Platelet-independent adhesion of calcium-loaded erythrocytes to von Willebrand factor

Adhesion of erythrocytes to endothelial cells lining the vascular wall can cause vaso-occlusive events that impair blood flow which in turn may result in ischemia and tissue damage. Adhesion of erythrocytes to vascular endothelial cells has been described in multiple hemolytic disorders, especially in sickle cell disease, but the adhesion of normal erythrocytes to endothelial cells has hardly been described. It was shown that calcium-loaded erythrocytes can adhere to endothelial cells. Because sickle erythrocyte adhesion to ECs can be enhanced by ultra-large von Willebrand factor multimers, we investigated whether calcium loading of erythrocytes could promote binding to endothelial cells via ultra-large von Willebrand factor multimers. We used (immunofluorescent) live-cell imaging of washed erythrocytes perfused over primary endothelial cells at venular flow rate. Using this approach, we show that calcium-loaded erythrocytes strongly adhere to histamine-stimulated primary human endothelial cells. This adhesion is mediated by ultra-large von Willebrand factor multimers. Von Willebrand factor knockdown or ADAMTS13 cleavage abolished the binding of erythrocytes to activated endothelial cells under flow. Platelet depletion did not interfere with erythrocyte binding to von Willebrand factor. Our results reveal platelet-independent adhesion of calcium-loaded erythrocytes to endothelium-derived von Willebrand factor. Erythrocyte adhesion to von Willebrand factor may be particularly relevant for venous thrombosis, which is characterized by the formation of erythrocyte-rich thrombi.


Effects of ADAMTS13 on erythrocyte binding to ECs under flow
Ionomycin-treated erythrocytes (10 8 cells/ml) were perfused with 100 μM histamine over ECs for 10 minutes. Without flow interruption, perfusion was continued with or without 5% autologous platelet-poor-plasma or 5 μg/ml recombinant ADAMTS13 (purified as described before [16]) for 10 minutes. At each time point images were acquired using an Axiovert 200M microscope with a 20x air objective.

Erythrocyte sorting
Erythrocyte ultra-purification was performed by fluorescence-activated cell sorting (FACS) using a FACSAria III cell sorter (BD). Erythrocyte concentrates (10 7 cells/ml) were stained using anti-CD235a-FITC (Sanquin) and anti-CD42b-APC (Life Technologies) for 30 minutes. Erythrocytes were purified by positive sorting for CD235a and negative sorting for CD42b with doublet exclusion by using FSC-H and FSC-W.

Western blot
Cell lysates were run on a 12.5% polyacrylamide gel and transferred onto Amersham™ Protan™ 0.2 μm nitrocellulose blotting membranes (GE Healthcare). Membranes were stained using anti-ICAM1 (Santa Cruz Biotechnology), anti-actin (Sigma Aldrich), anti-VWF (Dako) overnight at 4˚C and counterstained with anti-mouse HRP (Dako) or anti-rabbit HRP (Dako) for 1 hour. Immobilized antigens were detected by chemiluminescence using Pierce 1 ECL Western Blotting Substrate (Thermo Scientific) and exposed on Fuji Medical X-Ray films (Fujifilm).

VWF ELISA
ECs were stimulated with HEPES(+) buffer containing 100 μM histamine for 10 minutes or lysed in HEPES(+) buffer with Triton X-100 (1.0% v/v). Intracellular or secreted VWF was assayed by sandwich enzyme-linked immunosorbent assay (ELISA) as described previously [18] using rabbit polyclonal anti-VWF (DAKO) as coating antibodies and HRP-conjugated rabbit polyclonal anti-VWF (DAKO) for detection of bound VWF. Standard curves were made from dilutions of culture supernatants of HEK293 cells stably producing recombinant VWF. Optical density levels measured below 2 times the blank were set to 0 pM.

