Polyphosphate/platelet factor 4 complexes can mediate heparin-independent platelet activation in heparin-induced thrombocytopenia

Heparin-induced thrombocytopenia (HIT) is a common drug-induced autoimmune disorder characterized by arterial and venous thromboembolism.¹  The thromboembolic complications (TECs) have been attributed in part to activation of platelets by immune complexes composed of platelet factor 4 (PF4), heparin, and immunoglobulin G (IgG) antibodies.^2,3  Thrombus formation is enhanced by monocytes and endothelial cells activated by HIT immune complexes, which are induced to express tissue factor and to generate thrombin, reinforcing immune-mediated platelet activation and procoagulant pathways.^4-6 

Recurrent TECs can occur even in the presence of thrombin and factor Xa inhibitors given in doses that predispose to bleeding,^7,8  and the risk of recurrent thrombosis can extend for weeks after heparin therapy has been stopped.⁹  Heparin¹⁰  and PF4¹¹  are cleared from the circulation and catabolized^12,13  within hours and heparin/PF4 complexes are endocytosed by monocytes and delivered to lysosomes in a similar time frame.¹⁴  Circulating antigenic complexes would be expected to fall soon after the cessation of platelet activation. Further, inhibition of thrombin occurs within hours after institution of antithrombotic therapy. Anti–heparin-PF4 antibody can persist for many months,¹⁵  but these antibodies rarely cause TECs in the absence of PF4 or heparin.^16,17  Therefore, the basis for the severe and protracted prothrombotic state is incompletely defined. Together, these findings suggest that HIT immune complexes and thrombin might initiate additional and as-yet unrecognized host responses that exacerbate and perpetuate the risk of TECs.

One clue toward identifying the pathways that might predispose to delayed thrombotic complications in patients with HIT is the finding that PF4 forms antigenic complexes with a variety of polyanions, including glycosaminoglycans, sulfated anticoagulants, lipid A from gram-negative bacteria, RNA, and inorganic polyphosphates (polyPs).^2,4,18-22  PolyPs are highly anionic linear polymers of orthophosphate linked by phosphoanhydride bonds.^23-25  Although found in all mammalian cells, polyPs are present in the dense granules of human platelets at millimolar concentrations and are released following activation.^26,27  In platelets, polyP polymers range in length from ∼60 to 120 orthophosphate units,^26-30  and concentrations may exceed 1 to 3 µM in platelet-rich thrombi.²⁷  PolyPs provide an anionic surface to assemble factor XII, prekallikrein, and high-molecular-weight kininogen and trigger contact activation of coagulation.³¹  PolyPs are prothrombotic and pro-inflammatory in in vivo mouse models^27,32  by affecting multiple steps in the coagulation cascade.^27,33-39  PolyPs also dampen activation of the complement system by interfering with assembly of the terminal membrane attack complex⁴⁰  and by binding to and potentiating the activity of C1-esterase inhibitor, which helps to control initiation of the classical pathway⁴¹ ; thus, changing the function of polyPs could modulate their participation in hemostatic and inflammatory pathways.

The biological effects of PF4 on the prothrombotic and pro-inflammatory effects of polyPs have not been fully investigated to our knowledge. Here we characterize the biophysical, antigenic, and platelet-activating properties of PF4/polyP complexes and ask whether activated platelets can generate endogenous polyP-containing antigenic complexes capable of exacerbating and perhaps perpetuating HIT in the absence of exogenous PF4 and heparin.

