Xenotropic and polytropic retrovirus receptor 1 regulates procoagulant platelet polyphosphate

Inorganic polyphosphate (polyP) is a linear polymer of tens to hundreds of phosphate residues linked via high-energy phosphoanhydride bonds. This evolutionary ancient polyanion is ubiquitous in nature, in bacterial, fungal, plant, animal, and human cells.¹  Despite its structural simplicity, polyP regulates several distinct, intracellular activities. The polymer contributes to energy homeostasis and functions as a storage pool for inorganic phosphate (P_i).^2-4 

Regulation of polyP has remained poorly understood in metazoans but has been characterized in detail, in prokaryotes and yeast. Saccharomyces cerevisiae express 4 distinct P_i transporters (Pho84, Pho87, Pho90, and Pho89)⁵  that mediate uptake of extracellular P_i into the cytosol; deficiency in individual P_i transporters reduces polyP formation.⁶  Genetic analyses in yeast have confirmed a correlation between intracellular P_i levels and polyP content: strains with low imported P_i also exhibit reduced intracellular polyP,⁷  and, vice versa, polyP levels are reduced when budding S cerevisiae are switched from P_i-rich to P_i-starved medium.^8,9  Some bacteria such as Acinetobacter johnsonii and the parasite Trypanosoma cruzi (the etiologic agent of Chagas disease) accumulate up to 30% of their dry cell weight as intracellular granules referred to as acidocalcisomes, which are calcium- and polyP-rich organelles.^10,  In some mammalian cells, polyP is stored in organelles that function similarly to acidocalcisomes, including mast cell and basophil granules,¹²  astrocyte vesicles,¹³  and platelet-dense granules¹⁴  that allow for rapid release of preformed polyP into the interstitial and vascular spaces. However, only small amounts of soluble short-chain polymers are found in the supernatant of activated platelets, whereas most is retained on the plasma membrane as insoluble, calcium-rich nanoparticles.^15,16  Platelet polyP has been recognized as a procoagulant mediator with a multitude of functions in the coagulation system that include activation of factor XII (FXII), which in turn triggers the “intrinsic” pathway of coagulation.^15,17-21  The sum of polyP-mediated procoagulant activities leads to enhanced clotting in human plasma and increased thrombosis in animal models.^22,23 

Xenotropic and polytropic retrovirus receptor 1 (XPR1) is an 8-pass transmembrane molecule of 696 amino acids that was initially recognized as a cell surface receptor for xenotropic and polytropic murine leukemia retroviruses (X- and P-MLV).²⁴  Recent work in mammalian cells revealed that XPR1 is involved in P_i export and leads to the depletion of intracellular P_i.^25,26  In yeast and excavate eukaryotes, XPR1 expression has been linked to polyP metabolism. T cruzi expresses an XPR1 homologous protein, TcPho91, and modulation of TcPho91 activity influences intracellular polyP levels.²⁷  Deficiency in the XPR1 homolog Pho91 in yeast consistently increases intracellular P_i and polyP levels.⁶ 

Despite its fundamental functions, polyP regulation in mammalian cells has remained largely undefined. In this study, we tested the hypothesis that polyP levels are regulated in response to P_i homeostasis. Using omics data resources, we identified XPR1 as a major P_i transporter in platelets. Pharmacologic and genetic interference with XPR1 led to intracellular polyP accumulation. Similarly, targeting XPR1 in platelets increased polyP content, augmented activated platelet-driven coagulation, and led to thrombus formation under flow in an FXII-dependent manner. Conditional ablation of Xpr1 in platelets accelerated vascular occlusion in murine models of venous and arterial thrombosis without affecting hemostasis. Taken together, the experimental results of the present study provide evidence that XPR1 is a key regulator of platelet polyP and reveal an unidentified role of P_i homeostasis in thromboembolic disease.

Detailed description of antibodies, reagents, and additional methods can be found in the supplemental Methods (available on the Blood Web site).

Xpr1^fl/fl mice carrying loxP sites on either side of Xpr1 exon 2 have been described.²⁸  Xpr1^fl/fl mice were bred with Pf4-Cre transgenic animals to delete Xpr1, specifically in the megakaryocyte/platelet lineage. All mice were treated according to national guidelines for animal care at the animal facilities of University Medical Center Hamburg-Eppendorf, and the protocol was approved by local authorities (no. 76/16). For animal experiments, 8- to 14-week-old mice of either sex (1:1 ratio) were used. All animal procedures were conducted in accordance with the 3R (replacement, reduction, refinement) regulations.

