Tissue Plasminogen Activator in Central Nervous System Physiology and Pathology: From Synaptic Plasticity to Alzheimer’s Disease

Abstract

Tissue plasminogen activator’s (tPA) fibrinolytic function in the vasculature is well-established. This specific role for tPA in the vasculature, however, contrasts with its pleiotropic activities in the central nervous system. Numerous physiological and pathological functions have been attributed to tPA in the central nervous system, including neurite outgrowth and regeneration; synaptic and spine plasticity; neurovascular coupling; neurodegeneration; microglial activation; and blood-brain barrier permeability. In addition, multiple substrates, both plasminogen-dependent and -independent, have been proposed to be responsible for tPA’s action(s) in the central nervous system. This review aims to dissect a subset of these different functions and the different molecular mechanisms attributed to tPA in the context of learning and memory. We start from the original research that identified tPA as an immediate-early gene with a putative role in synaptic plasticity to what is currently known about tPA’s role in a learning and memory disorder, Alzheimer’s disease. We specifically focus on studies demonstrating tPA’s involvement in the clearance of amyloid-β and neurovascular coupling. In addition, given that tPA has been shown to regulate blood-brain barrier permeability, which is perturbed in Alzheimer’s disease, this review also discusses tPA-mediated vascular dysfunction and possible alternative mechanisms of action for tPA in Alzheimer’s disease pathology.

Introduction

The serine protease tissue-plasminogen activator (tPA), classically known for its role in fibrinolysis, is also expressed on the abluminal side of the vasculature in the central nervous system. Seminal in vivo studies demonstrated that brain tPA is upregulated following activity-dependent events in the hippocampus and cerebellum, suggesting a physiological role for tPA in synaptic plasticity and learning and memory^{1, 2}. These reports, however, were followed by others that showed a pernicious role for tPA in neurodegeneration, microglial activation, and blood-brain barrier (BBB) permeability^3–5. Moreover, a variety of plasminogen-dependent and plasminogen-independent substrates and receptors have been reported to effectuate tPA’s actions in these processes, including N-methyl-D-aspartate receptors (NMDAR), platelet-derived growth factor-CC (PDGF-CC), low-density lipoprotein (LDL) receptor-related protein (LRP1), and pro-brain derived neurotrophic factor (pro-BDNF)^5–8. To contextualize these varied reports, this review uses learning and memory as a paradigm to examine and analyze tPA’s actions in the central nervous system. It summarizes the foundational studies implicating tPA in synaptic plasticity and synaptic transmission, and it re-examines those studies considering more recent reports demonstrating a role for tPA in Alzheimer’s disease (AD) pathogenesis.

AD is a degenerative brain disease characterized by amyloid-β (Aβ) plaques, tau neurofibrillary tangles, and progressive cognitive decline⁹. Though Aβ and tau are hallmark pathologies of AD, accumulating evidence indicates that vascular dysfunction also plays a significant contributory, if not causal, role in disease progression¹⁰. Indeed, more contemporary studies in AD patients and in AD mouse models demonstrate that targeting vascular pathologies slows the clinical and pathological manifestations of AD^{11, 12}. Accordingly, this review discusses the putative roles for tPA in AD, including Aβ clearance and neurovascular coupling, and how tPA provides a possible mechanistic link between Aβ and vascular pathology. In addition, it proposes alternative pathways through which tPA could act on the vasculature to promote AD pathology.

Role of tPA in synaptic transmission and synaptic plasticity

Upregulation of tPA following activity-dependent events

A role for tPA in learning and memory was first suggested when tPA was shown to be an immediate-early gene that was induced by neuronal activity¹. In that study, tPA was identified in a differential screen of 30,000 clones from a complementary DNA library; using three activity-dependent events – seizure, long-term potentiation (LTP), and kindling – tPA gene expression was upregulated in vivo within 1 hour in the granule and pyramidal cell layers of the hippocampus in the adult rat brain. Consistent with these data, tPA was found to be induced in the cerebellum of rats after learning a complex motor task². Using a pegged runway of regular and irregular patterning to test cerebellar-dependent motor learning, tPA mRNA expression was upregulated in Purkinje cells during the most active phase of learning. These spatial and temporal data suggested that tPA may have a role in learning and memory by regulating neuronal plasticity.

Modulation of basal synaptic transmission by tPA

Further studies deconstructing tPA’s role in hippocampal plasticity ex vivo, however, have yielded somewhat varied results. While some groups have observed defects in basal synaptic transmission¹³ and early¹⁴ and late LTP (L-LTP) in the hippocampal CA1 region of tPA^−/− mice, others have not^{15, 16}. Frey and colleagues¹³ were the first to report deficits in synaptic efficacy at the Schaffer collateral-to-CA1 synapse in tPA^−/− mice under basal conditions. In tPA^−/− mice, a larger stimulus was needed to evoke synchronous somatic action potentials (so called “pop-spike”) of similar amplitude to that seen in wild-type mice. In addition, paired-pulse facilitation of the pop-spike was significantly reduced in the tPA^−/− mice, while paired-pulse facilitation of the field excitatory post-synaptic potential was largely unaltered. These results suggest that CA3-CA1 synapses in tPA^−/− mice are under enhanced γ-Aminobutyric acid (GABA)-ergic inhibition, especially since increased facilitation was observed when slices were treated with the GABA_A receptor competitive antagonist bicuculline. Consistent with the hypothesis that tPA^−/− mice have altered GABAergic transmission, no differences in L-LTP between tPA^−/− mice and wild-type mice were found, but significant differences were observed when GABAergic transmission was blocked with the noncompetitive GABA_A channel blocker picrotoxin¹³.

