Packaging of supplemented urokinase into alpha granules of in vitro–grown megakaryocytes for targeted nascent clot lysis

There is an unmet need for safe and effective therapeutics to treat or prevent thromboembolism in settings such as postextensive surgery or trauma. Yet, the risk of bleeding often delays the introduction of anticoagulation in the immediate postevent timeframe. In these settings, there is a need for therapies that distinguish preexisting hemostatic clots, leaving them intact, while targeting for fibrinolysis nascent, potentially occlusive, thrombi.

Fibrinolytic agents, such as the plasminogen (PLG) activators (PA), tissue-type PA (tPA), and urokinase PA (uPA) are effective at lysing thrombi^1-3 but permeate preexisting hemostatic fibrin⁴ and can degrade circulating fibrinogen.⁵ Both PAs have short half-lives, necessitating continuous intravenous infusions for prophylaxis,⁶ and have vasoactive and neurotoxic side effects after a stroke (reviewed in Medcalf⁷). Therefore, the clinical use of fibrinolytic agents has been limited primarily to the management of a subset of patients with acute ischemic stroke or life- or limb-threatening thromboembolism.^8-10 

To address these limitations, we found that PAs linked to red blood cells or platelets (PLTs) provide protracted and safe thromboprophylaxis in several animal models.^11,12 However, red blood cells are not rapidly and selectively recruited to sites of incipient arterial thrombus formation,¹³ and linking the PAs to the surface of the PLTs can lead to thrombocytopenia and bleeding in mice and nonhuman primate models (M. Poncz, S. Zaitsev, and D. Myers, University of Michigan, unpublished observations, September 2015).

As an alternative approach, we ectopically expressed single-chain (sc) murine (m) uPA in megakaryocytes (MKs)/PLTs of transgenic mice. The m-scuPA was stored in PLT alpha granules (α-granules) without causing systemic fibrinolysis. The transgenic mice had a mild bleeding diathesis during parturition,¹⁴ ameliorated by tranexamic acid, which inhibits plasmin and uPA activity.¹⁵ These mice simulated human Quebec platelet disorder (QPD), which is caused by a duplication of the uPA gene (PLAU), also characterized by storage of uPA in PLT α-granules,^16-19 absence of systemic fibrinolysis, and a mild bleeding diathesis mitigated by tranexamic acid.¹⁴ We have previously shown that infusion of the uPA-containing PLTs (uPA-PLTs) into wild-type mice prevented nascent thrombus growth in a microemboli infusion model.¹⁴ This raises the possibility that delivering uPA packaged within PLT α-granules inverts PLT function from prothrombotic to fibrinolytic and does so in a manner that does not lyse mature clots that are no longer recruiting PLTs.¹⁴ 

Recent advances in the generation of PLTs from in vitro–grown MKs have been made, especially beginning with induced pluripotent stem cells.^20-22 A recent paper describing the infusion of PLTs generated from induced pluripotent stem cells into a patient as a PLT transfusion had been described.²³ We wondered whether in vitro–grown MKs can endocytose uPA from the media as the MKs are grown in defined media and thus lack the normal array of endocytosed proteins. Prior studies have shown that these in vitro–grown MKs can endocytose added fibrinogen and factor V (FV).²⁴ Here, we show that scuPA or a thrombin- but not plasmin-activatable uPA (uPAT)¹¹ added to the media were readily endocytosed by maturing MKs. We define receptors involved in their uptake and the interaction of these uPAs with other endocytosed and with endogenous α-granular proteins. Using an in vivo murine carotid artery photochemical injury model, we show that the uPAs studied can effectively prevent thrombus development. We discuss potential uses of this approach for targeted thromboprophylactic fibrinolysis and the application of this strategy for delivering other ectopic therapeutics via PLTs.

Please see supplemental Methods for additional methods, including western blotting (WB) and agonist activation studies, and a description of reagents and antibodies.

Granulocyte colony-stimulating factor–mobilized human CD34⁺ hematopoietic progenitor cells, purchased from the Fred Hutchinson Cancer Research Center Hematopoietic Cell Processing and Repository Core, were differentiated into MKs (CD34⁺ MKs), as we have previously described.²⁵ Day 10 or 11 CD34⁺ MKs were incubated for various times in complete growth medium containing various combinations of the following fluorophore-conjugated proteins: scuPA, uPAT, noncleavable PLG (ncPLG),²⁶ recombinant receptor-associated protein linked to human immunoglobulin G Fc fragment (FcRAP),²⁷ and FV. The CD34⁺ MKs were washed once with phosphate-buffered saline and immobilized on microscopic slides using Shandon CytoSpin III Cytocentrifuge with the M964-20FW-CytoSep for 15 minutes. Please, see details of immunofluorescent staining and confocal laser-scanning microscopy in the supplemental Methods.

