During storage, erythrocytes undergo changes that alter their clearance and function after transfusion and there is increasing evidence that these changes contribute to the complications observed in transfused patients. Stored erythrocytes were incubated overnight at 37°C to mimic the temperature after transfusion. After incubation, several markers for erythrocyte damage were analysed. After overnight incubation, stored erythrocytes showed increased potassium leakage, haemolysis, PS exposure and vesicle formation, and all these effects increased with increasing storage time. Furthermore, we demonstrated that long-term stored erythrocytes develop decreased flippase activity and increased scrambling activity after overnight incubation, leading to PS exposure and the release of vesicles. Reduced intracellular potassium was identified as the cause of the decreased flippase activity. Lastly, we provide evidence that erythrocytes can return to a PS-negative state by shedding parts of their membrane as PS-containing vesicles and that these vesicles can serve as a platform for the coagulation cascade. These findings reveal that potassium leakage, a well-known phenomenon of prolonged erythrocyte storage, primes erythrocytes for PS exposure. PS exposure will lead to vesicle formation and might have an important impact on the post-transfusion function and side effects of stored erythrocytes.
During storage, erythrocytes undergo changes that affect their in vivo survival and function, which are collectively termed the “storage lesion” (Luten et al, 2008). Besides affecting the function and viability of the transfused erythrocytes, evidence exists that the storage lesion of red blood cells has deleterious side-effects for the recipient. Currently, the only two requirements to the storage of erythrocytes is an overall haemolysis during storage of >0·8% for Europe and 1·0% for North-America and a post-transfusion survival of at least 75% of the total amount of transfused erythrocytes (Dumont & AuBuchon, 2008; Luten et al, 2008). Although the exact mechanisms by which transfused erythrocytes are rapidly cleared are unknown, there is growing evidence that storage time negatively influences erythrocyte viability and clinical outcome of the transfused recipients (Hu et al, 2012; Sparrow et al, 2006; Leal-Noval et al, 2008; Tinmouth et al, 2006). Recent studies have shown that the negative impact of long-term stored erythrocytes on clinical outcome might not be a direct effect of the transfused erythrocytes, but the presence of cell-free haemoglobin, iron and the formation of vesicles during storage (Vlaar et al, 2010; Gladwin & Kim-Shapiro, 2009; Donadee et al, 2011; Hod et al, 2010; Kozuma et al, 2011).
One of the markers commonly used to determine the quality of stored erythrocytes is the presence of phosphatidylserine (PS) on the outside of the membrane (Burger et al, 2010). In healthy cells, the phospholipids in membranes are asymmetrically distributed with, most notably, PS on the inner leaflet of the membrane (Kahlenberg et al, 1974). The asymmetrical distribution of the lipid bilayer of cells is controlled by three different mechanisms (Daleke, 2003). First, an inward-directed flippase transports several phospholipids to the inside of the cell. Secondly, an outward-directed floppase transports other phospholipids to the outer layer of the membrane. Lastly, scrambling activity facilitates bidirectional movement of all phospholipids, thereby disturbing the normal asymmetrical distribution of the lipid bilayer. The most pronounced effect of scrambling activity and the loss of bilayer asymmetry is the exposure of PS on the outer leaflet of the membrane. PS has been described to be a so-called “death-signal” or “eat-me-signal”, because once externalized it can lead to phagocytosis, either directly via PS-recognizing receptors (Kobayashi et al, 2007; Lee et al, 2011; Fadok et al, 1992; Henson et al, 2001) or via opsonization by PS-bridging proteins, such as lactadherin (Nandrot et al, 2007; Hanayama et al, 2004; Fens et al, 2008) or Gas6 (Wu et al, 2006). On the other hand, PS exposure has also been shown to be important for activation of the coagulation cascade and has been claimed to play a role in membrane vesicle formation as well (Owens & Mackman, 2011). Thus, PS exposure on erythrocytes could serve several functions, ranging from being an “eat me” signal to being a factor to support coagulation.