Anion-exchange columns were used to isolate PolyP from HEK293 or MEG-01 cells.²⁹  In brief, cells were incubated with sodium chloride (4 M) for 5 minutes, diluted with 50 mM Tris (pH 8.1), incubated with DNase I (200 U/mL) and proteinase K (750 μg/mL; Sigma-Aldrich) for 1 hour at 37°C, and centrifuged at 14 000g for 8 minutes to remove debris. Lysates were diluted in buffer QG (Qiagen) and applied to QIAquick Spin Columns (Qiagen). Columns were washed twice with washing buffer, and polyP was eluted in 50 mM Tris buffer. Isolated material was subjected to a 1-hour incubation with 10 μg/mL recombinant Escherichia coli exopolyphosphatase (PPX) at 37°C, which ablated purified polyP procoagulant activity.³⁰  Released orthophosphate was estimated by using a Phosphate Assay Kit (Abcam, Cambridge, United Kingdom), according to the manufacturer’s instructions. P_i concentrations attributed to polyP were quantified by calculating the difference between PPX-treated and untreated samples.

Washed human platelets were incubated with XVDL and/or PVDL for up to 24 hours before polyP measurements. Flow cytometry analyses probing for surface expression of P-selectin (CD62P, 1:100 dilution; clone Psel.KO.2.7; Bio-Rad) indicated that platelets were not activated during the incubation period. Quantification of soluble and membrane-associated insoluble polyP was performed as previously described.¹⁶  In brief, washed human or Xpr1^fl/fl, Xpr1^fl/+ Pf4-Cre, and Xpr1^fl/fl Pf4-Cre mouse platelets were incubated for 20 minutes with Horm collagen (10 μg/mL) in HEPES-Tyrode’s buffer. Supernatants containing soluble polyP were incubated for 2 hours with buffer or PPX (50 μg/mL) at 37°C, and P_i concentrations were determined with a malachite green–based phosphate assay kit (Abcam).³⁰  To measure insoluble, membrane-associated polyP, collagen-stimulated platelets were incubated with 250 mM EDTA for 15 minutes at 37°C, to dissociate membrane-retained Ca²⁺-polyP nanoparticles,¹⁶  followed by PPX digestion.

Ten distinct phosphate transporters including XPR1, SLC17A1-4, SLC20A1-2 (P_iT-1 and P_iT-2), and SLC34A1-3 have been described in eukaryotic systems.^25,31  To predict P_i transporter expression in platelets, we sourced RNA sequencing (RNAseq) data from unfractionated human bone marrow, generated as part of the Human Protein Atlas project (https://www.proteinatlas.org/). XPR1, SLC20A1, and SLC20A2 were the only P_i transporters with detectable messenger RNA (mRNA) expression in the bone marrow (Figure 1A). To confirm transporter expression in circulating cells in an independent data set, RNAseq data from unfractionated human whole-blood samples (n = 407) was downloaded from GTEx Analysis V7 (https://gtexportal.org/home/). In all samples, XPR1, SLC20A1, and SLC20A2 mRNA was detected (mean fragments per kilobase of exon model per million reads mapped [FPKM], 2.6, 25.7, and 1.74, respectively). We next examined which of those transporters are expressed in platelets by using the Platelet Web database (http://plateletweb.bioapps.biozentrum.uni-wuerzburg.de/plateletweb.php).³²  XPR1 expression was detected in human platelets, on both the mRNA³³  and protein levels,³⁴  but SLC20A1 and SLC20A2 were not. To validate our findings, we probed for XPR1 protein in membrane fractions of human and murine platelets. Antibodies against the N and C termini of XPR1 detected a single band in human (Figure 1B) and mouse (Figure 1C) platelet membrane fractions, migrating at an apparent molecular weight of ∼65 kDa (supplemental Figure 1A-D). Recombinant human XPR1 expressed in HEK293 cells migrated with a similar apparent molecular weight and served as the control. Consistent with cell surface localization of XPR1 in adherent cells,²⁶  immunofluorescence microscopy visualized XPR1 at the plasma membrane in nonpermeabilized human platelets (Figure 1D). To assess the subcellular localization of platelet XPR1 on an ultrastructural level, the platelets were analyzed by transmission electron microscopy. Immunogold labeling showed that XPR1 was enriched at the plasma membrane (Figure 1E). These data indicate that XPR1 is a major phosphate transporter in platelets.