More recently, it was proposed that tPA modulates pre-synaptic transmission through its effects on synaptic vesicle cycling¹⁷. From immunoblots of neuronal membrane extracts and isolated synaptic fractions from synaptoneurosomes, tPA treatment was found to recruit βII-spectrin, a cytoskeletal protein implicated in synaptic vesicle release, to the active zone and induce the binding of synaptic vesicles to βII-spectrin. Treatment with tPA also induced phosphorylation of synapsin I, a synaptic vesicle membrane protein, presumably via tPA-mediated increases in voltage-gated calcium channel expression. In keeping with these expression and localization studies, tPA treatment was shown to increase the frequency of miniature excitatory post-synaptic currents in CA1 pyramidal neurons from rat brain slices¹⁷. Mechanistic studies investigating whether tPA increases the release probability of individual synaptic vesicles or the number of synaptic vesicles released were not performed. Moreover, it is unclear if this increase in miniature excitatory post-synaptic current frequency is due to tPA-mediated increases in the expression/recruitment of presynaptic vesicle cycling proteins to the active zone. Further studies, therefore, are necessary to determine if these reported effects of tPA are linked and if they are responsible for the defects in basal synaptic transmission first observed in tPA^−/− mice¹³.

Alternatively, in experiments utilizing neuronal cortical cultures, tPA treatment was shown to increase p-35 cyclin-dependent kinase-5/Ca²⁺ calmodulin-dependent kinase IIα (Cdk5/CaMKIIα) activation, and attenuate NMDA-induced decreases in post-synaptic density-95 and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) levels^{18, 19}. Changes to post-synaptic AMPAR have long been known to contribute to synaptic efficacy and synaptic plasticity²⁰. Though, roscovitine, a broad-range purine inhibitor of multiple cyclin-dependent kinases, including Cdk5, appeared to abrogate the effects of tPA treatment¹⁸, more mechanistic studies are needed to demonstrate if the observed effects of tPA are within the same pathway and if they cause changes in synaptic transmission. Moreover, decreased, not increased, Cdk5 activity has been reported to be responsible for NMDAR-induced AMPAR endocytosis²¹, further highlighting the importance of future experiments that directly test tPA-mediated Cdk5 signaling and its potential downstream effects on synaptic transmission.

Loss of tPA results in impairments in long-term potentiation

Despite these studies, however, other groups have reported no differences in basal synaptic transmission in tPA^−/− mice^14–16. Specifically, these reports show no differences in basal synaptic transmission in the CA1 hippocampal region of tPA^−/− compared to wild-type mice, but significant defects in L-LTP, with or without blocking GABAergic transmission^14–16. It has also been shown that maintenance of L-LTP at the Schaffer collateral-to-CA1 synapse can be diminished by application of the protease inhibitor tPA-Stop²². Conversely, potentiation can be enhanced when tetanization is coupled with either pharmacologic¹⁶ or genetic²³ increases in tPA. It was further demonstrated that this potentiation in CA1 requires activation of the cAMP/protein kinase A (PKA)-mediated pathway^{15, 22}, as analogs of cAMP and activators of cAMP/PKA-dependent signaling cascades can induce L-LTP in wild-type, but not tPA^−/− mice. Consistent with this, the tPA gene, Plat, has previously been shown to house a cAMP responsive element promoter²⁴.

The endocytic signaling receptor, LRP1, on neurons has also been implicated in mediating tPA’s actions in L-LTP¹⁶. Blocking LRP1 with an inhibitor, receptor-associated protein (RAP), caused deficits in L-LTP that were similar to what was reported previously for tPA^−/− mice. Moreover, RAP blocked Schaffer collateral-to-CA1 synaptic potentiation in tPA^−/− hippocampal slices that had been treated with tPA. PKA, a kinase known to play a key role in the induction and maintenance of L-LTP²⁵, was also shown to be activated upon tPA binding to LRP1. Together, these data support a LRP1-mediated mechanism of action for tPA in L-LTP. In addition, they suggest that tPA’s protease activity is not necessary for the full expression of L-LTP, as proteolytically active tPA is not required to interact with and signal through LRP1²⁶.

While the involvement of LRP1 in tPA-mediated L-LTP implicates tPA as having a plasminogen-independent mechanism of action, tPA/plasmin-induced cleavage of BDNF has also been demonstrated to be essential for long-term hippocampal plasticity⁸. Previously, genetic or pharmacologic ablation of BDNF or its receptor TrkB was shown to inhibit L-LTP^27–30. It was also previously demonstrated that secreted proBDNF can be cleaved extracellularly by plasmin into mature BDNF (mBDNF)³¹. It was not known, however, if the tPA/plasmin/mBDNF pathway formed a common pathway important for hippocampal L-LTP.

