The plasminogen system developed at the root of vertebrate evolution as an important mediator of fibrin surveillance securing vascular patency in species with a circulatory system. A plasminogen molecule cognate to that found in mammalians arose 550 million years ago (mya) with cyclostomes (i.e., lampreys), while its prime activators, tPA and uPA, appeared 450 mya with jawed-vertebrates (i.e., cartilaginous fishes). The most primitive orthologue known to resemble mammalian uPAR appeared 370 mya at the branch between lungfish and tetrapods; two lungfish genes were found to encode an uPAR-like protein with three consecutive LU domains, each having all 10 consensus cysteine residues (Chana-Muñoz et al., 2019). The unique loss of the 7–8 disulfide bond in the N-terminal LU domain occurred later with the radiation of tetrapods (amphibians, snakes, lizards, turtles, crocodilians, and mammals). Intriguingly, the elimination of that particular disulfide bond co-evolved with the acquisition of an uPA sequence compatible with receptor binding, as deducted from studies on mammals (Lin et al., 2010; Bager et al., 2012; Leth et al., 2019b). Within the avian lineage, the gene encoding uPAR was lost by chromosomal rearrangements, while uPA with a receptor-binding competent growth-factor like domain (GFD) was conserved (Aimes et al., 2003). The inability of a subgroup of avian species to focus uPA-mediated plasminogen activation on their cell surfaces was further exacerbated by the subsequent loss of a gene encoding a prominent plasminogen binding membrane protein (Plg-R_KT) in the galliform lineage (Galliformes), which includes chicken (Sharma et al., 2020).

The uPA•uPAR interaction represents a very high affinity binding with a K_D of 19 pM and the complex is long-lived (k_off = 2 × 10^–4 s^–1) as assessed by surface plasmon resonance with purified components (Leth et al., 2019b). This tight binding is replicated on cells with a K_D of 55 pM using ¹²⁵I-labeled uPA and isolated monocytes (Nykjaer et al., 1990) and is well aligned with a plasma concentration of 20 pM uPA in healthy donors² (Zeitler and Schuster, 1999). The β-hairpin of the growth factor-like domain (GFD) in uPA represents the key uPAR-binding region, with the hot-spot residues (Tyr²⁴, Phe²⁵, Ile²⁸, and Trp³⁰) being buried deeply within the ligand binding cavity (Figure 1C; Huai et al., 2006; Lin et al., 2010). In contrast, more than 15 residues distributed along the surface of uPAR’s ligand binding cavity contribute to the binding affinity for uPA, but none acts as prominent hot spot residues (Gardsvoll et al., 2006). Comparing the atomic structures of unoccupied and uPA-bound uPAR revealed additional flexibility in the multidomain organization. Loop 2 in uPAR DII (residues 130–140) undergoes profound structural shifts to partly cover the entrance of the ligand binding cavity by directly interacting with and covering the β-hairpin of the bound GFD (Barinka et al., 2006; Xu et al., 2012; Zhao et al., 2015). This region, which was dubbed the ligand loading/unloading loop in uPAR, further adds to the complexity of the conformational changes that occurs in the assembly of the LU domains upon uPA binding.

The first evidence that uPAR facilitates cell adhesion to vitronectin-coated surfaces was reported in 1994 (Waltz and Chapman, 1994; Wei et al., 1994). These early studies reported that uPA-binding stimulated the uPAR-mediated cell adhesion to vitronectin. We now know that vitronectin primarily binds uPAR via its small somatomedin B (SMB) domain and that this interaction is of relatively weak affinity with a K_D of 2 μM (Gardsvoll and Ploug, 2007). In agreement with the potentiation of cell adhesion by uPA-binding, the affinity of the uPAR•SMB interaction increases approximately 3-fold by either uPA-binding or by introducing a non-natural disulfide bond in uPAR^H47C–N259C. Both mechanisms drive uPAR into the closed conformation, lead to robust lamellipodia formations (Gardsvoll et al., 2011a; Gardsvoll et al., 2011b), and increase cell migration on vitronetin-coated matrices (Madsen et al., 2007). Although a 3-fold increase in the affinity of uPAR•uPA for SMB may at first sight appear incremental, its biological impact is likely driven by pronounced avidity effects originating from the binding between uPA–uPAR clusters in lipid rafts and vitronectin molecules embedded in the provisional matrix. Again, inherent flexibility in the assembly of the LU domains in uPAR is required for the allosteric regulation of vitronectin-binding by uPA.