In inframammalian systems, The XPR1 homologous proteins TcPho91 and Pho91 regulate polyP levels.^6,27  We hypothesized that XPR1 similarly controls polyP in platelets and examined whether there is a correlation between XPR1 expression and intracellular polyP levels in human cells. Transient overexpression of XPR1 in HEK293 cells (Figure 2A) and MEG-01 megakaryocytes (Figure 2B) increased XPR1 signal intensity in western blots by 8- and 45-fold, respectively, compared with the mock control. The increased XPR1 expression dose-dependently reduced total polyP content to 65% ± 3% and 58% ± 9%, respectively, compared with levels in mock vector–transfected HEK293 (Figure 2C) and MEG-01 (Figure 2D) cells (set to 100% each). In contrast, interference with XPR1 expression using siRNA-mediated knockdown decreased XPR1 mRNA, leading to increased polyP levels in HEK293 (up to 138% ± 3%; Figure 2E) and MEG-01 cells (up to 129% ± 5%; Figure 2F). Small interfering RNA (siRNA) treatment did not affect cell morphology or levels of the housekeeping vasodilator-stimulated phosphoprotein (not shown). The chain length of polyP molecules ranged from 50 to 400 and from 40 to 200 P_i moieties in control-treated HEK293 and MEG-01 cells, respectively (Figure 2G-H). XPR1 overexpression in those cell types only slightly reduced polyP size (50-150 and 40-140 mers; Figure 2G). Accordingly, XPR1 knockdown had only a minor impact on average polyP size or degree of dispersion in HEK293 and MEG-01 cells on day 3 (Figure 2H). We used PPX, an enzyme that specifically degrades polyP,^30,35  and confirmed that the cell-purified material was indeed polyP. Together, the data showed that polyP content is inversely proportional to XPR1 expression levels.

XPR1 is the entry receptor for X-MLV^36,37  and an XPR1 ligand derived from the X-MLV envelope receptor-binding domain XVDL inhibits the export of XPR1-mediated P_i, leading to increased P_i in adherent cells.²⁶  To pharmacologically target XPR1 activity, we cloned and expressed in E coli XVDL and the corresponding P-MLV envelope ligand (PVDL, negative control; schematically shown in supplemental Figure 2A-B). Although X- and P-MLV differ in their activities, they share high sequence homology.^26,38  Coomassie brilliant blue–stained polyacrylamide gels and immunoblot analyses confirmed the purity of recombinant XVDL and PVDL ligands (supplemental Figure 2C-D).

Washed human platelets were incubated with increasing concentrations (2-200 μg/mL) of XVDL or PVDL for up to 24 hours. Virus-derived ligands did not activate platelets (data not shown). PolyP was subsequently isolated from treated cells at various time points by anion-exchange chromatography and was quantified as monophosphate units after PPX digestion by a malachite green–based colorimetric assay.³⁰  XVDL treatment dose dependently and significantly increased platelet polyP content (up to 134% ± 7% polymer level) compared with buffer-incubated controls (set at 100%; Figure 3A). PolyP levels increased after 4 hours of incubation with XVDL and remained elevated for 24 hours. PVDL was inactive in altering platelet polyP (Figure 3B). X- and P-MLV compete with each other for binding to XPR1.³⁹  We tested PVDL for its interference with an XVDL-mediated increase in polyP. PVDL competed with an XVDL-induced increase in platelet polyP in a dose-dependent manner. PVDL levels as high as 600 μg/mL completely abolished the effects of XVDL (200 μg/mL) and restored polyP content to baseline levels (Figure 3C). Platelets contain soluble short-chain polyP that, upon cell activation, is released into the supernatant, as well as long(er) chain polyP that is insoluble and retained on the plasma membrane of procoagulant platelets as calcium-rich nanoparticles.^16,21  We examined the impact on each of these distinct polyP pools through XPR1 inhibition. Platelets were preincubated for 4 hours (sufficient for inducing maximum polyP increase) in solution with XVDL, PVDL, or buffer before activation with Horm collagen. Collagen treatment activates platelets and potently leads to polyP release.^16,30,40  Pretreating platelets with XVDL, dose dependently increased soluble polyP in the cell supernatant by threefold over polyP levels isolated in supernatant from buffer-treated platelets (1.8 ± 0.2 vs 0.6 ± 0.1 nmol P_i per 10⁸ platelets; Figure 3D); PVDL pretreatment did not influence soluble polyP levels (Figure 3E). Divalent cations mediate the formation and membrane association of insoluble polyP nanoparticles. EDTA (a chelator of divalent ions) disrupts polyP nanoparticles and allows for specific analysis of membrane-associated polyP.¹⁶  Preincubation of platelets with XVDL dose dependently increased membrane-associated insoluble polyP by more than twofold (6.3 ± 0.4 vs 2.6 ± 0.1 nmol per 10⁸ platelets) compared with buffer-pretreated platelets (Figure 3F). In contrast, even at maximum PVDL concentrations, no effect was noted in the amount of cell surface–exposed polyP (Figure 3G).