The following were purchased from commercial sources: thrombin receptor agonist peptide (TRAP), prostaglandin E1 (PGE1), and chondroitinase ABC from Proteus vulgaris (Sigma); calf intestinal alkaline phosphatase (CIP; New England Biolabs, Ipswich, MA); unfractionated heparin (UFH; Becton-Dickenson, Franklin Lakes, NJ); Hanks balanced salt solution (HBSS) and phosphate-buffered saline (Invitrogen/Life Technologies and GIBCO-Life Sciences, Grand Island, MI); gelatin veronal buffer, normal human serum (NHS) and human complement factors (Complement Technology, Inc., Tyler, TX); 96-well microplates (Corning, Amsterdam, Netherlands); 2,2′-Azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) (Roche Diagnostics, Indianapolis, IN); 0.22-µM filters (Millipore); plastic cuvettes (Fisher); horseradish peroxidase (HRP)-conjugated goat anti-human IgG Fc and HRP-conjugated goat anti-mouse IgG Fc antibodies (Jackson ImmunoResearch; West Grove, PA); HRP-conjugated goat anti-human C3 antibody (MP Biomedicals; Santa Ana, CA) that recognizes native, hydrolyzed C3b, iC3b, C3c, and C3d; and chicken erythrocytes (Colorado Serum Company, Denver, CO). CIP was buffer exchanged using Micro Bio-Spin Columns with Bio-Gel P-6 (Bio-Rad). Human PF4 was expressed in Schneider 2 cells (Invitrogen, Carlsbad, CA)⁴²  and was purified and characterized as described.^2,43  KKO, an IgG2bκ monoclonal heparin/PF4 binding antibody, and RTO, an isotype-matched monoclonal antibody that binds PF4 in the absence of heparin, were purified from hybridoma cell media.⁴⁴  TRA purified from hybridoma media PFHM-II (Gibco) and MOPC 141 (Sigma) is an isotype-matched murine monoclonal IgG control.

Heterogeneous, size-fractionated polyP preparations containing an estimated average of 14, 60, or 130 phosphate monomers (NaPO₃) (designated P₁₄, P₆₀, and P₁₃₀) were the kind gift of Toshikazu Shiba (Regenetiss Inc., Tokyo). All other polyP polymers were prepared by differential isopropanol precipitation of long chain polyP as previously described.³⁶  Their integrity was confirmed by electrophoresis on a 10% tris(hydroxymethyl)aminomethane-Borate-EDTA-urea gel, stained with 0.05% (wt/vol) toluidine blue for 20 minutes, followed by destaining overnight in water (not shown). Quantification was determined using the malachite green assay.³⁶  Final concentrations of polyPs are expressed in terms of monophosphate units. In some experiments, the effect of PF4 on the susceptibility of phosphates to digestion was studied (see the supplemental Methods for details).

A plasmid containing the yeast exopolyphosphatase (PPX1) gene was a gift from Adolfo Saiardi, University College, London, United Kingdom. Biotinylated polyP binding domain (PPXbd) and 6xHis-tagged PPX1 were prepared as previously reported⁴¹  (see the supplemental Methods for details).

Chicken red blood cells (cRBCs) (3.3 × 10⁸ cells/mL) were diluted in gelatin veranol buffer containing 10 mM EDTA to prevent upstream complement activation and generation of endogenous C5b,6. Hemolysis was initiated by adding purified C5b,6 and varying concentrations of PF4 and/or polyP for 30 minutes in the presence of 2% NHS as the source of C7, C8, and C9. Cells were pelleted at 600g and lysis, reflected by hemoglobin in the supernatant, was quantified by measuring absorbance at 405 nm. Percent lysis is expressed relative to 100% lysis with H₂O. The concentration of C5b,6 was determined in pilot studies to yield ∼75% lysis at 30 minutes.

Platelets were isolated from citrated platelet-rich plasma (PRP) by centrifugation at 800g followed by gradual deceleration. The platelet pellet was washed twice with 140 mM NaCl, 5 mM KCl, 12 mM trisodium citrate, 10 mM glucose, 12.5 mM sucrose, pH 6.0, and 1 µM PGE1 (Sigma, Oakville, Canada), resuspended in 10 mM N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid (HEPES), pH 7.4, and apyrase (Sigma) and allowed to rest for 30 minutes at 37°C.

