Ligation of Glycophorin A Generates Reactive Oxygen Species Leading to Decreased Red Blood Cell Function

Acute, inflammatory conditions associated with dysregulated complement activation are characterized by significant increases in blood concentration of reactive oxygen species (ROS) and ATP. The mechanisms by which these molecules arise are not fully understood. In this study, using luminometric- and fluorescence-based methods, we show that ligation of glycophorin A (GPA) on human red blood cells (RBCs) results in a 2.1-fold, NADPH-oxidase-dependent increase in intracellular ROS that, in turn, trigger multiple downstream cascades leading to caspase-3 activation, ATP release, and increased band 3 phosphorylation. Functionally, using 2D microchannels to assess membrane deformability, GPS-ligated RBCs travel 33% slower than control RBCs, and lipid mobility was hindered by 10% using fluorescence recovery after photobleaching (FRAP). These outcomes were preventable by pretreating RBCs with cell-permeable ROS scavenger glutathione monoethyl ester (GSH-ME). Our results obtained in vitro using anti-GPA antibodies were validated using complement-altered RBCs isolated from control and septic patients. Our results suggest that during inflammatory conditions, circulating RBCs significantly contribute to capillary flow dysfunctions, and constitute an important but overlooked source of intravascular ROS and ATP, both critical mediators responsible for endothelial cell activation, microcirculation impairment, platelet activation, as well as long-term dysregulated adaptive and innate immune responses.

Introduction overlooked source of intravascular ROS and ATP, both critical mediators responsible for endothelial cell activation, and microcirculation impairment.

Preparation of Red Blood Cells (RBCs)
Blood was obtained from healthy adult volunteers in accordance with the guidelines of, and approved by the Institutional Review Board of Beth Israel Deaconess Medical Center. Donors from multiple diversities were recruited by local advertising and written, informed consent was obtained from each volunteer in accordance with the Declaration of Helsinki. Approved study staff were responsible for all blood collection, processing, and documentation of consent files. Red blood cells were isolated as previously described [19]. For flow cytometry and microscopy experiments, 20-80 μL RBCs were harvested by finger prick and collected into a microcentrifuge tube containing HBSS ++ with 0.05% IgG-free BSA. RBCs were centrifuged (5,000 x g, RT, 1 min), and supernatant was removed by aspiration. This step was repeated 2 times.
Criteria for Sepsis and Septic Shock Patient Samples-For this study, "sepsis" was defined as meeting systemic inflammatory response criteria in the setting of infection, whereas severe sepsis was defined as sepsis plus organ failure. These definitions are based on the American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference [20].

ATP release measurement from RBC
RBCs (2% Hct in HBSS ++ , 0.05% BSA) were placed in 1.5 mL reaction tubes and mixed with inhibitors or activators already placed in the tubes to minimize stimulation. Following incubation (RT x 10 min), RBCs were centrifuged (500 x g, RT, 2 min), the supernatant was discarded and RBCs were resuspended in buffer using an electronic pipette with the appropriate inhibitor or activator or buffer alone and again centrifuged (500 x g, RT, 2 min). Five replicate aliquots were placed into black, flat-bottom, 96-well plates, and ATP concentration was measured as described before [19].

Hemolysis
Remaining cell suspensions were centrifuged (5,000 x g, RT, 2 min) and 80 μL of supernatant was measured at 415 nm using a VersaMax spectrophotometer (Molecular Devices, Sunnyvale, CA, USA). Experiments were excluded from this study if RBC hemolysis exceeded 0.25%

Immunoblotting
RBCs were analyzed as previously described [21]. Nitrocellulose membranes were developed with SuperSignal 1 West Pico or Femto Chemiluminescent Substrate [22] and imaged using the LAS 4000 imaging system (Fujifilm).

Flow Cytometry
RBCs were analyzed using either FACScan or LSR II (BD Biosciences, San Jose, CA, USA). In all experiments 100,000 events were recorded and analyzed using FlowJo software Version 9.3.3. For each figure, a representative histogram displaying the mean fluorescent intensity [23] is shown, in addition to a cumulative graph for all experiments conducted.