The functional and structural epitope on uPAR for vitronectin-binding is assembled by residues located in uPAR DI (Trp³², Arg⁵⁸, Ile⁶³), DII (Gln¹¹⁴, Arg¹¹⁶) and the flexible linker region between DI and DII (Arg⁹¹, Tyr⁹²), and is well separated from the central uPA-binding cavity (Figure 1C; Gardsvoll and Ploug, 2007; Madsen et al., 2007; Huai et al., 2008). This binding interface is dominated by the burial of Arg⁹¹ within a small cavity in SMB where it forms a strong ionic interaction with Asp²² and is flanked by Phe¹³ and Tyr²⁸. This binding pose resembles the one found between SMB and Arg¹⁰¹ in PAI-1 (Zhou et al., 2003) and the one utilized by the neutralizing monoclonal antibody 8B12 where Asp⁹⁹ forms ionic interactions with Arg⁸⁹ and Arg⁹¹ in uPAR (Zhao et al., 2015). One puzzling observation is that uPAR-binding of vitronectin completely buries a linear sequence in the linker region between uPAR DI and uPAR DII (–Ser⁸⁸–Arg⁸⁹–Ser⁹⁰–Arg⁹¹–Tyr⁹²–), which has been implicated in uPAR-mediated chemotaxis, directional cell migration, and angiogenesis via the formyl-peptide receptor type 1 (Resnati et al., 2002; Bifulco et al., 2010; Minopoli et al., 2019).

In vitro studies with purified proteins and cultured cells provided an outline of the biochemical pathway orchestrating cell surface associated plasminogen activation (Figure 2A). The hallmark of this pathway is the separate docking of two zymogens on the cell surface by: (i) a specific and high-affinity interaction between uPAR and pro-uPA (K_D ∼ 20 pM); and (ii) a low-affinity binding of plasminogen to an array of broadly distributed membrane proteins with C-terminal lysine residues (K_D ∼ 1 μM), including Plg-R_KT (Miles et al., 2014). Of note, Plg-R_KT colocalizes with uPAR on cell surfaces (Andronicos et al., 2010). The concomitant binding of pro-uPA and of plasminogen to cell surfaces provides a template for enhanced plasminogen activation due to an increased efficiency of the reciprocal activation of the two zymogens i.e., pro-uPA activation by plasmin and plasminogen activation by receptor-bound uPA thus forming a positive feedback loop. This arrangement lowers the Km for plasminogen activation from 25 μM in solution to 0.7 μM for cell-bound reactants, which is well aligned with a plasma concentration of 2 μM plasminogen (Ellis et al., 1991). Of note, cell-bound plasmin is refractory to inhibition by α₂-antiplasmin. This differential sensitivity to α₂-antiplasmin mediated inhibition provides a further drive toward a focal confinement of plasminogen activation on cell surfaces.