Resting platelets were activated with the phorbol ester 12-myristate acetate (PMA; 100 nM) or with the calcium ionophore A23187 (1 μM) and plated onto glass coverslips, fixed immediately with 2% (vol/vol) paraformaldehyde, and permeabilized with 0.1% (vol/vol) Triton-X-100; unreactive sites were blocked with 1% (wt/vol) bovine serum albumin (BSA). Anti-PF4 antibody (Abcam, Cambridge, MA) was added for 1 hour at room temperature (RT) followed by fluorescein isothiocyanate (FITC)-conjugated secondary antibody for 1 hour. PolyPs were labeled with 40 µg/mL biotinylated PPXbd and 10 µg/mL tetramethylrhodamine isothiocyanate–conjugated streptavidin (DyLight, Life Technologies, Grand Island, NY). Confocal imaging was accomplished with a Nikon C2⁺ confocal microscope; images were processed with NIS-Elements software and captured and processed using a Zeiss spinning disk confocal microscope and SlideBook software (Intelligent Imaging Innovations, Denver, CO). Confocal images supplied as grayscale were colorized in green or red using Adobe Photoshop.

Citrated whole blood diluted 1/50 in calcium-containing HEPES, pH 7.5, was incubated with PF4 (10 µg/mL) followed by addition of polyP₁₃₀ (0-30 µM) and KKO (50 µg/mL) for 30 minutes at RT in the presence of allophycocyanin-labeled anti-CD41 and phycoerythrin-labeled anti–P-selectin antibodies (both BD Bioscience). Samples were diluted fivefold with calcium/HEPES buffer containing 5 µL FITC-Annexin V and assessed by flow cytometry (BD Fortessa) within 30 to 60 minutes. Platelets were identified based on their characteristic forward and side scatter profiles and CD41 fluorescence. Expression of P-selectin and binding of annexin V was expressed as the increase in mean fluorescence intensity over control.

Platelet aggregation was measured using light transmission aggregometry.⁴⁵  PRP was prepared by sequential differential centrifugation from citrated whole blood. Platelets were sedimented as described previously, washed once in HBSS supplemented with 0.1% BSA, pH 7.3, in the presence of 5 µM PGE1, and resuspended in HBSS supplemented with 0.1% BSA and fibrinogen (0.1 mg/mL) to a concentration of 250 × 10⁹/L.

Five sets of experiments were performed in sequence. First, PRP was incubated with PF4 (10 µg/mL) or PF4 plus UFH (0.1 U/mL), each alone or followed by the addition of KKO or RTO (100 µg). Second, PRP was incubated with PF4 ± predetermined optimal concentrations of polyP_14, polyP₆₀, or polyP₁₃₀ alone or followed by addition of KKO. Third, washed platelets were preincubated with chondroitinase ABC (2.5 U/mL) for 30 minutes at 37°C, shown previously to inhibit binding of KKO by 90% to 95%.²  Fourth, platelets were stimulated with various concentrations of TRAP for 250 seconds to identify the minimum concentration that initiated a first wave but not an irreversible second wave of aggregation (∼1 µM). Platelets were then stimulated with this subthreshold concentration of TRAP in the presence of 100 µg KKO and platelet aggregation was followed over the ensuing 250 seconds. Fifth, chondroitinase ABC, CIP (20 or 200 U/mL), PPX1 (2-10 µg/mL), or PPXbd (10-125 µg/mL) was added immediately before adding TRAP and KKO.

We asked whether PF4 and polyPs form complexes capable of binding HIT antibodies as determined by enzyme-linked immunosorbent assay (ELISA). When PF4 was held constant, addition of polyP₁₃₀ caused a concentration-dependent increase in the binding of KKO (Figure 1A), whereas binding of RTO was unaffected and the control monoclonal antibody MOPC1 did not bind under any conditions (data not shown). Binding of KKO to complexes of PF4 (3 µg/mL) and polyP₁₃₀ (3 µM) met or exceeded binding to PF4 and a predetermined optimal concentration of heparin (0.1 U/mL) (Figure 1A). Higher concentrations of polyP₁₃₀ impaired KKO binding to PF4, as is seen when concentrations of heparin exceed the optimal PF4/heparin ratio (Figure 1A).²  PF4 formed antigenic complexes with polyPs of diverse chain lengths (polyP₁₄, ₆₀, ₁₀₀, ₁₃₀, ₁₅₅, and ₆₇₅). Optimal binding was seen at the same molar ratio of PF4 to all polyPs with chain lengths ≥P60 (Figure 1B; other data not shown), whereas slightly higher concentrations of polyP₁₄ were required to optimize binding (Figure 1B). At optimal concentrations, UFH (0.1 U/mL) and polyP₁₃₀ (3 µM) increased binding of KKO to PF4 13.7 ± 2.7-fold and 30.9 ± 5.0-fold (mean ± SEM, n = 18), respectively. Addition of CIP to polyP for 30 minutes prevented formation of antigenic complexes but did not disrupt binding of KKO to preformed complexes of PF4 and polyP (Figure 1C). In contrast, when CIP was added before or after polyP was allowed to form complexes with PF4 in buffer containing dithiothreitol (to disrupt intramolecular disulfide bonds in PF4), binding of KKO was almost completely prevented (Figure 1C). These results show that antigen formation requires contact between polyP and PF4 and suggests PF4 partially protects polyP from digestion by phosphatases.

