Activators and inhibitors of the plasminogen system in Alzheimer’s disease

Abstract Accumulation and deposition of Aβ is one of the main neuropathological hallmarks of Alzheimer’s disease (AD) and impaired Aβ degradation may be one mechanism of accumulation. Plasmin is the key protease of the plasminogen system and can cleave Aβ. Plasmin is activated from plasminogen by tissue plasminogen activator (tPA) and urokinase-type plasminogen activator (uPA). The activators are regulated by inhibitors which include plasminogen activator inhibitor-1 (PAI-1) and neuroserpin. Plasmin is also regulated by inhibitors including α2-antiplasmin and α2-macroglobulin. Here, we investigate the mRNA levels of the activators and inhibitors of the plasminogen system and the protein levels of tPA, neuroserpin and α2-antiplasmin in post-mortem AD and control brain tissue. Distribution of the activators and inhibitors in human brain sections was assessed by immunoperoxidase staining. mRNA measurements were made in 20 AD and 20 control brains by real-time PCR. In an expanded cohort of 38 AD and 38 control brains tPA, neuroserpin and α2-antiplasmin protein levels were measured by ELISA. The activators and inhibitors were present mainly in neurons and α2-antiplasmin was also associated with Aβ plaques in AD brain tissue. tPA, uPA, PAI-1 and α2-antiplasmin mRNA were all significantly increased in AD compared to controls, as were tPA and α2-antiplasmin protein, whereas neuroserpin mRNA and protein were significantly reduced. α2-macroglobulin mRNA was not significantly altered in AD. The increases in tPA, uPA, PAI-1 and α2-antiplasmin may counteract each other so that plasmin activity is not significantly altered in AD, but increased tPA may also affect synaptic plasticity, excitotoxic neuronal death and apoptosis.


Introduction
The abnormal accumulation and deposition of amyloid ␤ (A␤) peptide is one of the key neuropathological hallmarks of Alzheimer's disease (AD) and is thought to initiate a series of processes that cause synaptic dysfunction and neuronal death [1,2]. In recent years, reduced activity of enzymes capable of degrading A␤ has been suggested as a potential contributor to AD pathogenesis [3][4][5]. These enzymes include angiotensin-converting enzyme [6,7], neprilysin [8], endothelin-converting enzymes [9], insulin-degrading enzyme [10] and plasmin [11]. Plasmin is the key protease of the plasminogen system, the primary function of which is fibrinolysis, but has also been shown to be important in cell matrix degradation and cell migration [12,13]. Plasmin is activated from its inactive precursor plasminogen by two plasminogen activators, tissue plasminogen activator (tPA) and urokinase-type plasminogen activator (uPA) which are, in turn, regulated by inhibitors that include plasminogen activator inhibitor-1 (PAI-1) and neuroserpin. Plasmin itself is inhibited by ␣2-antiplasmin and ␣2-macroglobulin [14][15][16].
A␤ activates the plasminogen activators both in vitro and in vivo [11,[17][18][19]. This has the potential to be a protective mechanism to limit the accumulation of A␤. Plasmin cleaves A␤ at multiple sites, is capable of degrading A␤ fibrils and reduces A␤ deposition [11,20]. Plasmin protects cultured neurons from A␤-induced cell death [11,21,22] and enhances clearance of A␤ in vivo [23].
We previously found that plasmin protein and activity were not significantly altered in the human AD brain compared to controls [24]. This suggests that the activating influence of A␤ on the plasminogen system may be counterbalanced by changes affecting other activators and inhibitors of the system. Apart from plasmin activation, these activators and inhibitors mediate a range of additional effects that may also be of relevance to the development of AD. Neuroserpin, for example, was shown to interact directly with A␤ and reduce A␤ fibril formation and toxicity to cultured neurons [25]. tPA was shown in several studies to influence synaptic plasticity, a process essential in learning and memory [26][27][28][29][30], but is also a mediator of excitotoxic neuronal death [31][32][33] and apoptosis [34]. The activators and inhibitors of the plasminogen system have not been much studied in human brain tissue and in the few published reports, the findings are somewhat contradictory. One immunohistochemistry-based study showed increased tPA in AD, with highest levels in amyloid-rich regions of the brain [35]; a further study found no significant alteration in tPA or uPA proteins, but decreased tPA activity in AD, possibly resulting from an increased neuroserpin level [36]; yet another study reported no significant alteration of tPA activity in AD but did show a negative correlation between tPA activity and the level of A␤ [37].
Our aim in this study was to investigate tPA, uPA, neuroserpin, PAI-1, ␣2-antiplasmin and ␣2-macroglobulin in AD; to look at their distribution in regions of human brain tissue relevant to AD pathology; to compare the expression of the genes encoding them in AD and control brain tissue; and to measure tPA, neuroserpin and ␣2-antiplasmin proteins, in order to identify differences in AD that may contribute to the disease.

