’s modulus values from individual FD curves. Whole RBC elasticity was analyzed from Slow indentation experiments in FD-AFM mode, by fitting the repulsive part of FD curves with the Hertz model. To analyze the contribution of the plasma membrane and cytoskeleton to the cellular mechanical behavior, two different fit ranges were defined in the repulsive part of FD curves obtained in Fast indentation experiments (PeakForce QNM mode). Plasma membrane elastic contribution was estimated for indentations δ < 50 nm and cytoskeleton elasticity was extracted from δ between 50 and 100 nm. Data are expressed as means ± SEM when the number of independent experiments was n ≥ 3 or means ± SD if n ≤ 2. For all experiments, statistical tests were performed only when n ≥ 3. Tests were non-parametrical Mann–Whitney test or Kruskal–Wallis followed by Dunn’s comparison test, except for AFM and membrane curvature data for which two sample t-tests were applied as RBCs were analyzed individually and data obtained for all RBCs are presented. To evaluate the effect of chemical agents, the paired data were analyzed by paired non-parametrical tests (Wilcoxon matched-pairs signed rank tests). ns, not significant; ^∗p < 0.05, ^∗∗p < 0.01, ^****p < 0.0001.

The patient pHypoβ is a 49 year-old adult. Due to idiopathic thrombocytopenic purpura, his spleen has been removed at the age of 17 as a curative treatment. He has no additional treatment. The transaminases are increased (Supplementary Table 1) but he has no anemia and is otherwise healthy. He exhibits since more than 30 years a high number of acanthocytes on blood smears (Figure 1A), which was suspected to result from hypobetalipoproteinemia. The patient also presented a vitamin E level close to the lower reference limit (Supplementary Table 1). Despite of acanthocytosis, reticulocyte count, RBC mean corpuscular volume and Hb levels are within the normal range (Supplementary Table 1).

Next-Generation Sequencing of pHypoβ revealed a heterozygous single nucleotide deletion in APOB, the gene encoding ApoB. It concerns the deletion of adenosine at nucleotide 2,534 in exon 17, which leads to a shift in the reading frame and consequent premature end of translation at residue 862 (Human Gene Mutation Database, CD051293). Nonsense-mediated mRNA decay likely will prevent translation of such mutant APOB mRNA into a severely truncated protein. This mutation has already been identified (Fouchier et al., 2005), but consequences for RBC morphology, cytoskeletal and biophysical properties as well as functionality were not evaluated.

We started by confirming the presence of spiculated RBCs detected on a blood smear (Figure 1A) by optical microscopy of living RBCs in suspension and by SEM of fixed RBCs on filters (Figures 1B,C). Quantification indicated that spiculated RBCs represented ∼50% of all pHypoβ RBCs in blood smear and optical microscopy images and ∼30% in electron microscopy images while they accounted only for a minority of RBCs from healthy donors (Figures 1E–G). Discrepancies regarding the proportion of spiculated RBCs detected by electron and optical microscopy could result from the shear stress induced by filtration during RBC preparation for SEM, eventually leading to hemolysis of spiculated RBCs.

When RBCs were spread on PLL-coated coverslips, spiculated RBCs could still be distinguished from other RBCs as they presented big dark patches and smaller clearer vesicles at their surface (arrowheads and arrows at Figure 1D). The proportion of such RBCs was significantly increased in pHypoβ (Figure 1H) and corresponded exactly to the proportion determined above for non-spread RBCs. RBC morphological alterations were accompanied by an increased membrane area of spread RBCs (Figure 1I) and by the presence of Howell-Jolly bodies on blood smears (Figure 1A), compatible with splenectomy. In contrast, the increase of membrane area did not result from splenic absence since a healthy splenectomised donor did not show such a membrane surface area increase (Supplementary Figure 1A).

