Phosphoproteomics and morphology of stored human red blood cells treated by protein tyrosine phosphatases inhibitor

During ex vivo aging, red blood cells (RBCs) suffer from lesions^1-4 that are characterized, among other parameters, by changes in cell morphology and increased cell rigidity.^5,6 Alterations of cell shape and deformability result from different features influenced by vesiculation, osmolarity, and ion transport, as well as membrane proteins/lipids reorganization in active or passive events.^7-9 

The membrane/cytoskeleton interactions (rupture of which is observed in several diseases)⁷ regulate the RBC shape change in an adenosine triphosphate (ATP)-dependent manner.¹⁰ To soften its cytoskeleton, RBCs expend energy in the form of ATP, which induces enhanced membrane fluctuations and cell shape change.¹¹ The nonequilibrium dynamism is due to the dissociation of the cytoskeleton at the spectrin junctions powered by ATP. Once dissociated from the cytoskeleton, the membrane fluctuations are thermally governed.¹² Several groups evoked the hypothesis of protein phosphorylation in the regulation of membrane deformability.^10,11,13-16 

Phosphorylation (eg, band 3) is known to regulate enzymatic activity in RBCs,^17,18 and the phosphorylation of cytoskeleton and membrane proteins has been reported to modulate protein-protein interactions.^15,16 These modifications involve adducin, β-spectrin, protein 4.1, as well as band 3. A few proteomic studies focusing on Plasmodium falciparum–infected RBCs and sickle cell disease identified several phosphoproteins.^19-21 They pointed out their roles in signaling and RBC deformability (through the phosphorylation of proteins belonging to the junctional complex).²² Lately, another one highlighted the role of a set of phosphoproteins involved in deformability experiments.²³ As for transfusion medicine, Rinalducci et al quantified by label-free phosphoproteomics a few membrane phosphoproteins, including β-spectrin, ankyrin-1, α-adducin, dematin, glycophorin A, and glycophorin C, and observed the upregulation of targeted phosphoserines during storage.²⁴ 

Phosphorylation of band 3 has been studied since the 1980s.²⁵ Low et al detected the presence of 2 kinases in RBCs, p72^syk (Syk) and p56/53^lyn (Lyn), and demonstrated that Syk was responsible for band 3 phosphorylation under pervanadate treatment.²⁶ Later, Brunati et al reported the sequential phosphorylation of band 3 by Syk on Tyr8 and Tyr21 and Lyn on Tyr359 and Tyr904 (independent of Syk in the case of chorea-acanthocytosis).^27,28 They also discovered the role of SHP-2 protein tyrosine phosphatase (PTP) in the dephosphorylation of band 3 and Syk-mediated phosphorylation.²⁹ 

Protein phosphorylation could favor morphological changes in response to stress.^15,23,30,31 Ferru et al proposed a phosphorylation-induced oxidation mechanism in which Syk binds to oxidized band 3, phosphorylates it, and causes the weakening of the cytoskeleton and the formation of band 3 clusters. Those aggregates are released in microvesicles in thalassemia and G6PD deficiency.^30,31 Conversely, Minetti et al suggested that band 3 phosphorylation was not a prerequisite in Ca²⁺/ionophore-induced microvesicles.³² Under blood banking conditions, in which oxidative processes play a central role,^33-37 microvesicles accumulate during storage. So far, the link to phosphorylation remains unknown.

Several questions remain open regarding all these mechanisms of regulation. Indeed, the phosphorylation of proteins along the storage of RBCs has mainly been explored regarding band 3. Our goal here was to investigate both the Tyr-phosphorylation (pY)–dependent shape change as well as to relate them to the effect of aging under blood banking conditions. For this purpose, RBC morphology was studied by digital holographic microscopy (DHM)^38-40 using inhibitors of PTPs and protein tyrosine kinases. In addition, protein phosphorylation was quantified and characterized by Western blot (WB) and phosphoproteomics. This multidisciplinary approach comprehensively depicts the role of protein phosphorylation on RBC morphology and the impact of storage on this mechanism.

