Skip to main content
NCBI home page
Brain logoLink to Brain
. 2019 Feb 15;142(4):1077–1092. doi: 10.1093/brain/awz011

Selective loss of cortical endothelial tight junction proteins during Alzheimer’s disease progression

Yu Yamazaki 1, Mitsuru Shinohara 1, Motoko Shinohara 1, Akari Yamazaki 1, Melissa E Murray 1, Amanda M Liesinger 1, Michael G Heckman 2, Elizabeth R Lesser 2, Joseph E Parisi 3,4, Ronald C Petersen 4, Dennis W Dickson 1, Takahisa Kanekiyo 1,✉, Guojun Bu 1,✉
PMCID: PMC6439325  PMID: 30770921

Impairment of blood-brain barrier integrity has been implicated in Alzheimer’s disease. Yamazaki et al. show that the loss of cortical tight junction proteins is a common event in Alzheimer’s disease, and is associated with synaptic degeneration. The effects of compromised tight junctions may be both additive and synergistic to amyloid-β and tau pathologies.

Keywords: blood–brain barrier, claudin-5, neurovascular unit, occludin, tight junction

Abstract

While the accumulation and aggregation of amyloid-β and tau are central events in the pathogenesis of Alzheimer’s disease, there is increasing evidence that cerebrovascular pathology is also abundant in Alzheimer’s disease brains. In brain capillaries, endothelial cells are connected closely with one another through transmembrane tight junction proteins forming the blood–brain barrier. Because the blood–brain barrier tightly regulates the exchange of molecules between brain and blood and maintains brain homeostasis, its impairment is increasingly recognized as a critical factor contributing to Alzheimer’s disease pathogenesis. However, the pathological relationship between blood–brain barrier properties and Alzheimer’s disease progression in the human brain is not fully understood. In this study, we show that the loss of cortical tight junction proteins is a common event in Alzheimer’s disease, and is correlated with synaptic degeneration. By quantifying the amounts of major tight junction proteins, claudin-5 and occludin, in 12 brain regions dissected from post-mortem brains of normal ageing (n = 10), pathological ageing (n = 14) and Alzheimer’s disease patients (n = 19), we found that they were selectively decreased in cortical areas in Alzheimer’s disease. Cortical tight junction proteins were decreased in association with the Braak neurofibrillary tangle stage. There was also a negative correlation between the amount of tight junction proteins and the amounts of insoluble Alzheimer’s disease-related proteins, in particular amyloid-β40, in cortical areas. In addition, the amount of tight junction proteins in these areas correlated positively with those of synaptic markers. Thus, loss of cortical tight junction proteins in Alzheimer’s disease is associated with insoluble amyloid-β40 and loss of synaptic markers. Importantly, the positive correlation between claudin-5 and synaptic markers, in particular synaptophysin, was present independent of insoluble amyloid-β40, amyloid-β42 and tau values, suggesting that loss of cortical tight junction proteins and synaptic degeneration is present, at least in part, independent of insoluble Alzheimer’s disease-related proteins. Collectively, these results indicate that loss of tight junction proteins occurs predominantly in the neocortex during Alzheimer’s disease progression. Further, our findings provide a neuropathological clue as to how endothelial tight junction pathology may contribute to Alzheimer’s disease pathogenesis in both synergistic and additive manners to typical amyloid-β and tau pathologies.

Introduction

Alzheimer’s disease is the most common type of dementia in the elderly. Alzheimer’s disease pathologically is defined by the aggregation and accumulation of amyloid-β in the brain parenchyma as senile plaques and in the vasculature as cerebral amyloid angiopathy (CAA), as well as abnormally phosphorylated tau in neurons as neurofibrillary tangles (Serrano-Pozo et al., 2011; Nelson et al., 2012). In addition to these central pathological hallmarks, significant vascular pathologies such as circle of Willis atherosclerosis, cerebral infarcts, microbleeds, and white matter changes, frequently are detected in Alzheimer’s disease brains (Yarchoan et al., 2012; Kapasi and Schneider, 2016), suggesting a link between the cerebrovascular dysfunctions and the pathogenesis of Alzheimer’s disease.

In the CNS, capillary endothelial cells have distinct barrier properties referred to as the blood–brain barrier (Deeken and Loscher, 2007; Daneman and Prat, 2015). Since the blood–brain barrier controls entry of blood components into the brain and elimination of toxic molecules, it maintains brain homeostasis (Zlokovic, 2008; Daneman, 2012; Daneman and Prat, 2015; Zhao et al., 2015). It is increasingly recognized that blood–brain barrier impairment may contribute to the pathogenesis of various neurological diseases including late-onset Alzheimer’s disease (Zlokovic, 2005; Erickson and Banks, 2013; Montagne et al., 2017; Yamazaki and Kanekiyo, 2017; Zenaro et al., 2017). Dysfunction in the blood–brain barrier has been shown to contribute to amyloid-β accumulation and neuronal loss in Alzheimer’s disease mouse models of amyloidosis (Deane et al., 2004; Cirrito et al., 2005; Winkler et al., 2015; Storck et al., 2016). Conversely, ageing and amyloid-β accumulation cause blood–brain barrier dysfunction by affecting multiple properties of endothelial cells (Kook et al., 2012; Montagne et al., 2015; Yamazaki et al., 2016). Thus, current evidence from animal studies suggests that blood–brain barrier impairment might contribute to Alzheimer’s disease pathogenesis in a synergism to amyloid-β accumulation (Erickson and Banks, 2013; Yamazaki and Kanekiyo, 2017). However, since Alzheimer’s disease is a complex multifactorial disease that is difficult to recapitulate in mouse models (Drummond and Wisniewski, 2017; Jankowsky and Zheng, 2017), clarification is needed regarding the association of markers of blood–brain barrier integrity and Alzheimer’s disease-related pathology in humans.

Region-specific appearances of Alzheimer’s disease pathologies, including amyloid-β accumulation and neurofibrillary tangles, have provided important insights into disease pathogenesis (Klunk et al., 2004; Buckner et al., 2005; Jack et al., 2008; Vlassenko et al., 2010; Wang et al., 2016). Indeed, by analysing biochemical markers in multiple brain regions dissected from post-mortem brains of normal ageing, pathological ageing and Alzheimer’s disease, we have found novel evidence regarding pathomechanisms of amyloid-β accumulation and disease development (Shinohara et al., 2013, 2014, 2017). In this study, we quantified the amounts of tight junction proteins, which uniquely constitute the para-cellular barrier in the blood–brain barrier, in 12 brain areas, compared these among three disease groups, and assessed their associations with the Braak neurofibrillary tangle stage, CAA severity, vascular pathology severity, and other Alzheimer’s disease-related features. Our results indicate that the amounts of tight junction proteins selectively are decreased in the cortex of Alzheimer’s disease brains in association with insoluble amyloid-β accumulation and loss of synaptic markers.

Materials and methods

Subjects

Post-mortem tissues were obtained through the Mayo Clinic Brain Bank for neurodegenerative diseases, whose operating procedures are approved by the Mayo Institutional Review Board. In addition to 19 patients with sporadic Alzheimer’s disease, we evaluated 10 neurologically normal elderly subjects without amyloid-β accumulation (normal ageing) and 14 neurologically normal elderly subjects with extensive cortical amyloid-β deposits (pathological ageing) (Dickson et al., 1992; Murray and Dickson, 2014). All normal ageing, pathological ageing and sporadic Alzheimer’s disease patients were subjects in the Mayo Clinic Alzheimer Disease Research Center (ADRC; P50 AG016574) or Mayo Clinic Study of Aging (MCSA; U01 AG006786) and had standardized neurological and neuropsychological assessments. Neuropathological findings were reported using recommendations from the Consortium to Establish a Registry for Alzheimer’s disease (CERAD) (Mirra et al., 1991). A modified Bielschowsky stain was used to determine the Braak neurofibrillary tangle stages (Braak and Braak, 1991). The cohort and enzyme-linked immunosorbent assay (ELISA) measurements of Alzheimer’s disease-related molecules other than tight junction proteins overlapped with that of our previous studies (Shinohara et al., 2013, 2014, 2017).