The inherent power of arming cells with such a potent protease system on their cell-surface is clearly demonstrated in vivo by the severe cutaneous phenotypes (phemphigoid lesions) that develop in bitransgenic mice, in which Plaur and Plau transcription in the skin is controlled by the keratin 5 promoter (Zhou et al., 2000). Assembly of a functional cell-surface template with plasminogen and uPAR-bound pro-uPA is required to drive this dermal pathogenesis, since neither of the single transgenic mice nor the bitransgenic mouse crossed into a Plg^–/–background develop these lesions (Bolon et al., 2004). The necessity for an intact and functional template to activate plasminogen is also found under non-pathological conditions, since mice with single deficiencies in either Plaur, Plau, or Plg are resistant to an engineered anthrax toxin that requires proteolytic activation by surface bound uPA (Liu et al., 2003). That dependency is further underscored by the lack of toxicity in the Plau^GFDhu/GFDhu mouse strain, which produces a catalytic proficient pro-uPA that fails to bind uPAR (Connolly et al., 2010). It is therefore beyond any reasonable doubts that cell-surface plasminogen activation driven by the assembly of pro-uPA•uPAR complexes is operational in vivo. The mild overt phenotypes of Plaur^–/– mice and their late-onset is most likely a consequence of functional redundancy, since mice with combined deficiencies for uPAR/tPA or for uPA/tPA (the two prime plasminogen activators) exhibit exacerbated hepatic fibrin deposition compared to Plaur^–/– mice (Bugge et al., 1996a).

The membrane-bound template for pro-uPA and plasminogen activation is also implicated in the proteolytic activation of an oncogenic transmembrane receptor denoted CUB domain containing protein 1 (CDCP1) (Kryza et al., 2021). A proteomic search in three different cancer cell lines with a protease-reactive warhead build on CDCP1 sequences, identified uPA and plasmin as the lead candidates for cleaving CDCP1 at Arg³⁶⁸ or Lys³⁶⁹. Proteolytic processing of CDCP1 at these sites leads to potentiation of its pro-metastatic effect. Subsequent mechanistic studies demonstrated that uPA needed to be tethered on the cell surface via uPAR binding to promote the cleavage of CDCP1 (Kryza et al., 2021).

While the impact of uPAR and uPA•uPAR complexes on cell adhesion and migration is well documented in vitro by a plethora of different cell culture experiments (Madsen et al., 2007; Petzinger et al., 2007; Salasznyk et al., 2007; Smith and Marshall, 2010; De Lorenzi et al., 2016), their impact (if any) under normal physiologic conditions in vivo is less clear. Like Plaur^–/– mice, Vtn^–/– mice have no overt phenotypes that would support a vital role for vitronectin in cell adhesion and migration during normal development (Zheng et al., 1995). Nonetheless, one study found that Vtn^–/– mice had slightly slower dermal wound healing, reduced dermal microvessel density, and focal sites with delayed hemorrhage (Jang et al., 2000). The origin of this mild phenotype was presumably endorsed by an unbalanced activity of tPA and uPA in the wound field allegedly due to a faster latency transition of their shared inhibitor PAI-1 in the absence of vitronectin. An intriguing rendezvous between uPAR-dependent cell adhesion and cell-surface associated plasminogen activation that is mediated by uPA•uPAR complexes provides a possible regulatory mechanism to control the adhesion between uPAR-positive cells and vitronectin-rich provisional matrices. This interplay involves three sequential steps: (i) focal adhesion initiated by the binding between pro-uPA–uPAR complexes on the cell surface and SMB in vitronectin that is deposited as part of the provisional extracellular matrix (Figure 2B; Madsen et al., 2007), (ii) manifest cell adhesion and migration after relay of signals via indirect coupling to integrins, receptor tyrosin kinases, or G-protein-coupled receptors (Smith and Marshall, 2010), and finally (iii) attenuation of cell adhesion and migration by activation of pro-uPA to uPA (De Lorenzi et al., 2016). Of note, zymogen activation within the pro-uPA•uPAR•plasminogen template thus provides a negative feedback loop on cell adhesion by promoting the cleavage of either uPAR in the linker region between DI and DII (Figure 2C) or at the R↓GD-motif in the linker to the SMB domain (Hoyer-Hansen et al., 1992, 1997; De Lorenzi et al., 2016). Both cleavages dismantle the ternary uPA•uPAR•vitronectin complex and are mediated by either uPAR-bound uPA or by cell surface associated plasmin.