Blood collected from healthy donors was prepared in top-bottom bags system (CompoFlow CQ32250, Fresenius Kabi). A total of 450 ± 50 mL of whole blood mixed with 63 mL of CPD (citrate-phosphate-dextrose) anticoagulant was centrifuged, the components were separated, RBC fraction was filtrated to remove residual leucocytes, and 100 mL of SAGM (saline-adenine-glucose-mannitol) additive solution was added for storage. RCCs were kept at 4°C for up to 6 weeks. Six RCCs were used for the phosphoproteomic and WB experiments under orthovanadate (OV) treatment, 6 others to evaluate the effect of kinase inhibitors (KIs), and 3 more for DHM with KIs. The project was accepted by the institutional review board of Transfusion Interrégionale Croix-Rouge suisse (CRS), in compliance with local legislation, and was conducted according to the Declaration of Helsinki.

Samples were drawn from RCC using sterile needles and syringes through a sampling site. To retrieve RCC supernatant for hemolysis measurement, samples were centrifuged for 10 minutes at 2000g and 4°C. RBC pellets were washed 1 to 2 times in 0.9% NaCl and resuspended in HEPA buffer (15 mM Hepes, 1 g/L bovine serum albumin, 130 mM NaCl, 5.4 mM KCl, 0.5 mM MgCl₂·6H₂O, 10 mM glucose, and pH 7.4). Chemicals were purchased from Sigma-Aldrich (Steinheim, Germany), MSD Merck Sharp & Dohme (Luzern, Switzerland), Laboratorium Dr. G. Bichsel (Interlaken, Switzerland), and MP Biomedicals (Illkirch, France). KIs were from Selleckchem (Houston, TX).

RBCs were treated with 2 mM of phosphatase inhibitor sodium OV (Na₃VO₄, OV)¹⁵ or dH₂O (control) for 1 hour or 4 hours at 37°C either in tubes (under agitation) or in a 96-well plate for microscopy experiments (see below). To test the impact of KIs, RBCs were first treated with 10 μM of KI (Saracatinib, S1006, against Src; PRT062607, S8032, against Syk; bafetinib, S1369, against Lyn; PD153035, S6546, against EGFR; H89, S1582, against protein kinase A [PKA]; Go 6983, S2911, against PKC; and D4476, S7642, and Silmitasertib, S2248, against casein kinase) or dimethyl sulfoxide (DMSO; control) for 1 hour at 37°C before incubation with OV.

Hematological data were recorded with an automated hematology analyzer (KX-21N, Sysmex). Microvesicles were quantified by flow cytometry (FACScalibur flow cytometer with CellQuest pro software; BD Biosciences, Franklin Lakes, NJ).⁴⁰ Supernatant hemoglobin concentration was determined according to the Harboe method⁴¹ by spectrophotometry (NanoDrop 2000c, Thermo Scientific, Wilmington, DE). Samples were diluted in dH₂O if required (ie, A₄₁₅ above 1.5 arbitrary unit (A.U.)).

For WB and phosphoproteomic analyses, RBCs were lysed by incubation in hypotonic buffer (0.1× phosphate-buffered saline) for 1 hour at 4°C on a roller, in the presence of 2 mM OV or dH₂O (control) plus 10 μM KI or 0.1% DMSO (control). Cell lysates were centrifuged at 21 500g in 4°C for 75 minutes to isolate membranes. Membranes were washed in 0.1× phosphate-buffered saline (centrifugation cycles of 30 min) until white pellets were obtained. Aliquots were saved at −80°C.