Brain tissue
This study had local Research Ethics Committee approval. The tissue was obtained from the Human Tissue Authority-licensed South West Dementia Brain Bank, University of Bristol. The tissue was dissected from brains that had been removed from patients within 72 hrs of death. The left cerebral hemisphere had been sliced and frozen at Ϫ80ЊC. The right cerebral hemisphere had been fixed in 10% formalin for approximately 3 weeks before tissue was taken, processed and paraffin sections cut for neuropathological assessment and diagnosis. The area fractions of cerebral cortex immunopositive for phospho-( load) and A␤ after excluding A␤laden blood vessels (A␤ plaque load) had been measured as previously described [38,39] and APOE genotype characterized. These studies involved immunoperoxidase staining and measurement of gene expression and proteins for which different cohorts were used, as indicated later. The AD cases were selected on the basis of a diagnosis according to CERAD [40] of 'definite AD' and a Braak tangle stage of IV-VI.

Western blots
The specificities of the antibodies used for the immunoperoxidase studies and for protein measurements by ELISA were confirmed by Western blot. Brain tissue from the frontal cortex (BA6) was homogenized in 1 ml 1% SDS lysis buffer in a Precellys homogenizer (Stretton Scientific, Derbyshire, UK). Total protein levels were quantified using Total Protein Kit (Sigma-Aldrich, Dorset, UK) according to the manufacturer's instructions. Recombinant protein standards were loaded (as summarized in Table 1), along with 30 g (total protein) homogenate per sample onto a 4-20% Tris-HCL pre-cast gel (Bio-Rad, Hercules, CA, USA) and electrophoresed at 150 V for 1 hr. The proteins were transferred to nitrocellulose membrane overnight at 4ЊC. Non-specific binding was blocked by incubation of the membrane with 10% non-fat milk in 0.05% tris-buffered saline-Tween-20 (TBST) for 1 hr at room temperature, with agitation. The membrane was incubated with the appropriate primary antibody, diluted (as summarized in Table 1) in 5% milk/TBST, for 1 hr at room temperature with agitation. After three washes in TBST, the membrane was incubated for 1 hr at room temperature with the appropriate peroxidase-conjugated secondary antibody, diluted in 5% milk/TBST. The membrane was washed three times in TBST then ECL detection reagents (Amersham Biosciences, Buckinghamshire, UK) were applied for 1 min. before exposure to photographic film in the dark for 7 min. then development.

Immunoperoxidase staining
We performed immunoperoxidase staining to show the distribution of tPA, uPA, PAI-1, neuroserpin, ␣2-antiplasmin and ␣2-macroglobulin in brain tissue. Antibody specificity was assessed by Western blot as described earlier. We used 3-amino-propyl-triethoxy silane-coated slides to collect 7-m-thick paraffin sections of temporal lobe (including Brodmann areas BA20-22). Sections were incubated overnight at 60ЊC, dewaxed, dehydrated and incubated in 3% hydrogen peroxide in methanol for 30 min. to block endogenous peroxidase. Antigen-retrieval step were performed as outlined in Table 2, then the sections were washed in cold running water, and non-specific antibody binding was blocked by incubation with blocking serum (Vecatastain Universal Elite Kit; Vector Labs, Burlingame, CA, USA) for 20 min. at room temperature. Sections were washed twice in PBS and the primary antibodies were applied as outlined in Table 2, overnight at room temperature. Sections were again washed twice in PBS, and the biotinylated Universal Antibody (Vector Labs) applied for 20 min. at room temperature. After two washes in PBS, VectaElite ABC complex (ABC), made up previously and incubated at room temperature for 30 min., was applied for 20 min. After 2 washes in PBS 3,3'-diaminobenizidine was applied for 7 min., the sections were washed in running water, then immersed in copper sulfate for 4 min. and counterstained in Harris' haematoxylin for 15 sec. The sections were rinsed, dehydrated, cleared and mounted. Negative controls comprised sections treated as above apart from omission of the primary antibody.

