Elevated expression of urokinase plasminogen activator in rodent models and patients with cerebral amyloid angiopathy

The aim of this work is to study the association of urokinase plasminogen activator (uPA) with development and progression of cerebral amyloid angiopathy (CAA).


INTRODUCTION
Cerebral amyloid angiopathy (CAA) is a highly prevalent form of cerebral small vessel disease characterised by the progressive accumulation of amyloid-β (Aβ) peptide species in the vasculature of the brain.
CAA can occur across various calibres of cerebral vessels (small arteries, arterioles, capillaries, veins and venules) [1]. Aβ deposition occurring in arteries, arterioles and capillaries is designated as CAA type-I, while amyloid deposition solely occurring in arteries and arterioles but not in cerebral capillaries is referred to as CAA Type-II [2]. Due to the underlying amyloidotic pathology, CAA often coincides with Alzheimer's disease (AD), a common form of dementia in which Aβ peptides are deposited in the parenchyma of the brain in the form of senile plaques [3]. As a clinical syndrome, CAA can cause vascular cognitive impairment and dementia and is an important risk factor for the development of lobar haemorrhage [4].
CAA can occur as a sporadic entity (sCAA), but mutations in the APP gene can also predispose to the development of CAA. One of the most common autosomal mutations predisposing for this hereditary form of CAA is the so-called Dutch variant of CAA (D-CAA), caused by the mutation E693Q in the APP gene [5]. D-CAA causes symptoms similar to sCAA (including cognitive deterioration and repeated haemorrhages), although the clinical onset is approximately 20 years earlier and disease course is usually more aggressive. Because of the similarities in Aβ composition and clinical symptoms, D-CAA can be considered a model for sCAA [6][7][8].
Because a brain biopsy is rarely performed for definitive CAA identification, the clinical diagnosis of sCAA is highly dependent on magnetic resonance imaging (MRI) biomarkers to provide either a probable or a possible clinical diagnosis. These MRI biomarkers form the basis for the modified Boston Criteria to diagnose CAA and include lobar (micro)haemorrhages and cortical superficial siderosis [9]. However, these criteria are not optimal: The imaging biomarkers often appear only after significant progression of the disease and reflect mid-to end-stages of CAA, especially in the case of cerebral (micro)bleeds [10]. In addition, the presence of microbleeds in nonlobar locations precludes the diagnosis of CAA, whereas CAA and arterioloslerotic small vessel disease may in fact co-exist.
Timely CAA diagnosis could significantly benefit from novel specific biomarkers at earlier disease stages. Clinical relevance of a timely and correct CAA diagnosis is reflected in the dilemma of deciding to treat, or not to treat, atrial fibrillation with (oral) anticoagulants. Treatment with anticoagulants lowers the odds of a patient suffering of ischaemic stroke, whereas it increases the odds of inducing CAAinduced intracerebral haemorrhages [11].
Proteins present in cerebrospinal fluid (CSF) may offer an excellent source of such novel biomarkers, given the close contact between the CSF and brain interstitial fluid, and thus, reflecting the pathological processes occurring in the brain and its blood vessels. For example, Aβ40 levels in serum and CSF of a transgenic rodent model of CAA were found to be significantly decreased (and decreasing over time) when compared with levels in control rats [12]. Additionally, concentrations of Aβ42 and Aβ40 peptides are known to be lowered in the CSF of CAA patients, when compared with control subjects and AD patients [13][14][15]. However, given the substantial overlap of CSF Aβ42 and Aβ40 levels between CAA and AD patients, other biomarkers more specific to CAA, either as a single marker, or a combination of markers, would facilitate more accurate and timely diagnosis of the disease.
A potential candidate biomarker for CAA is urokinase plasminogen activator (uPA), a serine-proteinase that cleaves plasminogen into plasmin in the fibrinolytic pathway. uPA and its tissue counterpart, tissue plasminogen activator (tPA), have been linked to (vascular) amyloidosis occurring in AD and CAA [16][17][18][19]. Also, fibrillar Aβ accumulation can induce (over)expression of uPA and tPA in vitro in cultured rat neurons [20]. Additionally, it has been reported that Aβ peptides are cytotoxic to smooth muscle cells and pericytes and induce expression of the proteolytically active uPA (and its receptor) by cerebral vascular smooth muscle and endothelial cells [21][22][23].
Lastly, uPA and other components of the plasmin system (including plasmin and tPA) have been reported as mediators of Aβ degradation [24,25]. This myriad of relationships between vascular Aβ-amyloidosis and uPA implies a potential role for uPA in CAA pathophysiology, supporting its potential as a diagnostic biomarker for CAA.
Here, we analyse the potential of uPA in CSF to support the diagnosis of CAA. First, we researched amyloid deposition in brain tissue of a transgenic rat model of CAA (rTg-DI) compared with wild-type (WT) rodent controls and of a CAA patient compared with a human control subject. Additionally, we researched the extent to which uPA mRNA is produced by cerebral vessels in rTg-DI rat models at different ages, in comparison with cerebrovascular expression in WT rodent controls. Lastly, we determined uPA levels in CSF of both rTg-DI rodent models and WT rodents, and in human CAA patients and controls in two separate cohorts and researched their diagnostic performance, to determine whether CSF uPA can function as a diagnostic biomarker for CAA.