Vascular pathology scores

Haematoxylin and eosin stained sections were used to assess vascular pathology. Cerebrovascular pathology was assessed in each case using a modified ‘Kalaria vascular score’ as described by Craggs et al. (2013). A maximum score of 10 was calculated from six derived from all cortical regions and four from the basal ganglia. The modified ‘Kalaria vascular score’ (Craggs et al., 2013) provides an incremental scoring system based on the presence of the following vascular pathology: cortical region, vessel wall modifications (score 1–2), degree of perivascular and white matter modifications (score 3–4), microinfarcts (score 5), large infarcts (score 6); and basal ganglia region, vessel wall modifications (score 1), degree of perivascular space dilation (score 2), microinfarcts (score 3), and large infarcts (score 4). Vascular pathology scores were available for 32 patients in our cohort (six subjects with normal ageing, 10 subjects with pathological ageing and 16 patients with sporadic Alzheimer’s disease).

CAA scores

The amyloid-β immunostained section was used to assign a semi-quantitative CAA score in both the leptomeninges and brain parenchyma in cortical sections. The CAA severity was semi-quantitatively assessed on a 4-point scale in both the leptomeninges and the brain parenchyma: 0, no evidence of CAA; 1, mild involvement of vessels; 2, some circumferential amyloid deposition; and 3, severe, diffuse involvement (Love et al., 2014; Shinohara et al., 2016).

Sample preparation

Samples were prepared as described previously (Shinohara et al., 2013, 2014, 2017). In brief, grey matter from seven cortical areas (dorsolateral prefrontal, orbitofrontal, inferior temporal, inferior parietal, primary visual cortex, posterior cingulate, entorhinal) and five subcortical areas (amygdala, striatum, thalamus, hypothalamus, cerebellum) was dissected and kept frozen until protein extraction. Brain lysates were prepared by the three-step extraction method, based upon differential solubility in buffer alone, detergent (Triton™ X-100) and chaotropic agent [guanidine hydrochloride (GuHCl)]. After removal of meninges and blood vessels, 100–200 mg of frozen brain tissue were homogenized in ice-cold Tris-buffered saline (TBS) containing a protease inhibitor cocktail (Roche Diagnostics) by Polytron homogenization (KINEMATICA). After centrifugation at 100 000 g for 60 min at 4°C, the supernatant was aliquoted and stored at −80°C (referred to as ‘TBS fraction’ or ‘TBS’). The residual pellet was rehomogenized in TBS plus 1% Triton™ X-100 with protease inhibitor cocktail, incubated with mild agitation for 1 h at 4°C and centrifuged as above. The resultant supernatant was aliquoted and stored at −80°C (referred to as TX fraction or TX). The residual pellet was rehomogenized in TBS plus 5 M GuHCl, pH 7.6, and incubated with mild agitation for 6–12 h at 22°C. After centrifugation as above, the resultant supernatant (referred to as the GuHCl fraction) was diluted with nine volumes of TBS, aliquoted and stored at −80°C.

Quantification of tight junction proteins and the endothelial cell maker CD31

The amount of claudin-5 was determined by a mouse monoclonal capture antibody (Invitrogen) and a rabbit polyclonal detection antibody (Abcam), as described (Kazmierski et al., 2012). Goat anti-rabbit IgG (H+L) antibody (Invitrogen) was used as the secondary antibody for the claudin-5 ELISA. The amount of occludin was determined by using a rabbit polyclonal capture antibody (Invitrogen) and horseradish peroxidase (HRP)-conjugated mouse monoclonal capture antibody (Invitrogen). Recombinant claudin-5 and occludin proteins (Novus Biologicals) were used as standards. Colorimetric quantification was performed using a Synergy HT plate reader (BioTek) using HRP-linked streptavidin (Vector) or Poly-HRP 40 streptavidin (Fitzgerald) and 3,3′,5,5′-tetramethylbenzidine substrate (Sigma). The amount of endothelial cell marker, CD31 (also known as PECAM-1) was determined by a commercial ELISA kit according to the manufacturer’s instruction (R&D). The TX fraction was used to measure the amounts of these membrane proteins. To minimize the possibility that any observed associations with tight junction protein simply reflect associations with number of endothelial cells, the amounts of tight junction proteins in each brain-region were normalized (i.e. divided) by the CD31 amount in the given brain region; these normalized tight junction protein measures (hereinafter referred to as CLDN5 and OCLN, respectively) were used in all analyses. Of note, CLDN5 and OCLN were moderately to highly correlated, with Spearman’s correlation coefficients for the 12 different brain regions ranging from 0.32 to 0.71 (median = 0.61). The regional distribution of CLDN5 and OCLN in normal ageing brains is shown in Supplementary Fig. 1.

Quantification of Alzheimer’s disease-related proteins

The amounts of total amyloid-β40 (Aβx–40), total amyloid-β42 (Aβx–42), full-length amyloid-β40 (Aβ1–40), full-length amyloid-β42 (Aβ1–42), tau, postsynaptic density protein 95 (PSD-95, also known as DLG4), synaptophysin (SYP), and glial fibrillary acidic protein (GFAP) were determined by ELISAs as described previously (Shinohara et al., 2013, 2014, 2017). The amounts of N-terminally truncated amyloid-β40 (Aβt–40) and amyloid-β42 (Aβt–42) were calculated by subtracting values of Aβ1–40 and Aβ1–42 from Aβx–40 and Aβx–42 values, respectively (Shinohara et al., 2017). The TBS fraction was used to measure the amounts of cytosolic or secreted proteins, and molecules (GFAP, and PSD-95), while the TX fraction was used to measure the amounts of membrane proteins (i.e. SYP). The GuHCl fraction was used to measure the amounts of pathological aggregated proteins for Aβx–40, Aβ1–40, Aβx–42, Aβ1–42 and tau as described previously (Shinohara et al., 2013, 2014, 2017). Alzheimer’s disease-related proteins are summarized for each brain region in Supplementary Table 1 separately for the normal ageing, pathological ageing, and sporadic Alzheimer’s disease groups.

Outcome measures and calculation of average protein levels in cortical and subcortical regions

The outcome measures of the study were the values of two tight junction proteins (CLDN5 and OCLN), each of which were measured in seven cortical and five subcortical brain regions. In addition to these region-specific tight junction protein amounts, the average (i.e. mean) CLDN5 and OCLN levels across cortical regions and also across subcortical regions were also calculated. Similarly, averages of Alzheimer’s disease-related proteins across cortical and subcortical regions were calculated. Although brain-region-specific values were analysed independently, to present our results more succinctly given the large number of region-specific tight junction and Alzheimer’s disease-related protein values that were measured, the average values of cortical and subcortical regions were sometimes emphasized.

There was a small amount of missing data regarding CLDN5 and OCLN levels (0.6%), which has potential to bias average CLDN5 and OCLN levels across cortical and subcortical regions. Therefore, in calculating averages of cortical and subcortical tight junction protein values, if a patient did not have a given tight junction protein measure for a given brain region, we imputed the missing data using the mean of the values of patients who did have this information available in the given disease group (normal ageing, pathological ageing, sporadic Alzheimer’s disease) for that specific brain region. Any patient with missing data for >50% of the cortical or subcortical brain regions for a given tight junction protein measure was not included in any analysis, involving the given cortical or subcortical region. (Only one patient was missing data for CLDN5 for all 12 brain regions.) Missing data were handled in the same fashion when calculating cortical and subcortical averages of Alzheimer’s disease-related proteins (0.5% of Alzheimer’s disease-related protein data were missing).

Statistical analysis

Continuous variables were summarized with the sample median and range. Comparisons of characteristics [age at death, sex, post-mortem interval (PMI), Braak stage, CAA score, vascular pathology score, Alzheimer’s disease-related proteins] between normal ageing, pathological ageing, and sporadic Alzheimer’s disease groups were made using a Kruskal-Wallis rank sum test or a chi-square test.