Imaging and analysis
Images were taken using Zen 2012 software version 1.1.2.0 (Carl Zeiss). For all image processing ImageJ version 1.49g (https://imagej.nih.gov/ij/) was used. Weibel-Palade bodies were quantified using the functions "threshold" and "analyze particles" within regions of interest based on the VE-cadherin staining. Erythrocyte adhesion was quantified using the ImageJ plugin Cell Counter. Erythrocytes were qualified as firmly adherent when they remained immobile for 5 seconds before and 5 seconds after each time point otherwise they were qualified as loosely adherent. For each experiment, tile scans (3800 μm width) were made (S1 Fig).

Statistical analysis
Data shown are mean ±SD. Statistical comparisons were made using an unpaired, 2-tailed Student's t test, 1-way ANOVA followed by Bonferroni's post-hoc test, or 2-way ANOVA followed by Bonferroni's post-hoc test, where appropriate. Statistical analysis was performed using GraphPad Prism version 6.04 (GraphPad Software). For all tests, results were considered statistically significant at P < 0.05. Graphs showing individual data points are provided in S8 and S9 Figs.

Erythrocyte binding to ECs
To analyze the adhesion of normal erythrocytes to primary human ECs, we used in vitro flow chambers in which washed erythrocytes were perfused over cultured EC monolayers at a wall shear rate of 100 s -1 and wall shear stress of 0,72 dyne/cm 2 corresponding to the conditions in post-capillary venules [19] or veins [20]. First we confirmed that the ECs formed a confluent and stable monolayer within the perfusion chambers by staining for VE-cadherin. (Fig 1A). We also confirmed that the ECs kept their endothelial phenotype after culturing in the perfusion chambers by staining for VWF and VE-cadherin (S2 In contrast to SS erythrocytes, healthy erythrocytes do not bind to ECs [7], which we confirmed by perfusing freshly isolated normal erythrocytes over EC monolayers. An example of this experiment can be seen in S1 Video. Previously, the release of ULVWF multimers from ECs was shown to cause adhesion of normal erythrocytes to the venular endothelium in an ex vivo rat mesocecum model [12]. Thrombin and histamine may induce increased release of endothelial ULVWF multimers during episodes of vascular injury, inflammation, and infection [21,22]. Yet, in our flow-model, washed erythrocytes that contained virtually no platelets and white blood cells as shown in Table 1, showed little adhesion to ECs activated with 100 μM histamine or 1 U/ml thrombin (Fig 1C and 1D).
However, since it has been described that calcium-loaded erythrocytes can adhere to ECs [7], we tested the effect of calcium loading on erythrocyte-EC interaction. Erythrocyte treatment with 1 μM ionomycin induced a rapid calcium influx followed by phosphatidylserine exposure and cell volume decrease (S2 Fig). Interestingly, calcium-loaded erythrocytes adhered to non-treated ECs (Fig 1C and 1D and S1 Video). Moreover, a double hit experimental set-up in which erythrocytes were loaded with calcium and perfused over histamine-or thrombin-activated ECs, induced strong and abundant adhesion of erythrocytes to ECs ( Fig  1C and 1D and S1 Video). In contrast, the pro-inflammatory cytokine TNF-α, which is also secreted during injury, inflammation and infection, did not promote the binding of normal or calcium-loaded erythrocytes to ECs, despite the upregulation of EC adhesion molecules such Histamine-Ionomcyin □). Adhesion of erythrocytes from 3 independent donors was quantified at a 1 minute time interval for 10 minutes. For C-E the average number of adherent erythrocytes within 15 fields of view (FOV; 1 FOV = 419.23x319.41 μm) was used. For F the absolute number of adherent erythrocytes within 1 field of view was used. Values are means ± SD (n = 3; Each n represents single independent experiments using cells from different donors). Statistical analysis was performed using a 2-way ANOVA in which all means were compared with the double-blank control followed by Bonferroni's post-hoc test. ** P < 0.01; **** P < 0.0001. as intercellular adhesion molecule 1 (Fig 1E and S2 Fig). To test the effect of a different ionophore on erythrocyte adhesion to ECs we also stimulated erythrocytes with the potassium ionophore valinomycin. Although valinomycin caused a significant potassium efflux from erythrocytes, only minor adhesion of treated erythrocytes to ECs could be observed (S3 Fig).
To investigate the dynamics of the erythrocyte-endothelium adhesion process, ionomycin-or control-treated erythrocytes from 3 independent donors were perfused with or without histamine over ECs. Control treated erythrocytes showed both in the presence and absence of histamine no adhesion to ECs ( Fig 1F). On the contrary, the adhesion of calcium-loaded erythrocytes in the absence of histamine increased during the first minutes after which an equilibrium between attaching and detaching cells established (Fig 1F). The double hit experimental set-up induced strong and abundant adhesion of erythrocytes to ECs within the first minutes followed by a gradual increase in absolute numbers of adherent cells (Fig 1F).
We also looked at the ratio between firmly adherent erythrocytes and loosely adherent erythrocytes (S4 Fig). Although some erythrocytes detach after adhering to ECs (S1 Video), quantification showed that both in the ionomycin-treated condition as in the double hit experimental set-up the absolute number of loosely adherent erythrocytes is small (S4 Fig). The maximal amount of loosely adherent erythrocytes was observed after 10 minutes perfusion of ionomycin-treated erythrocytes over histamine-treated ECs. Here, the amount of loosely adherent erythrocytes peaked at 8% (S4 Fig).
These data show that adhesion of erythrocytes to ECs can be induced by calcium accumulation within the erythrocytes and is greatly enhanced upon the activation of ECs by either histamine or thrombin.