Fluorescence Recovery After Photobleaching
RBC lipid mobility was measured by incubating (RT x 30 min) RBC (10% Hct in HBSS ++ ) with DiO lipid dye (15 μg/mL) and washed prior to addition of antibodies or reagents. RBCs were imaged using a 60x objective on an Olympus microscope and bleached with Vector Controller Laser and analyzed as described before [19].
NY, USA), using a Retiga EXi (QImaging, Surrey, Canada) CCD camera controlled by iVision 4.7 software (BioVision) at a rate of 10 frames/s. Movies were analyzed off-line frame by frame, and RBCs that displayed unusual shapes, overlapped, or clustered were excluded from measurement. For each experimental condition at least 50 RBCs were counted.

Statistical Analysis
All flow cytometry experiments were analyzed using FlowJo version 9.6.4 (TreeStar) or higher. Statistical tests were performed using Prism version 4.0 (GraphPad). All error bars represent the mean of 5 readings + S.D. unless noted otherwise.

Results
Ligation of GPA stimulates ROS production DHR 123-loaded RBCs were incubated with tert-butyl hydroperoxide (t-BHP), a known ROS promoter. t-BHP (10, 100 μM) increased intra-RBC ROS production in a concentration dependent manner (S1 Fig), from MFI = 35.7 in control RBCs to 49.6 and 1164, respectively. For in vitro experiments, we elected to ligate GPA with monoclonal antibodies (mAb), which induce profound changes in human RBCs during pathologies associated with naturally occurring anti-GPA antibodies [9]. Ligation of GPA with mAb anti-GPA ( Fig 1A, top left panel) increased intra-RBC ROS production over control, from MFI = 7.9 to 16.5, and pretreating RBCs with the cell permeable ROS scavenger GSH-ME (1 mM) reduced GPA-mediated ROS increase below basal levels (MFI = 6.18, top right panel), indicating that at basal conditions circulating RBCs contain low but measurable levels of ROS. To confirm that ROS production was due to the direct and specific engagement of GPA, we ligated RBC CR1, which we have shown increases RBC membrane deformability [25] and triggers a robust ATP response [19]. Here, ligation of RBC CR1 did not promote any measurable increase in intra-RBC ROS production ( Fig 1A, lower right), suggesting that: a) GPA-mediated RBC ROS generation is not due to a non-specific response of RBCs to engagement of membrane proteins by antibody, and b) the increase in partially inhibited by the cell permeable ROS scavenger, GSH_ME ( Fig 1A, lower left panel). Recently, NADPH oxidase [26] was identified as a critical enzyme in RBC ROS production, both in healthy and sickle RBCs [27]. We used DPI, a potent NOX inhibitor, to test if GPA-mediated ROS production is dependent on NOX activity. Inhibition of NOX activity by preincubation of RBCs with 10 μM DPI decreased the intracellular ROS concentrations in GPA-ligated RBCs from MFI = 35.4 to MFI = 19.5 (IgM control MFI = 12.2) (Fig 1B) suggesting a significant role of NOX in GPA-mediated ROS production. In addition, the GTP-ase protein Rac1 was recently implicated as an activator of NOX in human RBCs [27], and therefore critical to RBC ROS production. To test the involvement of Rac1 in GPA-mediated ROS production we next treated RBCs with a cell permeable Rac1 inhibitor (NSC23766) prior to GPA ligation, and observed a similar decrease in ROS levels, from MFI = 26.3 to MFI = 14.8 ( Fig  1C), suggesting that GPA ligation induces an increase in RBC ROS production that is mediated by Rac1 and NOX.