Highlights
• Expression of urokinase plasminogen activator (uPA) is elevated in rodent models and human patients suffering from sporadic cerebral amyloid angiopathy (sCAA).
• uPA shows high degrees of co-localisation with vascular amyloid-β peptides in rodent model and human patients CAA tissue.
• uPA levels are significantly elevated, after correction for age, in the cerebrospinal fluid (CSF) of CAA rodent models and human CAA patients, compared with levels in control subjects.
• uPA is not significantly elevated in CSF of either asymptomatic or symptomatic patients suffering from E693Q Dutch-type CAA (D-CAA).

rTg-DI transgenic rat model of CAA type-I
The rTg-DI rat model expresses low levels of human amyloid precursor protein (APP) harbouring the E693Q/D694N Dutch/Iowa familial CAA mutations, under the control of the neuronal-specific Thy1.2 promoter, as well as induced overexpression through double K670N/ M671L Swedish APP-mutations [26]. Cerebral vascular Aβ fibril deposition starts at approximately 3 months of age throughout the cortex, hippocampus and thalamus and progresses dramatically with advancing age. Nontransgenic littermates served as WT control rats. For this study, heterozygous rTg-DI and WT rats were used and analysed at

Brain tissue collection and preparation
The rTg-DI and WT rats were sacrificed at the designated ages and perfused with cold phosphate buffered saline (PBS). Forebrains were subsequently extracted and dissected through the midsagittal plane.

Rat CSF collection and preparation
Rat CSF was acquired from the cisterna magna of both rTg-DI and WT rats at 3 months of age. Rats were anaesthetised deeply through isoflurane inhalation and were subsequently mounted on a stereotaxic unit. An incision was made along the midline of the head and neck, starting between the ears and ending at about 2.5 cm caudally. The fascia was pulled back, and muscles were dissected, which opens up the atlanto-occipital membrane. A small incision was made along the membrane midline and its underlying dura, aided by a surgical microscope. CSF was sampled through the slit in the dura using a fine-tipped pipette. Samples were aliquoted into sterile, polypropylene (PP) tubes and were subsequently frozen at À80 C. This procedure makes it possible to acquire an average of 120-to 150-μl CSF per rat.