Membrane proteins were extracted in 4 volumes of 8 M urea, 2 M thiourea, 0.5% sodium dodecyl sulfate (SDS), and 10 mM dithioerythritol (DTE). Each sample was in-solution digested with Lys-C (1:50 enzyme/protein) and then with trypsin gold (1:50 enzyme/protein). Phosphopeptides were desalted and enriched on titania tips⁴² and analyzed by liquid chromatography (Dionex Ultimate 3000 RSLC nanoUPLC system; Thermo Scientific, Rockford, IL) coupled to an Orbitrap Q-Exactive HF (Exploris 480 Mass spectrometer for KIs, Thermo Scientific). Raw data were treated using MaxQuant 1.6.0.16 (1.6.10.43 for KIs)⁴³ with Andromeda as an internal database search engine.⁴⁴ Fragmentation spectra were searched against the human UniProt database (July 2017; 71 567 sequences). A t test analysis was performed using Perseus 1.6.0.7,⁴⁵ and graphs were generated with homemade programs (R environment).⁴⁶ Class 1 phosphosites with intensities in monophosphopeptides (multiplicity = 1), with at least 12 valid values in at least 1 group (OV or Control), were kept for the analysis (3 valid values in at least 1 group for KIs). The significantly differentially quantified phosphosites were defined according to a permutation-based multiple-testing analysis (250 permutations; false discovery rate (FDR) = 0.05; S0 = 1; and S0 = 0.1 for the KIs).

Membrane/cytoskeleton proteins were extracted from pelleted membranes with 4 to 6 volumes of deoxycholate (DC) buffer: 1% DC in 50 mM Tris-HCl, 150 mM NaCl, and pH 8.1.³⁴ Ten μg of proteins from each sample and 10 μL of BenchMark Pre-Stained Protein Ladder (Invitrogen, ThermoFisher Scientific) were loaded on sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE, Mini-PROTEAN TGX gels, 4%-15%; BIO-RAD). WB were performed on polyvinylidene fluoride (PVDF) membranes against pY-containing proteins (pY PY99 1/6000 from Santa Cruz Biotechnology, Dallas, TX). Washing and incubation with primary and secondary antibodies were done in tris-buffered saline (TBS-T: 1× TBS + 0.05% Tween-20) and Top Block buffer (Sigma-Aldrich, 4% in TBS-T), respectively. Secondary antibody (polyclonal goat antimouse immunoglobulins HRP, Dako) was diluted at 1/6000 or 1/10 000 in Top Block buffer. Bands of interest (from WB and Ponceau red staining) were quantified by densitometry (ImageQuant TL software 7.0; GE Healthcare, Uppsala, Sweden) and expressed as “volume.” pY signals were normalized by total protein loading.

Quantitative phase images were acquired using a DHM T1000 (Lyncée Tec SA, Lausanne, Switzerland). DHM provides label-free measurements of the optical path difference (OPD), a parameter proportional to the refractive index and thickness of the sample.^38,47 For imaging, RBCs were seeded (80 000 per well) in a 96-well plate coated with 0.1 mg per mL Poly-L-ornithine.⁴⁰ To speed up cell sedimentation, the plate was centrifuged at room temperature for 2 minutes at 140g. Then, the plate was placed in a Chamlide WP Incubator System (LCI Live Cell Instrument, Seoul, South Korea) set at 37°C with high humidity and 5% CO₂. Four images, obtained at a 20× magnification (Leica 20×/0.40 NA objective; Leica Microsystems GmbH, Wetzlar, Germany), were taken in each well, with 4 wells per condition. First, 4 baseline images were acquired at a 5-minute interval. The RBCs were then treated with 0.1% DMSO (control) or 10 μM KI and incubated for 45 minutes (10 acquisitions, each 5 minutes). Finally, the RBCs were treated with 2 mM OV or dH₂O (control). Timelapse images were recorded every half hour for ∼20 hours.

Images were analyzed in 2 different ways. A population analysis was performed by calculating the spatial standard deviation of the OPD (SD-OPD) value.⁴⁰ Single-cell phenotypic analysis was also performed using CellProfiler (Broad Institute, www.cellprofiler.org, 2.1.0 rev 0c7fb94)⁴⁸ and CellProfiler Analyst (2.0 r11710)⁴⁹ to sort RBCs as: “discocytes,” “echinocytes,” “spherocytes,” and “errors.” For analysis of timelapse experiments, the area under the curve (AUC) parameter was selected because it integrates information about the behavior of the RBCs during the whole timelapse period.

GraphPad Prism (version 8.4.2 [464], GraphPad Software, LLC, San Diego, CA) was used for data presentation and statistical analyses.