Despite the high proportion of pHypoβ RBCs with a modified morphology, their resistance to hemolysis was similar to healthy controls (splenectomised or not), as shown by Hb release after RBC incubation in isotonic and increasingly hypotonic media (Figures 2A,B and Supplementary Figure 1B). Nevertheless, the more sensitive osmotic gradient ektacytometry test revealed that the O_min value was slightly increased in pHypoβ (153 vs. 125–148 in healthy donors), indicative of a slight decrease in surface area-to-volume ratio and a slight increase in osmotic fragility (Figure 2C). Moreover, the EI_max, and therefore the surface area, were slightly decreased in pHypoβ RBCs (0.566 vs. 0.585–0.607). This decrease a priori contrasted with the increased surface area determined by optical microscopy on spread RBCs which could result from the increase of the outer leaflet membrane area specifically (Figure 1I). Those changes did not result from splenectomy since all osmotic gradient ektacytometry-derived parameters were normal in the healthy splenectomized control (Figure 2C). To further investigate pHypoβ RBC deformability, the CMST was performed to assess the RBC ability to shed membrane when exposed to prolonged supraphysiological shear stress force. This is a physiological property of healthy RBCs and the decrease in EI that occurs under these circumstances is proposed to be a measure of membrane health. pHypoβ RBCs showed a ΔEI of −0.144, which was considerably less than the ΔEI of −0.187 from the healthy splenectomized control, and even more less than the ΔEI of −0.208 seen in healthy donors (Figure 2D). These results indicated that splenectomy on itself was accompanied with loss of the ability to shed membrane upon shear stress, but that this loss was more pronounced in pHypoβ RBCs. We also evaluated the intracellular contents in ATP and calcium, two key parameters for RBC functionality and deformation (Bogdanova et al., 2013; Conrard et al., 2018). The intracellular ATP content was slightly increased but the calcium content remained unchanged (Figures 2E,F).

As acanthocytotic RBCs in abetalipoproteinemia and homozygous familial hypobetalipoproteinemia are associated with profound alterations of membrane lipid composition (Biemer, 1980; Barenholz et al., 1981; Gheeraert et al., 1988), we then determined whether RBC morphology alterations in pHypoβ could be associated with modified membrane composition in FAs (Figures 3A–E), cholesterol (Figure 3F), phospholipids and/or sphingolipids (Figure 4). The relative proportion of saturated (SFA), monounsaturated (MUFA) and polyunsaturated (PUFA) FAs was unchanged in pHypoβ RBCs (Figure 3A). Closer examination of major SFAs and MUFAs did not reveal more changes (Figures 3B,C). For PUFAs, however, the proportion of long chain C22 PUFAs appeared to decrease in favor of PUFAs with shorter C18 or C20 chains (Figure 3D). The increase of C18 and C20 PUFAs appeared to result from the higher relative contents in linoleic (C18:2) and arachidonic (C20:4) acids (Figure 3E).

Since the plasmatic cholesterol content was decreased in pHypoβ (Supplementary Table 1), we then evaluated the RBC membrane cholesterol content. Surprisingly, it was not modified in the patient as compared to the healthy donors (Figure 3F). Sphingomyelin species were also largely maintained (Figure 4A) whereas all ceramide and dihydroceramide species, whatever their fatty acid length and unsaturation number, were increased by 1.5- to 2-fold in pHypoβ (Figure 4B). Among phospholipids, no obvious change was detected for PS and phosphatidylinositol (PI) but phosphatidylcholine (PC) species were decreased and phosphatidylethanolamine (PE) species were very slightly increased (Figures 4C,E,G,I). In agreement with the heightened proportion of PE, four lysoPE species were also increased. The other lysophospholipids remained unaffected in pHypoβ (Figures 4D,F,H,J).

As the proportion of long chain PUFAs was decreased, we investigated whether pHypoβ RBCs could suffer from oxidative stress due to low circulating vitamin E levels (Supplementary Table 1), eventually leading to lipid and protein oxidation. We started by measuring ROS using H₂DCFDA. A significant twofold increase of free ROS was observed in pHypoβ RBCs (Figure 5A) but not in a healthy donor without spleen (Supplementary Figure 1C), indicating that the ROS increase in pHypoβ was not due to splenectomy.

The extent of lipid peroxidation, determined through the level of MDA, was not modified in pHypoβ RBCs (Figure 5B). This might seem surprising because increased lipid peroxidation could be detected in HDL and platelets of patients with abetalipoproteinemia (Calzada et al., 2013). Nevertheless, only a slight increase was observed even for healthy RBCs treated with H₂O₂ (Figure 5B) and might be explained by the fact that MDA is not the only end product of lipid peroxidation. Nevertheless, oxysterol species were also maintained at levels similar to those found in healthy RBCs (Figure 5C).

Besides lipids, Hb is also a major target of oxidative stress in RBCs, generating metHb and eventually hemichromes (Mohanty et al., 2014). By ELISA we showed that pHypoβ RBCs exhibited a very slight metHb increase, comparable to the one obtained upon treatment of healthy RBCs with H₂O₂ (Figure 5D). All those data suggested that, despite major morphological changes and increase of free ROS, no obvious alterations in lipid and Hb oxidation appeared to occur in pHypoβ RBCs.