During long-term storage, up to 42 d, of leucoreduced erythrocytes very little PS exposure is observed, but it is unknown whether subsequent transfusion of stored erythrocytes induces PS exposure on the cell surface. We hypothesized that stored erythrocytes show additional damage including PS exposure after transfusion, which might explain the adverse clinical outcomes of patients receiving long-term stored erythrocytes. To study this hypothesis, we incubated stored erythrocytes overnight at 37°C. This study showed that overnight incubation of long-term stored erythrocytes leads to an increase in haemolysis, potassium leakage, vesicle formation and PS exposure. We also observed that reduced intracellular potassium in stored erythrocytes has a negative effect on flippase activity, thereby priming stored erythrocytes for PS exposure. PS exposure on stored erythrocytes after overnight incubation was identified as a marker of erythrocytes that were prone to shed vesicles. Lastly, the vesicles shed by long-term stored erythrocytes were found to be PS positive and to support the coagulation cascade.
Materials and methods
Annexin V alexa fluor 647 conjugated (AV-alexa647) and goat anti-human IgG alexa568 (anti-human alexa568) were from Invitrogen (Paisley, United Kingdom). 1-Palmitoyl-2-[6-((7-nitro-2–1,3-benzoxadiazol-4-yl)amino)caproyl]-sn-glycero-3-phosphoserine (NBD-PS), 1-palmitoyl-2-[6-((7-nitro-2–1,3-benzoxadiazol-4-yl)amino)caproyl]-sn-glycero-3-phosphocholine (NBD-PC) were from Avanti Polar Lipids (Alabaster, AL, USA). Annexin V fluorescein isothiocyanate (FITC) was obtained from VPS-Diagnostics (Hoeven, The Netherlands). Human serum albumin (HSA) and human anti-human antibodies against the minor blood group antigens Fya, Fyb, Lua, Lub, Kpa and Kpb were obtained from Sanquin Reagents (Amsterdam, the Netherlands). Human anti-human antibodies against the minor blood group antigens s and S were from Ortho Clinical Diagnostics (Tilburg, the Netherlands). Valinomycin was obtained from Sigma-Aldrich (Zwijndrecht, The Netherlands). Factor Xa was from Enzyme Research (South Bend, IN, USA).
Leucoreduced erythrocyte units, stored in saline-adenine-glucose-mannitol (150 mmol/l NaCl, 1·25 mmol/l adenine, 50 mmol/l glucose, 29 mmol/l mannitol, pH 5·6) (SAGM) (Fresenius Kabi, The Netherlands), were obtained from the Dutch Sanquin Blood Supply Foundation, after obtaining informed consent from the donors. Blood studies were approved by the Sanquin Research institutional medical ethical committee in accordance with the standards laid down in the 1964 Declaration of Helsinki. The erythrocytes were stored at 2–6°C in a standard blood bank refrigerator.
Isolating erythrocytes from whole blood
Venous blood was collected from healthy donors, after obtaining informed consent. Blood studies were approved by the Sanquin Research institutional medical ethical committee in accordance with the standards laid down in the 1964 Declaration of Helsinki. Erythrocytes were isolated from fresh heparinized whole blood by centrifugation at 270 × g for 15 min. After removing the platelet-rich plasma and the peripheral blood mononuclear cells, the erythrocytes were washed two times with SAGM and resuspended in SAGM. Final cell concentration was determined with an Advia 2120 (Siemens Medical Solutions Diagnostics, Breda, The Netherlands).
Samples were obtained from stored erythrocytes and diluted to a haematocrit (Hct) of 40% in their own supernatant. To obtain the supernatant of stored erythrocytes, stored erythrocytes were centrifuged at 1000 × g for 5 min, after which the supernatant was collected. Erythrocytes isolated from fresh whole blood were diluted to a Hct of 40% in SAGM. The diluted erythrocytes were incubated overnight at 37°C and analysed the following day.
Haemolysis was determined as described previously (de Korte et al, 2008). Briefly, free haemoglobin was determined by absorbance measurement of cell supernatant at 415 nm or 514 nm by a spectrophotometer (Rosys Anthos ht3, Anthos Labtec Instruments GmbH, Salzburg, Austria), with correction for plasma absorption if necessary. Haemolysis was expressed as a percentage of total haemoglobin present in the red cell concentrate after correction for Hct.
Extracellular Potassium was measured with a Rapidlab 865 (Siemens Medical Solution Diagnostics).