Comparisons of CLDN5 and OCLN values among the three groups were made using linear regression models that were adjusted for age at death, sex, and PMI. In particular, we adjusted all of the analyses for PMI as we observed several nominally significant (P < 0.05) associations with PMI (where increased PMI was associated with a lower OCLN level) prior to correction by multiple testing. Specifically, increased PMI was associated with a lower OCLN level in the dorsolateral prefrontal region (P = 0.025), inferior temporal region (P = 0.028), striatum (P = 0.041), and also when examining the average of all cortical areas (P = 0.011) (Supplementary Table 2). It should also be noted that none of these associations remained significant [all false discovery rate (FDR)-corrected P ≥ 0.13] when correcting for multiple testing using a FDR correction as we did in our other analyses.

An overall test of difference of each tight junction protein measure between the three disease groups was first performed. To account for the fact that overall tests of difference were made for each of the 12 different brain regions and for averages of cortical and subcortical areas, an FDR correction for multiple testing was used. Assuming an FDR of 5%, FDR-corrected P-values were calculated and denoted as Pc. When a statistically significant overall difference between groups was observed (i.e. Pc ≤ 0.05), pair-wise comparisons were made, and FDR-corrected P-values were subsequently calculated to correct for the three pair-wise comparisons made between the three disease groups. Comparisons of un-normalized claudin-5, un-normalized occludin, and CD31 values between the three disease groups were made in a similar manner.

Linear regression models that were adjusted for age at death, sex, and PMI were used to examine associations of CLDN5 and OCLN values with Braak stage, CAA score, and vascular pathology score. Regression coefficients and 95% confidence intervals (CIs) were estimated and are interpreted as the increase in the mean tight junction protein level corresponding to a 1-unit increase in the given measure. Separately, for associations with Braak stage, CAA, and vascular pathology scores, FDR-corrected P-values (Pc) were calculated to correct for the 12 different brain regions that were examined, as well as the cortical and subcortical averages, assuming an FDR of 5%.

Associations between CLDN5 and OCLN values with Alzheimer’s disease-related proteins were evaluated using linear regression models that were adjusted for age at death, sex, and PMI. For Alzheimer’s disease-related proteins, because of the large number of measures that were considered (each measured in the 12 different brain regions), distributions varied widely and many were skewed or contained outlying/influential points. With this in mind, these Alzheimer’s disease-related proteins were divided into four-level categorical variables based on the sample quartiles (i.e. 1 = ≤25th percentile, 2 = 26th to 50th percentile, 3 = 51st to 75th percentile, 4 = >75th percentile) for use in linear regression analysis. Regression coefficients and 95% CIs were estimated and interpreted as the increase in the mean tight junction protein level corresponding to an increase of one quartile of the given Alzheimer’s disease-related protein. For associations of tight junction proteins with Alzheimer’s disease-related proteins, analysis of the cortical and subcortical averages was considered to be primary (with region-specific analysis considered to be secondary), and FDR-corrected P-values (Pc) were calculated to correct for the 10 Alzheimer’s disease-related proteins that were examined in the primary analysis, assuming an FDR of 5%. In the secondary region-specific analysis, no adjustment for multiple testing was made, and therefore uncorrected P-values are presented (denoted as P) and P ≤ 0.05 is considered as significant.

All statistical tests were two-sided. Statistical analyses were performed using SAS (version 9.4; SAS Institute, Inc. Cary, North Carolina) and R Statistical Software (version 3.2.3; R Foundation for Statistical Computing, Vienna, Austria).

Data availability

The data that support the findings of this study are available from the corresponding author, upon reasonable request.

Results

Characteristics of normal and pathological ageing, and sporadic Alzheimer’s disease patient groups

Table 1 details characteristics of each of the three disease groups. Median age was significantly higher (P = 0.003) for pathological ageing patients (median = 94 years) compared to the normal ageing (median = 83 years) and Alzheimer’s disease (median = 86 years) groups, while sex (P = 0.30) and PMI (P = 0.77) were relatively similar between groups. As expected, both Braak stage and CAA score were lowest for normal ageing patients, followed by pathological ageing, and then by Alzheimer’s disease (both P < 0.001). Vascular pathology scores did not differ noticeably between the three groups (P = 0.77).

Table 1.

Comparison of characteristics between normal aging, pathological ageing, and sporadic Alzheimer’s disease patients

Variable Normal ageing (n = 10) Pathological ageing (n = 14) Alzheimer’s disease (n = 19) P-value
Age at death, years 83 (57, 95) 94 (82, 101) 86 (62, 95) 0.003
Sex, male (%) 3 (30.0) 1 (7.1) 5 (26.3) 0.30
PMI, h 12 (5, 18) 12 (2, 23) 11 (2, 24) 0.77
Braak stage (%) <0.001
    0 2 (20.0) 0 (0.0) 0 (0.0)
    I 1 (10.0) 1 (7.1) 0 (0.0)
    II 4 (40.0) 2 (14.3) 0 (0.0)
    III 2 (20.0) 5 (35.7) 0 (0.0)
    IV 1 (10.0) 6 (42.9) 0 (0.0)
    V 0 (0.0) 0 (0.0) 4 (21.1)
    VI 0 (0.0) 0 (0.0) 15 (78.9)
CAA score (%) <0.001
    0 8 (80.0) 6 (42.9) 0 (0.0)
    0.5 2 (20.0) 3 (21.4) 3 (15.8)
    1 0 (0.0) 3 (21.4) 5 (26.3)
    1.5 0 (0.0) 0 (0.0) 4 (21.1)
    2 0 (0.0) 1 (7.1) 4 (21.1)
    2.5 0 (0.0) 0 (0.0) 1 (5.3)
    3 0 (0.0) 1 (7.1) 2 (10.5)
VP score (%) 0.77
    2–3 3 (50.0) 4 (40.0) 8 (50.0)
    4–5 1 (16.7) 3 (30.0) 5 (31.3)
    6–10 2 (33.3) 3 (30.0) 3 (18.8)
Open in a new tab

The sample median (minimum, maximum) is given for age at death and PMI. P-values result from a Kruskal-Wallis rank sum test or Fisher’s exact test. Information was unavailable regarding vascular pathology (VP) score for 11 subjects.

Tight junction proteins are selectively decreased in the cortical areas of patients with Alzheimer’s disease

To determine whether the expression of tight junction proteins is altered in Alzheimer’s disease, the levels of two major tight junction proteins, CLDN5 and OCLN, were quantified by ELISA in TX fractions of each of the 12 brain regions in each of the three groups. The average CLDN5 level across cortical areas was significantly lower in Alzheimer’s disease compared to normal ageing and pathological ageing (both Pc = 0.002; Fig. 1A and Supplementary Table 3). Loss of cortical CLDN5 in Alzheimer’s disease was confirmed by brain region-wise analysis; CLDN5 was significantly lower in multiple cortical areas in Alzheimer’s disease compared to those with normal ageing and/or pathological ageing (Fig. 1C and Supplementary Table 3). Specifically, when adjusting for age at death, sex, and PMI, CLDN5 was significantly lower for Alzheimer’s disease compared to both normal and pathological ageing in the dorsolateral prefrontal, orbitofrontal, and inferior temporal regions (all Pc ≤ 0.023). Additionally, CLDN5 in the inferior parietal region was significantly lower for Alzheimer’s disease compared to pathological ageing (Pc = 0.027), whereas in the primary visual cortex it was lower for Alzheimer’s disease compared to normal ageing (Pc = 0.033). In contrast, there were no significant differences in the level of CLDN5 in normal and pathological ageing in the cortical areas (Supplementary Table 3), and no differences in CLDN5 in the three groups for subcortical areas (Fig. 1A, C and Supplementary Table 3).

Figure 1.

Figure 1

Open in a new tab

Widespread decreases of tight junction proteins in cortical areas of Alzheimer’s disease brains. (A) Boxplot of average claudin-5/CD31 across cortical and subcortical areas for normal ageing (NA), pathological ageing (PA), and sporadic Alzheimer’s disease (AD) patients. Average claudin-5/CD31 across cortical and subcortical areas was not available for one pathological ageing patient. (B) Boxplot of average occludin/CD31 across cortical and subcortical areas for normal ageing, pathological ageing, and sporadic Alzheimer’s disease patients. (C) Heat map summarizing statistically significant differences in tight junction proteins between normal ageing, pathological ageing and sporadic Alzheimer’s disease patients in each brain region. FDR corrected P-values (Pc) were calculated. P-values from each cell in C are available in Supplementary Table 3. *Pc < 0.05, **Pc < 0.01, N.S. = not significant after FDR correction. Cortical areas: dorsolateral prefrontal, orbitofrontal, inferior temporal, inferior parietal, primary visual cortex, posterior cingulate and entorhinal cortex. Subcortical areas: amygdala, striatum, thalamus, hypothalamus and cerebellum. AD = Alzheimer’s disease; NA = normal ageing; PA = pathological ageing.