RNA isolation, cDNA production and Real-time PCR (RT-PCR)
We homogenized frontal and temporal neocortex in TRIzol reagent (Invitrogen, Carlsbad, CA, USA) in a Precellys homogenizer. Chloroform was added to the homogenates for 3 min., after which they were centrifuged at 13,000 ϫ g for 15 min. at 4ЊC then the aqueous phase removed and mixed with an equal volume of isopropyl alcohol and 30 g of glycogen. After a 10-min. incubation at room temperature, the samples were centrifuged at 13,000 ϫ g for 10 min. at 4ЊC to precipitate the RNA, which was washed then twice in 75% ethanol and re-suspended in water (Sigma-Aldrich, Gillingham, UK). The RNA was treated with DNaseI (Roche Diagnostics Ltd., West Sussex, UK) to remove any DNA and the RNA concentration determined using the RiboGreen quantification kit (Invitrogen).
High Capacity cDNA Archive Kit (Applied Biosystems, Foster City, CA, USA) was used, as per the manufacturer's instructions, to produce cDNA. Quantification of the cDNA was performed using the PicoGreen quantification kit (Invitrogen). Table 1 Summary of the recombinant protein, primary and secondary antibody dilutions used in Western blots to demonstrate antibody specificity Table 2 Summary of antigen retrieval steps and antibody dilutions used for immunoperoxidase studies *tPA ab51518 was biotinylated, so HRP-conjugated streptavidin (R&D Systems), diluted 1:500 in TBST, was used in place of a secondary antibody.

Antibody
Antibody supplier and product code Application in this study Real-time PCR was carried out using an ABI 7000 sequencing detection system (ABI Prism; Applied Biosystems). Assay on Demand Gene Expression Products for tPA, uPA, neuroserpin, PAI-1, ␣2-antiplasmin and ␣2-macroglobulin, and for calibrating genes glyceraldehyde 3-phosphate (GAPDH) and neuron-specific enolase (NSE) (Taqman probes; Applied Biosystems) were incubated with SensiMix dT (Quantace, London, UK) and 10 ng cDNA (total volume of 20 l) at 50ЊC for 2 min., 95ЊC for 10 min. followed by 40 cycles of 95ЊC for 15 min. and 60ЊC for 1 min. Samples were analysed in triplicate and gene expression relative to mean control tissue expression was calculated using the 2 Ϫ⌬⌬Ct method [41]. We used two calibrator genes for the assays: NSE (because in the brain all of the plasminogen system components are expressed predominantly in neurons) and GAPDH (expressed by all cell types). As 2 Ϫ⌬⌬Ct is an exponential function (the fold-increase or decrease in mRNA), the data for each cohort were expressed as the geometric mean and 95% confidence interval [42].