Building on the seminal discoveries that plasma levels of uPA (Duffy et al., 1990) and soluble uPAR (Stephens et al., 1999) are robust biomarkers for adverse disease progression, accumulating observational studies continue to underscore that high uPA and/or high uPAR levels predict poor prognosis for patients with solid tumors (Grunnet et al., 2014; Dohn et al., 2015; Brungs et al., 2017; Loosen et al., 2018, 2019; Lu et al., 2018; Madunic, 2018). The source(s) of shed soluble uPAR is not always known, but the activated tumor-stromal microenvironment is a possible culprit from which uPAR may be released by proteases and/or hydrolases cleaving the GPI-anchor e.g., GDE3 (Figure 2C). The linker region between uPAR DI and DII is highly susceptible to proteolytic cleavage by e.g., uPA or plasmin, which releases uPAR DI from the cell surface (Hoyer-Hansen et al., 1992). Time-resolved fluorescence immunoassays with high sensitivity and specificity were therefore developed and validated for quantification of intact uPAR and its cleavage products in plasma (Piironen et al., 2004; Thurison et al., 2010). The baseline expression levels of uPAR, uPAR DIIDIII, and uPAR DI in healthy subjects are low; 36.5, 13.9, and 18.6 pM, respectively (Thurison et al., 2015). Subsequent studies showed that these soluble uPAR fragments were independent prognostic biomarkers in pre- and post-operative plasma samples from patients with colorectal cancer (Rolff et al., 2019). For more detailed information on uPAR expression and cancer dissemination the reader is referred to the following reviews (Andreasen et al., 2000; Romer et al., 2004; Allgayer, 2010; Kriegbaum et al., 2011; Madunic, 2018; Li Santi et al., 2021).

The etiology of a rare hematological disorder termed paroxysmal nocturnal hemoglobinuria (PNH) is the clonal expansion of hematopoietic stem cells carrying somatic loss-of-function mutations in PIGA, which encodes an enzyme pivotal for the biosynthesis of GPI-anchors (Hill et al., 2017; Kinoshita, 2020). The overarching hallmark of the PNH syndrome is that affected blood cells fail to express GPI-anchored proteins on their cell membranes and instead secrete a soluble truncated variant lacking the C-terminal signal sequence for GPI-anchoring. Accordingly, monocytes and neutrophils affected by PNH are deficient in membrane-tethered uPAR (Ploug et al., 1992b) and they secrete a soluble uPAR (Ploug et al., 1992a) leading to sustained elevated plasma levels of soluble uPAR with an average plasma concentration of 120 pM (Ronne et al., 1995; Sloand et al., 2008). Due to the pleiotropic effects of PIGA deficiency, the molecular causality underlying the clinic manifestations of PNH are often complex and incompletely delineated. Notwithstanding this uncertainty, treatment with a neutralizing anti-C5 antibody (eculizumab) alleviates all of the major adverse clinical complications in PNH i.e., intravascular hemolysis, venous thrombosis, renal dysfunction, and pulmonary hypertension (Hill et al., 2017). It is therefore likely that the principal driver of these pathogenic complications in PNH is intravascular hemolysis due to CD59 deficiency and that eculizumab treatment compensates for the impaired complement regulation. Secondary to episodes of intravascular hemolysis, excessive renal resorption of hemoglobin dimers accompanied by heme cytotoxicity may thus drive the development of acute and chronic kidney disease (Kokoris et al., 2018).