Specific loss of cortical tight junction protein in Alzheimer’s disease also was evident when the level of OCLN was examined. When considering the average OCLN across cortical areas, this was significantly lower for Alzheimer’s disease compared to the other groups (both Pc ≤ 0.011; Fig. 1B and Supplementary Table 3). Similar to the results for CLDN5, brain region-wise analysis showed that OCLN was significantly lower in Alzheimer’s disease in several cortical areas. As shown in Fig. 1C and Supplementary Table 3, for cortical areas, OCLN was significantly lower in Alzheimer’s disease compared to both normal and pathological ageing for the inferior temporal, posterior cingulate, and entorhinal regions (all Pc ≤ 0.030). OCLN level did not differ noticeably between normal and pathological ageing in cortical areas (Supplementary Table 3), and also did not differ significantly between the three disease groups for subcortical areas (Fig. 1B, C and Supplementary Table 3). Collectively, these results indicate that the amounts of tight junction proteins are predominantly decreased in the cortical, but not in subcortical areas in Alzheimer’s disease.

Results from comparisons of un-normalized claudin-5, un-normalized occludin, and CD31 between normal ageing, pathological ageing, and sporadic Alzheimer’s disease patients are presented in Supplementary Tables 4 and 5.

Cortical tight junction proteins are decreased as Braak stage increases

To determine whether the amounts of tight junction proteins are decreased in association with severity of Alzheimer’s disease pathology, the associations between tight junction proteins and Braak stage were examined across all three groups. After adjusting for age at death, sex, and PMI, the average level of CLDN5 across cortical areas decreased significantly as Braak stage increased (both Pc ≤ 0.008; Fig. 2A and Supplementary Table 6). The association between tight junction protein amount and Braak stage in cortical areas was further confirmed by brain region-wise analysis. As shown in Fig. 2C and Supplementary Table 6, as Braak stage increased, CLDN5 level in the dorsolateral prefrontal, orbitofrontal, inferior temporal, and inferior parietal cortical regions significantly decreased (all Pc ≤ 0.025).

Figure 2.

Figure 2

Open in a new tab

Brain region-wise association between Braak stages and tight junction proteins across normal, pathological ageing and Alzheimer’s disease. (A) Boxplot of average claudin-5/CD31 across cortical and subcortical areas according to Braak stage (0–II, III–IV, or V–VI) when considering the combined series of normal ageing, pathological ageing, and patients with Alzheimer’s disease. Average claudin-5/CD31 across cortical and subcortical areas was not available for one Braak stage 0–II patient. (B) Boxplot of average occludin/CD31 across cortical and subcortical areas according to Braak stage (0–II, III–IV, or V–VI) when considering the combined series of normal ageing, pathological ageing, and patients with Alzheimer’s disease. (C) Heat map summarizing statistically significant correlations between tight junction proteins and Braak stage in each brain region when considering the combined series of normal ageing, pathological ageing, and patients with Alzheimer’s disease. Regression coefficients are interpreted as the increase in the mean tight junction protein measure corresponding to each 1-unit increase in Braak stage. FDR-corrected P-values (Pc) were calculated. P-values from each cell in C are available in Supplementary Table 6. **Pc < 0.01, N.S. = not significant after FDR correction. Cortical areas: dorsolateral prefrontal, orbitofrontal, inferior temporal, inferior parietal, primary visual cortex, posterior cingulate and entorhinal cortex. Subcortical areas: amygdala, striatum, thalamus, hypothalamus and cerebellum.

Similarly, a significant association between cortical tight junction protein and Braak stage was evident for OCLN. The average level of OCLN across cortical areas decreased significantly as Braak stage increased (Pc ≤ 0.008; Fig. 2B and Supplementary Table 6). In addition, brain region-wise analysis showed that as Braak stage increased, OCLN decreased in dorsolateral prefrontal, inferior temporal, and entorhinal cortical regions (all Pc ≤ 0.048; Fig. 2C and Supplementary Table 6). All of the aforementioned associations for Braak stage appear to be driven by the tight junction protein values in high Braak stage subjects (all Alzheimer’s disease patients), while tight junction protein values in Braak 0–II and III–IV patients were relatively similar (Supplementary Table 6). With the exception of an association between increased Braak stage and higher OCLN in the cerebellum (Pc = 0.020), no associations with Braak stage were evident for CLDN5 or OCLN in subcortical areas (Fig. 2A–C and Supplementary Table 6). Together, these results indicate that the amounts of cortical tight junction proteins are decreased as Braak stage increases, suggesting that the loss of cortical tight junction proteins is region-specific and correlates with progression of Alzheimer’s disease pathology.

Cortical tight junction proteins are decreased, in part, in association with higher CAA scores

To determine the relationship between decreases in tight junction proteins in Alzheimer’s disease and the degree of cerebrovascular pathology or CAA, associations of CLDN5 and OCLN with CAA and vascular pathology scores were examined. In the combined cohort of normal, pathological ageing, and Alzheimer’s disease, vascular pathology scores were not significantly correlated with CLDN5 or OCLN in specific brain regions or when considering cortical and subcortical averages in analysis adjusted for age at death, sex, and PMI (Supplementary Fig. 2 and Supplementary Table 7). In contrast, while CAA scores were not significantly correlated with CLDN5 when considering cortical averages (Fig. 3A and Supplementary Table 8), as CAA score increased, CLDN5 in the dorsolateral prefrontal cortical region significantly decreased (Pc = 0.031; Fig. 3C and Supplementary Table 8). The average level of OCLN across cortical areas decreased significantly as CAA score increased (Pc = 0.048; Fig.3B and Supplementary Table 8). In addition, brain region-wise analysis showed that as CAA score increased, OCLN decreased in inferior temporal cortical region (Pc = 0.006; Fig.3C and Supplementary Table 8). CAA score was not correlated with CLDN5 and OCLN in subcortical areas (Fig. 3A–C and Supplementary Table 8). Together, these results indicate that cortical tight junction proteins are decreased, at least in part, as CAA score increases, but are not associated with vascular pathology score. Thus, loss of tight junction proteins occurs independently of the overall structural abnormalities in cerebrovasculature, whereas loss of tight junction proteins and overall severity of CAA might be related in cortical areas.

Figure 3.

Figure 3

Open in a new tab

Association between tight junction proteins and CAA scores across normal, pathological ageing and Alzheimer’s disease. (A) Boxplot of average claudin-5/CD31 across cortical and subcortical areas according to CAA score (0–0.5, 1–1.5, 2–3) when considering the combined series of normal ageing, pathological ageing, and Alzheimer’s disease patients. Average claudin-5/CD31 across cortical and subcortical areas was not available for one CAA score 0–0.5 patient. (B) Boxplot of average occludin/CD31 across cortical and subcortical areas according to CAA score (0–0.5, 1–1.5, 2–3) when considering the combined series of normal ageing, pathological ageing, and Alzheimer’s disease patients. (C) Heat map summarizing statistically significant correlations between tight junction proteins and CAA score in each brain region when considering the combined series of normal ageing, pathological ageing, and patients with Alzheimer’s disease. Regression coefficients are interpreted as the increase in the mean tight junction protein measure corresponding to each 1-unit increase in CAA score. FDR-corrected P-values (Pc) were calculated. P-values from each cell in C are available in Supplementary Table 8. *Pc < 0.05, N.S = not significant after FDR correction. Cortical areas: dorsolateral prefrontal, orbitofrontal, inferior temporal, inferior parietal, primary visual cortex, posterior cingulate and entorhinal cortex. Subcortical areas: amygdala, striatum, thalamus, hypothalamus and cerebellum.