GPA ligation induces caspase-3 activation through increased intracellular ROS
In the past, others have shown a correlation between intracellular ROS accumulation and caspase-3 activation [28]. We therefore asked if GPA-induced ROS production activated caspase-3. RBCs pretreated with FLICA were incubated with either control IgM, anti-GPA (E4), or 10 μM PAC-1 (positive control) for 15 min at RT. Using flow cytometry, we observed a 4.3x increase in caspase-3 activation in GPA-ligated RBCs (Fig 2A) compared to IgM control, (MFI = 28.3 vs. MFI = 6.6), similar to our positive control (MFI = 23.1). Incubation of RBCs with known ROS promoter t-BHP generated a dose dependent increase in caspase-3 activation from basal MFI levels of 3.52 to 39.6 (1 mM) and 99.4 MFI (3 mM, Fig 2B, left). Similar to our previous results (Fig 1A), incubation of RBCs with anti-CR1 did not trigger activation of procaspase-3 ( Fig 2C), also suggesting that GPA-dependent activation of caspase-3 is not a non-specific response of RBCs to GPA ligation. We next investigated the pattern of caspase-3 activation induced by GPA ligation by fluorescence microscopy ( Fig 2D). Interestingly, the distribution of activated caspase-3 showed a punctuated pattern near the membrane following GPA ligation (top panels), and a uniform cytoplasmic staining following t-BHP-dependent activation (bottom panels). Taken together, our data suggest that ligation of GPA triggers an increased intracellular production of ROS that is responsible for activation of procaspase-3.

GPA-mediated ATP release depends critically on intracellular ROS
To determine if a pore large enough to allow ATP (375 Da) release opens following GPA ligation, we incubated RBCs with YO-PRO 1 -1 iodide, a cell impermeant nucleic acid stain that has been used to quantify gap junction channels, a family of pores that allow passage of small molecules such as ATP [29]. Ligation of GPA using the R10 epitope (Fig 3A, bottom panel) induced an increase in RBC-associated fluorescence from 58 to 78 AU when compared to control Ab (top panel). To ensure that this increase was not the result of antibody-induced cell aggregation, we measured the size distribution of RBCs before and after GPA ligation (Fig 3A,  right panel). Having determined that a membrane pore was opening following GPA ligation, we investigated whether ATP was simultaneously released. RBCs incubated with anti-GPA (R10 and E4) significantly increased ATP concentration in the supernatant from 4.6 ± 0.4 nM to 10.8 ± 0.8 nM (R10) and 10.0 ± 1.6 nM (E4) (p < 0.008) (Fig 3B). We next incubated RBCs with a fluorescent ATP probe 2-2 Zn(II), which binds to the extracellular side of the plasma membrane and fluoresces upon interaction with ATP as it is released. [19] GPA ligated RBCs showed a significant increase in ATP release compared to control, from MFI = 39.6 to MFI = 1282 (unstained MFI = 5.3) (Fig 3C, left panel). This dye promotes cell aggregation due to its lipophilic nature [19], however this did not prevent observing fluorescent ATP levels for each cell. A similar increase in fluorescence intensity was observed using wide field fluorescence 3-D acquisition followed by deconvolution (right panels). RBCs incubated with t-BHP (100 μM and 1 mM) also released increased amounts of ATP, from 5.4 ± 1.2 nM to 7.4 ± 1.1 nM and 10.0 ± 2.1 nM, respectively (p < 0.01) (Fig 3D). Pretreatment of RBCs using the cellpermeable glutathione monoethyl ester (GSH-ME, 1 mM) significantly decreased the amount of ATP following GPA ligation, from 10.8 ± 0.6 nM to 8.2 ± 1.0 nM (p < 0.008) (Fig 3E). However, pretreatment under similar conditions with cell impermeable glutathione (GSH) had no effect (anti-GPA = 13.1 ± 2.2 nM, anti-GPA + GSH = 13.4 ± 1.2 nM, p = 1) (Fig 3F), suggesting that the extracellular ROS-based autocrine effect on RBCs was not required for the GPA-mediated ATP release. Next, we asked if caspase-3 played a role in ATP release following GPA-ligation. Incubation of RBCs with caspase-3 activator PAC 1 (1 and 10 μM) triggered a significant increase in ATP release, from 5.5 ± 0.5 nM to 6.9 ± 0.5 nM and 11.4 ± 0.2 nM, respectively (p < 0.02) (Fig 3G). However, pre-incubation of RBCs with a caspase-3 inhibitor, FAM-DEVD-X, prior to GPA ligation did not significantly alter ATP release (p = 0.2) (Fig 3H). Our data suggest that GPA ligation-induced ATP release is dependent on intracellular but not  extracellular ROS levels, and although GPA-mediated ROS production activates caspase-3, this signaling may not be directly linked to GPA and ROS-induced ATP release.
Several pores and channels have been described to be functionally associated with ATP release in RBCs such as pannexins, MRPs, and CFTR [30][31][32]. In particular, CFTR expression and its involvement with ATP release was previously shown in murine RBCs [31]. Despite confirming the presence of CFTR on human RBCs by Western blot (S2 Fig), we were unable to confirm its role in GPA-mediated ATP release using the specific inhibitor GlyH-101. Similarly, MRP channels and purinergic receptors were not found to be independent conductors for ATP released from RBCs secondary to GPA ligation (S2 Fig).