Soluble uPAR levels in patients with idiopathic focal and segmental glomerulosclerosis (FSGS) has attracted a lot of attention after Reiser et al., proposed that soluble uPAR was a prime candidate for the elusive serum permeability factor responsible for recurrent FSGS after kidney transplantations (Wei et al., 2008, 2011). Among other findings, they based their proposition on (i) that the elevated plasma levels of soluble uPAR in primary and recurrent FSGS were independent of the estimated glomerular filtration rates (eGFR) and (ii) that glomerular uPAR deposits activated podocyte α_vβ₃ and thereby drove foot process effacement leading to manifest proteinuric glomerular disease. Although this hypothesis initially gained some support, it nevertheless had some inherent weaknesses and it remains highly controversial (Konigshausen and Sellin, 2016; Kronbichler et al., 2016; Saleem, 2018). First, it was questioned whether the increased levels of soluble uPAR in FSGS patients actually were unrelated to decreased eGFR (Meijers et al., 2014; Musetti et al., 2015). Second, the finding that elevated levels of circulating soluble uPAR should be a pathogenic factor per se causing proteinuria and onset of FSGS in mouse model systems could not be reproduced by several independent laboratories (Cathelin et al., 2014; Spinale et al., 2015; Harel et al., 2020). One independent group (Alfano et al., 2015) did, however, report that a single iv injection of 20 μg full-length soluble mouse uPAR induced proteinuria after 24 h, but these studies were conducted in Plaur^–/– mice and therefore have little bearings on clinical FSGS—in common with the original studies by Reiser et al. Third, it is difficult to reconcile a causative pathogenic role of soluble uPAR in humans with the fact that patients with PNH have life-long and marked elevations of their plasma uPAR levels, but the renal dysfunction found in PNH patients can be treated with eculizumab showing that it is secondary to complement-mediated hemolysis (Kokoris et al., 2018).

Adding to the complexity in the tale of the pathogenic effects of soluble uPAR in FSGS, Reiser et al., reported that an alternative splice variant of mouse uPAR was the more potent nephrotoxin compared to the conventional soluble uPAR with intact LU-domains (Wei et al., 2011, 2019). Building on our extensive knowledge from protein structure-function analyses on uPAR, this proposal is very surprising. The alleged protein product from this alternative splice variant would thus comprise the N-terminal LU domain, the linker region and only half of the second LU domain and contain an uneven number of cysteine residues. Generally, the folding of LU domain proteins are very sensitive to mutations of their consensus cysteine residues leading to misfolded and aggregated proteins (Leth et al., 2019a). Notwithstanding these concerns, Reiser et al., reported a low resolution (17 Å) structure of this particular uPAR splice variant based on single particle reconstructions from electron microscopy images (Wei et al., 2019). Whether this structure actually represents the elusive uPAR splice variant in question remains nonetheless unclear to us, since their protein preparations also contained at least one dominating protein contamination of 50–80 kDa (as judged by their SDS-PAGE under reducing conditions). Furthermore, their native PAGE reveals a lot of protein aggregation as would be predicted for this construct from our general knowledge of LU domain protein stability (Leth et al., 2019a). As a consequence, we still consider the existence of such a properly folded truncated uPAR splice variant for highly speculative both in vitro and in vivo.

Chronic rheumatoid arthritis (RA) is a progressive inflammatory disease that ultimately leads to irreversible joint destruction by an invading pannus causing cartilage degradation and bone erosion. Studies on biopsies from arthritic lesions in RA patients by in situ zymography revealed pronounced elevations in uPA activity in the hyperproliferative synovial lining (Busso et al., 1997). Accordingly, these lesions express markedly elevated levels of both uPA and uPAR by infiltrating neutrophils, macrophages, and fibroblast-like cells in the inflamed RA synovium (Almholt et al., 2018) and increased plasma levels of soluble uPAR correlate to disease activity (Slot et al., 1999; Enocsson et al., 2021).

Several mouse models have been developed as experimental surrogates of rheumatoid arthritis in humans; please consult Buckley et al. (2019) for a comprehensive review. One widely used mouse model that mimics non-septic systemic polyarthritis in humans is the autoimmune collagen type-II induced arthritis model (CIA). Elegant genetic dissections, provided strong evidence for the causal role of uPA-mediated plasminogen activation in driving inflammatory CIA. First, Plg^–/– mice were shown to be resilient to CIA and to a cocktail of mouse monoclonal anti-collagen type-II auto-antibodies, but daily administration of purified human plasminogen restored their sensitivity to these treatments (Li et al., 2005). Second, Plau^–/– and Plaur^–/– mice showed none or only very mild symptoms of arthritis after being challenged with collagen type-II (Thornton et al., 2017). In all three genotypes, the adaptive immune response was intact and mounted a similar response toward collagen type-II as did littermate wild-type control mice. With reciprocal bone marrow transplantations, the impact of uPAR to disease incidence and severity was traced to cells originating from the bone-marrow compartment with inflammatory monocyte/macrophages being the likely culprits (Thornton et al., 2017).