Rat CSF uPA ELISA
The levels of uPA in CSF of both rTg-DI and WT rats (

Human subject inclusion and sample collection
We analysed the diagnostic performance of uPA as a novel biomarker for sCAA in two separate groups, a discovery and a validation cohort.
In the discovery cohort, we obtained human CSF They neither had the suspected neurological disease, nor a neurodegenerative disease, known cognitive impairment, sepsis, a recent stroke (<6 months), or a malignancy in the central nervous system.
Additional inclusion criteria were the availability of a sufficient amount of CSF and a normal composition of the CSF for a number of routine parameters (e.g., cell count, blood pigments, total protein, glucose, lactate and oligoclonal IgG bands). Additionally, a minority of CSF samples (n = 10) was collected from patients who underwent thoracoabdominal aortic aneurysm repair, for which they had an external lumbar drain. They did not have known cognitive impairment or recent (<6 months) stroke, or traumatic brain injury. Details of the patient characteristics for these groups are shown in Table 1. (n = 17). All controls (n = 40) were obtained from the Radboud University Medical Center. sCAA patients from all centres were diagnosed similarly as described for the discovery group. Details on patient characteristics of these groups are shown in Table 2.
D-CAA patients were exclusively recruited through the (outpatient) clinic of the LUMC (asymptomatic: n = 11; symptomatic: n = 12). Inclusion criteria were age >18 years and a proven APP mutation, or a medical history of one or multiple lobar ICHs and ≥1 firstdegree relative with D-CAA. Symptomatic D-CAA was defined by a history of at least one symptomatic ICH. Age-matched control groups were constructed for the asymptomatic and symptomatic D-CAA groups (n = 22; n = 28, respectively). Details of the patient characteristics in the D-CAA groups and respective controls are shown in

Elevated vascular uPA expression in rTg-DI rodent models and WT controls
We performed immunolabelling experiments to evaluate the expression of uPA in the brains of 12 months old rTg-DI rats with abundant microvascular CAA. We observed significant perivascular expression of uPA in cerebral capillaries harbouring amyloid deposits in rTg-DI rats ( Figure 1A-C). In contrast, uPA expression was not detected in cerebral capillaries of WT rats ( Figure 1D). We next evaluated uPA mRNA levels in the brain tissue of rTg-DI rats at 3 and 12 months of age (Figure 2A). At 3 months of age, that is, the age of the onset of microvascular CAA, there was a highly significant (p < 0.01) sixfold increase in uPA mRNA expression compared with WT rats. The increased uPA mRNA levels were sustained with an eightfold increase in rTg-DI rats at 12 months of age compared with similarly aged WT rats (p < 0.01).
The early-onset uPA expression in rTg-DI rats prompted us to perform ELISA analysis to measure uPA levels in the CSF of these transgenic rats compared with WT controls at 3 months of age. Figure 2B shows elevated levels of uPA in CSF in rTg-DI rats at

Elevated uPA levels in CSF of human CAA patients and control subjects
We determined if we could confirm the increased perivascular uPA expression in our rTg-DI preclinical model of human CAA. Figure 1E,F shows that in the presence of CAA pathology, perivascular uPA expression was increased, whereas we found negligible uPA expression in vessels in a control subject lacking CAA pathology.
We next performed ELISA analysis to evaluate uPA levels in the CSF of human CAA patients and controls.
In the discovery groups ( Figure 3A;  Expression was significantly elevated at each age in rTg-DI rats compared with WT rats (t-test; p < 0.01). (B) CSF uPA levels were assessed using a rat uPA ELISA in rTg-DI and WT rats at 3 months of age. uPA levels were significantly elevated in rTg-DI rats (t-test, n = 6, p = 0.03) In the validation groups ( Figure 3B;

No significant increase in CSF uPA in (a)symptomatic D-CAA
Comparisons between asymptomatic D-CAA patients and agematched controls ( Figure 3B;