Annexin V labelling of RBCs
The amount of erythrocytes expressing PS on their outer membrane was determined as described elsewhere (Burger et al, 2010). Breifly, labelling with AV was performed by adding 0·25 μl of AV-alexa647 or AV-FITC to 1 × 106 erythrocytes in 50 μl buffer. The buffer was either HEPES buffer [(132 mmol/l NaCl, 20 mmol/l HEPES (N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid), 6 mmol/l; KCl, 1 mmol/l MgSO4, 1·2 mmol/l K2HPO4), pH 7·4 (all from Sigma-Aldrich)] supplemented with 10 mmol/l glucose, 2 mmol/l CaCl2 and 0·5% HSA as the positive sample, or HEPES buffer supplemented with 10 mmol/l glucose, 2 mmol/l ethylene glycol tetraacetic acid (EGTA) (Sigma-Aldrich) and 0·5% HSA as the negative sample. After incubation on ice for 30 min, cells were diluted four times with incubation medium, supplemented with CaCl2 or EGTA respectively, and analysed on a flow cytometer with high throughput system. Data analysis was performed with computer software. The percentage of PS positive erythrocytes was determined by comparison of the positive sample with the negative sample.
Flippase and scrambling activity
To determine flippase and scrambling activity, NBD-PS and NBD-PC translocations were determined respectively. Erythrocytes were diluted to a concentration of 2 × 108 cells/ml in HEPES+. Subsequently, 0·8 μmol/l NBD probe was added and samples were taken immediately (0 min), as background, and after 15 mins incubation at 37°C in a shaking heating block. The samples were diluted 1:10 in ice cold HEPES supplemented with 10 mmol/l glucose with or without 1% HSA. After washing the samples once, they were resuspended in dilution buffer and analysed on a flow cytometer with high throughput system.
In experiments where double-staining with AV labelling was performed, the staining for AV was performed as above with the following modifications: After taking samples for flippase or scrambling activity they were incubated for 30 min on ice in the presence or absence of HSA. After spinning the samples down and washing once with HEPES buffer at 600 g at 4°C, the samples with HSA were resuspended in HEPES buffer supplemented with 10 mmol/l glucose, 1% HSA, 2 mmol/l CaCl2 and 1:200 AV-alexa647. The samples without HSA were resuspended in HEPES buffer supplemented with 10 mmol/l glucose, 2 mmol/l EGTA and 1:200 AV-alexa647. After 30 min incubation on ice, the samples were four times diluted with HEPES buffer supplemented with 10 mmol/l glucose, 1% HSA, 2 mmol/l CaCl2 or HEPES buffer supplemented with 10 mmol/l glucose, 2 mmol/l EGTA, respectively and analysed on a flow cytometer with high throughput system. Scrambling and flippase activity were expressed as the ratio between NBD probe on the inner leaflet of the membrane, the sample with HSA, and total NBD probe uptake, the sample without HSA.
Valinomycin-induced potassium leakage
Potassium leakage was induced by incubating erythrocytes isolated from fresh whole blood with the potassium ionophore, valinomycin. Erythrocytes isolated from heparinized whole blood were washed once and diluted in HEPES+ to a final concentration of 4 × 109 cells/ml. After addition of 10 μmol/l valinomycin, the cells were incubated overnight at 37°C with end-over-end shaking.
To buffer the potassium leakage the extracellular potassium concentration was increased to 100 mmol/l. This was accomplished by replacing 91·6 mmol/l NaCl present in the HEPES buffer with KCl (HEPES+/K). Erythrocytes isolated from heparinized whole blood were washed once and diluted in HEPES+/K to a final concentration of 4 × 109 cells/ml. After addition of 10 μmol/l valinomycin, the cells were incubated overnight at 37°C with end-over-end shaking.
FXa generating assay
Activated Factor X (FXa) generating assays were performed as previously described (Bloem et al, 2012). Briefly, vesicles were isolated after incubation of long-term stored erythrocytes following an overnight incubation at 37°C by centrifugation at 270 × g for 15 min. Subsequently, the supernatant was collected and centrifuged at 21 000 × g for 30 min. After removing the supernatant, the pellet was resuspended in HEPES medium. The amount of vesicles was expressed in μmol/l phospholipids. As a positive control, sonicated phospholipid vesicles comprising 15% PS, 20% phosphatidylethanolamine (PE) and 65% phosphatydilcholine (PC) were used. The assay was performed by incubating 0·3 nmol/l recombinant B-domain deleted factor VIII (FVIII) and 0–16 nmol/l activated factor IX (FIXa) in a buffer containing 40 mmol/l Tris-HCl (pH 7·8), 150 mmol/l NaCl, 0·2% (w/v) bovine serum albumin (Merck, Darmstadt, Germany). The reaction was started with the addition of 1·5 mmol/l CaCl2 and 1 nmol/l thrombin and incubated for 2 min, after which 200 nmol/l FX was added for 1 min. The amount of generated FXa per minute was subsequently assessed as described (Bloem et al, 2012).