Through a series of L-LTP experiments using tPA^−/− mice, plasminogen deficient mice, and mice heterozygous for BDNF, it was demonstrated that tPA/plasmin-mediated cleavage of proBDNF is critical for the full expression of L-LTP in the CA1 region of the mouse hippocampus⁸. Subcellular localization studies support these functional findings. In cultured hippocampal neurons transfected with BDNF-mCherry and tPA-EYFP vectors, BDNF and tPA were shown to be co-packaged in presynaptic dense core vesicles³², suggesting that tPA and BDNF are proximally localized and likely concomitantly released to act through a common pathway.

Interestingly, while these experiments were focused upon long-term plasticity in the CA1 region of the hippocampus, both BDNF and tPA have been shown to be most highly expressed in giant mossy fiber boutons in the mossy fiber pathway in the CA3 region of the hippocampus^33–36. In addition to en passant terminals and filopodial extensions, giant mossy fiber boutons are one of three identified presynaptic specializations that emanate from the mossy fiber axons of dentate granule neurons^{37, 38}. The subcellular localization of both BDNF and tPA within giant mossy fiber boutons, therefore, not only indicates that these two proteins are proximally localized to act in concert but suggests that their role in synaptic plasticity is highly compartmentalized and possibly unique to mossy fiber-to-CA3 synaptic plasticity. While deficits in L-LTP have been reported in tPA^−/− mice at the mossy fiber-to-CA3 cell synapse^{15, 22}, functional consequences of tPA/plasmin-mediated mBDNF generation have not been tested in the CA3 region specifically.

Behavioral consequences of tPA deficiency

Mice deficient in tPA display defects in avoidance behavior

Given the deficits in hippocampal L-LTP in tPA^−/− mice, it was thought that they would exhibit comparable behavioral deficits in hippocampal-dependent learning and memory tasks. While there is conflicting data on whether loss of tPA causes overt cognitive defects^{14, 15, 39}, tPA^−/− mice have been consistently found to have impairments in avoidance tests^{14, 15, 39}. Avoidance tests require the mouse to learn to avoid an adverse stimulus, usually foot shock. Depending on the experimental design, avoidance tests can reveal differences in acquisition, working memory, consolidation, and long-term recall⁴⁰. In addition, methods by which the adverse stimulus is presented to the mouse - in an active or passive way - can alter the cognitive difficulty and effect the type of learning being tested and brain structures involved.

Mice lacking the tPA gene have been tested in both two-way shuttle box active avoidance^{15, 22} and step-down passive avoidance³⁹. In a two-way shuttle box active avoidance test, mice are trained to avoid a foot shock upon the presentation of a light cue. It is inferred that mice have successfully learned the task if they associate the light cue with the oncoming foot shock and avoid the shock by moving to the neighboring compartment. Using this paradigm, tPA^−/− mice were found to have significant defects in their ability to avoid the aversive stimulus. Similarly, in a step-down passive avoidance test, mice are placed on a raised platform and learn to not step-down onto the lower platform, which is formed by an electrical grid, thereby avoiding a foot shock. In the step-down passive avoidance test, tPA^−/− mice had significantly shorter latencies to step-down than wild-type mice. Taken together, these studies suggest that tPA^−/− mice have deficits in acquisition or working memory required for aversive associative learning. However, this interpretation is complicated by the role tPA has been shown to have in the amygdala function and anxiety-like behavior⁴¹ (see next section).

Key role for tPA in anxiety-like behavior

Mice deficient in tPA have been found to be resistant to stress-induced anxiety, with no changes in basal anxiety^{39, 42, 43}. Basal anxiety was tested using elevated-plus maze, which takes advantage of a rodent’s natural aversion to open spaces and preference for closed spaces; tPA^−/− mice were found to spend equivalent time in the open arms compared to wild-type mice. However, when tPA^−/− and wild-type mice were first subjected to bouts of chronic restraint⁴² or injections of corticotropin-release factor⁴³, tPA^−/− mice did not exhibit the expected stress-induced anxiety-like behavior (significantly less time in open arms) when compared with wild-type mice. Moreover, the increased stress-induced anxiety-like behavior observed in wild-type mice correlated with increased tPA activity in the central and medial amygdala, but not the basolateral amygdala⁴². In line with this phenotype and in-keeping with tPA promoting anxiety-like behavior, intraventricular injections of the stress hormone, corticotropin-releasing factor, were found to upregulate tPA activity in the central and medial amygdala⁴³.

The molecular mechanism for tPA’s role in facilitating anxiety-like behavior points to plasminogen-independent neuronal remodeling, with evidence demonstrating that: 1) plasminogen null mice do not phenocopy tPA^−/− mice⁴², 2) signaling cascades implicated in spine plasticity⁴⁴ are upregulated in wild-type, but not tPA^−/− mice⁴², and 3) tPA^−/− mice have significantly attenuated stress-induced spine retraction in the medial amygdala compared to wild-type mice⁴⁵. Given the behavioral data, it is possible that the impairments observed in tPA^−/− mice in avoidance tasks are due to tPA’s effect on stress-induced neuronal plasticity in the amygdala and not a direct effect of tPA on the neural substrates that underlie hippocampal-dependent learning and memory. Recognizing the importance of stress-susceptibility on behavioral performance⁴⁶, future studies that employ task-independent stressors to test learning and memory ⁴⁷ would be informative in discriminating between tPA’s role in stress versus learning and memory.