We next used real-time thrombin generation assays to examine whether XPR1 increases platelet reactivity in human platelet-rich plasma (PRP; Figure 3H-K). Consistent with previous data,^41,42  collagen-stimulated platelets initiated thrombin formation (Figure 3H-I). Preincubation with XVDL (Figure 3H), but not PVDL (Figure 3I), dose dependently increased collagen-induced platelet procoagulant activity, as seen by the increase in total and maximum (peak) thrombin generation by 56% and 143%, respectively, compared with buffer-preincubated activated platelets (875 ± 95 to 1361 ± 49 nM per minute and 47 ± 11 to 114 ± 6 nM at 200 μg/mL XVDL; Figure 3J-K). Targeting XPR1 also consistently shortened the lag time (11.3 ± 1.0 vs 9.1 ± 0.8 minutes) and time to peak (20.3 ± 1.3 vs 14.7 ± 1.1 minutes) thrombin formation, compared with the buffer control (supplemental Figure 3). To confirm that an increase in platelet polyP rather than other XPR1-mediated events underlies the increase in platelet procoagulant activity, we targeted released polyP in XVDL-pretreated platelets by using 2 strategies: (1) recombinant E coli PPX, which specifically degrades polyP, but not other naturally occurring polymers,³⁰  and (2) an enzymatically inactive PPX mutant lacking domains 1 and 2 (PPX_Δ12) that interferes with polyP procoagulant activities without degrading it.³⁰  Both PPX and PPX_Δ12 reduced the thrombin generation induced by synthetic and platelet-derived polyP, but did not interfere with tissue factor- or nucleic acid–driven thrombin formation.³⁰  Addition of PPX and PPX_Δ12 (500 μg/mL each) to XPR1-inhibited platelets abolished platelet hyperreactivity, and thrombin formation triggered by collagen activation was similar to that of unstimulated controls (total, 319 ± 71 and 239 ± 6 vs 299 ± 31 nM per minute. Peak thrombin, 15 ± 4 and 11 ± 1 nM vs 15 ± 1 nM). PolyP initiates coagulation in plasma via contact activation of FXII.^15,43  3F7 (650 nM),⁴⁴  a neutralizing antibody to activated FXII (FXIIa) and rHA-infestin-4 (500 μg/mL),⁴⁵  a recombinant FXIIa inhibitor, both blocked the increased procoagulant activity of XVDL/collagen–stimulated platelets. Neutralizing FXIIa and targeting polyP similarly interfered with excess thrombin generation in XVDL/collagen-treated cells (total, 527 ± 5 and 547 ± 197 nM per minute; peak thrombin, 22 ± 1 and 25 ± 9 nM; Figure 3J-K). Combining 3F7 and PPX did not result in further reduction of procoagulant platelet activity (total, 522 ± 129 nM per minute; peak thrombin, 23 ± 7 nM). Together, the data show that XVDL-treated, activated platelets trigger excess coagulation by increased polyP/FXII activity in plasma.