Prompted by the beneficial impact of Plau^–/– and Plaur^–/– on the incident and severity of CIA, Almholt et al. (2018) used neutralizing murine monoclonal antibodies toward mouse uPA (mU1) or mouse uPAR (mR1) to treat progression of arthritis in the CIA model. Aligned with the genetic data, inhibition of uPA catalytic activity with mU1 markedly alleviated the pathology and the adverse disease progression of CIA. Unexpectedly, treatment with mR1 showed no beneficial effect, which is unexpected given the beneficial effects observed in either (i) Plaur^–/– mice (Thornton et al., 2017) or (ii) in a gene therapy approach competing the uPA•uPAR interaction with a hybrid molecule between human serum albumin (HSA) and the receptor binding amino-terminal fragment (ATF) of uPA (Apparailly et al., 2002). Although speculative, it is possible that the inhibitory properties of mR1 is inadequate for such long-term treatment studies. The inhibitory mechanism of mR1 is unique in the sense that it traps a partially open conformation of unoccupied uPAR by binding to DI, but in the presence of elevated concentrations of uPA a ternary complex will form and this will drive uPAR into the closed conformation and thereby expel the bound mR1 (Gardsvoll et al., 2011a). The reverse scenario, where mR1 displaces uPAR-bound uPA, is not an option, since mR1 does not recognize the closed conformation of uPAR in the uPA•uPAR complex (Pass et al., 2007).

Position emission tomography (PET) combined with computed tomography (CT) constitute the central imaging platform in clinical oncology (Duclos et al., 2021). The overarching virtue of PET/CT is that it is (i) non-invasive, (ii) provides an overview of the global expression of a given molecular target or function shortly after injection of the radionuclide tracer, (iii) has no issues with sampling bias, and (iv) is highly versatile in the sense that the collection of molecular targets that can be visualized depends only on the availability of a suitable PET probe. Although ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) is the unopposed toiler of PET imaging in nuclear medicine, new peptide-based PET probes are continuously being developed thus widening the range of molecular targets that can be visualized (Askari Rizvi and Zhang, 2021). As alluded to in the previous sections, the increased uPAR-expression in the tumor-stromal microenvironment of invading cancer lesions makes it an attractive target for molecular imaging in the clinical assessment of tumor invasion and metastatic dissemination. As listed in Table 1, a plethora of PET-probes, targeting uPAR via the AE105 moiety, have been designed and explored in preclinical animal models bearing various human cancers (subcutaneous, orthotopic, and patient derived xenografts). These studies include xenografts from different origins (e.g., brain, prostate, lung, and breast) and use PET tracers combining various macrocyclic chelators (DOTA, NOTA, CB-TE2, CB-TE2A-PA) with different radionuclides (⁶⁴Cu, ⁶⁸Ga, Al¹⁹F) - as shown in Table 1 (Price and Orvig, 2014). All studies showed specific uptake in tumor lesions and the uptake levels correlated to uPAR levels as determined by ELISA or immunohistochemistry in resected subcutaneous tumor xenografts (Persson et al., 2012a, 2013a). It should be emphasized that the endogenous expression of mouse uPAR will not be detected in these imaging studies as the AE105 targeting moiety is strictly specific for human uPAR (Ploug et al., 2001). Building on these encouraging preclinical mouse studies, the first phase-1 clinical trials were conducted in human cancer patients with ⁶⁴Cu-DOTA–AE105 (Persson et al., 2015) and ⁶⁸Ga-NOTA–AE105 (Skovgaard et al., 2017). Both studies showed no adverse pharmacological effects and internal radiation burdens were equivalent to that of standard ¹⁹F-FDG PET scans. Importantly, specific tracer uptake was noted in primary lesions as well as in several lymph node metastases and subsequent immunohistochemistry on resected tumor tissue confirmed that all these lesions were indeed uPAR positive. Circumstantial observations from these phase-1 clinical trials demonstrate the power of such global and unbiased uPAR–PET imaging (Figure 3). First, one patient enrolled with an indication of breast cancer revealed an unexpected additional tracer uptake in the brain, which was later diagnosed as a primary meningioma (Persson et al., 2015). Second, another breast cancer patient had a robust tracer uptake in two axillary lymph nodes despite previous preoperative staging by ultrasound, fine-needle biopsy, and CT failed to detect any malignant lymph node involvement (Skovgaard et al., 2017). A small prospective phase-2 clinical trial on ⁶⁸Ga-NOTA–AE105 uptake in the primary prostate lesions of 27 cancer patients found a positive correlation between standard uptake values (SUV_max) and Gleason score based on histopathological grading of biopsy specimens (Fosbol et al., 2021a). In a small follow-up study, baseline SUV_max of ⁶⁸Ga-NOTA–AE105 in bone metastasis index lesions of 14 prostate cancer patients were measured before two cycles of ²²³RaCl₂ therapy. Despite the low number of patients enrolled in this study, baseline SUV_max correlated to overall survival (Fosbol et al., 2021b). Notwithstanding these promising results, large scale clinical trials are needed to precisely define the clinical utility of uPAR-PET scans in cancer patient management.