Role of tPA in Alzheimer’s disease pathogenesis

Mechanisms of tPA-mediated neuropathology and cognitive deficits

Despite the cumulative work demonstrating that tPA modulates synaptic transmission and plasticity, most studies of tPA in AD have not focused on synaptic pathology; though one group has shown that inhibition of plasminogen activator inhibitor-1 (PAI-1), an endogenous inhibitor of tPA, ameliorates hippocampal LTP deficits in the amyloidogenic Tg2576 AD mouse model⁴⁸ (See Table 1 for a summary mouse models of AD-related neuropathology discussed in this review). In fact, most studies on tPA’s role in AD pathogenesis have concentrated on tPA-mediated clearance of Aβ plaques^49–55. One of the hallmark neuropathologies associated with AD is the deposition of Aβ plaques in the brain parenchyma and cerebrovasculature; an imbalance in amyloid precursor protein (APP) production and Aβ clearance is thought to contribute to AD pathogenesis⁵⁶.

Table 1.

Mouse models of Alzheimer’s disease-related neuropathology

Mouse model	Genes	Mutations	Genetic background	Primary reference	References
5×FAD	APP, PSEN1	Human APP770 KM670/671NL (Swedish), APP I716V (Florida), APP V717I (London), PSEN1 M146L (A > C), PSEN1 L286V	C57BL/6 × SJL	135	12, 132
J20 FAD	APP	Human APP770 KM670/671NL (Swedish), APP V717F (Indiana)	C57BL/6	136	86
APP/PS1	APP, PSEN1	Human APP770 KM670/671NL (Swedish), PSEN1:deltaE9	C57BL/6 × C3H	137	78, 85, 108
APPs1	APP	Human APP751 isoform KM595/596NL (Swedish), V642I (London)	C57BL/6 × CBS	138	72
Tg2576	APP	Human APP770 KM670/671NL (Swedish)	B57BL/6 × SJL	139	48, 53, 55, 77, 79
Tg4510	MAPT	Human MAPT P301L	129S6 × FVB	140	73
PSAPP	APP, PSEN1	Human APP770 KM670/671NL (Swedish), PSEN1 M146L (A > C)	B57BL/6 × D2 × Swe × SJL	141	48

APP, amyloid precursor protein; FAD, familial Alzheimer’s disease; PS, presenilin 1; PS1, presenilin 1; PSEN1, presenilin 1; MAPT, microtubule-associated protein tau.

Through its proteolytic activity, the tPA/plasmin system has been shown to promote α-secretase cleavage of APP⁵⁰, shuttling the APP production pipeline toward the non-amyloidogenic pathway, and degrade Aβ in vitro^48–52. Moreover, Aβ appears to act like a fibrin(ogen)-mimic, catalyzing plasminogen activation by tPA, and stimulating tPA activity via feedforward mechanisms in vitro^57–60.

Other studies have focused on tPA’s interaction with LRP1 to modulate Aβ uptake and influence AD pathogenesis. Immunohistochemistry of LRP1 and tPA, along with other LRP1 ligands like apolipoprotein E, α−2 macroglobulin, urokinase plasminogen activator, PAI-1, lipoprotein lipase, and lactoferrin, have been observed in Aβ plaques from human AD brain tissue⁶¹. Though no direct studies have investigated the effects of tPA on LRP1-mediated clearance of Aβ, other LRP1 ligands, like α−2 macroglobulin and lactoferrin, have been shown to enhance clearance of soluble Aβ in vitro⁶². Indeed, LRP1 is a well-established contributor to AD pathology. It is expressed on neurons, microglia, astrocytes, endothelial cells, choroid plexus epithelial cells, pericytes, and vascular smooth muscle cells, where it mediates Aβ uptake and trafficking to the lysosome for degradation or across the BBB for clearance^63–70.

This apparent protective effect of tPA/plasmin activity on direct Aβ degradation or LRP1-mediated clearance, however, appears to be blunted in vivo by upregulated expression of PAI-1. PAI-1, which increases with age and diseases of the central nervous system in humans⁷¹, is also upregulated in amyloidogenic AD mouse models, including Tg2576⁵³, APPs1⁷², and PSAPP⁴⁸, and the tauopathy AD mouse model Tg4510⁷³. Concomitant with the increased expression of PAI-1 is a reduction in tPA activity, but not mRNA expression or total protein, in vivo⁷², suggesting a post-translational inhibitory mechanism of action. This seeming discrepancy between the in vitro and in vivo effects of the tPA/plasmin pathway may be due to the in vivo microenvironment (e.g., chronic inflammation), comorbidities (including age), and/or genetic and lifestyle factors that influence AD susceptibility^74–76. Any one or more of these factors, therefore, may alter the ability of the tPA/plasmin system to exert regulatory control over Aβ aggregation and AD pathogenesis.