Measurement of neuroserpin protein by direct ELISA
A seven-point homogenate standard curve (total protein range 50-500 g), brain homogenates (100 g total protein per sample) diluted in PBS and blanks of PBS were incubated (in duplicate) in a clear 96-well microplate (Fisher Scientific, Loughborough, UK) for 2 hrs at room temperature with agitation. The plate was then washed five times with 0.05% PBS-Tween20 (PBST), tapped dry on the final wash and incubated for 2 hrs at room temperature with the mouse monoclonal anti-neuroserpin antibody (Abcam, Cambridge, UK) (the specificity of which was confirmed by Western blot) diluted 1:400 in PBS. After five washes in PBST, the plate was incubated with the peroxidase-conjugated anti-mouse secondary antibody (Vector Labs) diluted 1:100 in PBS for 30 min. at room temperature, in the dark. The plate was washed five times with PBST, tapped dry on the final wash, then 100 l of peroxidase substrate (R&D Systems, Minneapolis, MN, USA) was added to all wells for 3 min., followed by 50 l Stop solution, then absorbance measured at 450 nm using a multidetection microplate reader (FLUOstar OPTIMA, BMG Labtech, Aylesbury, UK). Relative neuroserpin levels were interpolated from the standard curve.

Measurement of tPA protein by luminescent sandwich ELISA
Black 96-well microplates (Fisher Scientific) were coated with capture goat polyclonal anti-tPA (Santa Cruz Biotechnology, CA, USA) antibody diluted 1:100 in coating buffer (10 mM sodium carbonate, 30 mM sodium bicarbonate, pH 9.6) at 4ЊC overnight. After five washes, the plate was tapped dry and incubated with 1% BSA/PBS for 90 min. at room temperature with agitation. The plate was washed five times, tapped dry and the recombinant tPA protein (ProSpec-Tany TechnoGene Ltd., Ness-Ziona, Israel) standard curve of 5 five-fold dilutions (concentration range 0.08-50 ng/ml), homogenate samples (100 g total protein per sample), blanks and controls (where either capture antibody, sample of detection antibody were omitted) were diluted in PBS, loaded in duplicate and incubated for 2 hrs at room temperature with agitation. The plate was again washed five times, tapped dry and incubated in the dark for 2 hrs at room temperature with biotinylated sheep polyclonal anti-tPA antibody (Abcam) diluted 1:100 in 1% BSA. Following five further washes, the plate was again tapped dry and incubated with peroxidase-conjugated streptavidin (R&D Systems) diluted 1:500 in 1% BSA (Sigma-Aldrich) for 30 min. in the dark at room temperature. The plate was again washed five times and 100 l luminescent peroxidase substrate (R&D Systems) added for 5 min., after which luminescence was measured using a multidetection microplate reader (BMG Labtech). Absolute protein levels were interpolated from the standard curve.

Measurement of ␣2-antiplasmin by sandwich ELISA
Clear, 96-well microplates (Fisher Scientific) were coated with capture goat polyclonal anti-␣2-antiplasmin antibody (Santa Cruz) diluted 1:1000 in coating buffer at 4ЊC overnight. After five washes in PBST, the plate was tapped dry and incubated with 1% BSA (Sigma-Aldrich) for 90 min. at room temperature with agitation. The plate was washed five times, tapped dry and the samples incubated for 2 hrs at room temperature with agitation. The samples that made up the standard curve consisted of 6 five-fold serial dilutions of full length human purified ␣2-antiplasmin protein (Haemotologic Technologies, VT, USA) diluted in PBS (protein range 0.003-1.7 g/ml). For the homogenate samples, 100 g total protein diluted in 50 l PBS was loaded and for the blanks, 50 l PBS. Additional controls where either capture antibody, sample or detection antibody were omitted were also included. All samples were loaded in duplicate. After 5 washes, the plate was tapped dry and incubated with the detection rabbit polyclonal anti-␣2-antiplasmin antibody (Abcam) diluted 1:1000 in 1% BSA (Sigma-Aldrich) for 2 hrs at room temperature with agitation. The plate was again washed five times, tapped dry and peroxidase-conjugated secondary antibody (Vector Labs) was applied for 30 min. in the dark at room temperature. The plate was again washed five times and 100 l peroxidase substrate (R&D Systems) added for 5 min. after which 50 l Stop solution was added and the absorbance measured using a multidetection microplate reader (BMG Labtech) at 450 nm. Absolute protein levels were interpolated from the standard curve.