Due to the short half-lives of traditional positron emitting radionuclides [⁶⁸Ga (67.7 min), ¹⁸F (109.7 min), and ⁶⁴Cu (12.7 h)], the majority of PET probes rely on small molecules or peptides as their targeting moiety due a fast pharmacokinetic profile with rapid contrast development. The clinical implementation of immunotherapy created, however, an unmet need for immuno-PET imaging to stratify and select patients eligible for antibody-based therapy (Liberini et al., 2021). To create a better match with the slow pharmacokinetics of antibodies, most immuno-PET studies use ⁸⁹Zr as positron emitting nuclide with a half-life of 3.3 days. So far this technology has not been applied to uPAR-PET imaging, but was used to image its high-affinity protease ligand uPA in subcutaneous tumors using an ⁸⁹Zr-labeled mouse monoclonal antibody (ATN291) that is specific for human uPA (Yang et al., 2016). It is unclear whether this imaging represents uPA in complex with uPAR on the cells surface of the transplanted human cancer cells. Two different laboratories conducted low resolution imaging with single photon emission computed tomography (SPECT) to visualize uPAR expression in xenografts using the ¹¹¹In-labeled monoclonal anti-huPAR antibodies ATN658 (Boonstra et al., 2015, 2017) and 2G10 (LeBeau et al., 2013, 2014). Both approaches provided clear primary tumor delineations 72 h post tracer iv injection and ¹¹¹In-labeled 2G10 detected occult bone metastasis in mice with triple negative human breast cancer xenografts (LeBeau et al., 2013).

The Holy Grail in cancer surgery is the radical removal of all malignant tissue creating tumor free resection margins while preserving as much of the healthy tissue as possible—ultimately leading to improvements in progression-free survival. While PET/CT and MRI imaging platforms are important for staging and localization of disseminated disease, they cannot easily guide surgical procedures in real-time. In the last decade, design of targeted near-infrared (NIR) probes for fluorescence-guided cancer surgery enabled the rapid development of emerging imaging tools that could potentially assist surgical navigation in real-time (Debie and Hernot, 2019). Although still in its infancy, targeted imaging in fluorescence-guided cancer surgery is gaining momentum especially with antibodies that are already applied in therapy or have a strong biomarker potential. But proof of concept studies demonstrating the added beneficial effects for cancer patient survival are still lacking (Hernot et al., 2019).