CD34⁺ MKs differentiated for 10 to 11 days were incubated in complete growth medium containing a combination of the following fluorophore-conjugated proteins: scuPA, uPAT, and FV. Unlabeled FcRAP²⁷ or the blocking αIIbβ3 monoclonal antibody ReoPro^28,29 were used as lipoprotein receptor–related protein 1 (LRP1) and αIIbβ3 antagonists, respectively. The uptake of endocytosed fluorescent proteins was analyzed in CD42b⁺ population using a CytoFlex S flow cytometer (Beckman). Please see supplemental Methods for more details.

Nonobese diabetic-severe combined immunodeficiency IL2rγnull (NSG) mice were originally purchased from The Jackson Laboratory. The transgenic NSG/VWF^R1326H mice, homozygous for a CRISPR/CRISPR-associated protein 9 mutation in murine von Willebrand factor (VWF), has been described previously.^30,31 Before infusion, MKs were or were not preloaded with recombinant scuPA or uPAT as follows: day-10 CD34⁺ MKs were left intact or incubated with recombinant scuPA or uPAT (600 nM each) for 24 hours in the culture media. The cells were washed and resuspended in phosphate-buffered saline at a concentration of 3 × 10⁷ CD34⁺ MKs per mL. Please see supplemental Methods for details of the thrombosis model.

scuPA protein concentration in the untreated and scuPA-treated MK lysates was measured using the huPA Quantikine enzyme-linked immunosorbent assay kit (R&D Systems) per manufacturer. Please see supplemental Methods for details.

Means ± standard deviation are shown. For data obtained by flow cytometry, differences were analyzed by ordinary 1-way analysis of variance. For in vivo studies, differences in carotid blood flow in the photochemical carotid injury studies were analyzed by ordinary 1-way analysis of variance using GraphPad Prism 5. For all studies, differences were considered significant with a P value of < .05 compared with the indicated control.

Approval was obtained from the institutional internal review board for the volunteer blood samples. These studies were conducted in accordance with the Declaration of Helsinki. Murine studies had institutional animal care and use committee approval. Euthanasia was performed as approved by the Panel on Euthanasia of the American Veterinary Medical Association.

We, and others, have shown that in vitro–generated MKs differentiated in defined media can endocytose fibrinogen, FVIII, and FV until the late stages of differentiation^24,32,33; however, given the lack of these proteins in the growth media, these proteins would be absent from the α-granules. Perhaps if exogenous uPA, not normally found in PLTs,¹⁴ were added to the media, the MKs would endocytose and store uPA in their α-granules, and be useful to generate uPA-PLTs for targeted thrombolysis. We, therefore, added to the media scuPA. This uPA contains the N-terminal uPA-receptor (uPAR)-binding growth factor-like domain, a Kringle domain, and a C-terminal proprotease domain (Figure 1A, top). scuPA is dependent on uPAR as well as LRP1 for uptake into cells.³⁴ We also studied a low–molecular weight variant we developed,¹¹ designated uPAT, which lacks the 2 N-terminal domains (Figure 1A, bottom), making it incapable of binding to uPAR. uPAT is activated by thrombin, for example, generated at the site of clot formation, but not by plasmin,³⁴ preventing uPAT activation by plasmin to avoid α-granular protein degradation seen in QPD.³⁵ 

WB of cell lysates showed that control MKs were devoid of uPA, whereas cells incubated with scuPA showed the presence of the protein (Figure 1B). Both Alexa-568–conjugated scuPA and Alexa-488–conjugated uPAT were internalized by day-10 CD34⁺ MKs in a concentration-dependent (Figure 1C and supplement Figure 1A, respectively) and in a time-dependent (Figure 1D and supplemental Figure 1B, respectively) manner. Both appeared as punctate particles in MKs (Figure 1E). In contrast, isolated human PLTs did not take up either uPA (supplemental Figure 2).