Measurement of post-mortem stability of tPA, neuroserpin and ␣2-antiplasmin
Tissue from the frontal cortex of an AD brain removed 4 hrs after death and a control brain removed 10 hrs after death was dissected and divided into 10 aliquots. These were incubated for 0, 6, 12, 18, 24, 48 or 72 hrs at room temperature or 24, 48 or 72 hrs at 4ЊC. The aliquots were then homogenized in 1 ml 1% SDS lysis buffer and tPA, neuroserpin and ␣2-antiplasmin protein measurements made as described earlier.

Statistical analysis
Differences in mRNA expression and protein level between the AD and control groups were analysed by Mann-Whitney U-test. The relationships of APOE genotype and Braak stage to mRNA and protein level were assessed by Kruskall-Wallis test, followed by Dunn's post-tests if appropriate. Relationships between individual mRNAs and the corresponding proteins, proteins and A␤ load, and between age or post-mortem delay and the level of individual activator and inhibitor mRNAs and proteins were assessed by Spearman's correlation. P-values Ͻ 0.05 were considered significant.

Localization of the plasminogen system activators and inhibitors in the human brain
The distribution of tPA, uPA, PAI-1, neuroserpin, ␣2-antiplasmin and ␣2-macroglobulin in sections from human brain is shown in Figure 1. The images are representative of the findings across the cohorts. The specificities of all the antibodies used were confirmed by Western blot (Fig. S1) and no labelling was present in the negative control sections included in each experiment. tPA and uPA were predominantly in neurons but occasional small clusters of immunopositive granular material were present in microglia or astrocytes (Fig. 1A-D). PAI-1 and neuroserpin were both present in neurons (Fig. 1E-H). ␣2-Antiplasmin was present predominantly  E, F), neuroserpin (G, H), ␣2-antiplasmin (I, J) and ␣2-macroglobulin (K, L) in AD and control brain tissue. All proteins were present predominantly in the neurons. ␣2-antiplasmin was also associated with swollen processes in the peripheral part of some A␤ plaques in AD brain tissue (arrows in I). No labelling was present in negative control sections. Scale bar ϭ 50 m for all sections.
in neurons but was additionally found in swollen cell processes at the periphery of some A␤ plaques (Fig. 1I-J). ␣2-Macroglobulin was predominantly neuronal (Fig. 1K and L). Our results did not replicate previous findings of co-localization of tPA and PAI-1 with A␤ plaques [36,43]. None of the proteins was associated with neurofibrillary tangles and, with the exception of co-localization of ␣2-antiplasmin with A␤ plaques in AD, there were no obvious differences in the distribution of any of the proteins in AD and control brains.

