CD59 signaling and membrane pores drive Syk-dependent erythrocyte necroptosis

Mature erythrocytes (red blood cells (RBCs)) undergo the programmed cell death (PCD) pathway of necroptosis in response to bacterial pore-forming toxins (PFTs) that target human CD59 (hCD59) but not hCD59-independent PFTs. Here, we investigate the biochemical mechanism of RBC necroptosis with a focus on the mechanism of induction and the minimal requirements for such RBC death. Binding or crosslinking of the hCD59 receptor led to Syk-dependent induction of vesiculated morphology (echinocytes) that was associated with phosphorylation of Band 3 and was required for Fas ligand (FasL) release. FasL-dependent phosphorylation of receptor-interacting protein kinase 1 (RIP1) in combination with plasma membrane pore formation was required for execution of RBC necroptosis. RIP1 phosphorylation led to the phosphorylation of RIP3, which was also critical for RBC necroptosis. Notably, RBC necroptosis was mediated by FasL and not by other candidate inducers, including tumor necrosis factor alpha (TNF-α) and TNF-related apoptosis-inducing ligand (TRAIL). Other types of RBC damage, such as eryptotic damage, failed to induce necroptosis when combined with hCD59 crosslinking. This work sheds light on the requirements for this recently discovered PCD in RBCs and provides a clear picture of the biochemical mechanism of induction of RBC necroptosis.


Results
RBC necroptosis induced by hCD59-specific PFTs is mediated by FasL but not by other known mediators/ stimuli of necroptosis. We previously demonstrated that RBC necroptosis, induced by the hCD59-specific PFTs vaginolysin (VLY) and intermedilysin (ILY), is dependent on Fas-FasL signaling, as a monoclonal antibody targeting FasL (mAb NOK-1) significantly inhibited RBC death. 1 Recombinant FasL (rFasL) enhances RBC death by near sublytic quantities of PFTs in a necrostatin-1-inhibitable manner (Figure 1a). The hCD59-specific PFTs, VLY and ILY, induce FasL-dependent phosphorylation of RIP1 in RBCs, 1 and addition of rFasL (but not TNF-α or TRAIL) to RBCs induces the phosphorylation of RIP1 (Figure 1b). Due to the fact that rFasL is sufficient to trigger RIP1 phosphorylation, its inclusion allows hCD59-independent PFTs, pneumolysin (PLY) and Staphylococcus aureus alpha toxin (A-tox), to induce necroptosis in RBCs (Figure 1a). 1 As FasL triggers necroptosis in some nucleated cells, [5][6][7] we asked whether two other defined stimuli of necroptosis, TNF-α and TRAIL, 6,7 have a role in this PCD in RBCs. In contrast to rFasL, addition of rTNF-α or rTRAIL to RBCs failed to induce RIP1 phosphorylation ( Figure 1b). Additionally, while neutralization of FasL with a mAb inhibits RBC necroptosis by VLY and ILY, 1 similar neutralization of TNF-α or TRAIL had no effect on necroptosis by these hCD59-specific PFTs (Figures  1c and e). Moreover, while the addition of exogenous rFasL endows PFTs with the ability to induce RBC necroptosis, regardless of intrinsic hCD59 specificity, similar addition of rTNF-α or rTRAIL had no effect on RBC death (Figures 1d  and f). These results indicate that FasL is a critical mediator of RBC necroptosis.