A label-free quantitative phosphoproteomic analysis of RBC membrane proteins treated or not with OV was carried out to identify and quantify phosphoproteins. It revealed the presence of 1949 phosphopeptides (1752 Class 1 phosphosites, ie, localization probability > 0.75, and 801 quantified Class 1 monophosphosites) distributed over 609 phosphoproteins (Figure 1A; supplemental Material 2, SM-2; additional data available ProteomeXchange).⁵⁰ The proportion of pS, pT, and pY Class 1 monophosphosites is 84.3%, 8.1%, and 7.6%, respectively. Upon OV treatment (PTP inhibition), 41 of 61 pY-sites belonging to 22 different proteins were significantly upregulated (Figure 1B) and 2 downregulated (see list in supplemental Material 3, SM-3). The first 20 upregulated pY-sites belong to proteins mainly involved in the cell structure: band 3 (gene name SLC4A1), α- and β-spectrin (SPTA1 and SPTB), α- and β-adducin (ADD1 and ADD2), protein 4.1 (EPB41), ankyrin-1 (ANK1), dematin (DMTN), and tensin-1 (TNS1), which is coherent with WB presented in Figure 2A. A network analysis (String-db.org, version 11.5, December 2022; Figure 1C) shows a cluster of these proteins involved in band 3–ankyrin and junctional complexes, in connection to kinase activities with nonspecific protein tyrosine kinase (YES1, LCK, SRC, or FYN, shared identified peptide) and Lyn kinase (LYN), as well as other proteins including flotillin-2 (FLOT2), aquaporin-1 (AQP1), receptor protein tyrosine kinase (EPHB4) and phosphatidylinositol 3,4,5-trisphosphate 5-phosphatase 1 (INPP5D).

In addition to pY modifications triggered by OV, pS and pT were also significantly regulated (see black dots; Figure 1B). Altogether, 60 and 123 phosphosites were significantly upregulated and downregulated, respectively, which indicated a global downregulation when pY were upregulated.

Accumulation of pY under OV treatment was confirmed by WB analyses. Figure 2A shows the pY signal after 1 or 4 hours of OV treatment in RBCs stored for 1, 21, or 42 days. The pY signals were observed on specific protein bands, corresponding to (top to bottom) α-spectrin, β-spectrin, ankyrin-1, α-adducin, β-adducin, band 3, and protein 4.1. Two bands around 70 kDa could correspond to Lyn and Syk kinases. During RBC storage, the effect of OV decreased progressively (Figure 2A, right). At day 42, only a weak signal was detected.

OV treatment induced the release of microvesicles and hemolysis (Figure 2B). This effect is visible after 4 hours of incubation but not after 1 hour. Consistent with the phosphorylation levels, RBCs stored for 1 day released a higher number of microvesicles and hemolyzed more (not significant) than RBCs stored for 21 or 42 days.

In the control group (0.1% DMSO), the SD-OPD signal (linearly correlated to the percentage of spherocytes)⁴⁰ slightly increased throughout the 20-hour timelapse, whereas it increased rapidly until reaching a plateau for OV-treated RBCs (Figure 3A). The AUC value, corresponding to integration of SD-OPD over time, was equivalent in both storage age of RCCs, with an upward trend at storage day 40.

Analysis of single-cell phenotype by automated analysis of phase images (Figure 3C) confirmed that the increase of SD-OPD signal was due to the rapid formation of echinocytes (intermediate shape) and spherocytes, concomitant to a drop-down of discocytes (Figure 3B). Morphological changes after OV treatment appeared sooner in older RCCs that were also more sensitive to the experimental conditions, even in the control samples.