mRNA expression
The mRNA expression of the genes encoding the plasminogen activators tPA and uPA is shown in Figure 2 and all the data for the mRNA expression studies are summarized in Table 3. tPA expression relative to both GAPDH and NSE was higher in AD than controls; this reached statistical significance in the temporal cortex (P ϭ 0.001 tPA relative to GAPDH, P ϭ 0.002 tPA relative to NSE) and in the frontal cortex when tPA was calibrated against NSE (P ϭ 0.007; Fig. 2A and B). uPA expression was also increased in AD compared to control in the temporal cortex (P ϭ 0.03 uPA relative to GAPDH, P ϭ 0.032 uPA relative to NSE) and in the frontal cortex when uPA was calibrated against NSE (P ϭ 0.017; Fig. 2C-D). tPA and uPA mRNA expression did not vary significantly with APOE genotype, or A␤ plaque load in the AD cohort. tPA mRNA relative to GAPDH varied significantly with Braak stage (P ϭ 0.028). Post-testing revealed the expression to be higher in temporal cortex from brains of Braak tangle stages V-VI than Braak stages 0-II (P Ͻ 0.05). uPA mRNA showed no significant correlation with Braak stage.
The mRNA expression of the inhibitors of the plasminogen system is shown in Figure 3. PAI-1 expression was increased in AD compared to controls in both the frontal and temporal lobes but this was only statistically significant in the temporal cortex (P ϭ 0.026 PAI-1 relative to GAPDH, P ϭ 0.029 PAI-1 relative to NSE; Fig. 3A and B). Neuroserpin expression in relation to GAPDH was significantly decreased in AD compared to controls in both brain regions (P ϭ 0.001 in the temporal cortex, P ϭ 0.007 in the frontal cortex) but there was very little alteration in the expression of neuroserpin in relation to NSE in either brain region ( Fig. 3C and D). ␣2-Antiplasmin mRNA in relation to NSE was significantly higher in AD than controls in both regions (P ϭ 0.001 Fig. 2 The fold change in tPA mRNA expression in the frontal (A) and temporal cortex (B) and in uPA mRNA expression in the frontal (C) and temporal (D) cortex in AD and control brains calibrated against endogenous GAPDH and NSE. The geometric mean and 95% confidence intervals are shown for each group on a logarithmic scale to the base 2. tPA mRNA expression was significantly increased in AD compared to control in the temporal cortex when calibrated against GAPDH (P ϭ 0.001) and NSE (P ϭ 0.002) and in the frontal cortex relative to NSE (P ϭ 0.007). uPA mRNA expression was significantly increased in AD compared to control in the temporal cortex relative GAPDH (P ϭ 0.03) and NSE (P ϭ 0.032) and in the frontal cortex relative to NSE (P ϭ 0.017). Ø in the temporal cortex, P ϭ 0.0004 in the frontal cortex); in relation to GAPDH, the increase was significant in the temporal cortex only (P ϭ 0.012; Fig. 3E and F). ␣2-Macroglobulin mRNA showed a non-significant increase in AD compared to controls in both brain regions (Fig. 3G and H). PAI-1, neuroserpin and ␣-macroglobulin mRNA all showed no significant relationship to APOE genotype or A␤ load in the AD cohort, or with Braak stage. ␣2-antiplasmin mRNA did not vary significantly with APOE genotype or A␤ load but, in the temporal cortex, was significantly altered with Braak stage (␣2-antiplasmin mRNA in relation to GAPDH P ϭ 0.003, in relation to NSE P ϭ 0.018) and post-tests showed increased expression in Braak stages V-VI compared to 0-II (␣2-antiplasmin mRNA in relation to GAPDH P Ͻ 0.01, in relation to NSE P Ͻ 0.05).