One can suggest that scuPA and uPAT, based on their structural differences, enter MKs through different pathways and are stored in distinct granules. Alexa-568 scuPA and Alexa-488 uPAT were coincubated with CD34⁺ MKs for 24 hours. Both scuPA and uPAT resided in the same granules as visualized by confocal microscopy, with representative data shown in Figure 2A, supplemental Figure 3, supplemental Video 1, and Table 1. To prove that the endocytosed proteins reside in the granules within the cell rather than positioned on the cell surface, we acquired serial images at different focal depths and created 3-dimensional representations to provide a composite image with a greater depth of field. Supplemental Figure 3 and supplemental Video 1 show that the endocytosed proteins reside in the same granules.

FV is endocytosed by human MKs and stored in α-granules.⁴¹ To help elucidate whether uPAs endocytosed by MKs are also stored in α-granules, CD34⁺ MKs were incubated with exogenous Alexa-488 scuPA or uPAT for 18 hours, followed by incubation with Alexa-568 FV for an additional 2 hours. By confocal microscopy, there was extensive, but not as complete, overlap as seen in Figure 2A, indicating that endocytosed scuPA or uPAT are stored in many of the same α-granules as endocytosed FV (Figure 2B and C, respectively; supplemental Figure 4; supplemental Video 2; Table 1). Furthermore, added Alexa-568 FV competed for the uptake of Alexa-488 scuPA and uPAT in in vitro–grown CD34⁺ MKs in a concentration-dependent manner (Figure 2D and E, respectively), supporting that uPAs are stored in α-granules that contain physiologically endocytosed proteins.

The finding that uPAT localizes within the same α-granules as scuPA and extensively colocalizes with FV led us to investigate the involvement of LRP1, which mediates FV endocytosis.⁴¹ Prior studies showed that LRP1 is expressed during late megakaryopoiesis but is lost before PLT release,⁴² and we confirmed this by WB (Figure 3A). Flow cytometry analysis also confirmed LRP1 expression on day-12 CD34⁺ MKs (supplemental Figure 5A). Internalized scuPA is colocalized with LRP-1 in day-11 CD34⁺ MKs (Figure 3B; supplemental Figure 6). Alexa-488 scuPA and the LRP chaperone/antagonist Alexa-555 FcRAP²⁷ overlapped after their endocytosis (Figure 3C; Table 1), and partially inhibited the uptake of Alexa-488 scuPA, uPAT, and FV (Figure 3D).

We asked whether PLTs released in vivo from injected MKs into NSG mice express LRP1. Nontreated and scuPA-loaded day-12 CD34⁺ MKs were injected into NSG mice, aliquots of blood were withdrawn 6 hours after injection of MKs, and LRP1 was measured on human PLTs in whole mouse blood by flow cytometry. Like infused donor-derived PLTs, released human PLTs do not express LRP1 in the circulation whether the MKs had been exposed to scuPA or not (supplemental Figure 5B-D).

Fibrinogen is endocytosed via integrin αIIbβ3 in PLTs and MKs.^32,44 On confocal microscopy, CD34⁺ MKs coincubated with Alexa-568 scuPA and Alexa-488 fibrinogen showed extensive overlap of the proteins (Figure 4A; supplemental Figure 7; supplemental Video 3). The αIIbβ3-blocking antibody ReoPro^28,29 reduced uptake of both fibrinogen and scuPA but did not significantly affect endocytosis of FV (Figure 4B). Coincubation with FcRAP and ReoPro did not further reduce uptake of Alexa-488 scuPA than FcRAP alone (Figure 4C), suggesting a common point in the pathway of uptake blocked by FcRAP and ReoPro. In support of this shared pathway, we observed extensive colocalization of the endocytosed scuPA, both with LRP1, which carries out endocytosis of its ligands via clathrin-coated pits⁴⁵ and with IFITM3 (Figure 4D; supplemental Figure 8; supplemental Video 4; Table 1), which mediates endocytosis of fibrinogen in CD34⁺ MKs and PLTs through clathrin-coated pits via clathrin and αIIbβ3.⁴⁶ 

Next, we asked whether scuPA and/or uPAT endocytosed by CD34⁺ MKs colocalized with endogenously expressed proteins stored in α-granules.⁴⁷ We found very limited colocalization of endocytosed scuPA and uPAT with VWF (Figure 4E; supplemental Figure 9; Table 1; supplemental Video 5) and platelet factor 4 (PF4) (supplemental Figure 10), suggesting that endocytosed uPAs and endogenous VWF are nearly completely segregated into different α-granules in MKs.