Binding or crosslinking of hCD59 leads to phosphorylation of RIP1 in human RBCs. hCD59-specific PFTs (VLY and ILY) induce RBC necroptosis while hCD59-independent PFTs do not, 1 but it is not known whether hCD59 ligation itself induces necroptosis signaling pathways. Treatment of RBCs with either an hCD59-specific mAb or a histidine-tagged version of the VLY binding domain (VLYD4) triggered RIP1 phosphorylation in RBCs (Figure 2a). Cross-linking of the hCD59-binding reagents with an anti-mouse IgG pAb (for the mAb) or an anti-His mAb (for VLYD4) resulted in more robust Figure 1 Necroptosis depends on FasL but not TNF-α or TRAIL in RBCs. (a) Addition of exogenous rFasL (10 ng/ml) with low doses of PFTs (0.02 HU) leads to necroptosis in RBCs that is prevented by the RIP1 inhibitor, necrostatin-1 (nec-1, 50 μM). (b) RIP1 IPs showing phosphorylation of RIP1 (p-RIP1) in response to rFasL (1 μg/ml) but not rTNFα (10 μg/ml) or rTRAIL (10 μg/mL) following 1 h of exposure. (c) Neutralization of TNF-α with the mAb J1D9 (1 μg/ml), (d) addition of rTNF-α (5 μg/ml), (e) neutralization of TRAIL with the mAb RIK-2 (1 μg/ml) or (f) addition of rTRAIL (5 μg/ml) has no effect on RBC death by VLY, ILY, PLYor A-tox. Time points for all hemolysis assays were 30 min. Error bars in hemolysis assays represent standard deviation. All hemolysis assays are the results of three independent experiments. Two-way ANOVA with Bonferroni posttest, ***Po0.001 CD59 drives erythrocyte necroptosis TJ LaRocca et al RIP1 phosphorylation (Figure 2a). Thus, signaling through hCD59 is sufficient to induce RIP1 phosphorylation in human RBC. The hCD59-specific PFT VLY induces RIP1 phosphorylation in RBCs, while the related but non-hCD59-specific PLY does not. 1 To determine whether toxin-induced RIP1 phosphorylation depends entirely on PFT binding to hCD59, we generated binding domain-swapped mutant versions of these PFTs (Figure 2b). When VLY was modified with the binding domain of PLY (VLY-PLYD4), it did not induce RIP1 phosphorylation ( Figure 2b). On the other hand, when PLY was modified with the binding domain of VLY (PLY-VLYD4) such that it was capable of binding hCD59, it did result in RIP1 phosphorylation ( Figure 2b). The phosphorylation of RIP1 is specific to CD59 signaling in RBCs as crosslinking of a related membrane protein, CD55, did not result in RIP1 phosphorylation ( Figure 2b). Likewise, when we crosslinked the binding domain of PLY (PLYD4), which does not bind hCD59, there was no phosphorylation of RIP1 (Figure 2b). These results demonstrate that PFT-mediated RIP1 phosphorylation depends on hCD59 ligation in RBCs.
The induction of RIP1 phosphorylation by mAb crosslinked hCD59 (CL-CD59) in RBCs was retained when combined with the hCD59-independent PFTs, PLY or A-tox (Figure 2c), or with two distinct inducers of eryptosis, hyperosmotic stress or excess calcium (Figure 2d). In addition, phosphorylation of RIP1 induced by hCD59 signaling via hCD59-specific mAb or VLYD4 was inhibited by both nec-1 and mAb against FasL (Figure 2e).
Crosslinking of hCD59 as well as hCD59-specific PFTs induces the phosphorylation of RIP3, which is required for RBC necroptosis. Similar to the phosphorylation of RIP1 (Figure 2), 1 RIP3, which is also critical for necroptosis, 7 is phosphorylated in response to crosslinking of hCD59 or hCD59-specific PFTs, VLY and ILY in RBCs (Figure 3a). This The p-RIP1 induced by CL-CD59 is retained in combination with (c) the hCD59-independent PFTs, PLY or A-tox, or (d) the eryptosis stimuli, hyperosmotic stress (osm) or excess calcium (cal). (e) The p-RIP1 induced by CL-CD59 or CL-VLYD4 is prevented by the RIP1 inhibitor nec-1 (50 μM) or neutralization of FasL (NOK-1, 1 μg/ml). Treatments for all experiments except for eryptosis were 1 h. Eryptosis stimuli were added to RBCs for a period of 24 h. When nec-1 or mAb NOK-1 was used, RBCs were treated with these inhibitors for 1 h before the addition of stimuli CD59 drives erythrocyte necroptosis TJ LaRocca et al is consistent with necrosome formation in RBC necroptosis. 1 hCD59-independent PFTs, PLY and A-tox, did not induce RIP3 phosphorylation. The phosphorylation of RIP3 was prevented by the recently characterized inhibitor of RIP3, GSK'872, 8 or by RIP1 inhibition with nec-1 (Figure 3b). That nec-1 prevents RIP3 phosphorylation indicates that this kinase is phosphorylated due to the action of RIP1 as is the case in nucleated cells. 7 Inhibition of RIP3 with GSK'872 also reduced RBC death by VLY and ILY but not by the hCD59-independent PFTs, PLY and A-tox, demonstrating a critical role for RIP3 in the necroptotic death of RBCs ( Figure 3c).