We next evaluated by immunofluorescence the distribution of spectrin, another major target of oxidative stress (Voskou et al., 2015). Although the spectrin network was homogeneous in ∼99% of healthy RBCs, this proportion was significantly reduced to ∼75% in pHypoβ (population 1 at Figures 6A,B). Closer examination of this population indicated a tendency to decrease of the spectrin occupancy per RBC surface combined with a lower variance of spectrin occupancy (Figures 6C,D), suggesting that the spectrin network at the surface of pHypoβ RBCs was less dense than in healthy RBCs. Besides, the patient exhibited two additional populations with differential spectrin patterns, namely spectrin-enriched patches and vesicles (populations 2 and 3 at Figure 6A). The proportion of those two populations together increased by ∼25-fold as compared to healthy RBCs (populations 2 and 3 at Figure 6B). In conclusion, the spectrin cytoskeleton was altered in pHypoβ RBCs, ∼25% of them showing a patchy and vesiculated pattern and the remaining ∼75% exhibiting a less dense network.

To determine whether spectrin cytoskeleton modifications were accompanied by altered membrane biophysical properties, we analyzed RBC membrane stiffness and curvature. AFM at SLOW indentation revealed that pHypoβ RBCs were stiffer than healthy RBCs (Figure 7A). Since our previous experiments on RBCs treated with cytoskeleton-depolymerizing drugs using the same indentation approach revealed changes in the Young’s modulus, we inferred that the main contribution in the whole Young’s modulus originated from the cytoskeleton even if some contribution from the cytoplasmic viscosity cannot be discarded. To further test the contribution of the cytoskeleton in the whole RBC Young’s modulus vs. the plasma membrane stiffness, RBCs were analyzed at Fast indentation. Plasma membrane elasticity was similar in CTL and pHypoβ RBCs, while a stiffening of the cytoskeleton was observed for pHypoβ RBCs (Figure 7B). Hence, a higher variability of measurements was observed for the patient than healthy donors, which might reflect the presence of both acanthocytes and discocytes that were difficult to discriminate after RBC spreading. Interestingly, for both control and pHypoβ RBCs, the highest plasma membrane elastic modulus corresponded to the highest cytoskeleton stiffness (Supplementary Figure 2), suggesting that in diseased RBCs both the plasma membrane and the cytoskeleton were altered.

Because it was not possible to discriminate between spiculated and non-spiculated RBCs in AFM due to spreading we then used chambers compatible with RBC morphology preservation and quantified the membrane curvature of the non-spiculated RBCs (Figure 7C, population 1). Although the differential curvature between high (HC) and low curvature (LC) areas was preserved in the patient, curvature in both HC and LC areas was increased (Figures 7C,D). This latter increase was in good agreement with the lower spectrin occupancy per RBC surface in pHypoβ. Of note, no modification of membrane curvature was detected for RBCs from a healthy splenectomised donor (Supplementary Figure 1D). Altogether our data indicated that the cytoskeleton alteration in pHypoβ RBCs was accompanied by increased stiffness and curvature. To further analyze the potential alteration of the plasma membrane in the disease, we determined the membrane asymmetry both at the transversal and lateral levels.

RBC PS exposure, a measure of membrane transversal asymmetry integrity loss, was not increased in pHypoβ RBCs as compared to healthy RBCs. As internal control, we used blood stored for 2 weeks in K⁺/EDTA tubes at 4°C, which we previously validated as a model of accelerated RBC aging in vitro based on morphological, biophysical and biochemical storage lesions, including PS surface exposure (Cloos et al., 2020). Thus, according to our previous work, a strong increase of PS surface exposure was observed in this condition (Figure 7E).

We then took benefit from our expertise in submicrometric lipid domain organization at the RBC outer plasma membrane to evaluate the respective abundance of cholesterol-, GM1- and sphingomyelin-enriched domains, which contribute to the RBC deformation process (Leonard et al., 2017b; Conrard et al., 2018). Indeed, global quantitative analyses at the whole cell level may not reveal more localized alterations of acanthocyte membranes. Moreover, lipid domain abundance depends on membrane:cytoskeleton anchorage (Conrard et al., 2018). To visualize lipid domains, RBCs were labeled with a mCherry-toxin fragment specific to endogenous cholesterol or BODIPY fluorescent analogs of GM1 and sphingomyelin. Although the three types of lipid domains showed a tendency to increase in pHypoβ, only the abundance of sphingomyelin-enriched domains per hemi-RBC was significantly increased by 2.5-fold (Figures 8A,B). In contrast, no modifications of lipid domain abundance could be observed for RBCs from healthy splenectomised donors (Supplementary Figure 1E).