Data was analysed using Graphpad Prism 5·01 for Windows (GraphPad Software, La Jolla, CA, USA). Except where more than two groups were compared, the student's t-test was used to compare means. Where more than two groups were compared, One-Way analysis of variance (anova) was used, with the Bonferroni post-test to compare all pairs. To compare means over time, statistical analysis was performed by 2-way anova tests with the Bonferroni post-tests.
Stored erythrocytes are primed for K+-leakage, haemolysis, PS exposure and vesicle formation
To be able to identify changes occurring in stored erythrocytes after transfusion, we incubated samples of stored erythrocytes overnight at 37°C in their own supernatant. Upon overnight incubation, we observed a clear effect of storage time on potassium leakage (Fig 1A), PS exposure (Fig 1B) and haemolysis (Fig 1C). Furthermore, we observed the presence of vesicles when erythrocytes that had been stored for 42 d were incubated overnight at 37°C (Figs 1D and E).
Storage of erythrocytes is associated with changes in PS flippase and scrambling activities
As PS exposure is determined by both flippase and scrambling activity (Daleke, 2003), we investigated whether these activities were changed in long-term stored erythrocytes. The flippase and scrambling activity were determined by incubation with NBD-PS or NBD-PC (Verhoeven et al, 2006), respectively. As shown in Fig 2A, the flippase activity was lower in long-term stored erythrocytes, which is in line with a previous study (Verhoeven et al, 2006). There was no difference in scrambling activity when the erythrocytes were directly analysed from the storage bag. Clearly, this result was consistent with the lack of PS exposure on long-term stored erythrocytes when they were analysed directly from the storage bag as described above [see also Verhoeven et al (2006)].
Next, we determined the flippase and scrambling activities after overnight incubation at 37°C in long-term stored erythrocytes and erythrocytes freshly isolated from whole blood. As an internal control, both long-term stored and freshly isolated erythrocytes were diluted in ABO-matched whole blood, upon which the erythrocytes present in whole blood itself were used as a standard. The freshly isolated erythrocytes and the erythrocytes in whole blood showed the same amount of flippase activity (Figs 2B and C). The difference in flippase activity between long-term stored and freshly isolated erythrocytes prior to overnight incubation was increased after overnight incubation. There was also no difference in scrambling activity between freshly isolated and whole blood erythrocytes, while long-term stored erythrocytes showed increased scrambling activity (Figs 2D and E). These findings are consistent with the observed PS exposure on long-term stored erythrocytes after an overnight incubation.
Given that not all long-term stored erythrocytes were PS positive, this indicated that a subgroup of erythrocytes had a higher scrambling activity or lower flippase activity, as either of these changes would lead to PS exposure. By performing a double staining for PS exposure and either flippase or scrambling activity, we observed that there was no difference in flippase activity (Fig S1A), but that PS-positive erythrocytes had a higher scrambling activity (Fig S1B)
The role of potassium leakage in the regulation of flippase and scrambling activity
Previous reports have shown that erythrocytes treated with the potassium ionophore, valinomycin, are more prone to PS exposure (Wolfs et al, 2009). As potassium is also lost by erythrocytes during storage and after overnight incubation, we investigated the effect of intracellular potassium concentration on flippase and scrambling activity. Therefore, we incubated fresh isolated erythrocytes with valinomycin. As expected, based on previous findings (Wolfs et al, 2009), treating erythrocytes with valinomycin led to potassium leakage (Fig 3A). By increasing the extracellular potassium concentration, we were able to inhibit the potassium leakage induced by valinomycin. The potassium leakage induced by valinomycin treatment led to a decrease in flippase activity (Fig 3B), which could be blunted by a higher extracellular potassium concentration. Valinomycin treatment did not lead to a higher scrambling activity (Fig 3C) and, in line with this observation, did not induce PS exposure (Fig 3D). As intracellular ATP levels can also affect the flippase activity, we also determined the ATP levels after valinomycin treatment. However, we did not observe any differences in ATP levels between erythrocytes treated with or without valinomcyn (Fig S2). These results show that potassium levels have a direct effect on flippase activity, but have no effect on scrambling activity.