Binding or crosslinking of hCD59 leads to necroptosis of RBCs when combined with PLY. Although crosslinking of hCD59 with antibody or VLYD4 induces RIP1 phosphorylation, it is not sufficient to induce necroptotic death in RBCs ( Figure 4a). However, similar to the effect of rFasL, hCD59 ligation enhances RBC death by hCD59-independent PLY (Figures 4b and c). This suggests that, at minimum, RBC necroptosis requires hCD59 signaling and a PFT. RBC death by necroptosis induced by PLY combined with CL-CD59 is prevented by nec-1 (Figure 4d) or FasL neutralization ( Figure 4e). Moreover, inhibition of the known downstream effector pathways of RBC necroptosis, including formation of ceramide, reactive oxygen species, and AGEs, 1 results in inhibition of enhanced RBC death by PLY and CL-CD59 (Figures 4f and h).
The induction of RBC necroptosis by hCD59 crosslinking and pore formation depends on the size and composition of the membrane pore. As hCD59 binding/crosslinking led to RBC necroptosis when combined with PLY, we hypothesized that hCD59 signaling might induce RBC necroptosis when combined with any agent that forms membrane pores. Therefore, we examined the effect of CL-CD59 on RBC death caused by the cholesterol-dependent cytolysins arcanolysin and listeriolysin O, the complement MAC, the cholesterolindependent PFT S. aureus α-toxin (A-tox), or the polyene PFT of Streptococcus agalactiae β-hemolysin/cytolysin ( Figure 5). When CL-CD59 was combined with the cholesterol-dependent cytolysins, ALN or LLO, or the MAC it resulted in a level of RBC necroptosis similar to CL-CD59 with PLY (Figures 5a and c). However, when combined with A-tox, which forms pores of 1-2 nm in diameter 4 in contrast to the large pores (≥10 nm) formed by cholesterol-dependent cytolysins 9 and the MAC, 10 CL-CD59 induced a much lower level of necroptosis in RBCs ( Figure 5d). When combined with the non-protein PFT β-hemolysin/cytolysin, CL-CD59 failed to induce RBC necroptosis ( Figure 5e). The pore size for β-hemolysin/cytolysin has not been formally determined but is thought to be large. 11 Thus, RBC necroptosis induced by CL-CD59 combined with PFTs depends on pore size and perhaps nature (protein versus non-protein). RBC necroptosis induced by hCD59 signaling appears specific to damage by membrane pore formation, as CL-CD59 did not enhance death via eryptosis ( Figure 5f).