Intracellular calcium accumulation potentiates erythrocyte adhesion to ECs
Before we tested whether ULVWF multimers were involved in the adhesion of calcium-loaded erythrocytes to ECs, we determined the role of calcium in this process. While intracellular calcium is important for cell signaling [23], extracellular calcium is important for maintaining a functional conformation of adhesion molecules [24]. Preventing the ionomycin-induced intracellular increase of calcium (Fig 2Ai) by 5 mM EGTA pre-incubation (Fig 2Aii) caused an almost complete (~92%) reduction of erythrocyte adhesion to histamine-activated ECs ( Fig  2B). When extracellular calcium was removed by adding 5 mM EGTA to the flow buffer after the ionomycin-induced calcium influx (Fig 2Aiii), only a 32% reduction in erythrocyte-EC adhesion was observed (Fig 2B). Significantly less erythrocytes adhered to ECs when intracellular calcium was kept low compared with a reduction in extracellular calcium (Fig 2B). When erythrocyte calcium influx was induced in the presence of varying calcium concentrations, increasing erythrocyte-EC adhesion could be observed at a calcium concentration of 25 μM and higher (Fig 2C). A significant increase of erythrocyte-EC adhesion was observed at intracellular calcium levels of 50 μM and higher. (Fig 2C). These results suggest that the adhesion of erythrocytes to activated ECs depends on intracellular calcium accumulation whereas extracellular calcium is not required for erythrocyte-EC binding.