In QPD and in PLT uPA-transgenic mice, α-granular proteins are proteolyzed by plasmin activated by the stored uPA, potentially compromising hemostasis.^14,16-19 Therefore, we investigated whether exogenously added uPAs also trigger proteolysis of proteins stored in the α-granules of in vitro–grown CD34⁺ MKs. Because we did not detect PLG in the lysates of in vitro–grown CD34⁺ MKs (supplemental Figure 11), we inferred that these cells would only contain PLG in vivo if it were endocytosed from plasma. To study the effects of PLG on uPA-loaded MKs, we used CD34⁺ MKs preloaded with Alexa-488 scuPA and Alexa-568 FV for 18 hours and then added for 18 hours Alexa-647 ncPLG, which cannot be converted to active plasmin²⁶ to avoid activation/degradation of endocytosed scuPA or uPAT and FV. We found significant colocalization of endocytosed uPAs, FV, and ncPLG in the α-granules of CD34⁺ MKs (Figure 5A; supplemental Figure 12).

We asked whether endocytosed uPAs would be activated by the presence of endocytosed native PLG and lead to degradation of endocytosed FV and/or endogenous VWF. We found that the copresence of endocytosed scuPA and PLG led to degradation of FV but not VWF, on western blot (Figure 5B). A similar study, but with added uPAT instead of scuPA, resulted in less FV degradation and, again, no cleavage of endogenous VWF (Figure 5B).

Next, we examined in vivo in mice whether endocytosed uPA by in vitro–grown CD34⁺ MKs affects thrombopoiesis and subsequent PLT half-life and agonist responsiveness. We used a model for studying the biology of released human PLTs that we had described previously (Figure 6A).³¹ This model recapitulates the pulmonary release of PLTs from MKs⁴⁸ and in which the PLTs closely resemble donor-derived PLTs.³¹ We found that both unmodified and uPA-exposed MKs released a similar number of PLTs per infused MK, time to maximal PLT count, and PLT survival at 24 hours (Figure 6B). The uPA content per PLT appeared to remain stable over the study (Figure 6C) and uPA-MKs and uPA-PLTs had similar responsiveness to human-specific, thrombin receptor activating peptide 6 (Figure 6D and E, respectively).

In vivo, we also examined whether the uPA-PLTs, released from uPA-MKs, mediate thrombolysis in mice. NSG mice, which tolerate infusion of human MKs or PLTs,³¹ formed carotid artery thrombi and blood flow reduction in response to Rose Bengal photochemical injury (Figure 7B, orange bar). To focus on uPA released by human PLTs, we studied NSG transgenic mice that are homozygous for a mutant murine VWF^R1326H that binds to the human glycoprotein Ib/IX but not mouse glycoprotein Ib/IX on PLTs. These NSG/VWF^R1326H mice exhibit a mild to moderate bleeding diathesis characterized by the absence of clot formation, as evidenced by greater blood flow after the photochemical injury than that of NSG mice (Figure 7B, dark grey bar vs orange bar). After injection of unmodified human CD34⁺ MKs into NSG/VWF^R1326H mice, the amount of blood flow was reduced to that seen in NSG mice (Figure 7B, blue bar vs orange bar). In contrast, injection of CD34⁺ MKs preloaded with either scuPA or uPAT, provided brisk fibrinolysis nearly to the extent seen with no human MK infusion in NSG/VWF^R1326H mice, consistent with the released uPA-PLTs preventing nascent thrombus formation after induction of a vascular injury (Figure 7B, red and green bars vs blue bar).³¹ 

We compared fibrinolysis by scuPA, packaged in MKs with nontreated MKs and with nontreated MKs plus infused scuPA (40 μg per 20 g mouse) in NSG/VWF^R1326H mice. Infusion of unpackaged scuPA plus nontreated MKs had a trend to increase fibrinolysis, but compared with nontreated MKs the difference was not significant (Figure 7C, pink vs orange bar), whereas fibrinolysis in mice receiving the scuPA-MKs was significantly different from the mice receiving the nontreated MKs (Figure 7C, brown vs orange bar). To estimate how much uPA is delivered into the mouse by 3 × 10⁶ scuPA-MKs, we lysed a fixed number of scuPA-MKs and measured the concentration of uPA by enzyme-linked immunosorbent assay, and then calculated the total amount of scuPA present in 3 × 10⁶ scuP-MKs. We calculated 8.8 ± 5.1 ng per 20 g mouse, a dose ∼4500 times lower than the dose of free scuPA infused along with nontreated MKs (Figure 7C). Therefore, <10 ng scuPA packaged within MKs was ∼4500 times more effective then free scuPA infused at a dose 2 mg/kg to provide similar (Figure 7C, pink vs brown bars) clot lysis effect in NSG/VWF^R1326H mice. These data support our hypothesis that uPA-MKs–derived uPA-PLTs release a concentrated wave of the drug locally at the nascent thrombus.