DISCUSSION
This study demonstrates an elevation of CSF uPA in patients suffering from CAA pathology. In two separate cohort studies, CSF uPA levels, after adjustment for age, were significantly elevated in sCAA patients, when compared with respective controls. In contrast, no elevation of uPA levels was observed after adjustment for age in (a)symptomatic D-CAA patients, compared with their respective age-matched controls. Perivascular uPA in amyloid-affected cerebral vessels was overexpressed in rTg-DI rats and a sCAA patient. Moreover, uPA (mRNA) levels were increased in CSF from the rTg-DI CAA rodent model.
Earlier studies have demonstrated a decreased of uPA levels in plasma, but not CSF, in mild-cognitive impairment (MCI) patients with decreased Aβ42/Aβ40 ratios (indicative of amyloid pathology) compared with controls [28]. Similarly, no differences in the CSF concentrations of plasminogen, tPA or plasminogen activator inhibitor-1 between AD patients and controls were discovered. Moreover, in this study, zymography of CSF revealed no difference in tPA activity between AD and controls, whereas uPA activity could not be detected in CSF of either AD or control cases [29]. However, in one other F I G U R E 3 Comparison of human CSF uPA levels in CAA patient and control groups. Scatter plots of CSF uPA levels in our (A) discovery and (B) validation studies. In our discovery study, uPA levels were significantly elevated in sCAA (in dark grey; n = 27) compared with control subjects (in light grey; n = 40) (MU p < 0.001), even after adjusting for age (MU p = 0.05). In our validation studies, uPA levels in sCAA (in dark grey; n = 38) were borderline nonsignificant (MU p = 0.08), but significantly elevated after adjusting for age (MU p = 0.03) compared with controls (in light grey; n = 40). Asymptomatic D-CAA patients (in dark purple; n = 11) presented elevated uPA levels in CSF compared with their respective controls (in light purple; n = 22; MU p = 0.01; after adjusting for age, MU p = 0.09), whereas symptomatic D-CAA patients (in dark blue; n = 12) did not show an elevation (MU p = 0.46; after adjusting for age, MU p = 0.44) compared with age-matched controls (in light blue; n = 28). All data presented as median with IQR study, increased levels of CSF tPA were observed in AD versus patients with subjective cognitive impairment, but not in AD versus MCI patients [30]. In summary, these studies show that uPA, as an influential factor in CAA, is barely being researched and that uPA CSF levels are not increased in AD, whereas inconsistent results were obtained for tPA levels in AD. Despite the partial overlap in pathology of AD and CAA, elevated levels of uPA in CSF seem to be restricted to CAA. Future studies should reveal if uPA levels are increased in AD patients with severe CAA compared with AD cases without CAA [3].
Endothelial cells are a major source of vascular uPA expression, thereby suspected of major contributors to the activation of the plasminogen activation system, through the conversion of plasminogen into plasmin [31,32]. Because of this role in plasmin production, uPA and other components of the plasminogen activation system are often associated with tissue remodelling processes, such as cellular migration and metastasis, haemostasis, fibrinolysis and angiogenesis [33].
Earlier studies have shown elevated expression of both uPA and its receptor, uPAR, after stimulation of human smooth muscle cells with Aβ [23]. Whereas the co-localisation of uPA with Aβ in CAA patients and rTg-DI models we describe here has not been described before, tPA, the tissue-counterpart of uPA, has been found to co-localise with Aβ in the cerebral vasculature of Tg2576 mice, an AD mouse model [34]. The co-localisation of tPA and Aβ may be explained by the observations that both tPA and plasminogen bind to lysine-rich structures, which are present both in fibrin and in Aβ [19].
These previous studies also revealed decreased CSF Aβ42 levels in sCAA patients in comparison with controls. Also, in D-CAA, decreased CSF Aβ40 and Aβ42 levels were found in asymptomatic and symptomatic mutation carriers in comparison with age-matched controls [14]. However, CSF Aβ42 levels were similarly decreased in AD patients as well, limiting the specificity of CSF Aβ42 as a biomarker for CAA. Other protein biomarkers have been associated with sCAA, but again, with limited specificity (e.g., CSF Apolipoprotein D) [50].
We observed no significant differences in CSF uPA levels between symptomatic D-CAA patients (either unadjusted or when adjusted for age of subjects) and asymptomatic D-CAA patients (borderline insignificant, after adjustment for age of subjects). We did however observe a difference in unadjusted CSF uPA levels between asymptomatic D-CAA patients and controls. This could be induced by age effects (as the difference is not present in our ageadjusted model) but could also have a physiological cause. We can only speculate on the origins of this difference, but it is likely that

DATA AVAILABILITY STATEMENT
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.