tPA, neuroserpin and ␣2-antiplasmin proteins
Data from the protein measurements are summarized in Table 4. tPA protein was measured in the temporal cortex only. tPA was increased in the AD group compared to controls but this did not reach statistical significance (Fig. 4A). The mean tPA concentrations Ϯ S.E.M. were: AD 0.431 ng/ml Ϯ 0.112, controls 0.277 ng/ml Ϯ 0.0573. Neuroserpin was significantly reduced in Fig. 3 The fold change in PAI-1 mRNA expression in the frontal (A) and temporal cortex (B), in neuroserpin mRNA expression in the frontal (C) and temporal (D) cortex, in ␣2-antiplasmin mRNA in the frontal (E) and temporal (F) cortex and in ␣2-macroglobulin mRNA in frontal (G) and temporal (H) cortex in AD and control brains calibrated against endogenous GAPDH and NSE. The geometric mean and 95% confidence intervals are shown for each group on a logarithmic scale to the base 2. PAI-1 mRNA expression was significantly increased in AD compared to control in the temporal cortex with GAPDH calibration (P ϭ 0.026) and with NSE calibration (P ϭ 0.029). The increase in PAI-1 in the frontal region did not reach statistical significance. Neuroserpin mRNA expression was reduced in AD compared to control when calibrated against GAPDH (P ϭ 0.001 in the temporal cortex, P ϭ 0.007 in the frontal cortex). ␣2-Antiplasmin mRNA was increased in both brain regions in AD compared to control; this was significant when calibrated against NSE (P ϭ 0.001 in the temporal cortex and P ϭ 0.0004 in the frontal cortex) and against GAPDH in the temporal cortex (P ϭ 0.012). ␣2-Macroglobulin mRNA was not significantly altered in AD compared to control.
both the frontal and temporal cortex in AD compared to controls (P ϭ 0.003) in the frontal cortex and P ϭ 0.035 in the temporal cortex; Fig. 4B and C). The mean relative neuroserpin levels Ϯ S.E.M. were: in the frontal region, AD 1.647 relative units Ϯ 0.091, controls 2.099 relative units Ϯ 0.107; in the temporal region AD 0.616 relative units Ϯ 0.048, controls 0.771 relative units Ϯ 0.052. ␣2-antiplasmin was increased in AD compared to controls but this only reached significance in the temporal region (P ϭ 0.018; Fig. 4D and E). The mean ␣2-antiplasmin concentrations Ϯ S.E.M. were: in the frontal region, AD 0.151 g/ml Ϯ 0.057, controls 0.074 g/ml Ϯ 0.017; in the temporal region, AD 0.245 g/ml Ϯ 0.069, controls 0.063 g/ml Ϯ 0.012.
The protein concentrations all correlated positively with the levels of the corresponding mRNAs; the correlation was significant for temporal neuroserpin (neuroserpin mRNA in relation to GAPDH, P ϭ 0.004) and for ␣2-antiplasmin in both brain regions (␣2-antiplasmin mRNA in relation to GAPDH: frontal cortex P ϭ 0.019, temporal cortex P ϭ 0.048).
tPA and ␣2-antiplasmin protein concentrations increased with increasing number of APOE e4 alleles and with Braak stage but these correlations did not reach statistical significance. Table 4 Summary of changes in tPA, neuroserpin and ␣2-antiplasmin proteins in AD compared to controls *P Ͻ 0.05; **P Ͻ 0.01 ( : significantly increased; Ø : significantly decreased); ns: not significant; NA: not assessed. and temporal (E) cortex. The bars indicate the mean values and standard errors. tPA protein was increased in AD compared to control but this did not reach statistical significance. Neuroserpin was significantly reduced in both brain regions in AD compared to control: in the frontal cortex P ϭ 0.003, in the temporal cortex P ϭ 0.035. ␣2-Antiplasmin protein was increased in AD compared to control; this reached statistical significance in the temporal cortex (P ϭ 0.018).
Neuroserpin protein decreased with increasing Braak stage, a relationship that was significant in the frontal region (P ϭ 0.028); post hoc testing showed the difference to be significant between Braak stages V-VI and 0-II (P Ͻ 0.05). Neuroserpin did not vary significantly with APOE genotype. None of the proteins correlated with A␤ plaque load in the AD samples.

Age, gender and post-mortem delay
The protein and mRNA measurements showed no relationship to age or post-mortem delay and did not vary significantly with gender, either in the individual cohorts or when the cohorts were combined. tPA, neuroserpin and ␣2-antiplasmin protein concentrations did not alter significantly when unfixed brain tissue was incubated for up to 72 hrs at room temperature or 4ЊC (Fig. 5). The protein measurements are therefore unlikely to have been affected by variations in post-mortem delay between the different cases or phenotypic groups.