At sites of vascular injury, collagen is exposed from the subendothelial matrix, resulting in platelet activation and subsequent thrombus formation in an FXII-dependent manner.⁴⁶  Therefore, we next investigated whether interference with XPR1 augments thrombus formation on collagen-coated surfaces under flow (Figure 4). Citrate-anticoagulated human whole blood preincubated with increasing doses of XVDL (2-200 μg/mL), was recalcified before being perfused at arterial (Figure 4A) and venous (Figure 4B) shear rates of 1000 and 100 per second, respectively. Consistent with earlier data,^44,46  in the absence of XVDL platelets adhering to collagen fibers, aggregation and thrombus formation were observed within 4 minutes of the time of perfusion (16.5% ± 1.9% and 9.3% ± 0.8% surface coverage by thrombi at arterial and venous shear, respectively). Preincubation with XVDL dose dependently increased thrombus formation by >50%, compared with the untreated control (26.4% ± 2.5% and 17.0% ± 2.5%, respectively). In contrast, targeting polyP with PPX almost completely abolished thrombosis, at both arterial (<5%) and venous (<2%) shear rates. Immunofluorescence microscopy of formed thrombi confirmed the polyP-mediated procoagulant activity of XVDL. Fibrin deposition assessed by the fibrin-specific antibody 59D8⁴⁷  and platelet accumulation increased in thrombi after XPR1 inhibition (Figure 4C-D). To confirm that XVDL treatment led to increased polyP in platelets, we used fluorescently labeled PPX_Δ12 as a polyP-specific probe.⁴⁸  FITC-PPX_Δ12 signal intensity was >70% higher in thrombi produced by XPR1-inhibited platelets vs buffer control (10.8 ± 0.2 vs 18.5 ± 1.0 reflectance units at 1000 per second and 7.0 ± 0.3 vs 12.2 ± 0.7 reflectance units at 100 per second; Figure 4E-F).

To characterize the functions of XPR1 in platelets, we targeted its expression selectively in the megakaryocytic/platelet lineage by breeding Xpr1^fl/fl to Pf4-Cre transgenic mice. As littermate controls for homozygous conditional Xpr1 deficiency (Xpr1^fl/fl Pf4-Cre), we used Xpr1^fl/fl animals that lacked Cre expression. In addition, we produced heterozygous Xpr1^fl/+ Pf4-Cre mice having only a single Xpr1 allele in platelets. Polymerase chain reaction confirmed each genotype (Figure 5A). Xpr1 expression in megakaryocytes prepared from a primary bone marrow cell culture of Xpr1^fl/fl Pf4-Cre mice, was reduced to 45.7% ± 7.9% of levels seen in Xpr1^fl/fl mice (Figure 5B). Furthermore, flow cytometry and western blot analysis confirmed XPR1 protein reduction to 38.2% ± 7.8% and 55.2% ± 7.9%, respectively, in Xpr1^fl/fl Pf4-Cre megakaryocyte preparations (Figure 5C). In line with these data, Xpr1 expression was virtually absent in Xpr1^fl/fl Pf4-Cre mouse platelets (7.4% ± 1.8% of Xpr1^fl/fl platelets; Figure 5D). XPR1 protein level in platelets from Xpr1^fl/fl Pf4-Cre mice was reduced to 12.0% ± 5.1% as compared with Xpr1^fl/fl controls (Figure 5E), and flow cytometry confirmed the loss of XPR1 in Xpr1^fl/fl Pf4-Cre mouse platelets (supplemental Figure 4). We next asked whether reduced platelet XPR1 alone can affect polyP content. In platelets from Xpr1^fl/fl Pf4-Cre mice, total polyP level was increased by >75% over controls (2.60 ± 0.57 vs 1.42 ± 0.34 nmol per 10⁸ platelets), whereas a single Xpr1 allele in Xpr1^fl/+ Pf4-Cre mice was sufficient to maintain polyP close to baseline levels (1.41 ± 0.48 nmol per 10⁸ platelets; Figure 5F). PolyP-specific, negative 4′,6-diamidino-2-phenylindole (DAPI) staining confirmed increased polyP content in isolated platelets of Xpr1^fl/fl Pf4-Cre mice (not shown). Collagen-activated platelets from Xpr1^fl/fl Pf4-Cre mice released slightly more soluble polyP into the supernatant, compared with platelets from mice with intact or heterozygous XPR1 expression (0.77 ± 0.07 nmol vs 0.66 ± 0.02 and 0.66 ± 0.03 nmol polyP per 10⁸ platelets; Figure 5G), although the differences did not reach significance. Quantification of membrane-associated nanoparticle polyP showed that activated platelets of Xpr1^fl/fl Pf4-Cre mice, exposed ∼70% more polyP on their cell surface than did animals with heterozygous or normal XPR1 expression (1.91 ± 0.09 vs 1.12 ± 0.10 nmol or 1.09 ± 0.10 nmol polyP per 10⁸ platelets; Figure 5H). Hematologic analysis of peripheral blood from Xpr1^fl/fl Pf4-Cre, Xpr1^fl/+ Pf4-Cre, and Xpr1^fl/fl mice did not reveal significant differences in platelet counts, platelet morphology (size and volume), production, and life span when assessed by platelet distribution width or percentage of platelets among other circulating cells (plateletcrit). Similarly, total leukocyte, leukocyte subpopulation, and red blood cell counts and morphology did not differ among Xpr1^fl/fl Pf4-Cre, Xpr1^fl/+ Pf4-Cre, and Xpr1^fl/fl mouse lines (Table 1). To confirm the specificity of XPR1 toward polyP as the driver of increased platelet procoagulant activity, we characterized platelets from Xpr1^fl/fl and Xpr1^fl/fl Pf4-Cre mice. Luminescence assays revealed that intracellular ATP levels were indistinguishable in platelets from Xpr1^fl/fl and Xpr1^fl/fl Pf4-Cre mice (not shown). Real-time thrombin formation assays and conversion of thrombin substrate S2238 indicated that activated FX-triggered thrombin generation was indistinguishable in PRP from Xpr1^fl/fl and Xpr1^fl/fl Pf4-Cre mice (not shown). Moreover, annexin V-staining in flow cytometry showed comparable phosphatidylserine exposure on Xpr1^fl/fl Pf4-Cre and Xpr1^fl/fl mouse platelets upon activation with thrombin and collagen-related peptide (supplemental Figure 5A). CD31, CD41, and CD42b expression also remained unchanged (supplemental Figure 5B-D). These studies indicate that polyP-independent platelet functions and platelet surface properties were not different between Xpr1^fl/fl and Xpr1^fl/fl Pf4-Cre mice. Furthermore, in vitro studies demonstrated that aggregation in response to ristocetin (71.7% ± 3.1% vs 69.3% ± 1.4%; P = .45; Figure 5I) and low (71.0% ± 5.0% vs 77.0% ± 3.2%; P = .32; Figure 5J) or high concentrations of collagen (87.0% ± 2.1% vs 86.5% ± 2.1%; P = .89; Figure 5K) were similar in Xpr1^fl/fl and Xpr1^fl/fl Pf4-Cre platelets. Consistent with XPR1 inhibition in human platelets (Figure 4), thrombus formation on collagen-coated surfaces under arterial and venous shear rates was increased in blood from Xpr1^fl/fl Pf4-Cre mice vs that in blood from Xpr1^fl/+ Pf4-Cre and Xpr1^fl/fl animals (supplemental Figure 6).