As GM1- and sphingomyelin-enriched domains were proposed to be implicated in calcium exchanges necessary for RBC deformation, we investigated their functionality in pHypoβ by stimulation of RBCs with Yoda1 and EGTA in a calcium-free medium to induce calcium entry and intracellular depletion, respectively. RBC incubation with Yoda1 led to an increased abundance of GM1-enriched domains in healthy RBCs, as previously shown (Conrard et al., 2018), and the same amplitude of response could be observed for pHypoβ RBCs (Figure 8C). In contrast, after intracellular calcium depletion, a heightened abundance of sphingomyelin-enriched domains was detected for healthy RBCs, as expected (Conrard et al., 2018), but not for pHypoβ RBCs (Figure 8D).

Besides well-defined submicrometric lipid domains, larger lipid-enriched patches (arrowheads at Figure 8A) and peripheral spicules (open arrows at Figure 8A) were seen in pHypoβ RBCs. We quantified the proportion of those lipid-enriched patches and spicules and their potential relationship with the patches and vesicles observed by contrast phase microscopy (see Figure 1D). Although none of the patches were enriched in cholesterol or GM1 ganglioside (white arrowheads at Figure 9A), some sphingomyelin-enriched patches were observed (red arrowheads at Figure 9A). However, their abundance was non-significantly different from the one in healthy RBCs (Figure 9B) and represented a low proportion of the patches evidenced at Figure 1D.

Intrigued by the increased content of ceramide and dihydroceramide species in pHypoβ (Figure 4B), we wondered if the patches could be enriched in ceramide. A ∼30-fold increase of the number of RBCs presenting ceramide-enriched patches was observed (red arrowheads at Figure 9A and quantification at Figure 9C). A heightened proportion of RBCs with ceramide-enriched patches was also observed for a healthy splenectomised donor but those RBCs accounted only for ∼10% of all RBCs compared to ∼35% for pHypoβ (Figure 9C and Supplementary Figure 1F). This could suggest that those patches are normally eliminated by the spleen.

Besides patches, ceramide-enriched vesicles can also be found but they did not significantly increase in abundance as compared to healthy RBCs (Figure 9C). The same was true for sphingomyelin-enriched vesicles (Figure 9B). In contrast, cholesterol-enriched spicules were observed at the edges of pHypoβ RBCs. As for patches, GM1 was enriched neither in vesicles nor in spicules (Figure 9A, white thin and thick arrows).

Altogether, our data indicated that pHypoβ RBCs were spiculated and showed enhanced ROS content, altered spectrin cytoskeleton integrity as well as increased membrane stiffness and curvature and abundance of sphingomyelin-enriched domains. These elements are consistent with the RBC morphological and biochemical storage lesions we previously revealed upon blood storage in K⁺/EDTA tubes for up to 4 weeks at 4°C as a model of accelerated RBC aging in vitro (Cloos et al., 2020). On the other hand, pHypoβ had increased membrane surface area and proportion of ceramide-enriched patches, rather consistent with reticulocyte properties (Ney, 2011). To therefore evaluate whether morphological, biochemical and biophysical alterations of pHypoβ RBCs could result from acceleration of RBC aging, RBCs from pHypoβ and healthy donors were stored for up to 3 weeks at 4°C and compared for morphology, fragility, functionality and biophysical properties. After 1 week of storage, the proportion of pHypoβ RBCs with patches did not increase any more (Figure 10A) and the differences in RBC membrane area and Hb release between pHypoβ and healthy donors seen in fresh RBCs were both abrogated (Figures 10B,C). The increase of intracellular calcium content measured after 1 week of storage in untreated and Yoda1-treated conditions were similar in pHypoβ and healthy donors (Figure 10D). Quite surprisingly, the extent of PS surface exposure measured after 3 weeks of storage increased three times less in pHypoβ than in healthy donors (Figure 10E). Regarding GM1-enriched domains, they increased in healthy RBCs after 1 week as expected from our previous data (Cloos et al., 2020), but this increase was even more pronounced and significant in pHypoβ RBCs. However, the increase of GM1-enriched domain abundance resulting from Yoda1 stimulation was lower in pHypoβ than healthy RBCs which might be due to the fact that GM1-enriched domain abundance was already higher at the surface of 1 week-stored pHypoβ RBCs than healthy RBCs stored for the same time (Figure 10F). We conclude that pHypoβ RBCs did not appear to represent an accelerated model of RBC aging.