PS exposure on long-term stored erythrocytes is linked to vesicle formation
As stated in the introduction, the PS exposure on long-term stored erythrocytes described above could have several implications. As we already showed that overnight incubation of long-term stored erythrocytes led to vesicle formation, we hypothesized that the PS exposure on long-term stored erythrocytes might be a sign or perhaps even a prerequisite of vesicle formation. Given that the loss of membrane is expected to be limiting, vesicle formation was anticipated to be a transient phenomenon. To determine whether this was indeed the case, long-term stored erythrocytes were incubated overnight at 37°C after which the PS-positive cells were sorted using a cell sorter (Fig 4A). To remove the bound annexin V from the erythrocytes, the cells were washed with buffer containing EGTA and resuspended in HEPES+. Subsequently, the erythrocytes were incubated for 4 h and again analysed for PS exposure and vesicle formation. After the 4-h incubation, a high percentage of PS-positive sorted erythrocytes were no longer PS-positive, while not all PS-negative sorted cells were PS-negative (Fig 4A). Strikingly, the PS-positive sorted samples contained significantly more vesicles than the PS-negative sorted samples after the 4-h incubation (Figs 4B-D). Together, these data show that the observed PS exposure is indeed transient and is likely to mark the process of vesicle formation. To confirm that at least some of the PS exposed on long-term stored erythrocytes was removed by vesicle shedding, we analysed the PS exposure on the vesicles. Indeed, a considerable percentage of vesicles was PS-positive (Fig 4E). We observed no difference in percentage of PS-positive vesicles from the PS-positive sorted erythrocytes and vesicles from the PS-negative sorted erythrocytes (Fig 4E). This suggests that a similar mechanism is likely to be present in all erythrocytes, irrespective of earlier differences in PS exposure, but that the extent to which vesicle shedding occurs is directly linked to the level of PS exposure on the outer leaflet of the erythrocyte membrane.
Erythrocyte vesicles act as a cofactor in the coagulation cascade
Several studies have demonstrated that phospholipid vesicles with PS in their outer leaflet can form a pro-coagulant platform to support the coagulation cascade (Bloem et al, 2012). Long-term stored erythrocytes may therefore serve this role as well. To this end, we assessed whether isolated erythrocyte vesicles support FXa generation by the activated FVIII (FVIIIa) –FIXa complex assembly, the activity of which critically depends on the presence of PS in the outer leaflet of the membrane. The results showed increased FXa generation in the presence of increasing concentrations of the erythrocyte vesicles (Fig 4F and G). In addition, the FIXa concentration required to reach half-maximum FXa generation on the erythrocyte vesicles was comparable to that obtained for FXa generation on the synthetic phospholipid vesicles (approximately 2 nmol/l). Taken together, these findings imply that the vesicles isolated from erythrocytes can form a platform for the correct formation of the FVIIIa-FXa complex assembly. This finding suggests in turn that these vesicles may indeed play a role in the coagulation cascade.
Many studies have addressed the changes occurring in erythrocytes during storage by analysing erythrocytes directly from the storage bag. However, it is highly likely that stored erythrocytes undergo many more changes once transfused. With this study we attempted to uncover hidden storage lesions in donor erythrocytes that become expressed only after transfusion. By incubating long-term stored erythrocytes overnight at 37°C, we were able to identify several additional changes which might also occur after transfusion of long-term stored erythrocytes. First of all, we showed that overnight incubation of stored erythrocytes induced additional potassium leakage, haemolysis and PS exposure in the long-term stored erythrocytes. Moreover, we observed clear vesicle formation. These observations are in line with several mouse models, where both free haemoglobin and vesicle formation were observed after transfusion and have been proposed as the mechanisms by which long-term stored erythrocytes induce damage after transfusion (Donadee et al, 2011; Hod et al, 2010; Kozuma et al, 2011). In particular, the amount of haemolysis in our model was striking, as it exceeded the international limit for stored erythrocytes, which is 0·8% for Europe and 1% for North-America. The pathophysiological consequences of cell free haemoglobin are extensive and mostly related to the ability to scavenge NO, a potent vasodilator and anti-thrombotic agent (Owens & Mackman, 2011). Prolonged exposure to cell free haemoglobin may even cause kidney damage, although in the case of transfusions, that would only become a problem when a patient needs repeated transfusions over a long period of time.