Despite a limited number of replicates, the data were consistent with the literature and preliminary data.¹⁵ It has to be noticed that the same phenomena were present with Ca²⁺-containing HEPA buffer. OV treatment transformed RBCs into spherocytes, and the phosphorylation level in membranes also decreased during storage (supplemental Figures 1 and 2 in SM-1, respectively).⁵¹ 

Based on the top-overexpressed phosphoproteins and the literature, the effect of 8 KIs was investigated on the morphology of stored RBCs using a DHM-based screening. Different variations of the SD-OPD were reported during the incubation in the presence of OV+KIs compared with that of OV only and control (SM-1, supplemental Figure 4). Only 3 appeared to decrease SD-OPD values and therefore counteracting the effect of OV. The advanced analyses are presented in Figure 4.

The incubation with PRT062607 and bafetinib before the OV treatment drastically abolished the pY process (Figure 4A). However, it was not the case with H89, in which only a weak effect on band 3 was observed. The level of pY on bands corresponding to α-adducin and spectrins was slightly affected.

Phosphoproteomics of KI- and OV-treated RBCs confirmed the differences at the protein level (SM-4). Expressions of phosphosites show shifts in favor of OV, meaning both KIs inhibited the effect of OV treatment (Figure 5). Notably, not only pY sites were significantly downregulated (red dots) when treated with KIs compared with OV, but also pS and pT (blue and gray dots) (SM-5 to 7). As expected, PRT062607 and bafetinib behaved the same, with 12 phosphosites in common, of which 9 were pY. As for H89, pY were not significantly downregulated, and the main modulations concerned pS and pT. Of note, pT378 on protein 4.1 was the only site commonly regulated site to both KIs.

When examining cell morphology, the effects of KIs can be seen through the AUC calculated from the integration of SD-OPD over time (Figure 4B). The presence of KIs on their own increased the AUC value because they favored the stomatocyte phenotype, as shown by image-analysis (Figure 4C). On day 6 of storage, the addition of KIs counterbalanced the effect of OV with a baseline level (Figure 4B). Nonetheless, the inhibitory effect of KIs was less pronounced with increasing storage time (SM-1, supplemental Figure 5). The OV + KIs conditions favored the presence of echinocytes compared with OV only (Figure 4C) and decreased the proportion of spherocytes, suggesting that the KIs delayed the transition from one phenotype to the other.

In summary, the 3 KIs enabled relative cell shape recovery after stimulation of pY, but only bafetinib and PRT062607 entirely abolished the pY phosphorylation, and H89 affected pS and pT.

Two-thousand proteins are known in mature RBCs.^52,53 In their study on RBC deformability, Moura et al detected 63 and 226 phosphoproteins in mature RBCs and reticulocytes, respectively.²³ Phosphoproteomics on sickle cells identified 155 phosphoproteins and 527 phosphosites,¹⁹ and in the case of infected RBCs by P falciparum, 1083 phosphosites were detected on 553 proteins.²¹ Our data are consistent with these previous sets of data. The higher number of phosphosites in this study is most likely related to the use of a phosphatase inhibitor that exacerbated pY formation.

Several phosphoproteins were remarkably upregulated upon OV treatment, of which a few have already been studied.^26,29 Band 3, α- and β-spectrin, α- and β-adducin, protein 4.1, ankyrin-1, dematin, tensin-1, aquaporin-1, and plasma membrane calcium–transporting ATPase 4 are present within a shared cluster. Indeed, these proteins are part of the well-known band 3–ankyrin and junctional complexes.⁵⁴ In addition, Fyn and Lyn were identified. The latter is known to participate in the sequential phosphorylation of band 3 in which Lyn phosphorylates Tyr359 and Tyr904, in a second phase 2²⁷; both pY modifications were reported here. Consistently, no phosphorylation on band 3 Tyr8 and Tyr21was detected, which could be explained by direct phosphorylation on Tyr359 and 904.²⁸ In addition, deletions or modifications (ie, methylation) near tyrosine residues such as aspartate and glutamate in the N-terminal part⁵⁵ might affect the detection of phosphosites. Alteration of band 3 is associated with storage lesions and changes RBC properties.⁵⁶ As for Fyn, this enzyme is known to phosphorylate spectrins and β-adducin.

Regarding pS and pT, 12 phosphosites are known to be upregulated during RBC storage.²⁴ Over these phosphosites, 6 were detected in this study (only 2 presenting significant variations). pS122 in glycophorin C was increased in both studies, whereas pS152 in dematin was downregulated here.