We therefore evaluated the alternative possibility of a maturation defect of pHypoβ RBCs by labeling with a fluorescent mitoTracker. The proportion of pHypoβ RBCs presenting mitoTracker-positive patches at their surface increased by ∼40-fold as compared to healthy RBCs (Figures 10G,H). Importantly, the presence of those patches cannot be explained by the absence of spleen (Supplementary Figure 1G).

Acanthocytes provide a typical clinical feature of abetalipoproteinemia, representing 50–90% of total RBCs (Biemer, 1980). In contrast, the proportion of acanthocytes and their existence are more debated for homozygous and heterozygous familial hypobetalipoproteinemia. In homozygous familial hypobetalipoproteinemia, while it has been stated that acanthocytes are as numerous as in abetalipoproteinemia Biemer (1980) and Tamura et al. (1988) reported two patients suffering from this disease with acanthocytes representing only 10–15% of all RBCs. For the heterozygous form of the disease, their presence is less frequently detected but not excluded (Biemer, 1980; Hardie, 1989; Clarke et al., 2006). Actually, splenectomy could be partly responsible for the unusual high proportion of spiculated RBCs detected for pHypoβ as (i) in some cases, acanthocytes can be observed post-splenectomy (Shah and Hamad, 2020); and (ii) the absence of the spleen in this patient could prevent hemolysis of acanthocytes normally provoked through splenic sequestration. Even if the post-splenectomy hypothesis could be supported by Howell–Jolly bodies observed on pHypoβ blood smears, it seems more likely that reduced RBC elimination as consequence of the absence of a spleen contributes to the abundance of acanthocytes in pHypoβ as no spiculated RBCs could be detected for a heathy splenectomised donor and acanthocytosis is part of the usual clinical picture of lipoprotein disorders and vitamin E deficiencies (Wong, 2004).

Vitamins E and A are important antioxidants and their supplementation is part of therapy for patients with homozygous familial hypobetalipoproteinemia and abetalipoproteinemia in order to prevent neurological and ophthalmological disorders (Granot and Kohen, 2004). This is not the case for heterozygous familial hypobetalipoproteinemia even if vitamin E levels below the normal range were detected (Burnett and Hooper, 2015). Regarding pHypoβ, although vitamin A was within the normal range, vitamin E was at the lower level. Clarke et al. actually detected reduced plasma vitamin E levels combined with reduced alpha-tocopherol content in RBCs of 9 patients suffering from heterozygous hypobetalipoproteinemia (Clarke et al., 2006). For patients suffering from abetalipoproteinemia and supplemented with liposoluble vitamins, observations differ. According to Granot and Kohen (2004), no signs for oxidative stress can be detected in the plasma of those patients while Calzada et al. (2013) reported decreased alpha-tocopherol levels in both platelets and HDLs of these patients combined with increased lipid peroxidation, consistent with the fact that vitamin E is a potent inhibitor of lipid peroxidation (Burnett and Hooper, 2015).

A high amount of free ROS was measured in pHypoβ RBCs but was not accompanied by increase in lipid peroxidation and metHb. However, it remains unclear where these ROS are coming from. Actually, vitamin K is not the typical antioxidant but it has nevertheless been shown to prevent oxidative stress in neurons (Li et al., 2003). Plasma levels of the latter were very low in pHypoβ. Taken together with low vitamin E levels, oxidative balance could be disturbed in pHypoβ RBCs because of reduced amount in circulating antioxidant vitamins eventually sufficient to avoid important lipid peroxidation but not to totally prevent RBC oxidative stress.

Besides membrane lipids and Hb, spectrin and several membrane:cytoskeleton anchorage proteins such as Band3 also represent targets of oxidative stress (Voskou et al., 2015). In pHypoß RBCs, we provided several lines of evidence supporting the role of the altered cytoskeleton in RBC morphology alteration. First, the abundance of RBCs with a homogenously dense spectrin network was decreased at the benefit of RBCs presenting spectrin-enriched patches and/or vesicles. Second, in the RBCs with an homogeneous spectrin network, the spectrin membrane occupancy was decreased, which could reflect a reduced membrane:cytoskeleton anchorage and lead to increased stiffness. Third, the latter increase in pHypoβ RBCs was confirmed by AFM.