Intracellular ROS production decreases membrane deformability
We have previously reported that during chronic inflammations such as in SLE and trauma, RBCs undergo alteration in the phosphorylation levels of β-spectrin and band 3 [5,6]. Importantly, band 3 phosphorylation was recently reported to depend critically on protein kinase p72 Syk, an enzyme that is activated by increased intra-RBC ROS [33]. We next investigated the effect of RBC GPA ligation on tyrosine phosphorylation of band 3, and its dependence on ROS. We used eosin-5-maleimide (EMA) as an indirect marker for band 3 phosphorylation. The efficiency of EMA binding to lysine 430 on the extracellular loop of band 3 was shown to parallel the levels of band 3 tyrosine phosphorylation, although the details of the binding are currently poorly understood [34]. Our data show that anti-GPA mAb (E4) induced band 3 phosphorylation (E4 MFI = 25.5, control MFI = 14.6) (Fig 4A, top left panel), an increase that was significantly reduced when RBCs were pretreated with GSH-ME (E4 MFI = 14.9). Also, incubation of RBCs with 100 μM t-BHP triggered an increase in band 3 phosphorylation from MFI = 14.6 to 17.7 (top right panel). In both experiments, band 3 phosphorylation was reduced following pretreatment with GSH-ME (100 μM t-BHP MFI = 6.4), suggesting that the baseline phosphorylation levels of band 3 depend partially on ROS. This was further confirmed by Western blotting (bottom right panel) using anti-phosphotyrosine, pY20 (Biosource) as a developing antibody. As a positive control, RBCs were incubated with sodium ortho-vanadate. We next asked if the effect of GPA ligation on membrane deformability is a result of excessive ROS generation, and if ROS scavengers can prevent it. We used 2-D microchannels to assess RBC deformability, measured as the time required for RBCs to pass through a capillary-like array (Fig 4B, top panel). Ligation of GPA increased the average time required for RBCs to pass through the capillary-like channel, from 2.7 ± 0.2 s to 3.7 ± 0.5 s (p < 0.01). However, GPA-ligated RBCs in the presence of GSH-ME required less time to deform and move through the channels (2.6 ± 0.2 s, p < 0.001), suggesting that intracellular ROS is in part responsible for the GPA-mediated decrease in membrane deformability (Fig 4B, bottom panel).

GPA-mediated ROS production inhibits lipid mobility
We next investigated the role of GPA-mediated ROS production on plasma membrane lipid mobility using fluorescence recovery after photobleaching (FRAP). Fig 4C shows serial images of a representative RBC during a FRAP experiment, while a corresponding recovery rate of the membrane at the specified region is shown in Fig 4D. GPA-ligated RBCs resulted in a significant reduction in the mobile fraction from 87.5 ± 3.7% in control IgM-treated RBCs down to 77.4 ± 3.4% (p < 0.0001) (Fig 4E). Our results also show that pretreatment with GSH-ME restored the GPA-mediated loss in lipid mobility, with a mobile fraction of 91.0 ± 4.0% compared to 77.4 ± 3.4% (p < 0.0001).