Pore formation is critical for the induction of RBC necroptosis by hCD59 signaling. PFTs may activate host cell signaling pathways via pore-dependent or -independent mechanisms. 4 We tested whether RBC necroptosis triggered by CL-CD59 and PFTs required functional pores. RBC death caused by the hCD59-specific PFTs, VLY and ILY, was abolished by dextran osmoprotection (Figures 6a and b). Additionally, mutant versions of PLY and A-tox that are capable of binding but are defective in membrane pore formation (PdB and A-tox toxoid, respectively) 8-10 did not induce RBC death with or without CL-CD59 ( Figure 6c). Moreover, RBC necroptosis induced by the combination of CL-CD59 with PLY, ALN, LLO, or A-tox was completely inhibited under osmoprotective conditions (Figures 6d and g). These results collectively demonstrate that functional membrane pore formation is required for hCD59-induced necroptosis in RBCs.  RBC necroptosis is associated with echinocyte transformation. We next examined the morphology of RBCs exposed to hCD59-specific PFTs or to mAb-based hCD59 crosslinking (Figure 7). When exposed to VLY or ILY, which induce RBC necroptosis, RBCs underwent a morphological transformation to echinocytes (Figure 7). Echinocytes are a phenotype of RBCs that actively form and release vesicles and can be induced by a number of different forms of stress. 12 Echinocyte transformation was not observed in response to PLY and A-tox, which do not induce RBC necroptosis. The basis for echinocyte transformation of RBCs during necroptosis depended on hCD59 binding, as VLY-PLYD4 did not induce echinocyte transformation while PLY-VLYD4 did (Figure 7). hCD59 crosslinking alone was sufficient to induce echinocyte transformation (Figure 7). Phosphorylation of Band 3 by Syk kinase leads to FasL release, echinocyte transformation, and RBC necroptosis. Echinocyte formation in RBCs is mediated by Syk kinase-dependent phosphorylation of the integral membrane protein Band 3. [13][14][15][16] The phosphorylation of Y21 of Band 3 by Syk disrupts the binding of this protein to the adaptor protein Ankyrin, thereby severing the anchoring of the RBC membrane to the spectrin cytoskeleton. [13][14][15][16] Due to this disruption, membrane vesiculation proceeds in RBCs resulting in the appearance of the echinocyte morphology. 13,14 As echinocyte formation was associated with RBC necroptosis (Figure 7) we asked whether Syk-dependent Band 3 phosphorylation occurred during this PCD. Indeed, phosphorylated Band 3 was observed in RBCs in response to the necroptosis-inducing hCD59-specific PFTs, VLY and ILY, but  Figure 8a). Furthermore, hCD59 signaling alone was sufficient to induce p-Band 3 as observed by mAb crosslinking of hCD59 (Figure 7a). Phosphorylation of Band 3 was also dependent on Syk, consistent with the molecular mechanism of echinocyte formation in RBCs ( Figure 8b). As Syk-dependent echinocyte formation is associated with RBC necroptosis, we considered a potential role for RBC vesiculation in this PCD pathway. FasL-dependent death in other cell types requires the release of FasL in membrane vesicles, which is followed by FasL binding to its receptor, Fas. 7,17-20 We observed FasL release from RBCs in response to sub-lytic quantities of VLY and ILY but not PLY or A-tox (Figures 8c and d). Crosslinking of hCD59 alone, which does not induce RBC death (Figure 4a), also resulted in the release of FasL from RBCs (Figures 8c and d). The release of FasL from RBCs in response to these necroptotic stimuli was dependent on Syk, suggesting that RBC vesiculation/echinocyte formation is the basis for FasL release during RBC necroptosis (Figures 8c and d). This is necessary for RBC necroptosis to proceed as the inhibition of Syk also prevented the phosphorylation of RIP1 and RIP3 induced by hCD59specific PFTs or CL-CD59 (Figures 8e and f). As predicted, RIP1 and RIP3 phosphorylation was downstream of Sykdependent Band 3 phosphorylation as nec-1 had no effect on this event (Figure 8g).
Syk inhibition on RBCs prevented CL-CD59-induced echinocyte transformation, but nec-1 had no effect (Figure 8h) supporting the concept that Syk-dependent Band 3 phosphorylation leads to vesiculation in RBC necroptosis and that RIP1 and RIP3 phosphorylation are downstream of this event. Syk activity was required for necroptotic death of RBCs as its inhibition resulted in a decrease of RBC death caused by VLYand ILY while having no effect on RBC death by PLY (Figures 8i and k). Inhibition of Syk also reduced RBC necroptosis caused by the combination of CL-CD59 and PLY (Figure 8l). Src family kinases interact with CD59 and phosphorylate cytosolic targets such as Syk in response to extracellular signals, 21 which may subsequently drive vesiculation. This is indeed the case in the induction of RBC necroptosis as inhibition of Src family kinases prevented the phosphorylation of Syk induced by VLY, ILY, or CL-CD59 (Figure 8m). The hCD59-independent PFTs, PLY and A-tox, did not induce phosphorylation of Syk, consistent with the results of Band 3 phosphorylation and echinocyte formation. These results suggest that hCD59 signaling activates Src family kinases, which lead to Syk-dependent vesiculation that is required for FasL-dependent RBC necroptosis.