Cortical tight junction proteins decrease in association with higher insoluble amyloid-β and lower synaptic marker levels

The observed negative correlations between the amount of cortical tight junction proteins and Braak stage or CAA score (see above) suggest that tight junction pathology may occur synergistically with the development of Alzheimer’s disease pathology. To determine potential molecular mechanisms underlying the loss of tight junction proteins in Alzheimer’s disease, correlations between the amounts of tight junction proteins and Alzheimer’s disease-related molecules were measured in normal ageing, pathological ageing, and Alzheimer’s disease. After adjusting for age at death, sex, and PMI, significant negative correlations with cortical CLDN5 and OCLN levels for insoluble Aβx–40 and Aβx–42 in the GuHCl fraction (all Pc ≤ 0.046; Table 2) were found. In addition, significant positive correlations were observed for average cortical CLDN5 and OCLN with PSD-95 and SYP (all Pc ≤ 0.019). There were no significant correlations between averaged tight junction protein amounts and those of insoluble tau in the GuHCl fraction or GFAP in cortical areas. Thus, the inverse relationship observed between tight junction protein levels and Braak stage (Fig. 2A–C and Supplementary Table 6) might be confounded by the presence of amyloid-β pathology. For illustration, the strongest five of these associations (Pc < 0.01) are shown in Fig. 4A–E. Together, the results indicate that there are quantitative correlations between the amount of tight junction proteins and both insoluble amyloid-β and synaptic markers in cortical areas.

Table 2.

Association between tight junction proteins and Alzheimer’s disease-related proteins when examining averages across cortical and subcortical regions

Analysis of average values of cortical areas Analysis of averages of subcortical areas
Association with claudin-5/CD31 Association with occludin/CD31 Association with claudin-5/CD31 Association with occludin/CD31
Alzheimer’s disease-related protein Regression coefficient (95% CI) P-valuec Regression coefficient (95% CI) P-valuec Regression coefficient (95% CI) P-valuec Regression coefficient (95% CI) P-valuec
Aβx–40 −0.17 (−0.28, −0.06) 0.006 −2.35 (−3.55, −1.14) 0.002 −0.13 (−0.28, 0.01) 0.47 −0.99 (−2.78, 0.78) 0.52
Aβx–42 −0.13 (−0.24, −0.01) 0.046 −1.73 (−3.03, −0.42) 0.019 0.01 (−0.14, 0.16) 0.92 1.11 (−0.58, 2.79) 0.48
Tau −0.07 (−0.19, 0.05) 0.27 −1.33 (−2.67, 0.02) 0.068 −0.10 (−0.25, 0.04) 0.23 0.61 (−1.15, 2.37) 0.62
PSD-95 0.19 (0.08, 0.29) 0.005 2.15 (0.83, 3.47) 0.007 0.16 (0.01, 0.31) 0.098 1.20 (−0.61, 3.01) 0.48
SYP 0.18 (0.08, 0.29) 0.005 1.73 (0.41, 3.10) 0.019 0.18 (0.03, 0.32) 0.097 0.71 (−1.11, 2.53) 0.62
GFAP −0.07 (−0.18, 0.05) 0.27 −0.39 (−1.79, 1.02) 0.58 −0.16 (−0.30, −0.02) 0.097 0.33 (−1.43, 2.09) 0.79
Open in a new tab

Regression coefficients, 95% CI, and P-values result from linear regression models that were adjusted for age at death, sex, and PMI. Alzheimer’s disease-related proteins were divided into four-level categorical variables based on the sample quartiles (i.e. 1 = ≤25th percentile, 2 = 26–50th percentile, 3 = 51–75th percentile, 4 = >75th percentile). Regression coefficients are interpreted as the increase in the mean tight junction protein measure corresponding to an increase in the given Alzheimer’s disease-related protein from one quartile to the next highest quartile (i.e. from the first quartile to the second quartile). All P-values have been adjusted for multiple testing using FDR correction, assuming an FDR of 5%. Statistically significant P-values are given in bold. Aβx–40 = total amyloid-β40; Aβx–42 = total amyloid-β42.

Figure 4.

Figure 4

Open in a new tab

Cortical tight junction proteins are decreased for patients with higher insoluble amyloid-β and lower synaptic marker levels. (A) Boxplot of average claudin-5/CD31 across cortical areas according to average total amyloid-β40 (Aβx–40) across cortical areas. Average Aβx–40 across cortical areas was divided into four groups based on sample quartiles (i.e. 25th percentile, 50th percentile, 75th percentile). (B) Boxplot of average occludin/CD31 across cortical areas according to Aβx–40 across cortical areas. Average Aβx–40 across cortical areas was divided into four groups based on sample quartiles (i.e. 25th percentile, 50th percentile, 75th percentile). (C) Boxplot of average claudin-5/CD31 across cortical areas according to average PSD-95 across cortical areas. Average PSD-95 across cortical areas was divided into four groups based on sample quartiles (i.e. 25th percentile, 50th percentile, 75th percentile). (D) Boxplot of average occludin/CD31 across cortical areas according to average PSD-95 across cortical areas. Average PSD-95 across cortical areas was divided into four groups based on sample quartiles (i.e. 25th percentile, 50th percentile, 75th percentile). (E) Boxplot of average claudin-5/CD31 across cortical areas according to average SYP across cortical areas. Average SYP across cortical areas was divided into four groups based on sample quartiles (i.e. 25th percentile, 50th percentile, 75th percentile). Regression coefficients are interpreted as the increase in the mean tight junction protein measure corresponding to an increase in the given Alzheimer’s disease-related protein from one quartile to the next highest quartile [i.e. from the first quartile (Q1) to the second quartile (Q2)]. FDR-corrected P-values (Pc) were calculated. **Pc < 0.01, N.S. = not significant after FDR correction. Cortical areas: dorsolateral prefrontal, orbitofrontal, inferior temporal, inferior parietal, primary visual cortex, posterior cingulate and entorhinal cortex. Aβx–40 = total amyloid-β40.

Quantitative correlations in cortical areas were investigated by a brain region-wise analysis. Noticeably, as shown in Fig. 5 and Supplementary Table 9, significant correlations between the amount of tight junction proteins and insoluble Aβx–40 and synaptic markers were almost exclusively present in cortical areas harbouring loss of tight junction proteins (Fig. 1C). On the other hand, while quantitative correlations between the amount of tight junction proteins and insoluble Aβx–42 and tau were also detected in multiple cortical regions, they were less likely to be present in the areas harbouring loss of tight junction proteins (Fig. 5 and Supplementary Table 9). Collectively, these results indicate that loss of cortical tight junction proteins in Alzheimer’s disease is characterized by associations with insoluble total amyloid-β40 accumulation (when compared to total amyloid-β42) and loss of synaptic markers.

Figure 5.

Figure 5

Open in a new tab

Heat map summarizing statistically significant correlations between tight junction proteins and Alzheimer’s disease-related molecules. Correlations in each brain region were calculated when considering the combined series of normal ageing, pathological ageing, and Alzheimer’s disease patients. P-values from each cell are available in Supplementary Table 9. Aβx–40 = total amyloid-β40; Aβx–42 = total amyloid-β42; TJ = tight junction.

There were no significant associations of CLDN5 and OCLN and Alzheimer’s disease-related proteins when examining subcortical region (Table 2). This lack of association was also observed for the vast majority of individual brain regions in brain region-wise analysis (Fig. 5 and Supplementary Table 9).

Association between tight junction proteins and amyloid-β species

Total amyloid-β values (values in Aβx–40 and Aβx–42) in our measurements include those from the full-length amyloid-β (Aβ1–40 and Aβ1–42) and N-terminally truncated amyloid-β (Aβt–40 and Aβt–42), which may play different roles in Alzheimer’s disease pathogenesis (Bayer and Wirths, 2014; Shinohara et al., 2017). To determine the relationship between decreases in tight junction proteins and the accumulation of each of these amyloid-β species, correlations between the amounts of tight junction proteins and amyloid-β species measured in a combined analysis of normal ageing, pathological ageing, and Alzheimer’s disease were made. After adjusting for age at death, sex, and PMI, significant negative correlations with average cortical CLDN5 and OCLN for insoluble Aβ1–40, Aβt–40 and Aβt–42 in the GuHCl fraction (all Pc ≤ 0.035, Supplementary Table 10) were found. In contrast, there were no significant correlations between average tight junction protein and insoluble Aβ1–42 in GuHCl fractions in cortical areas. Thus, these results indicate that tight junction proteins correlate with insoluble full-length amyloid-β40 and N-terminally truncated amyloid-β40 and amyloid-β42 species in cortical areas.