Complement-altered RBCs from septic patients partially recapitulate the functional phenotype of GPA-ligated RBCs
RBCs from 27 sepsis patients and 22 controls were analyzed for complement deposition, ROS production, band 3 phosphorylation, ATP release, membrane deformability, and lipid mobility. Not all patient samples were investigated for all functional parameters. The results are presented as ratios between the measured value of septic over control RBCs. Fig 5A shows that in the case of C4d fragments (right panel), deposition of RBCs from septic patients was more pronounced in severe septic and septic shock when compared to controls, whereas C3d fragments in (left panel) septic shock patients had levels not different than uninfected controls. RBC band 3 phosphorylation was not uniformly increased in sepsis patients; only a few patients displayed increased levels of band 3 phosphorylation, without a clear correlation between the RBC Treatment of RBC with GPA decreases lipid recovery rate, which is partially restored by GSH-ME. RBCs were pretreated with 1μM GSH-ME for (RT x 5 min) and then incubated with control or anti-GPA antibody (1 μg/ml, RT x 10 min).
doi:10.1371/journal.pone.0141206.g004 complement levels and the intensity of band 3 staining (Fig 5B). Conversely, intra-RBC ROS concentrations were increased in all septic samples studied (Fig 5C). When measuring the ATP concentration in the supernatant, we observed a gradual decrease in the amount of ATP that correlated inversely with the severity of sepsis (Fig 5D). Consistent with our in vitro FRAP data RBC lipid mobility of 4 patients with septic shock (Fig 5E) was significantly decreased (Mf control = 87.5%, Mf sepsis = 80.6%, p < 0.0008), when compared to matched controls. Consistent with previous reports [35][36][37], RBCs from septic shock patients also showed a significant decrease in membrane deformability (Fig 5F), as measured by 2-D microchannel arrays (t control = 3.2 seconds, t sepsis = 4 seconds, p < 0.0001). Moreover, RBCs isolated from septic patients treated with 1 mM GSH-ME for 15 minutes prior to functional analyses partially restored both membrane deformability (from 4 to 3.2 seconds) as well as lipid mobility (from Mf control = 80.6% to Mf sepsis = 89.6%), suggesting that the beneficial microcirculatory effects of antioxidant therapy reported in sepsis [38] could also be attributed to its effect on the mechanical properties of circulating RBCs.