The cytoskeleton alterations found in pHypoβ were in agreement with the literature. First, spectrin enrichment in the thorny projections of acanthocytes was already described by electron microscopy (Siegl et al., 2013). Second, spectrin modification through treatment with urea combined with lysophosphatidylcholine membrane insertion has been shown to induce the transformation of discocytes into acanthocytes (Khodadad et al., 1996). Third, cytoskeletal alterations and abnormalities of Band3 immunolabeling have been revealed in neuro-acanthocytosis (Wong, 2004; Adjobo-Hermans et al., 2015). Acanthocytes could be actually formed by alterations in Band3 conformation similar to echinocytes in which the ratio of outward and inward facing Band3 is responsible for the morphological alteration (Wong, 2004). Thus, spectrin but also Band3 could be the target of ROS in pHypoβ RBCs, leading to decreased membrane:cytoskeleton anchorage and to acanthocytes with spectrin-enriched projections. However, the role of Band3 in the disease remains to be explored.

Analysis of the whole pHypoβ RBC population revealed no modification in the total cholesterol content in agreement with findings from other groups obtained on RBCs from patients with heterozygous hypobetalipoproteinemia (Biemer, 1980; Gheeraert et al., 1988). In abetalipoproteinemia and homozygous hypobetalipoproteinemia, observations are less consistent. Some groups reported increased cholesterol-to-phospholipid ratios (McBride and Jacob, 1970; Biemer, 1980; Barenholz et al., 1981) while others detected increased total cholesterol content without any increase in the cholesterol-to-phospholipid ratio (Iida et al., 1984) and still others described no alterations at all of the total cholesterol amount (Simon and Ways, 1964). Fatty acid composition was also largely maintained in pHypoβ RBCs, except for the slight increase of the linoleic and arachidonic acids, which could be related to specific dietary intake (Takkunen et al., 2013; Ford et al., 2016). To the best of our knowledge, the fatty acid content has not been studied in heterozygous hypobetalipoproteinemia. Abetalipoproteinemia and homozygous hypobetalipoproteinemia are instead generally associated with decreased linoleic acid content together with lower membrane fluidity. Besides, an increased sphingomyelin/lecithin or sphingomyelin/phosphatidylcholine ratio is also seen in RBCs of those patients (Ways et al., 1963; Simon and Ways, 1964; Cooper et al., 1977; Biemer, 1980; Barenholz et al., 1981; Iida et al., 1984). An increased sphingomyelin/phosphatidylcholine ratio resulting from a reduced abundance of phosphatidylcholine species was also evidenced in pHypoß. A higher (dihydro)ceramide content was also seen in pHypoβ RBCs in agreement with the increased ceramide-enriched patches. More surprisingly, a specific increase of lysophosphatidylethanolamine species was revealed while lysophosphatidylcholine is generally described to be associated with RBC morphological transformation into acanthocytes or echinocytes (Fujii and Tamura, 1984; Khodadad et al., 1996).

Lipid distribution at the RBC surface was analyzed in parallel to evaluate whether local changes in lipid domains could be revealed. We showed three types of membrane lipid distribution alterations. First, the abundance of sphingomyelin-enriched domains increased by 2.5-fold at the surface of pHypoβ RBCs. As membrane sphingomyelin content was not altered, the decreased membrane to cytoskeletal anchorage in pHypoβ RBCs could be responsible for this increase. This hypothesis is based on the observation by Conrard et al. that impairment of the membrane:cytoskeleton anchorage by PKC activation leads to an increase of sphingomyelin-enriched domains (Conrard et al., 2018). Second, ∼35 and ∼7% of pHypoβ RBCs presented ceramide- and sphingomyelin-enriched patches, respectively. Third, a high number of RBCs presented cholesterol-enriched spicules. The evident changes in membrane lipid organization stressed the importance of combining quantitative lipidomic analyses with qualitative lipid imaging to better understand diseases characterized by a variable- not systematically inventoried- amount of acanthocytes.

Quite surprisingly at first glance, only slight modifications of pHypoβ RBC functionality were observed. First, although no modification was seen in the poorly sensitive osmotic fragility test, osmotic gradient ektacytometry indicated a slight increase in surface area-to-volume ratio and therefore of osmotic fragility. Those observations were in accordance with the literature as for other models of acanthocytotic RBCs no modification in osmotic fragility is detected (Cooper, 1969; McBride and Jacob, 1970). Second, the intracellular calcium and ATP contents were normal. Third, the response of GM1-enriched domains to PIEZO1 activation through Yoda1 in fresh RBCs was similar in the patient and healthy donors.