We also provide evidence that potassium leakage, a well-known effect of prolonged storage, leads to a decrease in flippase activity, as observed in long-term stored erythrocytes after overnight incubation. Previous studies had already shown that PS exposure could be induced by treatment of erythrocytes with a calcium-ionophore (Wolfs et al, 2009; Allan et al, 1980; Lang et al, 2003). In combination with the potassium-ionophore, valinomycin, this effect was increased (Wolfs et al, 2009; Kerbiriou-Nabias et al, 2011). Blocking K+ leakage by adding extracellular potassium or potassium channel blockers neutralized the additional effect of valinomycin on calcium-ionophore treated erythrocytes, showing that potassium leakage itself makes the cells more prone to expose PS. Whether this is a direct effect on the scrambling activity or the flippase activity had until now not been investigated. Our results show that decreased intracellular K+ levels directly inhibited flippase activity and that this was independent from the effect ATP levels have on flippase activity (Verhoeven et al, 2006). Although a decreased flippase activity will not directly result in PS exposure, it does make the cells more prone to expose PS once scrambling is activated. After an overnight incubation, long-term stored erythrocytes showed decreased flippase activity and a subset of these erythrocytes also had increased scrambling activity. We propose that during storage, erythrocytes sustain damage, but as they are stored at 4°C the scrambling activity will be very low and will be difficult to detect. However, upon overnight incubation at 37°C scrambling activity will be induced. Based on our current findings we propose that it is this latter subset of erythrocytes with enhanced scrambling activity that has suffered most of the damage due to storage, and that the decreased flippase activity that occurs as a result of the increased K+ leakage further promotes the exposure of PS on these cells upon transfusion.
We were also interested in the observed PS exposure when long-term stored erythrocytes were incubated overnight at 37°C. First of all, PS exposure is a hallmark of apoptotic cells and an “eat-me” signal (Wu et al, 2006). However, there are no reports showing directly that erythrocytes are phagocytosed as soon as they expose PS in a physiological manner. Secondly, vesicles from erythrocytes were isolated in previous studies. Sadallah et al (2008) showed that the vesicles from erythrocytes had anti-inflammatory properties and that this was due to the PS exposure on these vesicles. And finally, several reports have shown that PS exposure also has a function in cell signalling and is therefore not always a death signal (Elliott et al, 2006). In line with these reports, we showed that PS exposure can be transient on erythrocytes. Moreover, we provide evidence that the PS positive cells are most prone to shed vesicles. The PS exposure that was present on the erythrocytes was mostly transient and accompanied by vesicle formation. Furthermore, a considerable amount of these vesicles was PS-positive, indicating that shedding vesicles contributes to restoring the erythrocytes to a PS-negative state.
As vesicle formation is believed to be at least one of the mechanisms that contribute to complications in the recipients of blood transfusion (Amabile et al, 2010;Jy et al, 2011), we investigated the possible role of erythrocyte vesicles in coagulation. Our findings clearly demonstrate that erythrocyte vesicles derived from long-term stored erythrocytes incubated at 37°C can trigger coagulation and these findings are in line with previous reports (Jy et al, 2011; Rubin et al, 2010). Furthermore, erythrocyte vesicles have been implicated in influencing inflammation (Sadallah et al, 2008), angiogenesis (Rhee et al, 2004; Deregibus et al, 2007), and the regulation of vascular tone (Donadee et al, 2011). This growing body of evidence highlights the potential complications of vesicle release.
In conclusion, potassium leakage primes stored erythrocytes for PS exposure and subsequent shedding of vesicles. The shedding of vesicles, which promote coagulation and other adverse phenomena, may contribute to the deleterious side effects associated with transfusion. In the future, strategies to produce safer and more effective blood products should take the prevention of potassium leakage and vesicle formation from stored erythrocytes into account.
The research was supported by Grant PPOC-07-20 from Sanquin, Amsterdam, the Netherlands.
Contribution: P.B. performed research and wrote the manuscript; E.B., E.K. and P.H. performed experiments; A.V., S.M., D.d.K. and T.K.v.d.B. suggested key experiments and supervised the writing of the manuscript; and R.v.B. designed the study and supervised the writing of the manuscript.
Disclosure of conflicts of interest
Conflict-of-interest disclosure: The authors declare no competing financial interests.