The amyloid-β species-dependent difference in its association with tight junction proteins was confirmed further by a brain region-wise analysis. As shown in Fig. 6 and Supplementary Table 9, significant correlations were found between tight junction proteins and insoluble Aβt–40, Aβt–42 and Aβ1–40 across most cortical areas that had loss of tight junction proteins (Fig. 1C). On the other hand, there were no significant associations of tight junction protein with Aβ1–42 in cortical areas (Fig. 6 and Supplementary Table 9). Collectively, these results indicate that loss of cortical tight junction proteins in Alzheimer’s disease is characterized by predominant associations with full-length amyloid-β40 and N-terminally truncated amyloid-β40 and amyloid-β42, but not with full-length amyloid-β42.

Figure 6.

Figure 6

Open in a new tab

Heat map summarizing statistically significant correlations between tight junction proteins and amyloid-β species. Correlations in each brain region were calculated when considering the combined series of normal ageing, pathological ageing, and Alzheimer’s disease patients. P-values from each cell are available in Supplementary Table 9. Aβx–40 = total amyloid-β40; Aβx–42 = total amyloid-β42; Aβ1–40 = full-length amyloid-β40; Aβ1–42 = full-length amyloid-β42; Aβt–40 = N-terminally truncated amyloid-β40; Aβt–42 = N-terminally truncated amyloid-β42; TJ = tight junction.

No associations were found for amyloid-β with CLDN5 or OCLN in subcortical areas (Supplementary Table 10). The lack of association in subcortical areas was also observed for most brain regions in brain region-wise analysis (Fig. 6 and Supplementary Table 9).

Positive correlation of CLDN5 and synaptic markers, in particular SYP, exists independently of insoluble amyloid-β and tau accumulation

The correlations of tight junction proteins with PSD-95 and SYP (when considering cortical averages) were evaluated by additionally adjusting for insoluble Aβx–40, Aβx–42 and tau to address any potential confounding influence of these three variables on synaptic loss. These additional model adjustments weakened associations of PSD-95 and SYP with CLDN5, but were still either significant or borderline-significant (Pc = 0.055 and Pc = 0.032, respectively). On the other hand, associations of PSD-95 and SYP with OCLN weakened and were no longer significant (Pc = 0.12 and Pc = 0.17, respectively). Thus, the correlation of CLDN5 and synaptic markers, in particular SYP, exists independent of insoluble amyloid-β and tau. This suggests that loss of cortical tight junction proteins is associated with synaptic degeneration, at least in part, independent of insoluble Alzheimer’s disease-related protein pathology.

Discussion

In this study, we analysed the amounts of endothelial tight junction proteins in multiple brain regions from neocortical, limbic and subcortical areas in post-mortem brains of normal ageing, pathological ageing and Alzheimer’s disease. Our results demonstrated that the amounts of two major tight junction proteins, CLDN5 and OCLN, are predominantly decreased in cortical areas of Alzheimer’s disease compared to normal ageing and pathological ageing, but not in subcortical areas. Supporting the interrelationship between loss of cortical tight junction proteins and Alzheimer’s disease, there were significant correlations of cortical tight junction protein reductions with accumulation of insoluble amyloid-β and synaptic loss in the combined series of normal ageing, pathological ageing, and Alzheimer’s disease patients. While the potential alternation of tight junction proteins in post-mortem brains from Alzheimer’s disease have been studied previously, only one or two brain regions were usually investigated in these prior studies (Romanitan et al., 2007, 2010; Carrano et al., 2011, 2012; Keaney et al., 2015). Moreover, the immunohistochemical approaches frequently used have limitations in determining quantitative endothelial pathology (Harik, 1992; Horwood and Davies, 1994; Mooradian et al., 1997; Deane et al., 2004; Romanitan et al., 2007; Miller et al., 2008; Carrano et al., 2011, 2012; Keaney et al., 2015). Therefore, to fill these gaps, we took advantage of established biochemical methods (Shinohara et al., 2013, 2014, 2017) that allow us to quantify tight junction protein in multiple brain regions and that were also studied for other Alzheimer’s disease-related molecules. Our results not only support previous reports showing an intimate relationship between cerebrovascular pathology and Alzheimer’s disease (Yarchoan et al., 2012; Toledo et al., 2013), but also extend these results by demonstrating how the changes in membrane-associated tight junction proteins, a critical measure of blood–brain barrier integrity, occurs during Alzheimer’s disease progression in a brain region-dependent manner.

One major finding of this study is that loss of cortical tight junction proteins occurs in association with accumulation of Alzheimer’s disease-related insoluble proteins, in particular amyloid-β. The associations between total amyloid-β40 and tight junction proteins were stronger than that between total amyloid-β42 and tight junction proteins in cortical areas, suggesting that the relationship of endothelial pathology to amyloid-β may differ between amyloid-β40 and amyloid-β42. Since higher amyloid-β40 levels are correlated with CAA severity (Shinohara et al., 2016), decreases in tight junction proteins in Alzheimer’s disease may reflect a change in endothelial cell properties associated with CAA pathology. Indeed, cortical tight junction proteins are decreased, at least in part, as CAA score increases, suggesting that loss of tight junction proteins and overall severity of CAA might be related in cortical areas. More specifically, since leptomeningeal vessels were removed from those brain samples, parenchymal rather than leptomeningeal CAA is reflected in these results (Carrano et al., 2011). Alternatively, it is also possible that endothelial cell damage during Alzheimer’s disease progression could predominantly interfere with the clearance of amyloid-β40, causing a significant regional association between decreases of tight junction proteins and amyloid-β40 accumulation in brain parenchyma. In either scenario, our results suggest a synergistic mechanism that promotes both endothelial and amyloid-β pathology during progression of Alzheimer’s disease pathology.

Of note, by focusing on amyloid-β isoforms, we also found significant associations of tight junction proteins for both N-terminally truncated amyloid-β40 and amyloid-β42. Thus, relationships between N-terminally truncated amyloid-β and endothelial pathology may be different from that of full-length amyloid-β with endothelial pathology, which is in line with the concept that truncated amyloid-β species may play distinct roles in Alzheimer’s disease pathogenesis (Bayer and Wirths, 2014; Shinohara et al., 2017). How these species-specific interactions of amyloid-β with endothelial pathology occur and contribute to Alzheimer’s disease pathogenesis require further clarification.

Since loss of synapses strongly correlates with a cognitive decline in Alzheimer’s disease (Terry et al., 1991; DeKosky et al., 1996; Coleman and Yao, 2003; Koffie et al., 2011), identifying biological processes related to a change in synaptic molecules in Alzheimer’s disease adds to our understanding of Alzheimer’s disease pathogenesis. In this context, it is noteworthy that loss of tight junction proteins was associated with loss of synaptic markers in cortical areas. Although further studies are needed, blood–brain barrier disruption due to tight junction protein reduction may exacerbate neuroinflammation, oxidative stress and structural changes in the neurovascular unit, resulting in synaptic loss (Zlokovic, 2011; Marques et al., 2013; Zenaro et al., 2017). Conversely, since neuronal function likely mediates angiogenesis and blood–brain barrier development in endothelial cells through co-patterning and mutual crosstalk (Walchli et al., 2015), neuronal damage during Alzheimer’s disease progression may reduce the tight junction protein expression through a loss of physiological crosstalk between these two cell types. Importantly, previous studies suggested that circle of Willis or intracranial atherosclerosis is significantly associated with cognitive decline regardless of Alzheimer’s disease pathology (Dolan et al., 2010; Arvanitakis et al., 2016). Moreover, recent human brain imaging studies suggested that cerebrovascular pathology affects cognitive impairment independent of amyloid-β deposition in both normal and cognitively impaired patients (Vemuri et al., 2015; Ye et al., 2015). Since the correlation of CLDN5 and synaptic markers, in particular SYP, was independent of insoluble Aβx–40, Aβx–42 and tau, it is reasonable to postulate that tight junction pathology may also contribute to cognitive decline in Alzheimer’s disease above and beyond amyloid-β and tau pathologies.