To determine whether platelet XPR1 contributes to thrombosis in vivo, we challenged platelet-specific XPR1-deficient mice in models of arterial and venous thrombosis. Thrombosis in the carotid artery was induced by topical application of 5% ferric chloride (FeCl₃), a well-established model of platelet-driven thrombus formation.⁴⁹  Platelet polyP-induced thrombosis is defective in FXII deficient (F12^−/−) mice^{30,42,45,50-52}  and consistent with the previous reports, F12^−/− mice were protected from arterial thrombosis, as seen by the lack of occlusive thrombi after FeCl₃ challenge. Vessel occlusion times were significantly shortened in Xpr1^fl/fl Pf4-Cre mice compared with times in Xpr1^fl/fl controls (320 ± 34 vs 473 ± 52 seconds), whereas occlusion times in heterozygous Xpr1^fl/+ Pf4-Cre mice were almost identical to those in control mice (472 ± 50 seconds; Figure 6A-B). To study venous thrombosis, we used a model of platelet-driven lethal pulmonary thromboembolism (PTE). In this assay, platelets were activated by IV infusion of collagen and epinephrine, and mice that survived >30 minutes were considered survivors (Figure 6C; bottom). In line with earlier data,^42,52  F12^−/− mice were largely protected from precipitate disease in this PTE model (3 of 4 animals survived). However, all Xpr1^fl/fl Pf4-Cre mice (5 of 5) universally succumbed within the first 3 minutes after challenge, with an average survival time of only 1.1 ± 0.2 minutes. Similarly, all Xpr1^fl/fl control mice (5 of 5) did not survive the collagen/epinephrine challenge but their survival was prolonged by ninefold (9.8 ± 1.7 minutes). Survival times of heterozygous mice were similar to those in control mice (9.8 ± 3.7 minutes) and 3 of 7 mice died within the first 5 minutes of the challenge. To confirm PTE formation, lung perfusion was studied in all challenged mice by using IV administration of Evans blue dye. Perfused lung areas turned blue, whereas occluded parts remained their natural pink color. The collagen/epinephrine challenge resulted in almost complete vascular thrombotic occlusion in Xpr1^fl/fl Pf4-Cre mice, visualized by disturbed perfusion of the dye. Heterozygous Xpr1^fl/+ Pf4-Cre and control mice showed incomplete perfusion of the dye, indicating partial vessel occlusion. In contrast, lungs of F12^−/− mice presented with uniform distribution of the dye, consistent with preserved vessel perfusion (Figure 6C; top). Histologic sections from challenged mice showed a significantly higher number of large occlusive thrombi per field of view in the lungs of Xpr1^fl/fl Pf4-Cre mice compared with the number in Xpr1^fl/+ Pf4-Cre and Xpr1^fl/fl mice. Virtually no thrombi were found in challenged F12^−/− mice (Figure 6D).