Von Willebrand factor mediates erythrocyte-EC binding
Next we determined whether ULVWF multimers from the ECs were involved in the adhesion of calcium-loaded erythrocytes to ECs. Interestingly, calcium-loaded erythrocytes bound at low levels in a random fashion to control-treated ECs (Fig 3A and 3B left panels), while calcium-loaded erythrocytes perfused over histamine-or thrombin-activated ECs adhered in high numbers and clusters of adherent cells aligned in the direction of the flow (Fig 3A and 3B right panels). The alignment of erythrocytes adherent to ECs resembled platelet-binding to VWF (Fig 3C and 3D). We therefore investigated the role of VWF in erythrocyte-EC interaction by perfusing monoclonal anti-VWF coupled beads (CLB-RAg20-beads) together with calcium-loaded erythrocytes over histamine-activated ECs to visualize the VWF strings. The CLB-RAg20-beads bound, together with the erythrocytes, to the ECs and aligned along the same track as the erythrocytes in the direction of the flow (Fig 3E and S5 Fig).
We also perfused PKH67-labeled calcium-loaded erythrocytes in combination with Alexa Fluor 1 568-conjugated polyclonal anti-VWF antibodies over histamine-activated ECs which confirmed that erythrocytes adhered to EC-derived VWF strings (Fig 4A; S2 Video).
The plasma metalloprotease ADAMTS13 (a disintegrin and metalloproteinase with a thrombospondin type 1 motif, member 13) can cleave VWF strings [25]. The binding of calcium-loaded erythrocytes to histamine-activated ECs did not change after subsequent perfusion of control HEPES buffer (Fig 4Bi), but was significantly reduced after perfusion of 5% Values are means ± SD (n = 4; Each n represents single independent experiments using cells from different donors). Statistical analysis was performed using a 1-way ANOVA (B) or a 1-way ANOVA in which all means were compared with 0 μM calcium (C) each followed by Bonferroni's post-hoc test. * P < 0.05; ** P < 0.01; *** P < 0.001; **** P < 0.0001. autologous plasma or 5 μg/ml recombinant ADAMTS13 over the adherent erythrocytes (Fig  4Bii and 4Biii respectively). This indicates that VWF cleavage prevents erythrocyte adhesion to activated ECs.
Finally, lentivirally transduced short hairpin RNA against VWF significantly reduced the endothelial VWF levels (Fig 4C), caused a loss of VWF-positive endothelial Weibel-Palade bodies (S6 Fig) and reduced VWF string formation upon histamine activation of ECs (S6 Fig). Importantly, VWF knockdown significantly reduced the adhesion of calcium-loaded erythrocytes to histamine-activated ECs (Fig 4C). Together, these results suggest that the adhesion of calcium-loaded erythrocytes to histamine-activated ECs depends on ULVWF strings.

VWF levels determine binding of calcium-loaded erythrocytes to ECs
In addition to HUVECs, we investigated erythrocyte adhesion to blood outgrowth ECs (BOECs), which are more closely related to microvascular endothelial cells [26], human aortic ECs (HAECs), and a human microvascular EC line (HMEC-1). As we observed with HUVECs (Fig 5A), the double hit experimental design induced the most abundant erythrocyte-EC adhesion to BOECs (Fig 5B), HAECs (Fig 5C), and HMEC-1 (Fig 5D), albeit that the absolute number of adherent erythrocytes differed between EC types with lowest binding to the HMEC-1 cell line. Next, we questioned whether differences between intracellular VWF levels of the various EC types could explain the different amounts of adherent erythrocytes. VWF stainings of the different EC types showed a reduction of Weibel-Palade bodies in the HAECs compared to the HUVECs and the BOECs and an almost complete absence in the HMEC-1 cell line (Fig 5E  and S7 Fig). This suggests reduced VWF levels in the HAECs and the HMEC-1 cell line. This was confirmed when we measured both intracellular VWF levels and histamine-induced, secreted VWF levels (Fig 5F and 5G). By plotting the absolute number of ionomycin treated erythrocytes adherent to histamine treated ECs against secreted VWF levels a positive correlation between VWF levels and erythrocyte adhesion could be observed ( Fig 5H). These data show that erythrocyte-EC adhesion is not specific to HUVECs and can be observed using ECs from different vascular beds, however the extent of erythrocyte adhesion depends on the VWF levels of the ECs.

Erythrocytes bind independent of platelets to VWF
Platelets are the main cells that bind VWF [27]. We questioned whether erythrocytes bind to VWF strings in a fashion independent of platelet binding. Calcium-loaded erythrocytes, mixed with platelets, and perfused over histamine-activated ECs, bound side-by-side to the same VWF strings (Fig 6A and 6B). Furthermore, in the presence of increasing numbers of platelets, there was no significant increase in the number of adherent erythrocytes (Fig 6B).
The platelet integrin α IIB β 3 has been described to bind to the erythrocyte membrane molecule ICAM-4 [28] thereby facilitating erythrocyte binding to VWF [29]. However, we observed platelet-independent binding of erythrocytes to VWF from histamine-activated ECs using live immunofluorescent imaging of anti-CD235a-FITC-labeled calcium-loaded erythrocytes mixed with CD42b-APC-labeled platelets and Alexa Fluor 1 568-labeled anti-VWF antibodies ( Fig 6B). Furthermore, erythrocytes purified by FACS, based on the expression of CD235a and lack of CD42b, bound to EC-derived VWF strings (Fig 6C). Together, these results suggest that calcium-loaded erythrocytes can bind, independent of platelets, to VWF strings and that platelets and erythrocytes do not show significant competition for binding sites on ULVWF strings.