PF4 (50 µg) and UFH form complexes that are sufficiently large (exceeding 5 nm) to be analyzed using dynamic light scattering (DLS). Therefore, we next used DLS analysis to compare the biophysical properties of complexes between PF4 (50 µg) and polyP₁₃₀ with those formed with UFH. The DLS profiles of PF4/UFH complexes were nearly identical to those previously described⁴⁶  (Figure 2A-B). At low concentrations of UFH (0.1 U/mL), PF4 complexes were of moderate size (∼220 nm) but with a high polydispersity index (PDI), indicating size heterogeneity. As the concentration of UFH was increased to 1.0 U/mL, which approached theoretic neutrality, the complexes enlarged up to 4500 nm, but heterogeneity persisted. Higher concentrations of UFH (10 U/mL) generated small negatively charged, highly unstable complexes (not shown).

The analogous experiments with polyP₁₃₀ plus PF4 showed similar trends, but with some notable differences (Figure 2A-B). PF4 incubated with 2 µM polyP₁₃₀ formed intermediate-sized complexes (∼800 nm) with a high PDI. As the concentration of polyP was increased to 20 μM, complexes enlarged (∼2300 nm) but retained a low PDI in clear contrast to UFH. Increasing the concentration of polyP further to 200 µM generated small (∼150 nm) particles with a low PDI. Consistent with their biophysical homogeneity, these latter complexes were extremely stable over 72 hours (Figure 2B).

We focused on complexes between 5 µg/mL PF4 and 3 µM polyP₁₃₀, the ratio that optimized antigen formation assessed by ELISA. The biophysical properties of these complexes followed a pattern similar to those between PF4 and UFH. PF4/polyP₁₃₀ complexes formed within 5 minutes, attaining a mean size of 295 nm and a PDI of ∼0.227, whereas complexes with 0.1 U/mL UFH had a mean size of 122 nm and a PDI of ∼0.208. The complexes increased only slightly in size during the ensuing 96 hours (Figure 2C).

Addition of KKO (7.5 µg/mL) to PF4/polyP₁₃₀ complexes for 1 to 120 hours caused a small stable increase in particle size from ∼70 nm to ∼90 nm (consistent with binding of multiple IgG antibody molecules) and lowered the PDI to ∼0.130, excluding nonspecific particle agglutination through crosslinking PF4 (Figure 2D). In contrast to the time-dependent enlargement in PF4/polyP₁₃₀ complexes, complexes preincubated with KKO maintained their size (∼70-100 nm) and PDI for up to 120 hours, indicating enhanced stability (Figure 2E).

In theory, complexes between PF4 and polyP₁₃₀ should be susceptible to degradation by plasma phosphatases unless “protected” by antibody; therefore, we examined degradation of preformed PF4/polyP₁₃₀ complexes in the absence of or following addition of KKO. Complexes were incubated with buffer or KKO for 1 hour and DLS analysis was performed. CIP (200 U/mL) was then added for 2 hours at 37°C and the DLS analysis was repeated. PF4/polyP₁₃₀ complexes were almost totally degraded by CIP in the absence of antibody (average size ∼10 nM), whereas complexes preincubated with KKO increased slightly in size from ∼70 nm to ∼100 nm and showed minimal increase in polydispersity following addition of CIP (Figure 2F). The isotype control monoclonal antibody TRA had no effect on complex size, stability at 72 hours, or susceptibility to phosphatases. These data suggest that KKO may provide additional protection to PF4/polyP₁₃₀ complexes from phosphatases present in biologic milieus.