Discussion
Minimal requirements for RBC necroptosis. Our biochemical analysis of RBC necroptosis has allowed us to determine its minimal requirements. The combination of phosphorylated RIP1/RIP3 and a functional membrane pore results in necroptosis of RBCs. RIP1 phosphorylation can be induced by rFasL (Figure 1c) 1 or by hCD59 signaling (mediated by endogenous FasL) in RBCs (Figure 2e). Additionally, hCD59 signaling leads to RIP3 phosphorylation mediated by RIP1 (Figures 3a and b). Despite the fact that these stimuli induce RIP1/RIP3 phosphorylation in RBCs they were not sufficient to cause death (Figures 1b and 4a). However, when combined with the addition of PFTs to RBCs these stimuli did induce necroptosis (Figures 1b,4d, h, 5a, and d). Osmotic protection and experiments with pore-deficient toxoids ( Figure 6) supported the hypothesis that the combination of phosphorylated RIP1/RIP3 and a functional membrane pore are minimally required for RBC necroptosis. Membrane poreinduced damage appears to be specifically required for RBC necroptosis, as hCD59 crosslinking had no effect on RBC death by eryptosis (Figure 5f).
Although functional pore formation is required for RBC necroptosis, the ability to stimulate necroptosis appears to depend on pore size and possibly toxin composition. hCD59 crosslinking stimulated robust necroptosis when combined with toxins that form large, protein-based pores. 22,23 In contrast, hCD59 crosslinking primed RBC necroptosis to a much weaker extent when combined with A-tox (Figure 5d), which forms smaller pores, and not at all with the polyene β-hemolysin (Figure 5e). This suggests that RBC necroptosis may be a specific response to protein-based membrane pores.
Biochemical induction of hCD59-induced necroptosis in RBCs. Our findings have shed light on the mechanism of hCD59-induced RBC necroptosis providing a fundamental understanding of RBC necroptosis caused by hCD59-specific PFTs (Figure 9). The process of RBC necroptosis begins with ligation of the hCD59 receptor. Crosslinking of hCD59 alone leads to vesicle formation in RBCs, transforming these cells into echinocytes (Figure 7). This begins as hCD59 signaling that leads to phosphorylation/activation of Syk via Src family kinases (Figure 8m). The induction of membrane vesiculation in RBCs downstream of hCD59/Src signaling is mediated by Syk-dependent phosphorylation of the integral membrane protein, Band 3 (Figures 8a, b, and 8h). The Syk-dependent formation of vesicles is the basis for the release of FasL from RBCs (Figures 8c and d). Following release from RBCs due to Syk-dependent vesiculation, FasL is free to bind its cognate receptor, Fas. The ligation of FasL with Fas leads to the phosphorylation of RIP1 (Figures 1b and 2e), which is critical for necroptosis in RBCs and other cell types. 1,7 It is clear that the interaction of FasL with Fas is upstream of RIP1 phosphorylation as neturalization of FasL prevents this phosphorylation downstream of hCD59 signaling (Figure 2e). Additionally, we have confirmed that FasL release is downstream of membrane vesiculation/echinocyte formation as inhibition of Syk prevents the release of FasL and the phosphorylation of RIP1 (Figures 8c and e). Moreover, FasL release occurs in the absence of RBC death, suggesting that this event occurs before the induction of necroptotic death (Figures 8c and d). The RIP1 inhibitor, nec-1, had no effect on the phosphorylation of Band 3 (Figure 8g) providing further confirmation that membrane vesiculation/echinocyte formation is upstream of RIP1 phosphorylation in RBC necroptosis. FasL-dependent RIP1 phosphorylation subsequently results in the phosphorylation of RIP3, which is required for RBC necroptosis (Figure 3). While hCD59 signaling alone causes FasL-dependent RIP1/RIP3 phosphorylation via Syk-dependent membrane vesiculation in RBCs, this is not sufficient to promote death by necroptosis (Figure 4a). Functional pore formation in the RBC membrane was shown to be necessary for necroptosis in these cells ( Figure 6). This leads to the conclusion that while RIP1/RIP3 phosphorylation and functional pore formation are each necessary for RBC necroptosis they are sufficient to promote this PCD only in combination.