Concomitant to the OV-induced phosphorylation, a transition from discocytes to spherocytes was observed (Figure 3). The phosphorylation of membrane proteins, particularly band 3 complexes, probably induced a rupture in the cytoskeleton-integral membrane proteins link, initiating shape change. In addition, the binding to protein 4.1 could be inhibited by the phosphorylation on band 3 Y359.⁵⁴ Moreover, the Y660 phosphorylation on protein 4.1 is known to disturb the spectrin/actin/4.1 complex.⁵⁷ These modifications trigger the formation of spherocytes, similar to spherocytosis with a rupture of the vertical link. The phenotype observed here is similar to previously reported OV-treated RBCs.^15,58 In both studies, the authors showed the formation of echinocytes after 20 or 60 minutes of incubation. In these data, ∼30% of echinocytes were detected after 1 hour of treatment. At the end of the incubation period (∼20 hours), RBCs were mostly spherocytes. The formation of the spicules could be due to band 3 phosphorylation and clustering, as described by Turrini at al.^15,30,31 

The morphological analyses of OV-treated RBCs showed that the addition of KIs resulted in the enrichment of echinocytes (Figure 4C). The final stage of cell transformation was not reached. It might be explained by the downregulation of phosphorylation on proteins, involved in cell structure, that preserve the link between integral membrane complexes and the cytoskeleton. Therefore, in this timelapse, the formation of spherocytes is partly inhibited.

The 3 KIs had a similar impact on the morphology, even though they caused varying phosphorylation levels. The 2 targeting band 3 (PRT062607 and bafetinib) abolished the pY signals on all proteins, including band 3, protein 4.1, ankyrin-1, and spectrins. In this situation, the similar effect on morphology may be attributed to shared pathways. Bafetinib targets Lyn, which is directly involved in Y359 and Y904 phosphorylations. As for PRT062607, its lower inhibition activity on Lyn could be enough to stop band 3 phosphorylation. In addition, and despite the absence of Syk-induced phosphorylation of band 3 in our proteomic data set, phosphorylation on Syk kinase (pY296) was downregulated (detected but not significantly in the case of bafetinib) and might also inhibit band 3 phosphorylation. The total decrease in pY was also observed on other proteins involved in band 3 complexes, such as protein 4.1. Indeed, the pY660 (detected as pY627 from a truncated form in the KI experiments) was downregulated by both bafetinib and PRT062607. On the contrary, H89 targets PKA involved in adducins and proteins 4.1 and 4.2 phosphorylations but not on Tyr residues, which explains the weak impact on WB.⁵⁹ The regulation of other phosphosites (as shown by the phosphoproteomic data, Figure 5C) might significantly impact cell morphology. Although protein 4.1 is expected to be phosphorylated on Y660 by epidermal growth factor receptor (EGFR), it has been reported that PKA can interact with EGFR. Y660 was not affected by H89, but 2 other phosphosites were downregulated by H89 (pS85 and pT378). Therefore, the KI H89 could also counteract the OV-induced phosphorylation of protein 4.1. Suppressing protein 4.1 phosphorylation could protect against the dissociation of spectrin/actin/4.1 complexes. Therefore, it can be speculated that the modifications of PKA-dependent proteins (such as the protein 4.1 and ankyrin-1) have an equivalent impact on morphology but through pS and pT phosphorylations. Additional experiments will be required to investigate such mechanisms.