The working model presented in Figure 1 is supported by in vivo evidence from PAI-1^−/− and tPA^−/− mice that have been crossed with the amyloidogenic APP/PS1 and Tg2576 AD mice, respectively^{54, 55}, and from pharmacologic studies that show activation of the tPA/plasmin system with PAI-1 inhibitor treatment^{48, 77, 78}. In brain homogenates from APP/PS1/PAI-1^−/− mice, both tPA and plasmin activity were shown to be elevated when compared with APP/PS1/PAI-1^+/− mice (birth and survival rates for APP/PS1/PAI-1^+/+ mice were too low for comparison), and this increased tPA/plasmin activity was associated with reduced Aβ burden⁵⁴. Similarly, when Tg2576 mice were crossed with tPA^−/− mice, Tg2576/tPA^+/− mice were found to have greater Aβ burden than their Tg2576/tPA^+/+ controls (birth and survival rates for Tg2576/tPA^−/− mice were too low for comparison)⁵⁵.

Figure 1. Working model of tPA-mediated pathways in AD pathogenesis.

This working model proposes that risk factors, including, age, genes, inflammation, and lifestyle, promote an environment of high susceptibility to AD development. One or more of these risk factors can initiate a cascade of pathological responses involving tPA that culminate in cognitive deficits. This cascade of tPA-mediated pathways, within a high AD susceptibility milieu, is initiated by the pathological interplay of elevated (plus sign) Aβ and vascular dysfunction, which contribute to increased levels (plus sign) of PAI-1. Thick black, dotted arrows represent physiological or pathological processes upregulated by increased levels of tPA activity, and thin black, dotted arrows represent physiological or pathological pathways downregulated by decreased levels of tPA activity. Gray, solid arrows represent the effect of tPA and/or plasmin on the corresponding protein or process. Abbreviations: Aβ, amyloid-β; PAI-1, plasminogen activator inhibitor-1; tPA, tissue plasminogen activator; LRP1, low-density lipoprotein receptor-related protein 1; MAC1, macrophage-1 antigen; PDGF-CC, platelet-derived growth factor-CC; NMDAR; N-methyl-D-aspartate receptor; NR2B, NMDAR subtype 2B; PDGFRα, platelet-derived growth factor receptor α; nNOS, neuronal nitric oxide synthase; NO, nitric oxide; mBDNF, mature brain-derived neurotrophic factor; BBB, blood-brain barrier; APP, amyloid precursor protein.

Reports on Aβ clearance via activation of the tPA/plasmin system with pharmacologic inhibition of PAI-1 are variable in AD mouse models: some show reduced Aβ burden, while others show no change^{48, 77–79}. It’s likely that differences in detection method (enzyme-linked immunosorbent assay [ELISA] versus immunohistochemical staining), tissue sample (whole brain homogenates versus cortical and/or hippocampal sections), and form of angiopathy (insoluble versus soluble and cerebral amyloid angiopathy versus parenchymal Aβ) are responsible for the discrepancy. Moreover, it’s possible that off-target effects of PAI-1 inhibition, not increased tPA/plasmin activity, are responsible for the increased Aβ degradation and clearance in AD mice with inhibitor treatment. Interestingly, oral administration of TM5275, a small molecule PAI-1 inhibitor, in APP/PS1 mice upregulated LRP1 expression on what appeared to be endothelial cells, though co-localization studies were not performed to determine cell type specificity⁷⁸.

Consistently, though, the homeostatic state of the tPA/plasmin system correlates with cognitive ability in AD mouse models. In the Morris water maze, Tg2576/tPA^+/− mice spent less time in the target quadrant during the probe trial than their Tg2576/tPA^+/+ controls, suggesting a deficit in their ability to form a spatial memory of platform location⁵⁵. Tg2576/tPA^+/− mice also had lower latency times than Tg2576/tPA^+/+ controls in a passive avoidance task with retention test, indicating impairments in their long-term (> 24 hours) memory recall. The learning in both the Morris water maze and the passive avoidance task used by Oh et al⁵⁵ is motivated by aversive stimuli. Prior studies, however, have demonstrated tPA^−/− mice to be resistance to stress-induced anxiety-like behavior^{42, 43}. Moreover, multiple studies have established the importance of stress on behavioral performance^{80, 81}; indeed, performance across a range of behavioral domains - valence, arousal, and cognition – (assessed using open-field, and object and social exploration tasks; reward-seeking operant learning tasks; and the Morris water maze) has been shown to be impaired in mice that are genetically susceptible to stress⁴⁶. Future cognitive tasks, therefore, like novel-object recognition, that are confounded to a lesser degree by stress and are appropriately controlled for, should be considered when assessing cognition in tPA^−/− mice. This is especially true for cognitive tasks that are trying to determine the relative contribution of tPA versus AD pathology in learning and memory.

Nonetheless, in agreement with these genetic studies, AD mice treated with pharmacologic inhibitors of PAI-1 have shown improvements in hippocampal-dependent cognitive tasks. Inhibiting PAI-1 appeared to ameliorate deficits in contextual fear conditioning (Tg2576 mice, PAZ-417 inhibitor [Aleplasinin]⁴⁸), Morris water maze (Tg2576 mice, PAI-039 inhibitor [Tiplaxtinin]⁷⁷; APP/PS1 mice, TM5275 inhibitor⁷⁸), and novel object recognition (Tg2576 mice, PAI-039^{77, 79}). Some of these studies implicated reduced Aβ burden from ELISA⁴⁸ or immunohistochemistry⁷⁸ analysis of amyloidogenic AD mice treated with PAI-1 inhibitors (via increased tPA/plasmin-mediated clearance) as causal for the observed cognitive improvements^{48, 78}. While other studies, despite observing decreases in Aβ plaque deposition⁷⁷ or cerebral amyloid angiopathy⁷⁹, in amyloidogenic AD mice treated with PAI-1 inhibitors, implicated pathways more ancillary to Aβ pathology (e.g. BDNF⁷⁷ or cerebrovascular⁷⁹). In human patients with AD, it is highly debated whether Aβ is uniquely responsible for cognitive decline or whether a spectrum of insults – Aβ-dependent and Aβ-independent – is responsible⁸²; clinical trials investigating drugs that target Aβ have largely failed⁸³ and the recent approval of the Aβ-directed monoclonal antibody aducanumab (Aduhelm) by the US Food and Drug Administration is controversial⁸⁴. Given the current treatment landscape for patients with AD, it’s critical to explore less researched, though viable hypotheses, on AD pathogenesis for future drug development.