In this study, normalized tight junction protein measures (the amount of tight junction protein in each brain-region normalized against that of CD31 in the given brain region) were used as surrogates to represent tight junction protein expression in endothelial cells. Accordingly, decreases in CLDN5 or OCLN can be explained by the relative increase in CD31, the relative decrease in claudin-5 and occludin, or both. In fact, in terms of the severity and number of regions involved, the differences in claudin-5 and occludin measures (without normalization by CD31) between the three groups were much more prominent than that of CD31 (Supplementary Table 4 and 5); this suggests that decreases in CLDN5 and OCLN were indeed primarily driven by the relative decreases in claudin-5 and occludin rather than changes in CD31 in Alzheimer’s disease brains. Thus, with the assumption that the amount of CD31 expression represents the number of endothelial cells, it is reasonable to argue that loss of CLDN5 and OCLN represents tight junction pathology in endothelial cells. Alternatively, considering CD31 is classified as an adhesion molecule in endothelial cells, loss of CLDN5 and OCLN might also represent the selective reduction of tight junction protein expression among other molecules vital for endothelial cell functions. Measurements of additional endothelial cell-specific molecules such as Tie2 and glucose transporter 1 (GLUT1) might help to distinguish these possibilities. In either scenario, however, our results highlight the presence of prominent endothelial tight junction pathology in Alzheimer’s disease.

Several limitations of our study need to be acknowledged. First, all subjects were Caucasian, and therefore our findings may not be directly extrapolated to other populations. Second, information regarding other possible covariates such as APOE genotype, hypertension, hyperlipidaemia, diabetes and medications as statins or anti-hypertensives was not assessed. Thus, possible contributions of these factors to endothelial pathology need to be determined in future studies. Third, considering the complexity and diversity of vascular pathobiology in Alzheimer’s disease (Hawkes et al., 2013; Ramirez et al., 2016; Banerjee et al., 2017), inclusion of subcortical white matter regions in the analysis might provide more comprehensive insight regarding how tight junction pathology in subareas of the brain contributes to Alzheimer’s disease. In addition, molecular profiling studies using human brain tissues have provided critical insight into the mechanism by which vascular pathology occurs in association with Alzheimer’s disease (Thomas et al., 2015; Miners et al., 2016, 2018). Thus, pericyte and astrocyte markers, plasma-derived surrogate markers for blood–brain barrier disruption, and cytokines and signalling molecules vital for maintaining vascular homeostasis represent key biomarkers to be analysed in parallel with endothelial tight junction pathology in future studies. Finally, because of the relatively small sample size, power to detect associations with tight junction protein is limited after adjustment for multiple testing. Therefore, the possibility of a type II error (i.e. a false-negative finding) needs to be considered when interpreting our results. For example, brain region-wise negative association between Braak stages (Fig. 2) or insoluble tau (Fig. 5) and tight junction proteins might represent the association between tau accumulation and tight junction pathology. On the other hand, there were no significant correlations between averaged tight junction protein amounts and those of insoluble tau. These seemingly inconsistent findings might be attributed to the relatively small sample sizes in our study. Considering that tau pathological changes could influence endothelial cell biology in vivo (Bennett et al., 2018), further studies where several forms of pathological tau are investigated in a larger number of samples would be warranted. Nonetheless, our study overall indicates that tight junction pathology increases with the severity of Alzheimer’s disease pathology.

In conclusion, we have shown that the tight junction proteins are selectively decreased in cortical areas in Alzheimer’s disease. Biochemical analysis in 12 brain regions revealed that loss of cortical tight junction proteins in Alzheimer’s disease is associated with insoluble amyloid-β40 accumulation and loss of synaptic markers. Furthermore, quantitative correlation between CLDN5 and synaptic markers, in particular SYP, was at least partially independent of insoluble amyloid-β and tau. Taken together, these results suggest that tight junction pathology may contribute to Alzheimer’s disease pathogenesis in both synergistic and additive manners to amyloid-β and tau.

As loss of tight junction proteins is not a complete measure of blood–brain barrier disruption, and correlation of readouts in post-mortem brain does not necessarily represent causality, the results of our study should serve as a premise for future prospective studies that aim at determining whether blood–brain barrier impairment per se causatively contributes to Alzheimer’s disease pathogenesis. It is also critical to test above mentioned mechanistic hypothesis generated from the neuropathological clue (i.e. interactions of amyloid-β and synaptic loss with tight junction pathology) in future studies perhaps using experimental model systems.

Supplementary Material

Supplementary Figures
Click here for additional data file. (725.8KB, pdf)

Acknowledgements

The authors would like to acknowledge the continuous commitment and teamwork offered by Linda G. Rousseau, Virginia R. Phillips, and Monica Castanedes-Casey for their dedicated efforts to Mayo Clinic Brain Bank. The authors thank Mr. Joshua A. Knight for the careful reading of this manuscript. Finally, the authors thank Drs Shunsuke Koga and Masaya Tachibana for helpful discussions.

Glossary

Abbreviations

CAA

cerebral amyloid angiopathy

PMI

post-mortem interval

PSD-95

postsynaptic density protein 95

Funding

This work was supported by grants from the National Institutes of Health (NIH) (RF1AG051504, P50AG016574, R37AG027924, R01AG035355, and R01AG046205 to G.B.; P50AG016574, and U01AG006786 to R.C.P.; R01AG054449 to M.E.M; and R01AG051574 to T.K.); a Cure Alzheimer’s Fund (to G.B.); the GHR Foundation (to R.C.P.); an American Heart Association (15SDG22460003 to T.K.); Florida Department of Health Ed and Ethel Moore Alzheimer’s Disease Research Program (7AZ22 to T.K.); and fellowships from the Japan Society for the Promotion of Science (JSPS), Mochida Memorial Foundation for Medical & Pharmaceutical Research, Mayo Clinic Alzheimer's Disease Research Center and American Heart Association (to Y.Y.).

Competing interests

The authors report no competing interests.