Monocytes endocytose ultralarge complexes of PF4/heparin in a heparin-dependent manner.⁴⁶  To determine if PF4/polyP₁₃₀ ultralarge complexes are of sufficient size and stability to be processed similarly, monocytes were incubated for 24 hours with AF647-labeled PF4 (25 µg/mL) alone or with 10 to 100 µM polyP₁₃₀ and uptake was visualized by confocal microscopy. Uptake of PF4 alone was negligible (Figure 3A), whereas endocytosis of PF4/polyP₁₃₀ complexes was readily evident (Figure 3A). Endocytosis of the complexes was concentration dependent, reaching a maximum at 50 µM polyP₁₃₀ (Figure 3B).

We previously showed that polyPs suppress the terminal pathway of complement activation in a concentration-dependent manner.⁴⁰  To examine the effect of PF4 on this property, exogenous purified C5b,6 (1250 pM) was added to cRBCs to achieve ∼75% lysis after 30 minutes. PF4 alone (1-200 µg/mL) had essentially no effect on lysis (not shown). PolyP₁₃₀ (50 µM) reduced C5b,6-induced lysis to ∼12%, which was used in subsequent experiments. PF4 at concentrations above ∼12 µg/mL dramatically reversed the activity of polyP₁₃₀, which entirely lost its complement inhibitory properties (Figure 4A), whereas monophosphates had no effect (not shown). Moreover, addition of complexes between PF4 and polyP₁₃₀ to normal plasma activated complement directly, as we previously reported for complexes between PF4 and UFH,⁴⁷  with complete loss of activity following preincubation of polyP₁₃₀ with the exophosphatase (Figure 4B).

We also previously reported that HIT antibodies fix complement to platelets and to endothelial cells.^6,45  We asked if the alterations in complement activity by polyP we described might be reflected in complement activation initiated by HIT antibodies. To do so, we incubated KKO with PF4/polyP₁₃₀ or with PF4/UFH complexes in the presence of normal human plasma as a source of complement. Addition of KKO to PF4/polyP₁₃₀ complexes led to fixation of C3 to at least the same extent, as did PF4/UFH complexes (Figure 4C). Plasmas from patients referred for evaluation of HIT and a positive PF4/UFH ELISA⁴⁸  showed a clear relationship between IgG binding to each form of PF4/polyanion complexes (Figure 4D top). In the presence of human plasma as a source of complement, C3 deposition was at least as extensive, and perhaps greater, with PF4/polyP₁₃₀ as the target antigen (Figure 4D bottom). In contrast, none of the 8 plasmas with a negative ELISA result showed IgG binding to PF4/polyP₁₃₀ above the normal range, and only 1 caused increased C3 (data not shown).

KKO binds to complexes of PF4 released from activated platelets with heparin or cell-surface chondroitin sulfate, which activate platelets through FcRγIIA.^49,50  We asked whether immune complexes composed of polyP, PF4, and KKO activate platelets in an analogous manner. PRP was preincubated with PF4 (10 µg/mL) and optimized concentrations of polyP₁₃₀ (0.5 µM) or polyP₁₄ (8 µM) followed by KKO (or RTO as a negative control in the case of UFH). Addition of polyP₁₃₀ along with PF4 and KKO caused platelets to aggregate (Figure 5A). Similarly, addition of polyP₁₃₀ to PF4 mediated KKO-induced platelet activation in whole blood in a dose-dependent manner, assessed by surface expression of P-selectin and binding of annexin V (Figure 5B). The extent of activation followed a bell-shaped curve as seen with UFH,²  with maximum activation at 3 µM polyP₁₃₀ (Figure 5B). PolyP₁₄ and polyP₆₀ also mediated platelet aggregation upon addition of PF4 and KKO; the concentration of polyP required was inversely related to chain length, similar to the relationship between chain length and the potency of heparin (Figure 5A).