We had pursued a promising approach to targeted thrombolysis using PLTs to deliver a fibrinolytic agent to sites of nascent thrombosis, avoiding lysis of mature thrombi. We called such PLTs “antithrombotic thrombocytes.” Our prior in vivo studies in mice showed that infusing uPA-containing PLTs isolated from transgenic mice that expressed scuPA during megakaryopoiesis into wild-type mice led to clot lysis without causing bleeding¹⁴; however, it was not clear how we could translate this genetic approach to clinical care and circumvent the excessive α-granule protein proteolysis by uPA-generated plasmin, seen in QPD.^16-19 The development of technologies to generate donor-independent in vitro–generated MKs^51,52 and potentially PLTs has led to the first description of infused in vitro–generated PLTs in a patient²³ and suggested a new strategy of generating antithrombotic thrombocytes. We envisioned that in vitro–differentiated MKs could be incubated with uPA before infusion to a patient, with the released PLTs containing sufficient uPA to achieve effective fibrinolysis of incipient thrombi.

We tested 2 different uPAs: (1) scuPA being a full-length uPA that retains plasmin activation but is known to be able to be taken up by uPAR^53,54 as well as other receptor systems on MKs,^55,56 and (2) uPAT, a low–molecular weight uPA that lacks a plasmin activation site but instead can be activated by thrombin.⁴⁰ Our reason to study uPAT was to limit proteolysis of uPA in the PLT α-granule pool that is seen in QPD.^16-19 Another reason is that requiring thrombin to be present would limit proteolytic activation of uPAT to actively thrombosing sites in which thrombin would be available as opposed to mature thrombus sites.

Surprisingly, scuPA and uPAT had similar endocytic characteristics into MKs: both were targeted to the same granular pools in in vitro–differentiated MKs. Uptake of these uPAs appears to overlap with uptake of naturally endocytosed proteins, FV and fibrinogen, with considerable overlap in the granules in which they are stored. Approximately half of FV endocytosis had been reported to occur via LRP1,⁴¹ and we show the same for the uPAs. We also show that the uPAs are also taken up via αIIbβ3 binding. Whether IFITM3, which participates in the uptake of fibrinogen in PLTs during inflammation⁴⁶ and colocalizes with endocytosed scuPA in CD34⁺ MKs (Figure 4D), is sufficient for uPA uptake by human PLTs in inflammatory states in the absence of LRP1, or whether inflammation induces reexpression of LRP1 in PLTs enabling them to endocytose uPA, needs be elucidated in future studies. The LRP1 and αIIbβ3 pathways were not completely independent because the sum total of uptake via these 2 pathways never exceeded approximately half of the total uptake of the uPAs. Whether the remaining uptake by MKs of the uPAs, FV and fibrinogen, involves a less-specific mechanism such as pinocytosis needs to be examined.^57,58 

We speculated that MKs grown in vitro in a specified cultured medium may have α-granules lacking the exogenous proteins normally endocytosed from the marrow stroma (ie, “naked α-granules”) and be especially able to endocytose other proteins as part of PLT-delivered targeted therapy. Our studies showed that FV and fibrinogen added to the medium competed with uPA uptake. The highest level of uPA uptake occurred when no competing protein was present. Whether uptake was limited by α-granule storage capacity or the availability of surface receptors and their limited capacity to recycle to the surface during megakaryopoiesis is untested. What is clear is that the capacity to endocytose uPAs was lost by released PLTs, likely because of the fact that receptors, such as LRP1, were no longer present on the PLT surface (Figure 3A; supplemental Figure 2). Thus, MKs can store functionally significant amounts of a protein normally not present in these cells. The released, modified PLTs are capable of performing biological functions distinct from their natural roles.