Discussion
In this study, we investigated the deleterious effects of RBC ROS production promoted by GPA ligation using both in vitro and ex vivo approaches. Our data demonstrate that ligation of GPA by antibody initiates a NOX-dependent increase in ROS production. In turn, elevated intracellular ROS triggers multiple signaling cascades culminating in loss of RBC mechanical functions and ATP release, both of which could partially explain the flow alterations described in sepsis. Others have shown that Rac GTP-ases and PKC contribute to greater oxidative stress in sickle cell RBCs by increasing erythrocyte NOX activity [27]. Our data show a similar signaling pathway whereby GPA ligation activates Rac 1, increasing NOX activity, and generating ROS. However, the GPA-mediated ROS production seems not to fall under the classical NOX activation pathway involving activation of PKC and phospholipase A2 [39]. Both steps require Ca ++ , yet engagement of GPA by antibody does not promote a measurable Ca ++ influx. Furthermore, inhibition of PKC with calphostin C prior to GPA ligation does not prevent ROS production, suggesting that additional mechanisms involved in activation of NOX play a role in GPAmediated ATP release.
Human RBCs contain large amounts of procaspase-3, considered a leftover erythropoiesis protein [40]. In nucleated cells, reports show a correlation between increased intracellular ROS and programmed cell death by direct activation of caspase-3 [28]. Our results show that GPA activates caspase-3, but in a different manner than t-BHP; whereas t-BHP generated cell-wide activation, resulting in uniform staining, GPA-ligation activated caspase-3 in a punctuated, well-defined pattern. Localized, transient, functional domains have been described in RBCs as a result of specific signaling pathway responses [41]. It is conceivable a similar mechanism is responsible for limited activation of caspase-3, although they did not co-localize with GPA molecules.
Recently, a novel role for ATP in RBC mechanical properties has been uncovered [42][43][44][45][46][47]. We here have showed using fluorescence microscopy and flow cytometry using a novel ATP probe [19] that GPA ligation by antibody induces a robust increase in ATP released by RBCs. This increase was significantly reduced in the presence of additional antioxidants such as the cell-permeable GSH-ME. While we observed a significant ATP release in RBCs treated with PAC-1, further experiments did not support our hypothesis that this release was causally linked to GPA ligation (Fig 2G and 2H). It is therefore likely that increased ROS levels induce multiple responses from RBCs including caspase-3 activation and ATP release.
We previously reported an inability to identify pannexin-1 in human RBCs using flow cytometry or immuno-blotting methods [19]. However, MRP1 and MRP4 were previously involved in ATP release, and both were identified in human RBCs by mass spectrometry [48]. Using the specific inhibitor MK-571, we showed RBC MRP1 is involved in CR1-mediated ATP release [19]. Moreover, previous reports show mice with CF having increased ATP concentrations in blood [31], while others have reported the importance of purinergic receptors as regulators of ATP release in osteoclasts [49]. In spite of promising data, we could not confirm the role of any of these channels in mediating ATP release following GPA ligation.
Our data obtained using RBCs isolated from septic patients paradoxically showed an opposite trend from our in vitro results. Even though there is an increased activity of plasma ectonucleases in sepsis that directly lowers plasma ATP concentrations due to bacterial presence and generalized inflammation [50,51], our experiments using isolated and purified RBCs suggest that another mechanism is responsible for diminished RBC ATP release in sepsis. Increased concentrations of extracellular ADP reversibly inhibit RBC ATP release by acting on P2Y13 receptors on RBC membranes through a c-AMP dependent mechanism [52]. During septic episodes, the blood concentration of ADP increases 6-8 times over control (from 1.2 to 8.4 μM/L), decreasing to near normal values 72 hours post-sepsis [53]. Patients in our study were enrolled within a 24-hour window post admission, providing a possible explanation for the observed, ex vivo low ATP release by complement-altered RBCs isolated from septic patients.
Our results regarding C3d levels on circulating RBCs from sepsis patients were indistinguishable from control samples. In humans, C3 fragments present on RBC membranes have been shown to decrease RBC membrane deformability [7] and significantly shorten the halflife of circulating RBCs compared to RBC-bound C4 fragments [54]. It is possible that, at the onset of the disease, complement is not fully activated, whereas after time, the less deformable RBCs are trapped in the spleen and liver or removed from circulation. This mechanism could provide an explanation as to why C4 fragments are present on RBC membranes from sepsis patients, while the C3b-RBCs are missing.
Septic patients display a wide array of microcirculatory flow disturbances [55][56][57][58]. Consistent with our in vitro data, our ex vivo functional results indicate that RBCs from septic patients show a significant decrease in RBC membrane deformability and lipid mobility. These results indicate that the marked changes in membrane deformability due to GPA ligation either by complement fragments C3b, C4b, or antibodies are the result of GPA-specific signaling cascades rather than passive, antibody-driven cross-linking of membrane proteins.
In summary, our findings regarding GPA engagement by complement fragments or autoantibodies reveal a novel and detrimental role of circulating RBCs during inflammatory conditions. Overwhelming intracellular ROS concentrations ultimately impair critical RBC mechanical capabilities, thus disrupting their movement and interactions within the microvasculature. The data point to complement-altered RBCs as possible effector cells, actively influencing nearby cells through release of bioactive molecules such as ROS and ATP. (B) CFTR Inhibitor raises baseline ATP levels in RBC. RBCs were pre-incubated with DMSO (control) or two concentrations of GlyH-101. ATP concentration was determined as described in Experimental Procedures. (C) Inhibition of CFTR does not reduce GPA-mediated ATP release. RBCs were pre-incubated with 0.1% DMSO (control) or 1 μM GlyH-101. RBCs were then washed twice with buffer containing1 μM GlyH-101 and ligated with anti-GPA mAb (R10) or IgG control mAb (5 μg/mL). Extracellular ATP concentration was determined as described in Experimental Procedures. (D) Indomethacin does not inhibit GPA-mediated ATP release. RBCs were pre-incubated with 0.1% DMSO (control) or 10 μM indomethacin. RBCs were then ligated with either control IgG or anti-GPA mAb. ATP concentration was determined as described in Experimental Procedures. (E) MRP channels are not responsible for GPA mediated ATP release. RBCs were pre-incubated with buffer or 10 μM MK-571, a known specific of MRPs. RBCs were then processed as outlined in methods with the antibodies given during the last resuspension step. ATP concentration was determined as described in Experimental Procedures. (F) Inhibition of P2 receptors does not reduce GPA-mediated ATP release. RBCs were pre-incubated (RT x 10 min) with buffer or either 10 μM PPADS (left graph), iso-PPADS (middle graph) or suramin (right graph). RBCs were then processed as outlined in methods with the antibodies given during the last resuspension step. ATP concentration was determined as described in Experimental Procedures. (JPG)