As alluded to in the previous sections, baseline expression of uPAR is low and scattered in normal healthy tissues while it is robustly upregulated in most active cancer lesions, particularly in the microenvironment of the invading tumor-stromal interface. This expression pattern makes uPAR an ideal target candidate in fluorescence guided intraoperative imaging and several strategies have accordingly been explored in preclinical mouse models with AE105 as the targeting core (Table 1). Using NIR fluorophores as reporter groups increases the depth of the surgical field that can be visualized due to a higher tissue penetration and a decreased auto-fluorescence of wavelengths in the NIR-I (650–900 nm) and NIR-II (1,000–1,700 nm) window. The first generation of uPAR-specific optical probes exploited the clinically approved fluorophore indocyanine green (ICG) as reporter group (Juhl et al., 2016). Although the conjugation of ICG to AE105 reduced the affinity for uPAR by 20-fold, initial optical imaging studies on subcutaneous xenotransplants of the U87MG glioblastoma cell line (Allen et al., 2016) led to a specific probe uptake reaching maximal tumor-to-background ratio (TBR) after 6–24 h post iv administration (Juhl et al., 2016). The ICG–AE105 conjugate also showed promise in the real-time guidance of preclinical surgery of xenotransplanted orthotropic mouse models of head-and-neck and pancreatic cancers (Christensen et al., 2017; Juhl et al., 2019). Building on these promising preclinical studies, a phase 1–2 clinical trial was launched in 2020 exploring the safety and use of ICG–AE105 in patients with glioblastoma (clinicaltrialsregister.eu, EudraCT 2020-003089-38). In second generation AE105-conjugates, ICG was replaced by fluorophores with more optimal spectral properties i.e., CH1055 (NIR-II) and IRDye800CW (Kurbegovic et al., 2018, 2021). Especially IRDye800CW–AE105 was optimized biochemically and analyzed extensively in a surrogate glioblastoma mouse model with U89MG cells as orthotropic transplants (Figures 4A,B) (Kurbegovic et al., 2021). Testing different linker sequences, the IRDye800CW-linker³–AE105 conjugate was selected (Table 1), which binds uPAR with only 3-fold lower affinity as compared to unconjugated AE105. Dependent on the amount of probe administered, maximal TBRs of 6.6 to 7.0 were reached within 1–3 h post iv injection. The fast kinetics in generating optimal contrast is remarkable and is most likely a consequence of the relatively small size and hydrophilicity of this probe as compared to ICG-AE105 (more hydrophobic) and the larger antibody based probes, where optimal contrast for the latter typically requires a much longer probe washout (3–5 days).

Besides peptide-based targeting of uPAR—the topic of this review—several macromolecular uPAR-targeting moieties have been developed for optical imaging and therapeutic targeting including (i) monoclonal anti-uPAR antibodies (Figure 4C; LeBeau et al., 2014; Boonstra et al., 2015; Baart et al., 2021) and (ii) various nanoparticles carrying receptor binding fragments of uPA (ATF) or peptide derivatives (AE105) (Hansen et al., 2007; Yang et al., 2013; Gao et al., 2017; Zuo et al., 2020). Due to the inherently long “washout times” (days rather than hours) needed for these probes to reach maximal TBR, they are probably less suited in the daily clinical workflow as imaging modalities, but they are optimally suited for targeted delivery of cytotoxic payloads.

In the last two decades, a variety of different targeted intervention strategies have been developed to eradicate uPAR-expressing cells. Specificity of the cytotoxic insult were devised by targeting uPAR directly or by exploiting the proteolytic activity of uPAR-bound uPA (Liu et al., 2003; Rustamzadeh et al., 2007; Schafer et al., 2011; Morodomi et al., 2012; LeBeau et al., 2013; Jing et al., 2017; Harel et al., 2019; Amor et al., 2020; Zuppone et al., 2020). Several of these treatment modalities have shown promising results in preclinical mouse models bearing human cancer cell xenografts. One small study showed beneficial effects on canine oral mucosal melanomas inducing stable disease in all five dogs treated with uPA- and MMP2-activated anthrax toxin (Nishiya et al., 2020). But none of these modalities have so far entered clinical trials in humans. We predict that the non-invasive uPAR–PET imaging platforms discussed in this review will greatly facilitate the future clinical translation of a given uPAR-targeted therapy regimen—assisting in both initial patient selection based of target availability as well as response marker monitoring treatment efficacy.