Discussion
Under physiological flow, erythrocytes do not adhere to the vessel wall. In contrast to erythrocytes, platelets do adhere to the vessel wall under flow, provided the endothelium is activated. Such activation, for example by histamine [22] or thrombin [21], induces the release of VWF from Weibel-Palade bodies. Platelets will adhere to unfolded VWF under flow and thereby participate in primary hemostasis (Fig 7A) [30]. The current study shows that calcium-loaded erythrocytes can also adhere, in a platelet-independent fashion, to VWF strings released by activated ECs (Fig 7B).
Accumulation of calcium within erythrocytes is associated with a number of hereditary anemias including sickle cell disease [23,31], however calcium influxes into normal erythrocytes   (1 hour) erythrocytes, mixed with platelets and perfused (10 minutes) with histamine (100 μM) over HUVECs. An example of an adherent platelet is marked by (*) and an adherent erythrocyte is marked by an arrowhead. (B) Ionomycin-treated erythrocytes labeled with anti-CD235a-FITC (green), mixed with anti-CD42b-APC-labeled platelets (magenta) and Alexa Fluor ® 568-labelled anti-VWF antibody (red) were perfused with histamine over HUVECs. Adherent erythrocytes were have also been described. For instance, calcium influxes were observed after erythrocyte cold storage [32]. Furthermore, the expression of a NMDA receptor channel on erythrocyte membranes has been reported for rat [33] and later in human erythrocytes [34]. It was shown that NMDA receptor agonists including glutamate, N-methyl D-aspartate (NMDA), homocysteic acid, and glycine could raise erythrocyte intracellular calcium levels [34]. Furthermore, lysophosphatidic acid (LPA), a second messenger generated by activated platelets that assists in clot formation, was shown to cause a calcium influx into erythrocytes [35,36]. Whether these stimuli would cause a sufficient raise in intracellular calcium to induce the binding of erythrocytes to VWF remains to be established. quantified as in Fig 1. Values are means ± SD (n = 3; Each n represents single independent experiments using cells from different donors). Statistical analysis was performed using a 1-way ANOVA in which all means were compared with 0 platelets ml -1 followed by Bonferroni's post-hoc test. (C) Erythrocytes were purified by FACS. The erythrocyte population based on cell size (i) was positive sorted for CD235a-FITC and negative sorted for CD42b-APC (ii). Purified erythrocytes (10 7 cells/ml) treated (1 hour) with ionomycin (1 μM) (green), mixed with Alexa Fluor ® 568-labelled anti-VWF antibody (red) were perfused with histamine (100 μM) over HUVECs. (iii) The dashed box corresponds to the zoomed region. Scale bars represent 50 μm.
doi:10.1371/journal.pone.0173077.g006  [22] or thrombin [21] leads to the release of VWF from Weibel-Palade bodies. Platelets adhere to VWF under flow and thereby contribute to primary hemostasis [27]. Normal erythrocytes do not adhere to VWF and remain in circulation. (B) Double hit model for adhesion of circulating cells to ECs. ECs, activated by histamine or thrombin, release VWF from Weibel-Palade bodies. Platelets adhere to the VWF strings. In addition, calcium-loaded erythrocytes can adhere independently of platelets to VWF strings. Micro-thrombi-like structures are formed. Besides the erythrocyte binding to ULVWF strings which we describe here, erythrocyte adhesion to endothelial cells can also be mediated by many other mechanisms. The endothelial proteins αVβ3 [37], and P-selectin [38] and the erythrocyte proteins BCAM/Lu [2], CD44 [37], and CD36 [39,40] were also shown to facilitate an EC-erythrocyte interaction. However, if any of these receptors were involved in the adhesion of calcium-loaded erythrocytes to ECs, this interaction should also occur in the absence of ULVWF multimers. Although our results do not rule out a possible partial contribution of adhesion mechanisms, other than through VWF, the ADAMTS13 cleavage-and the VWF-knockdown experiments showed that the majority of calcium-loaded erythrocytes adhere to endothelial cells via ULVWF strings.