One of those hypotheses involves the tPA/plasmin/BDNF pathway, and while evidence is currently limited, it is well-supported for future studies into learning and memory physiology and AD pathogenesis (tPA-mediated cerebrovascular contributions to AD pathogenesis are discussed in detail in a later section). Indeed, the mossy fiber-to-CA3 synaptic circuit (where tPA and BDNF are highly expressed in mossy fiber boutons and where tPA has been shown to modulate synaptic plasticity) is of particular interest as a more recent study demonstrated that forms of neuronal plasticity at the mossy fiber-to-CA3 synapse were impaired in APP/PS1 mice⁸⁵. Alterations to mossy fiber structure and organization have also been reported in the J20 FAD amyloidogenic AD mouse model, along with synaptic site changes to the thorny excrescences of CA3 dendritic spines⁸⁶. In addition, functional magnetic resonance imaging has revealed functional and structural changes to the CA3/dentate gyrus region in patients with mild cognitive impairment and AD. While AD patients show decreased activity in CA3/dentate gyrus⁸⁷, patients with mild cognitive impairment were more likely to have a hyperactive signal in CA3/dentate gyrus, indicating dysfunctional circuitry during disease progression^87–89; CA3/dentate gyrus morphology (shape and volume) was also altered in patients with mild cognitive impairment compared to control patients⁸⁸. It is thought that the cellular physiology underlying CA3 circuitry in the hippocampus is critical for rapid associative learning and contributes to episodic memory⁹⁰ and impairments in episodic memory are one of the distinctive clinical manifestations of AD in patients⁹¹. The effect of the tPA/plasmin/mBDNF cascade on regulating plasticity at the mossy fiber-to-CA3 synapse under physiological conditions, and in the context of AD pathology, remains to be investigated.

Mechanisms of tPA-mediated vascular dysregulation

In addition to the tPA/plasmin/BDNF pathway, there is evidence to suggest that tPA’s role in AD pathogenesis is not related to the clearance of Aβ, but rather, mediated by a plasminogen-independent neurovascular coupling pathway. A role for tPA in neurovascular coupling was demonstrated when mice deficient in tPA, but not plasminogen, were shown to have a reduced functional hyperemia response in the whisker barrel cortex following whisker stimulation⁹². This response in tPA^−/− mice could be restored with topical recombinant tPA treatment to the cortical surface⁹². As the NR2B subunit of the NMDAR is functionally coupled to neuronal nitric oxide synthase (nNOS) it was hypothesized that modulation by tPA altered nNOS-dependent nitric oxide (NO) synthesis and, in turn, cerebral perfusion. To test if this mechanism was responsible for tPA’s effects on blood flow, the cell permeable peptide inhibitor NR2B9c was used to uncouple NMDAR activity from NO production. With NR2B9c treatment, wild-type mice had an attenuated cerebral blood flow response to whisker stimulation, and recombinant tPA no longer rescued the functional hyperemia response in tPA^−/− mice.

Given emerging data which suggest a strong contributory, if not causal, role for vascular dysfunction in AD^{93, 94}, tPA’s actions on vascular function were recently investigated in Tg2576 mice⁷⁹. Similar to tPA^−/− mice, Tg2576 mice were found to have an attenuated cerebral blood flow response following whisker stimulation; this attenuated response could be restored with topical application of recombinant tPA. Altered tPA-NMDAR-NO signaling/production was implicated as Tg2576 mice were revealed to have reduced tPA activity and a diminished cerebral blood flow response to NMDA treatment. In addition, dissociated cortical neurons from Tg2576 mice also exhibited enhanced NO production upon co-treatment of NMDA and recombinant tPA, but no response to NMDA treatment alone. Moreover, acute or chronic inhibition of PAI-1, whose activity is upregulated in 3- to 4-month-old Tg2576 mice, with the PAI-1 inhibitor PAI-039 rescued the whisker stimulation-induced cerebral blood flow response in Tg2576 mice, and ameliorated the cognitive deficits typically observed in aged (11- to 12-month-old) Tg2576 mice. Finally, PAI-1^−/− mice treated with Aβ_1–40 phenocopied the Tg2576 mice treated with PAI-039. Though these data suggest a vasoregulatory role for tPA in AD, chronic PAI-1 inhibition with PAI-039 was also shown to reduce Aβ_1–40 levels in the cortex and hippocampus and cortical cerebral amyloid angiopathy in Tg2576 mice. It is unclear, therefore, if the improved cognition and vascular responses with PAI-1 inhibition are a result of reduced Aβ burden or enhanced tPA modulation of the NMDAR/nNos/NO signaling pathway.