We next asked if polyP must remain intact and form complexes with PF4 to permit antibody-induced platelet aggregation, as opposed to an indirect effect on platelet function. To do so, we used washed platelets and examined the effect of degradation of polyP by CIP. Platelet aggregation by PF4, polyP₁₃₀, and KKO was abolished in the presence of chondroitinase ABC or CIP (Figure 5C). This suggests that immune complexes containing exogenous polyP₁₃₀, exogenous PF4, and KKO activate platelets; activation requires persistent contact between the polyanion and PF4; and binding of PF4 to cell-surface chondroitin sulfate facilitates formation or stability of the resultant immune complexes.

PF4 is stored in platelet α granules, whereas polyPs are found primarily in dense granules (Figure 6A).^26,27  We asked whether PF4 and polyP form complexes potentially capable of mediating HIT antibody binding when platelets are activated. Using confocal microscopy and direct immunolocalization, we independently visualized PF4 and polyP. PolyP was detected using biotinylated PPXbd, the specificity of which was confirmed by loss of signal when platelets were incubated with PPX1 (not shown). PF4 and polyP were distributed separately in resting platelets, confirming their localization in different organelles. When platelets were activated by PMA or the calcium ionophore A23817, polyP and PF4 appeared to coalesce and colocalize (Figure 6B). Because the platelets were permeabilized, we could not establish whether this occurred intracellularly and/or on the cell surface.

To address the question of localization, we asked whether PF4 and polyP, when released from activated platelets, form antigenic complexes on the cell surface in the absence of exogenous PF4 or exogenous heparin. To do so, we activated washed platelets with a subthreshold concentration of TRAP (ie, sufficient to initiate a first wave of aggregation but not an irreversible secondary wave) under the assumption that this would permit polyPs and PF4 to form complexes that would be released and expressed on the platelet surface where they could be recognized by KKO. Indeed, priming of platelets with TRAP sensitized them to KKO-induced activation, and activation was abrogated by CIP (Figure 6C). Because CIP can cleave ADP and potentially other platelet agonists, the experiments were repeated in the presence of the specific inhibitor PPXI, which does not degrade ADP. Either degradation of polyP by PPX1 or inhibition of polyP by including PPXbd totally prevented KKO-induced aggregation, whereas chondroitinase ABC minimally delayed, but did not prevent, complete aggregation (Figure 6C). These studies show that exogenous platelet activation is followed by formation of endogenous antigenic complexes capable of forming immune complexes that can support platelet activation in the absence of exogenous PF4, exogenous heparin, and cell-surface GAGs.

These studies affirm that PF4 can form antigenic complexes with polyphosphates of diverse sizes ranging from those released by platelets to those released by bacteria²⁰  (Figure 1). The latter suggests a potential new link between the postulated binding of PF4 to bacteria and the genesis of anti-PF4 antibodies responsible for HIT.^51,52  The complexes were recognized both by the HIT-like monoclonal antibody KKO (Figure 1) that was generated by immunizing mice with PF4/UFH and by human HIT antibodies (Figure 4). PF4/polyP₁₃₀ complexes reach more than 1 μm in size, depending on the ratio of reactants, comparable in size and stability to those formed between PF4 and UFH,⁴⁶  but were more homogeneous (Figure 2), and were similarly susceptible to rapid internalization by monocytes (Figure 3). Binding of KKO protected these complexes from rapid degradation by phosphatases, such as those present in plasma, suggesting they may endure for sufficient times to initiate or augment immune complex–mediated cell activation in vivo. PF4 neutralized the inhibitory effect of polyphosphates on activation of the terminal complement pathway and immune complexes formed between PF4, polyphosphates, and KKO or human antibodies activated complement and incorporated C3 to at least the same extent as PF4/UFH complexes (Figure 4).