Nevertheless, a deeper analysis indicated that the number of GM1-enriched domains was increased- although not significantly- in fresh RBCs and significantly in 1 week-old RBCs. Likewise, the abundance of sphingomyelin-enriched domains was enhanced and their response to intracellular calcium depletion through EGTA was abrogated. Thus, a fine analysis of the RBC surface both upon resting state and upon stimulation of calcium exchanges revealed functionality impairments in pHypoβ. Those alterations were not accompanied by altered calcium content or large increases in fragility probably because those measurements were made on the whole RBC population without any distinction between acanthocytes and discocytes. Interestingly, Ohsaka et al. reported on a patient with acute myelodysplasia with myelofibrosis who presented acanthocytes a heightened calcium uptake in RBCs combined with normal calcium levels (Ohsaka et al., 1989). In agreement with slight but detectable functionality impairment in pHypoβ, a loss of deformability after prolonged exposure to high shear stress was observed. Since this loss could only be partly explained by splenectomy, this observation revealed a slight reduction of membrane health of pHypoβ RBCs. This conclusion was further supported by alterations of RBC membrane biophysical properties.

In this study, we showed that the proportion of pHypoβ RBCs presenting mitoTracker-enriched patches, and thus mitochondrial fragments, increased by ∼40-fold as compared to healthy RBCs and represented ∼15% of total RBCs in pHypoβ. Moreover, ∼35% of pHypoβ RBCs presented ceramide-enriched patches. We propose that those ceramide-enriched patches corresponded to mitochondrial fragments, since (i) the increased abundance of those two types of patches in pHypoβ was similar; (ii) ceramides are known to be enriched in the outer mitochondrial membrane (Siskind, 2005); and (iii) mitoTracker and BODIPY-ceramide double-labeling in RBCs from patients with hereditary spherocytosis showed high colocalization (our unpublished data).

Besides the high number of mitochondria-positive RBCs, the RBC maturation defect in pHypoβ is further supported by (i) the increased membrane area measured after RBC spreading, normally reduced by vesiculation upon reticulocyte maturation (Trakarnsanga et al., 2017); (ii) the higher oxidative stress, as lipid peroxidation and activity of antioxidant enzymes are increased in reticulocytes as compared to mature RBCs (Sailaja et al., 2003); (iii) the increased abundance of GM1-enriched domains, as those domains are related to calcium influx in RBCs and calcium uptake is heightened in reticulocytosis (Ohsaka et al., 1989); and (iv) the threefold reduction in stored RBCs of surface exposure of PS, a lipid normally found in autophagic vesicles upon reticulocyte maturation (Trakarnsanga et al., 2017). Actually, the resistance to PS surface exposure has also been reported by Siegl et al. (2013) in chorea-acanthocytosis after RBC stimulation with lysophosphatidic acid. Thus, pHypoβ has a high proportion of acanthocytes which could result from a maturation defect during the R1 reticulocyte stage known to undergo significant rearrangements in reticulocyte membrane and intracellular components via several mechanisms including exosome release and mitophagy (Minetti et al., 2018, 2020).