To test whether increased platelet polyP content alters hemostasis, we examined bleeding times and blood loss in Xpr1^fl/fl, Xpr1^fl/fl Pf4-Cre, and Xpr1^fl/+ Pf4-Cre mice. Time to cessation of bleeding after tail clipping was similar in all 3 groups (90 ± 15 vs 93 ± 17 and 128 ± 17 seconds; Figure 6E). The amount of blood loss, as measured by hemoglobin concentration, was also unaltered among the respective groups (0.16 ± 0.05 vs 0.11 ± 0.03 and 0.25 ± 0.06; Figure 6F). In an independent experimental approach, we monitored bleeding from a tail injury site by gently adsorbing blood on a filter paper without touching the wound. Bleeding times of Xpr1^fl/fl, Xpr1^fl/fl Pf4-Cre, and Xpr1^fl/+ Pf4-Cre mice were recorded in 5-second intervals and were similar (368 ± 8, 364 ± 6, and 372 ± 11 seconds; Figure 6G; supplemental Figure 7). In summary, these studies indicate that XPR1 deficiency in platelets leads to accumulation of polyP that results in increased thrombosis, but has no impact on hemostasis.

Our study identified XPR1 as an instrumental contributor to procoagulant polyP homeostasis in mammalian systems and, by extension, as a key regulator of thrombosis. Specifically, XPR1 was the most abundant phosphate exporter in platelets and its expression correlated inversely with polyP content. Increasing intracellular P_i levels by interfering with platelet XPR1 activity led to polyP accumulation, which increased thrombosis risk in vitro and in vivo, without affecting hemostasis.

Seminal studies have revealed that activated platelets trigger coagulation in an FXII-dependent manner^41,53-55  and work during the past decade identified platelet polyP as an FXII contact activator in human plasma in vivo.¹⁵  An array of experimental thrombosis models in mice and baboons confirmed that polyP/FXII–mediated coagulation plays a critical role in arterial and venous thrombo-occlusive diseases.^30,52,56-59  Recently, polyP release and deposition as calcium-rich nanoparticles on the surface of activated platelets has been visualized.^16,60  Consistent with these findings,^21,61  pharmacologic and genetic targeting of platelet XPR1 increased polyP stores and release, increased fibrin formation, and accelerated thrombus development under arterial and venous shear rates in human blood (Figure 4) and in vivo in mice (Figure 6). Conversely, mice deficient in inositol-hexakisphosphate kinase 1 (IP6K1), a key enzyme implicated in diphosphoinositol pentakisphosphate (an abundant inositol pyrophosphate) biosynthesis, displayed lower polyP levels in platelets, lengthened clotting times, altered clot architecture, and were protected against PTE.⁵⁷  However, gene disruption of murine IP6K1 also increased insulin sensitivity, reduced spermatogenesis, and, in contrast to platelet XPR1 deficiency (Figure 6), compromised hemostasis, indicating that the kinase is involved in a multitude of pathways across various tissues.⁵⁷  Patients with Hermansky-Pudlak syndrome also have reduced platelet polyP,⁶²  and, similar to their murine counterparts, FXII-dependent clotting is defective in PRP.¹⁵  In contrast, targeted inhibition of XPR1 led to platelet polyP accumulation and promoted thrombosis (Figure 3). Potent agents that neutralize polyP have been developed, including cationic polyethylenimine and polyamidoamine dendrimers,⁵⁶  crown ether–based universal heparin reversal agents,⁵⁹  and exopolyphosphatase mutants that either degrade (PPX) or bind (PPX_Δ12) the polymer.³⁰  These polyP inhibitors interfere with platelet-driven arterial and venous thrombosis, but they spare hemostasis, as evidenced by their limited impact on bleeding.^30,59  This selective role of polyP inhibitors on thrombosis is in line with our findings that polyP accumulation in XPR1-defective platelets differentially contributes to thrombosis while sparing platelet hemostatic functions (Figures 3,4, and 6). Thromboprotection in the absence of bleeding conferred by various polyP inhibitors also supports the notion that platelet polyP contributes to coagulation in vivo, in large part through FXII activation. In contrast to other potential polyP targets, including coagulation factors V and XI, tissue factor pathway inhibitor, von Willebrand factor or fibrinolytic pathway components,^{17-20,22,63,64}  FXII has no role in hemostatic coagulation mechanisms and thus interference with its activation does not increase bleeding.