Our studies also support that uPAT would have less risk of proteolysis of α-granular proteins than scuPA. We show that the risk is greatest for other endocytosed proteins and less so for endogenous proteins such as VWF. This is distinct from what is seen in QPD and in our transgenic mice expressing murine uPA in which both exogenous and endogenous proteins were proteolyzed.^14,35,59 We propose that PLG is initially endocytosed by MKs into an endocytosis α-granule compartment, but with time, these admixed with an endogenous α-granule compartment in the MKs and in the subsequent PLTs. Our studies with added scuPA and PLG were too short-lasting to see this admixing of the 2 original pools, and so although endocytosed FV was clearly digested, VWF in the endogenous α-granule pool was protected. In contrast to added scuPA plus PLG, added uPAT plus PLG had reduced FV proteolysis (Figure 5B).

Our murine in vivo data support that uPA-MKs retained their capacity to release functional PLTs after infusion (Figure 6). Both scuPA-PLTs and uPAT-PLTs appear to be equally effective in preventing significant thrombus growth at a site of photochemical injury in a carotid artery model. Although in the presence of added PLG, scuPA-MKs may proteolyze α-granule proteins, in the absence of added PLG to the media, scuPA as well as uPAT should not be activated, leaving all α-granule proteins intact. Whether having a thrombin-activatable site in uPAT, rather than a plasmin-activatable site in scuPA, results in more targeted fibrinolysis to the site of new thrombi and allows greater discrimination from mature thrombi need to be further tested.

Finally, besides targeted storage of uPAs in PLTs, a similar strategy of delivery of a protein of interest to a site of vascular injury has been proposed for FVIII in hemophilia A^60,61 and ADAMTS13 in thrombotic thrombocytopenic purpura.^62,63 It may also be useful to deliver antiangiogenic proteins in prevention of hematogenous cancer spread or anti-inflammatory proteins in a number of vasculitides. The further development of a practical system for in vitro–PLT preparation from MKs is the major limitation. Once such a system is accepted clinically, its use for targeted delivery of a wide variety of therapeutics could be pursued. Many of these uses may require much smaller numbers of modified PLTs than needed for maintaining hemostasis and PLT counts.

In summary, in vitro–grown MKs in defined media can endocytose ectopic proteins such as scuPA and uPAT. The uptake of these uPAs are by pathways that overlap with those used by naturally endocytosed proteins, such as FV and fibrinogen, and the addition of these proteins can compete with uPA uptake. uPA uptake can be partially blocked by inhibiting uptake through LRP1 using FcRAP and uptake via αIIbβ3 using abciximab. These 2 overlapping pathways only account for half of uPA uptake. The basis for the remaining uptake is unclear. The endocytosed uPAs are in an α-granule pool distinct from those containing endogenously expressed proteins such as VWF and PF4. PLG is also endocytosed by MKs, and its co-uptake with scuPA and FV leads to complete FV proteolysis, but its co-uptake with uPAT leads to limited FV degradation. In these studies, VWF is not proteolyzed, consistent with their segregation to a distinct α-granule pool. In vivo studies show that thrombopoiesis and PLT biology are not affected by MK incubation with these uPAs, and the resulting PLTs can prevent nascent thrombi formation in mice. These studies suggest a new strategy for targeted intravascular therapy using in vitro–generated PLTs that may be of benefit for selective fibrinolysis and serve as a model for other potential clinical uses. However, the clinical application remains to be determined in light of the potential differences between a murine model and clinical disease, and because in vitro–generated MKs and PLTs are still in an early stage of clinical development.

The authors thank the University of Pennsylvania Cell & Developmental Biology Microscopy Core for management of confocal microscopy experiments, and the CHOP Flow Facility Core for management of flow cytometry experiments.

Support for this study came from the following National Institutes of Health (NHLBI) grants: R35 HL150698 (M.P.), RO1HL141462 (V.S.), RO1HL159256 (D.B.C.), and P01 HL139420, Project 2 (R.M.C.), and Russian Science Foundation grant 23-15-00540 (K.V.D.).

Contribution: M.P. along with V.S. provided overall leadership and experimental design and manuscript preparation; S.V.Z. prepared and characterized uPAs and performed the confocal and western blot studies; H.A. carried out the cell growth studies and the flow cytometric and in vivo studies; S.V.Z. and H.A. contributed to the data analysis and writing; M.A.K., K.B., K.V.D., L.I., and R.M.C. contributed important ingredients for these studies; and D.B.C. made valuable insights to the studies and their interpretation, and to manuscript preparation.

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

Correspondence: Victoria Stepanova, Department of Pathology and Laboratory Medicine, University of Pennsylvania Perelman School of Medicine, 513B Stellar Chance Bld, 422 Curie Blvd, Philadelphia, PA 19104; email: vstepano@pennmedicine.upenn.edu.