Addition of PF4 and polyphosphates to PRP or whole blood initiated antibody-induced platelet activation (Figure 5). The polyphosphates had to remain intact for platelet activation to proceed, excluding an indirect or nonspecific effect on platelet function. The finding that chondroitinase ABC blocked platelet activation by exogenous PF4/polyphosphate complexes suggests that chondroitin sulfate may assist in the initial binding of PF4 to the cell and formation of the antigenic complexes. Endogenous PF4 and polyphosphates can also mediate antibody-induced platelet activation. We affirmed prior findings that PF4 and polyphosphates localize predominantly in discrete granules (Figure 6). When platelets are activated, these components coalesce, which we interpret as having the potential to form antigenic complexes. In support of this interpretation, platelets activated by a subthreshold concentration of TRAP released sufficient PF4 and polyphosphates to form complexes that mediate antibody-induced platelet activation in the complete absence of exogenous PF4 or exogenous heparin and following enzymatic digestion of cell-surface chondroitin sulfate (Figure 6). PolyPs released by other cell types as a result of inflammation or injury^53,54  might also complex with PF4 and participate in an autostimulatory loop involving antibody formation, platelet activation, perpetuation of thrombocytopenia, and the risk of thrombosis.

These findings suggest a model in which platelets initially activated by PF4/UFH-containing immune complexes in solution or PF4 bound to cell-surface chondroitin sulfate release additional PF4 and polyphosphates (Figure 7). Released PF4 and polyphosphates form antigenic complexes that bind to the cell surface and are capable of mediating platelet activation in the absence of exogenous PF4 or heparin. Polyphosphates released by platelets activated by immune complexes and later by thrombin or other agonists might thereby exacerbate PF4/UFH-mediated platelet activation and potentially perpetuate the prothrombotic state in patients with HIT after heparin has been metabolized and plasma levels of PF4 dissipate as the initiating inflammatory or postoperative state resolves. Additional studies are needed to elucidate how PF4/polyP binds to the platelet. Additional studies are also needed to determine whether polyPs released from bacteria or activated platelets predispose to formation of PF4/heparin antibodies upon exposure to high or low doses of heparin or after surgery⁵⁵  or might contribute to the development of “delayed” or “spontaneous HIT.”^16,56  Inhibitors of polyphosphates might provide a new approach to interrupting the prothrombotic phenotype in HIT.^22,57-59 

The authors thank Connor Paez and Victor Lei for assistance in performing key experiments.

This study was supported by grants from the National Institutes of Health National Heart, Lung, and Blood Institute (grant HL128895) (D.B.C.), (grant HL110860) (D.B.C., L.R., M.P., and G.M.A.), and (grant R01 HL047014) (J.H.M.); Canadian Institutes of Health Research (CIHR) Operating Grant (OMH-134130) and Clinician-Scientist Salary Award (H.K.); University of British Columbia Faculty of Dentistry Postdoctoral fellowship (C.G.); and grants from the CIHR, the Natural Sciences and Engineering Research Council of Canada, and the Canada Foundations for Innovation (E.M.C.). E.M.C. holds a CSL Behring Research Chair and a Tier 1 Canada Research Chair in Endothelial Cell Biology, is an Adjunct Scientist with the Canadian Blood Services, and is a Canadian Venous Thromboembolism Trials and Outcomes Research Investigator.

Contribution: R.J.T., S.A.S., and J.H.M. helped generate, characterize, and analyze experiments involving polyphosphates; L.R. and M.P. performed and interpreted the flow cytometry experiments; G.M.A. and S.K. designed and interpreted experiments involving complement activation and endocytosis by monocytes; A.C. identified and recruited patients with heparin-induced thrombocytopenia; H.K. and C.G. performed the confocal microscopy; S.V.Y., S.V.Z., V.S., T.L, and A.H.R. helped design and perform all other experiments; L.M.O. helped to design and implement and E.M.C. oversaw the complement inhibition experiments; D.B.C. is responsible for the overall design of project, integration of the results, initial preparation of the manuscript; and all of the authors read and contributed to preparation of the final manuscript.

Conflict-of-interest disclosure: R.J.T., S.A.S., J.H.M., and E.M.C. are coinventors on patents and patent applications on medical applications of polyphosphate and polyphosphate inhibitors. The remaining authors declare no competing financial interests.

Correspondence: Douglas B. Cines, Departments of Pathology and Laboratory Medicine and Medicine, Perelman School of Medicine, University of Pennsylvania, 513A Stellar-Chance Pavilion, 422 Curie Blvd, Philadelphia, PA 19104; e-mail: dcines@mail.med.upenn.edu.