The reduced adhesion of calcium-loaded erythrocytes to ECs from microvascular beds (HMEC-1) compared to ECs from macrovascular beds (HUVEC, and HAEC) seems contradictory to previous studies. Microcirculatory studies using ex vivo mesocecum vasculature of the rat showed a strong inverse correlation between adhesion of SS erythrocytes and the venular diameter [9,19]. However, in our experiments we applied a similar wall shear stress on the different EC types, while in the ex vivo model, the wall shear stress changes along the vasculature which thereby could interfere with erythrocyte adhesion. Furthermore, we showed a positive correlation between the amount of ULVWF produced by the different types of endothelial cells with the amount of adherent, calcium-loaded erythrocytes. Although the microvascular endothelial cell line HMEC-1 produced barely any VWF, the VWF levels of the primary HUVEC, BOEC, and HAECs were in line with previous results [41]. These differences could explain the apparent contradiction.
Platelets are the main cells to bind to VWF. Erythrocyte ICAM-4 can bind the platelet integrin α IIB β 3 [28]. Thus, platelets could potentially facilitate the erythrocyte-VWF interaction [29]. In this study we show that an ultra-pure erythrocyte population can in vitro adhere to VWF in a platelet-independent fashion. This does not exclude that platelet-erythrocyte complexes could enhance erythrocyte-VWF binding, but our data suggests that the contribution of platelets to erythrocyte binding is limited.
The function of erythrocyte binding to VWF is not known. However, increasing evidence shows that erythrocytes play an important role in hemostasis [29]. Their rheological properties cause platelets to be concentrated in the periphery of the blood vessels, close to the vessel wall [42]. In addition, the negative surface of PS-exposing erythrocytes facilitates thrombin generation [43]. Erythrocytes also play an important role in thrombosis. In contrast to white arterial thrombi that consist mainly of platelets, erythrocytes are the major cellular component of red venous thrombi [44,45]. Our in vitro results show that VWFassociated erythrocytes will readily form thrombus-like structures. In vivo, VWF -/mice were shown to be protected from thrombosis in a deep venous thrombosis model by preventing the adhesion of platelets and leukocytes to the vascular endothelium [46]. A similar erythrocyte-VWF interaction may therefore contribute to the formation or propagation of venous thrombi.
In summary, this study shows that erythrocyte adhesion to ECs is rapidly induced when (i) erythrocytes undergo a calcium-influx and (ii) VWF is released from ECs. Our work shows that the adhesion of erythrocytes to ECs is mediated by newly released ULVWF strings and is independent of platelets. The calcium-loaded erythrocytes interact with endothelial-bound ULVWF strings through unidentified receptors. Future work is needed to define the molecular characteristics of erythrocyte receptors that can bind VWF as well as the VWF domain that is responsible for this interaction. Finally, investigating whether the erythrocyte-VWF interactions could play a role in hemostasis and (venous) thrombosis and exploring ways to modulate these interactions will likely prove to be of clinical relevance. Values are means ± SD (n = 3; Each n represents single independent experiments using cells from different donors). Statistical analysis was performed using a 2-way ANOVA in which all means were compared with the double-blanc control followed by Bonferroni's post-hoc test. (B) Extracellular potassium levels of valinomycin-or control-treated erythrocytes measured by a blood gas analyzer. Values are means ± SD (n = 3; Each n represents single independent experiments using cells from different donors). Statistical analysis was performed using a 1-way ANOVA in which all means were compared with the DMSO control. ÃÃ P < 0.01; ÃÃÃ P < 0.001.