In this study, a wide range of research methods was used to evaluate morphological, biochemical and biophysical changes and their potential consequences for RBC functionality in hypobetalipoproteinemia. Moreover, thanks to a deep imaging analysis, we were able to reveal local changes in lipid composition and cytoskeleton distribution in acanthocytes. Nevertheless, our experimental strategy also presents some drawbacks that are inherent to the low prevalence of heterozygous hypobetalipoproteinemia with a high proportion of acanthocytes on one hand and to the study of membrane lipid lateral organization on the other hand. However, those limitations were minimized, based on the following evidences. First, although generalization of observations based on only one patient is difficult, our findings are in agreement with literature data on diseases associated with acanthocytosis. In fact, altered membrane composition and biophysical properties have also been demonstrated in a patient suffering from hereditary elliptocytosis and a series of patients suffering from hereditary spherocytosis, two diseases associated with increased membrane fragility due to cytoskeleton defects (Pollet et al., 2020; unpublished data), thereby supporting the importance of membrane:cytoskeleton interplay in RBC morphology and functionality. Second, although fluorescent lipid analogs were extensively used and validated in previous researches, their validity as bona fide surrogates of endogenous lipid counterparts is debated in view of the bulk fluorophore which can deeply modify biophysical properties. To minimize difficulties, we extensively validated all the lipid probes we used. For information regarding BODIPY-sphingomyelin and -GM1, we refer the reader to our previous papers (Tyteca et al., 2010; D’Auria et al., 2011, 2013; Carquin et al., 2016; Leonard et al., 2017b). For BODIPY-ceramide, the following lines of evidences supported a distribution as endogenous ceramides: (i) the fluorescent BODIPY-ceramide labeling correlated with the endogenous ceramide content measured by lipidomics both in pHypoβ and in RBCs upon blood storage at 4°C (Cloos et al., 2020); (ii) BODIPY-ceramide accumulates in the Golgi complex in nucleated cells (Tyteca et al., 2010) and in mitochondria remnants in RBCs from patients suffering from familial hypobetalipoproteinemia and from hereditary spherocytosis (data not shown), indicating that the lipid probe is able to join the intracellular compartments enriched in endogenous ceramide (Pagano et al., 1991); (iii) in the Golgi complex, BODIPY-ceramide can be efficiently converted into BODIPY-sphingomyelin, demonstrating the “one molecule at a time” conversion by the selectivity of the catalytic site of sphingomyelin synthase (Tyteca et al., 2010); (iv) BODIPY-ceramide accumulates in the inner plasma membrane leaflet, as revealed by resistance to back-exchange by BSA (Pollet et al., 2020) and demonstrating ability of the probe incorporated into the outer leaflet to flip to the inner one, as natural ceramide (Contreras et al., 2010); (v) BODIPY-ceramide-enriched domain abundance correlates with the ceramide content, as revealed by the decrease of ceramide-enriched domain abundance at the RBC membrane and the concomitant enrichment of ceramide species in RBC-derived extracellular vesicles at the end of the blood storage period in K⁺/EDTA tubes at 4°C (Cloos et al., 2020). Besides extensive validation of fluorescent lipid analogs, we developed the use of fluorescent Lysenin and Theta toxin fragments specific to endogenous sphingomyelin and cholesterol. However, it should be noted that, as fluorescent lipid analogs, toxin fragments also present drawbacks (Carquin et al., 2014, 2015, 2016) and it is therefore crucial to use complementary unrelated lipid probes to target one lipid, once possible. Third, regarding the imaging method used, although confocal microscopy offers several advantages such as vital imaging, multiple labeling and 3D-reconstruction, the size of observed lipid domains could have been overestimated.

In a case of familial hypobetalipoproteinemia, resulting from heterozygosity for the pathogenic Gln845Argfs^∗18 mutation in the ApoB gene, we showed that the resulting acanthocytes exhibited impaired cytoskeleton and membrane biophysical properties without significant loss of RBC functionality as assessed by functional tests applied on the total RBC population, without any distinction between acanthocytes and discocytes (e.g., osmotic fragility test, intracellular calcium measurement). In contrast, more sensitive tests (e.g., ektacytometry) and assays aimed at investigating specific RBC populations (e.g., lipid domains, membrane curvature) did demonstrate functional impairments in pHypoβ.

Although our findings were generated from only one patient, the observed cytoskeleton and membrane alterations were consistent with literature data on diseases associated with acanthocytosis. Indeed, cytoskeletal alterations and abnormalities of Band3 immunolabeling have been shown in neuro-acanthocytosis (Wong, 2004; Adjobo-Hermans et al., 2015). The resistance to PS surface exposure has also been reported in chorea-acanthocytosis after RBC stimulation with lysophosphatidic acid (Siegl et al., 2013). The modest alterations of pHypoβ RBC functionality agreed with the absence of modification in osmotic fragility in other models of acanthocytotic RBCs (Cooper, 1969; McBride and Jacob, 1970).

From a more global point-of-view, altered membrane composition and biophysical properties have also been demonstrated in two RBC disorders associated with altered cytoskeleton function and a large diversity of RBC morphological changes and deformability alterations (Pollet et al., 2020; unpublished data), indicating the close membrane:cytoskeleton interplay and its role in RBC morphology and functionality. More specifically, a parallelism could be established with the patients with hereditary spherocytosis exhibiting a high proportion of spiculated RBCs (our unpublished data).

Our study demonstrates that evaluation of membrane biophysical properties and membrane lipid distribution could be of benefit in the diagnosis and better understanding of RBC disorders.