PolyP is an evolutionary ancient molecule that serves as a storage form of P_i. The role of P_i in polyP synthesis and the function of P_i export in the regulation of polyP is evolutionarily conserved from prokaryotes to eukaryotes.²⁷  In prokaryotes, the conversion of P_i to polyP contributes to reducing intracellular osmotic pressure¹ ; however, it is unknown whether this function is preserved in higher eukaryotic cells. P_i transporters help maintain intracellular P_i concentration at significantly higher levels than extracellular P_i (60-100 vs 1 mM). Disturbance of P_i levels affects polyP content in HEK293 cells and similarly, influences the XPR1-regulated P_i/polyP equilibrium in megakaryocytes and platelets (Figures 2 and 5). XPR1 expression in Xpr1^fl/fl Pf4-Cre mouse platelets was below residual XPR1 levels in cultured megakaryocytes from these mice. These higher XPR1 levels may be related to XPR1-expressing cells in the primary megakaryocyte preparation and incomplete Cre-mediated Xpr1 ablation (Figure 5). P_i transporter activity controls P_i/polyP levels in bacteria and yeast.^6,8,27,65  Although an association of P_i accumulation and storage as polyP seems to be conserved from prokaryotes to metazoans, regulation of polyP metabolism in mammalians has remained enigmatic.¹  XPR1 activity is regulated by the phosphate-sensing domain SPX which responds to P_i levels.⁶⁶  Moreover, XPR1 trafficking to distinct subcellular compartments (Figure 1E)⁶⁷  restricts its impact on intracellular polyP content to specific locations and pools within cells.^16,30  Indeed, targeting IP6K1 reduces platelet polyP levels by ∼50%,⁵⁷  supporting that polyP accumulation occurs in specific sites within cells and in an orderly, regulated manner.

Beyond its role in platelets, impaired XPR1 function caused by missense mutations in humans has been linked to an adult-onset neurodegenerative disease, primary familial brain calcification. Although it is a rare disorder characterized by calcium-(poly)phosphate deposits in the basal ganglia and other brain regions,^68,69  its clinical phenotype highlights the impact of P_i/polyP regulation in health and disease. The number of missense variants in the human XPR1 gene is consistently far below the number of expected variants, with a significantly increased z score of 3.23 (https://gnomad.broadinstitute.org/) further confirming its vital role in homeostasis.

In summary, our data indicate that the ancient link between intracellular P_i and polyP levels are conserved in mammals. Consistent with their functions in prokaryotes, P_i transporters also regulate polyP homeostasis in platelets and influence thrombosis.

The authors thank Clément Naudin for critically reading the manuscript, and Anne Jämsä for valuable technical assistance.

This work was supported by Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) grants A11/SFB 877, B8/SFB 841, and P6/KFO306 (T.R.); and the Federal Ministry of Education and Research (BMBF) (A.S.).

Contribution: S.G., E.X.S., C.M., and T.R. conceived and/or designed the study; L.M.B. and J.O. provided the bioinformatics; R.K.M., M.A., G.P., and M.H. performed the in vitro experiments; R.K.M. performed the in vivo experiments; M.S. performed the transmission electron microscope analyses; D.F. provided the in vivo model; C.D., A.S., T.B.H., M.F., M.G., S.R.-J., S.B., and C.K. collected and discussed the data; R.K.M., M.A, E.X.S., and T.R. wrote the manuscript; and all authors edited the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Thomas Renné, University Medical Center Hamburg-Eppendorf, Institute of Clinical Chemistry and Laboratory Medicine (O26), Martinistrasse 52, D-20246 Hamburg, Germany; e-mail: thomas@renne.net.