These results are further complicated by what appear to be underlying vascular or neurovascular deficiencies innate to the Tg2576 mice in particular and/or AD pathology in general. When treated with topical application of acetylcholine, Tg2576 mice exhibited a decreased cerebral blood flow response compared to their wild-type littermate controls⁷⁹. Though acetylcholine treatment was intended to test endothelium-dependent vasodilation, acetylcholine receptors are not only present on endothelial cells, but vascular smooth muscle cells and astrocytes as well^95–97. Thus, topical treatment of acetylcholine, rather than intravenous infusion, abrogated the ability to test the role of endothelial cells specifically. As such, it is unclear if the deficit in vasoactivity from acetylcholine is specific to the endothelium, or more broadly indicative of altered cholinergic signaling and vascular dysregulation in Tg2576 mice. Moreover, as the authors themselves indicate, Aβ treatment is known to diminish acetylcholine-mediated vasodilation⁹⁸. Indeed, cholinergic dysregulation is a well-established aspect of AD pathology⁹⁹. Accordingly, potential changes in the molecular programming of acetylcholine receptors or other receptors/signaling pathways because of altered cholinergic innervation may obfuscate experimental results and make determination of casual actors difficult. Finally, we note that while acetylcholine has been reported to mediate tPA release from the endothelium, the concentrations required for this effect^{100, 101} are significantly higher than those reported by Park et al⁷⁹.

Changes in receptor expression and signaling are also important to consider with respect to the NMDAR, the proposed substrate and mediator of tPA’s actions. Like the deficits observed with acetylcholine treatment, there is an attenuated cerebral blood flow response in Tg2576 mice upon NMDA treatment. Park et al⁷⁹ implicated reduced tPA activity for this attenuation, but it is also possible that altered NMDAR signaling is responsible⁹². Numerous groups have reported that Aβ downregulates NMDAR surface levels and impairs NMDAR function^102–104. Moreover, reverse transcription-PCR analysis of AD brains revealed significantly lower mRNA levels of NMDAR-NR2B, the subunit thought to mediate tPA’s actions^{105, 106}. Complicating this narrative is the fact that Tg2576/tPA^+/− mice have decreased protein levels of the NR2B subunits⁵⁵, suggesting that tPA heterozygosity or that Tg2576 mice themselves display changes in NMDAR subunit expression. It is possible, therefore, that chronic Aβ burden or other underlying molecular changes associated with AD pathology, like altered acetylcholine receptors or NMDAR levels, establish a condition of vascular dysfunction in Tg2576 mice, possibly driving a proportion of the neurovascular coupling deficit attributed to tPA. Future studies examining tPA-mediated NMDAR/nNos/NO signaling should control for changes in NMDAR or acetylcholine receptor expression to determine tPA’s relative contribution to regulating neurovascular coupling.

Interestingly, circulating tPA has also been shown to play a part in potentiating the neurovascular coupling response induced by stimulation of the somatosensory cortex¹⁰⁷. Like Park et al⁹², Anfray and colleagues¹⁰⁷ also found tPA^−/− mice to have a diminished neurovascular coupling response, albeit to a lesser degree. However, administration of intravenous recombinant tPA, not topical administration to the cortical surface, restored the cerebral blood flow response in tPA^−/− mice to a level observed in 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES)-injected wild-type mice following whisker stimulation¹⁰⁷. Though the effects of circulating tPA on neurovascular coupling weren’t investigated in an AD mouse model, another group has shown that chronic intravenous administration of human recombinant tPA (Activase) in APP/PS1 mice correlates with reduced Aβ burden and improved cognition¹⁰⁸. It was suggested from in vitro assays that the effects of recombinant tPA were mediated through LRP1 on microglia to enhance mobility and phagocytosis¹⁰⁸. As tPA has been shown to cross a diseased BBB (e.g. cerebral ischemia) in vivo¹⁰⁹, and an intact BBB in vitro via LRP1¹¹⁰, additional pharmacologic or genetic studies would be informative in discerning circulating tPA’s mechanism of action on neurovascular coupling and AD pathology.

Cumulatively, evidence for tPA-mediated changes in vascular tone is intriguing, especially in the context of Aβ clearance and AD pathogenesis. Perivascular drainage is one of the pathways by which the brain clears metabolic waste products, like Aβ^{111, 112}, and there is accumulating evidence to suggest that the motive forces driving perivascular drainage are intrinsic to the cerebral arteries (i.e., vasocontraction and vasodilation)^{113, 114}. And, while perivascular drainage wasn’t explicitly investigated, vascular abnormalities, including altered blood vessel morphology, increased blood vessel number and density, and obstructed blood flow, were observed in the Tg4510 tauopathy AD mouse model⁷³. Increased endothelial cell PAI-1 expression was implicated in these vascular remodeling changes as PAI-1 mRNA and protein were found to be significantly upregulated⁷³. Targeting tPA-mediated signaling pathways – whether upstream or downstream or through direct or indirect interaction with Aβ and/or tau – may provide novel therapeutic benefit against AD-related pathologies, and possibly, slow disease progression.