Figure 9
Proposed mechanism for RBC necroptosis. (a) RBCs contain CD59, Fas, FasL, and the transmembrane protein Band 3 in their membrane. Band 3 anchors the membrane to the spectrin cytoskeleton via the adaptor protein Ankyrin. Within the RBC cytoplasm are RIP1 and RIP3 kinases. (b) RBC necroptosis is intitated through CD59 crosslinking/signaling, which may be caused by the binding of hCD59-specific PFT monomers (pictured). CD59 signaling leads to the phosphorylation of Syk kinase through the action of Src family kinases. Following activation by CD59/Src kinases, Syk phosphorylates Band 3. (c) Syk-dependent Band 3 phosphorylation disrupts anchoring of the RBC membrane to the spectrin cytoskeleton resulting in vesiculation/echinocyte formation. (d) FasL is released due to Syk-dependent vesiculation/echinocyte formation and subsequently binds its cognate receptor, Fas. Fas receptor signaling, following ligation of FasL, leads to the phosphorylation of RIP1 which in turn phosphorylates RIP3, but is not sufficient to induce necroptotic death of RBCs. (e) Creation of a membrane pore by PFTs combined with the phosphorylation of RIP1/RIP3 leads to RBC necroptosis and the death of the RBC CD59 drives erythrocyte necroptosis TJ LaRocca et al Although Syk-dependent membrane vesicle production/ echinocyte formation is not specific to RBC necroptosis, our findings indicate that this event is critical for necroptosis to proceed. This is evident as the inhibition of Syk resulted in a decrease in RBC death caused by the necroptosis-inducing hCD59-specific PFTs, VLY and ILY while having no effect on death by PLY, which does not induce RBC necroptosis (Figures 8i and k). Further evidence of the requirement of Syk-dependent vesiculation/echinocyte formation for RBC necroptosis was shown as Syk inhibition completely prevented this PCD induced by the combination of crosslinked hCD59 and PLY (Figure 8l).
In nucleated cell types, necroptosis is mediated by one of three ligands of the TNF superfamily, TNF-α, FasL, or TRAIL. [5][6][7] Our results suggest that in RBCs, necroptosis is mediated only by FasL. This is evident as addition of rFasL to RBCs resulted in the phosphorylation of RIP1 while addition of rTNF-α or rTRAIL did not (Figure 1b). Furthermore, while rFasL endows bacterial PFTs with the ability to induce RBC necroptosis regardless of intrinsic hCD59 specificity (Figure 1a), rTNF-α and rTRAIL do not (Figures 1d and f). It is also clear that RBC necroptosis induced by hCD59-specific PFTs is specifically mediated by endogenous FasL 1 and not TNF-α or TRAIL (Figures 1c and e).
Potential relevance of RBC necroptosis to pathological conditions. That the combination of phosphorylated RIP1/ RIP3 (induced by FasL alone or hCD59 signaling) and certain PFTs results in necroptosis of RBCs (Figures 1a, 4d, h, 5a, and d) may be of relevance to pathologic situations in vivo. In the context of infection by an organism such as S. pneumoniae, FasL has been shown to be present at high levels in vivo 24 and in vitro. 25 As infection with S. pneumoniae can progress to bacteremia 21 it is possible that the combination of circulating FasL and PLY leads to RBC necroptosis as a component of systemic infection. Moreover, as the natural function of hCD59 is to prevent the formation of the MAC through binding of complement components C8 and C9, we speculate that RBC necroptosis may occur during situations of autoimmune complement attack. In a situation such as this we would expect C8/C9 to be bound to hCD59 on RBCs, eliciting signal transduction by this receptor. This combined with formation of a functional MAC may produce RBC necroptosis as is the case for CL-CD59 and the MAC (Figure 5c). This idea is supported by the evidence that hCD59 signaling induced by C8/C9 mimicks the signal transduction caused by mAb crosslinking of this receptor in vitro. 26 Materials and Methods Human RBCs. Primary human RBCs were obtained from healthy volunteers under a protocol approved by the Columbia University Institutional Review Board.