Discussion
In recent years, there has been mounting interest in the contributions of enzymes capable of degrading A␤ to the clearance of this peptide from the brain. Plasmin is one such enzyme but the plasminogen system has not been much studied in the human brain. We recently found no significant alterations in protein concentration or enzyme activity of plasmin in AD and concluded, therefore, that alterations in plasmin are not generally responsible for A␤ accumulation, although they may contribute in some cases [24].
We have now shown that there are marked changes in the expression of the activators and inhibitors of the plasminogen system in AD: significant increases in tPA, uPA, PAI-1 and ␣2-antiplasmin mRNA and increased tPA and ␣2-antiplasmin protein, and significant reduction in neuroserpin mRNA and protein. Although the net effect may be little or no change in plasmin activity, altered expression of plasminogen activators and inhibitors could have other consequences in AD; this is particularly so for tPA. The stimulus responsible for the increase in tPA mRNA and protein in AD is unclear. Although the previous immunohistochemical study of Medina et al. [35] suggested that plaque-associated A␤ may induce tPA synthesis, we did not find tPA to co-localize with plaques, and tPA mRNA and protein levels did not correlate with A␤ plaque load in the AD samples. There was a significant increase in tPA mRNA in Braak stages V-VI compared to stages 0-II, which would suggest that tPA synthesis increases with progression of the disease; however, this relationship did not follow through to the protein expression. One possible explanation is the significant increase in PAI-1 mRNA expression in AD which, if also translated to the protein level, would result in increased inhibition and removal of tPA. Alternatively, tPA expression may be under translational control, as hypothesized by Salles and Strickland [44], such that the protein is synthesized only under certain conditions [45].
The activation of uPA by A␤ has been somewhat controversial, some studies reporting activation [11] and others no activation [17,19]. We found an increase in uPA mRNA in AD but this did not correlate with A␤ plaque load. ␣2-Antiplasmin has not previously been shown to interact directly with A␤. Our immunohistochemical demonstration of ␣2-antiplasmin around A␤ plaques raises the possibility that there is indeed interaction between the two but again, in the AD samples there was no correlation between ␣2-antiplasmin mRNA or protein level and A␤ plaque load. ␣2-Antiplasmin mRNA, like tPA mRNA, increased with Braak stage and was significantly higher in stages V-VI than 0-II; again, this increase was not translated to the protein level.
The balanced increases in the expression of the plasminogen activators tPA and uPA, and the inhibitors PAI-1 and ␣2-antiplasmin in AD may account for the lack of overall alteration in plasmin activity in AD, which we showed previously. However, these proteins have other actions that may be relevant in AD. On the positive side, tPA enhances synaptic plasticity in mice and restores cognition and improves performance in memory tasks [26]. However tPA is a mediator of excitotoxic neuronal death [31] and apoptosis [34]. uPA and PAI-1 have also been shown to be anti-apoptotic [46][47][48][49][50][51][52] so the observed increased expression of these proteins may be neuroprotective. The consequence of increased ␣2-antiplasmin expression may be limited to a reduction in plasmin activity as, to our knowledge, it has not been shown to have any other effects to date.
Our finding of reduced neuroserpin mRNA and protein differs from that in a previous study which showed an increase in neuroserpin protein in AD compared to controls [36]. The reasons for this disparity are not entirely clear but some of our other observations may suggest a partial explanation. We found neuroserpin gene expression to be reduced relative to GAPDH (constitutively expressed in all cells) but not relative to NSE (expressed solely by neurons). This suggests that the reduction in neuroserpin in our AD cohort was secondary to neuron loss, which may have been more pronounced than in the study by Fabbro and Seeds [36]. In the frontal region neuroserpin protein level was significantly lower in Braak stages V-VI than 0-II, in keeping with the interpretation that neuroserpin level falls late in the disease, after most neuronal damage has occurred. It is not clear at what stage of disease were the AD patients studied by Fabbro and Seeds or, indeed, what cri-teria were used to make the diagnosis. Other differences that may be relevant relate to cohort size (12 AD and 12 control samples in the previous study and 38 AD and 38 controls in this study), and the method for measuring neuroserpin (densitometry of Western blots calibrated with a single standard sample on each gel, rather than direct ELISAs with each plate calibrated against a seven-point standard). A decrease in neuroserpin could help to limit the continued accumulation of A␤, as tPA activity, and hence plasmin activity, will be prolonged. However, the findings of Kinghorn et al. [25] indicate that neuroserpin is also capable of reducing A␤1-42 toxicity, both in rat pheochromocytoma cells and in Drosophila.
One limitation of this study is that the enzymatic activities of tPA and uPA have not been assessed. A second limitation of this study is that the expression of the activator and inhibitor proteins has not been measured in corresponding plasma samples. The levels of these proteins are higher in the plasma than the brain parenchyma. Unfortunately, matched plasma and brain tissue samples are difficult to obtain, and in any case post-mortem measurements of plasmin-related proteins in blood samples are likely to be affected by coagulation of blood after death. The aim of this study was therefore limited to investigation of the plasminogen system in the brain, where its potential role in the clearance of A␤ deposition may be of relevance in AD.
Our finding of several significant alterations in the plasminogen system highlights the need for further investigation of this system in the CNS.