Future investigations into the vascular contribution to AD pathology

tPA-mediated BBB dysregulation as a therapeutic target

The data described above clearly indicate a vascular component to AD pathology. Indeed, BBB dysfunction is now a well-established feature in AD^{115, 116}. And, while tPA has not been directly implicated in BBB dysfunction in AD, there is evidence to suggest a critical role for tPA in regulating BBB permeability in other pathologies^{5, 117–120}. In the context of BBB regulation, tPA appears to act through cleavage of platelet-derived growth factor-CC (PDGF-CC) and subsequent activation of the PDGFR-α in the neurovascular unit. Blocking the tPA/PDGF-CC/PDGFRα signaling pathway with the tyrosine kinase inhibitor Imatinib preserves barrier function and improves outcome in models of stroke, seizure, and traumatic brain injury^{5, 117, 120}.

While this pathway has been validated in multiple models of BBB dysfunction, earlier efforts to discern tPA/plasmin’s actions on the BBB in stroke implicated matrix metalloproteinase 9^{121, 122}. Though the mechanism and cellular source of matrix metalloproteinase 9 in stroke and BBB damage is still controversial^{123, 124}, numerous studies have demonstrated that neutrophils infiltrating the brain parenchyma are the main source of matrix metalloproteinase 9 in the later stages of stroke and BBB damage^{125, 126}

The tPA/PDGF-CC/PDGFRα signaling pathway, however, appears to be activated during the early phases of BBB opening post-stroke^{5, 127}; moreover, it is now appreciated that the integrin MAC-1 (macrophage-1 antigen; also known as αMβ2 and CD11b/CD18) on resident microglia acts as a co-factor for tPA to accelerate cleavage of PDGF-CC¹²⁷. While earlier biochemical work had shown tPA to be an inefficient enzyme, compared to other serine proteases¹²⁸, the activity of tPA can be greatly enhanced by co-factors, like fibrin(ogen) or Aβ⁵⁷. Accordingly, Su et al¹²⁷ hypothesized that there existed some unknown co-factor in the neurovascular unit that was facilitating tPA’s cleavage of PDGF-CC; in a series of in vitro and in vivo model systems, both LRP1 and MAC-1 were shown to be necessary and sufficient in facilitating activation of PDGF-CC by tPA.

That tPA activity is significantly increased by MAC-1, fibrin(ogen), or Aβ is of note as these proteins intersect at the BBB in the context of AD. In post-mortem tissue from human AD patients and mouse models of AD, fibrin(ogen), which is normally compartmentalized to the vasculature, is detected in the intra- and extravascular space^{12, 129–131}. Extravasated fibrin(ogen) appears to induce microglia activation and generation of reactive oxygen species, which result in neurodegeneration and cognitive deficits in the 5×FAD amyloidogenic AD mouse model¹³². This neuroinflammatory response is mediated by MAC-1, as genetic loss of the cd11b integrin of MAC-1 disrupts fibrin(ogen)-induced microglia activation and prevents neurodegeneration and cognitive impairment. Indeed, treatment with the novel monoclonal antibody 5B8, which inhibits fibrin-induced inflammation and oxidative stress, protected 5×FAD mice from cholinergic neuron loss and aberrant inflammation¹². The role of tPA and the downstream PDGF-CC/PDGFRα signaling cascade in potentiating BBB opening, and contributing to further fibrin(ogen) extravasation, was not examined in these studies.

Preserving BBB function with Imatinib treatment has been shown to be therapeutically beneficial in other mouse models of neurodegenerative and neuroinflammatory disorders. In a model of amyotrophic lateral sclerosis Imatinib treatment restored integrity of the brain spinal cord barrier and delayed amyotrophic lateral sclerosis onset¹¹⁹, while rats and mice with autoimmune encephalomyelitis, a model of multiple sclerosis, had improved barrier function and autoimmune encephalomyelitis symptoms after Imatinib treatment^{118, 133}. Because of the promising preclinical data on Imatinib, clinical trials in the United States and Europe are currently investigating Imatinib treatment in patients with multiple sclerosis (ClinicalTrials.gov Identifier: NCT03674099) and stroke (EudraCT Number: 2010–019014-25). The congruity of the tPA/PDGF-CC/PDGFRα pathway across disease model systems suggests that this pathway is a conserved, common regulator of BBB permeability. Targeting this pathway to preserve barrier integrity, along with other downstream effectors of vascular dysfunction, could have an additive therapeutic benefit for the treatment of AD.

Conclusions

It has been 40 years since Krystosek and Seeds¹³⁴ were the first to demonstrate a non-fibrinolytic role for the plasminogen activation system in the CNS, and it has been almost 30 years since tPA was first implicated as an immediate-early gene involved in learning and memory¹. Though many questions remain, progress has been made in elucidating tPA’s physiological role in synaptic plasticity and its pathological role in AD. While early work on the tPA/plasmin system in AD models focused on Aβ clearance and degradation, more recent studies have explored mechanisms of tPA-mediated vascular dysregulation. In addition to these studies demonstrating a role for tPA in cerebrovascular pathologies associated with AD, they highlight a growing body of evidence and appreciation for how vascular dysfunction is a key pathogenic contributor to AD. Understanding how tPA and other proteins are involved in vascular dysfunction is critical for developing new druggable targets to treat AD and other neurodegenerative diseases.