The capacity of RBCs to phosphorylate proteins decreases with storage time. A rejuvenation process (that stimulates ATP production)^60-62 restores the ability of protein phosphorylation (supplemental Figure 3 in SM-1).⁵¹ It demonstrates that the enzymes are still functional and that the decrease in phosphorylation is a metabolic issue. Indeed, during storage, RBCs exhibit energy metabolism shifts in which ATP is produced during the first week and then decreases constantly.^4,63,64 

OV is known to induce protein phosphorylation and the formation of spherocytes.^15,58 Nevertheless, the morphology of long-stored RBCs was impacted by the OV treatment (Figure 3A), although no phosphorylation of proteins was detected (Figure 2A). In that case, the effect of OV could be related to (1) the impact of incubation conditions on weakened RBCs⁶⁵ or (2) side effects because OV can also inhibit different enzymes.^66,67 OV is a Ca²⁺ pump inhibitor that alters cell morphology in addition to phosphorylation events. Although these experiments were carried out without extracellular Ca²⁺, it cannot be excluded that changes in intracellular Ca²⁺ induce shape modification, especially in older RCCs.⁶⁸ Therefore, the impact on morphology is probably due to pY on short-term stored RBCs (because of phosphorylation capacity, Figure 2A) and ion issues in long-term stored RBCs (absence of phosphorylation capacity). The same effects on morphology and phosphorylation in function of storage time were observed in Ca²⁺-containing HEPA buffer (SM-1).⁵¹ 

As mentioned in the “Introduction,” the release of microvesicles^40,69 results from different mechanisms such as phosphorylation, Ca²⁺ entry, and oxidative stress during storage. According to these data and the decrease of phosphorylation capacity proportionally to storage time, the microvesiculation process is probably independent of phosphorylation events in stored RBCs. This effect is consistent with the negative correlation between ATP concentration and microvesiculation⁷⁰ as well as the band 3 phosphorylation-independent vesiculation process reported by Minetti et al.³² However, their microvesicles were Ca²⁺ induced, which is known to recruit specific proteins and have a different composition than storage microvesicles.⁷¹ 

The loss of protein phosphorylation ability might be an issue in the transfusion of long-stored RCCs.^3,72 Indeed, it was recently shown on RBCs, especially on reticulocytes, that inhibiting band 3 phosphorylation (using bafetinib and without chemical stimulation) decreases the cell velocity in a microfluidic system.²³ Therefore, problems in pY may render RBC less efficient in crossing capillaries and oxygenating tissues, and in the worst case, it may trigger phagocytosis and RBC clearance.^73,74 Even though this ATP-dependent mechanism was restored in vitro and will be restored in vivo a few hours after transfusion, it is essential to maintain a high phosphorylation capacity to ensure transfusion efficiency. This capacity could be conserved by boosting metabolism (particularly the lower part of glycolysis and the pentose phosphate pathway, because RBCs are devoid of mitochondria).⁷⁵ 

The data presented here focused on phosphorylation events of RBC membrane proteins and their relation to cell morphology. The membrane phosphoproteomics highlighted many phosphoproteins and phosphosites directly involved in cell morphology. Using KIs pointed out the central role of band 3 and other membrane proteins, such as protein 4.1, in the stability of integral membrane proteins and cytoskeleton linkage. The impact of phosphorylation (even exacerbated here) is central in transfusion medicine, because it is associated with RBC aging in blood bags and, when dysregulated, is known to be involved in many diseases. Metabolic pathways and kinase activity must be investigated further to decipher the reactions controlling these phosphorylation events. The reported results and the further studies they suggest will contribute to enriching our knowledge in hematology.

The authors thank E. Längst for complementary experiments on rejuvenation and the Foundation SRTS-VD for the financial support.

Contribution: M.P. and M.B. designed the experiments and wrote the manuscript; M.B. and D.C. prepared samples and carried out experiments; R.H. and F.A. performed phosphoproteomics and bioinformatics analyses, respectively, and they analyzed data and contributed to the writing of the manuscript; J.-D.T., B.R., and G.T. reviewed the data and manuscript; and all the authors read and approved the final manuscript.

Conflict-of-interest disclosure: B.R. is an employee of Lyncée Tec SA, which commercializes digital holographic microscopy. The remaining authors declare no competing financial interests.

Current affiliation: M.B., Biomolecular Screening Facility, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland.

Correspondence: Michel Prudent, Transfusion Interrégionale CRS, Biopôle Secteur Croisettes - Bâtiment Metio, Rte de la Corniche 2, 1066 Epalinges, Switzerland; email: Michel.Prudent@itransfusion.ch.