References

  1. Arvanitakis Z, Capuano AW, Leurgans SE, Bennett DA, Schneider JA. Relation of cerebral vessel disease to Alzheimer’s disease dementia and cognitive function in elderly people: a cross-sectional study. Lancet Neurol 2016; 15: 934–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Banerjee G, Kim HJ, Fox Z, Jager HR, Wilson D, Charidimou A, et al. MRI-visible perivascular space location is associated with Alzheimer’s disease independently of amyloid burden. Brain 2017; 140: 1107–16. [DOI] [PubMed] [Google Scholar]
  3. Bayer TA, Wirths O. Focusing the amyloid cascade hypothesis on N-truncated Abeta peptides as drug targets against Alzheimer’s disease. Acta Neuropathol 2014; 127: 787–801. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bennett RE, Robbins AB, Hu M, Cao X, Betensky RA, Clark T, et al. Tau induces blood vessel abnormalities and angiogenesis-related gene expression in P301L transgenic mice and human Alzheimer’s disease. Proc Natl Acad Sci USA 2018; 115: E1289–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Braak H, Braak E. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol 1991; 82: 239–59. [DOI] [PubMed] [Google Scholar]
  6. Buckner RL, Snyder AZ, Shannon BJ, LaRossa G, Sachs R, Fotenos AF, et al. Molecular, structural, and functional characterization of Alzheimer’s disease: evidence for a relationship between default activity, amyloid, and memory. J Neurosci 2005; 25: 7709–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Carrano A, Hoozemans JJ, van der Vies SM, Rozemuller AJ, van Horssen J, de Vries HE. Amyloid Beta induces oxidative stress-mediated blood-brain barrier changes in capillary amyloid angiopathy. Antioxid Redox Signal 2011; 15: 1167–78. [DOI] [PubMed] [Google Scholar]
  8. Carrano A, Hoozemans JJ, van der Vies SM, van Horssen J, de Vries HE, Rozemuller AJ. Neuroinflammation and blood-brain barrier changes in capillary amyloid angiopathy. Neurodegener Dis 2012; 10: 329–31. [DOI] [PubMed] [Google Scholar]
  9. Cirrito JR, Deane R, Fagan AM, Spinner ML, Parsadanian M, Finn MB, et al. P-glycoprotein deficiency at the blood-brain barrier increases amyloid-beta deposition in an Alzheimer disease mouse model. J Clin Investig 2005; 115: 3285–90. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Coleman PD, Yao PJ. Synaptic slaughter in Alzheimer’s disease. Neurobiol Aging 2003; 24: 1023–7. [DOI] [PubMed] [Google Scholar]
  11. Craggs LJ, Hagel C, Kuhlenbaeumer G, Borjesson-Hanson A, Andersen O, Viitanen M, et al. Quantitative vascular pathology and phenotyping familial and sporadic cerebral small vessel diseases. Brain Pathol 2013; 23: 547–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Daneman R. The blood-brain barrier in health and disease. Ann Neurol 2012; 72: 648–72. [DOI] [PubMed] [Google Scholar]
  13. Daneman R, Prat A. The blood-brain barrier. Cold Spring Harb Perspect Biol 2015; 7: a020412. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Deane R, Wu Z, Sagare A, Davis J, Du Yan S, Hamm K, et al. LRP/amyloid beta-peptide interaction mediates differential brain efflux of Abeta isoforms. Neuron 2004; 43: 333–44. [DOI] [PubMed] [Google Scholar]
  15. Deeken JF, Loscher W. The blood-brain barrier and cancer: transporters, treatment, and Trojan horses. Clin Cancer Res 2007; 13: 1663–74. [DOI] [PubMed] [Google Scholar]
  16. DeKosky ST, Scheff SW, Styren SD. Structural correlates of cognition in dementia: quantification and assessment of synapse change. Neurodegeneration 1996; 5: 417–21. [DOI] [PubMed] [Google Scholar]
  17. Dickson DW, Crystal HA, Mattiace LA, Masur DM, Blau AD, Davies P, et al. Identification of normal and pathological aging in prospectively studied nondemented elderly humans. Neurobiol Aging 1992; 13: 179–89. [DOI] [PubMed] [Google Scholar]
  18. Dolan H, Crain B, Troncoso J, Resnick SM, Zonderman AB, Obrien RJ. Atherosclerosis, dementia, and Alzheimer disease in the Baltimore Longitudinal Study of Aging cohort. Ann Neurol 2010; 68: 231–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Drummond E, Wisniewski T. Alzheimer’s disease: experimental models and reality. Acta Neuropathol 2017; 133: 155–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Erickson MA, Banks WA. Blood-brain barrier dysfunction as a cause and consequence of Alzheimer’s disease. J Cerebr Blood F Met 2013; 33: 1500–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Harik SI. Changes in the glucose transporter of brain capillaries. Can J Physiol Pharm 1992; 70: S113–7. [DOI] [PubMed] [Google Scholar]
  22. Hawkes CA, Gatherer M, Sharp MM, Dorr A, Yuen HM, Kalaria R, et al. Regional differences in the morphological and functional effects of aging on cerebral basement membranes and perivascular drainage of amyloid-beta from the mouse brain. Aging Cell 2013; 12: 224–36. [DOI] [PubMed] [Google Scholar]
  23. Horwood N, Davies DC. Immunolabelling of hippocampal microvessel glucose transporter protein is reduced in Alzheimer’s disease. Virchows Arch 1994; 425: 69–72. [DOI] [PubMed] [Google Scholar]
  24. Jack CR, Lowe VJ, Senjem ML, Weigand SD, Kemp BJ, Shiung MM, et al. 11C PiB and structural MRI provide complementary information in imaging of Alzheimer’s disease and amnestic mild cognitive impairment. Brain 2008; 131: 665–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Jankowsky JL, Zheng H. Practical considerations for choosing a mouse model of Alzheimer’s disease. Mol Neurodegener 2017; 12: 89. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Kapasi A, Schneider JA. Vascular contributions to cognitive impairment, clinical Alzheimer’s disease, and dementia in older persons. Biochim Biophys Acta 2016; 1862: 878–86. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Kazmierski R, Michalak S, Wencel-Warot A, Nowinski WL. Serum tight-junction proteins predict hemorrhagic transformation in ischemic stroke patients. Neurology 2012; 79: 1677–85. [DOI] [PubMed] [Google Scholar]
  28. Keaney J, Walsh DM, O’Malley T, Hudson N, Crosbie DE, Loftus T, et al. Autoregulated paracellular clearance of amyloid-beta across the blood-brain barrier. Sci Adv 2015; 1: e1500472. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Klunk WE, Engler H, Nordberg A, Wang Y, Blomqvist G, Holt DP, et al. Imaging brain amyloid in Alzheimer’s disease with Pittsburgh Compound-B. Ann Neurol 2004; 55: 306–19. [DOI] [PubMed] [Google Scholar]
  30. Koffie RM, Hyman BT, Spires-Jones TL. Alzheimer’s disease: synapses gone cold. Mol Neurodegener 2011; 6: 63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Kook SY, Hong HS, Moon M, Ha CM, Chang S, Mook-Jung I. Abeta(1)(-)(4)(2)-RAGE interaction disrupts tight junctions of the blood-brain barrier via Ca(2)(+)-calcineurin signaling. J Neurosci 2012; 32: 8845–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Love S, Chalmers K, Ince P, Esiri M, Attems J, Jellinger K, et al. Development, appraisal, validation and implementation of a consensus protocol for the assessment of cerebral amyloid angiopathy in post-mortem brain tissue. Am J Neurodegener Dis 2014; 3: 19–32. [PMC free article] [PubMed] [Google Scholar]
  33. Marques F, Sousa JC, Sousa N, Palha JA. Blood-brain-barriers in aging and in Alzheimer’s disease. Mol Neurodegener 2013; 8: 38. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Miller MC, Tavares R, Johanson CE, Hovanesian V, Donahue JE, Gonzalez L, et al. Hippocampal RAGE immunoreactivity in early and advanced Alzheimer’s disease. Brain Res 2008; 1230: 273–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Miners JS, Palmer JC, Love S. Pathophysiology of hypoperfusion of the precuneus in early Alzheimer’s disease. Brain Pathol 2016; 26: 533–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Miners JS, Schulz I, Love S. Differing associations between Ab accumulation, hypoperfusion, blood-brain barrier dysfunction and loss of PDGFRB pericyte marker in the precuneus and parietal white matter in Alzheimer’s disease. J Cerebr Blood F Met 2018; 38: 103–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Mirra SS, Heyman A, McKeel D, Sumi SM, Crain BJ, Brownlee LM, et al. The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD). Part II. Standardization of the neuropathologic assessment of Alzheimer’s disease. Neurology 1991; 41: 479–86. [DOI] [PubMed] [Google Scholar]
  38. Montagne A, Barnes SR, Sweeney MD, Halliday MR, Sagare AP, Zhao Z, et al. Blood-brain barrier breakdown in the aging human hippocampus. Neuron 2015; 85: 296–302. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Montagne A, Zhao Z, Zlokovic BV. Alzheimer’s disease: a matter of blood-brain barrier dysfunction? J Exp Med 2017; 214: 3151–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Mooradian AD, Chung HC, Shah GN. GLUT-1 expression in the cerebra of patients with Alzheimer’s disease. Neurobiol Aging 1997; 18: 469–74. [DOI] [PubMed] [Google Scholar]
  41. Murray ME, Dickson DW. Is pathological aging a successful resistance against amyloid-beta or preclinical Alzheimer’s disease? Alzheimers Res Ther 2014; 6: 24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Nelson PT, Alafuzoff I, Bigio EH, Bouras C, Braak H, Cairns NJ, et al. Correlation of Alzheimer disease neuropathologic changes with cognitive status: a review of the literature. J Neuropathol Exp Neurol 2012; 71: 362–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Ramirez J, McNeely AA, Scott CJM, Masellis M, Black SE. White matter hyperintensity burden in elderly cohort studies: The Sunnybrook Dementia Study, Alzheimer’s Disease Neuroimaging Initiative, and Three-City Study. Alzheimers Dement 2016; 12: 203–10. [DOI] [PubMed] [Google Scholar]
  44. Romanitan MO, Popescu BO, Spulber S, Bajenaru O, Popescu LM, Winblad B, et al. Altered expression of claudin family proteins in Alzheimer’s disease and vascular dementia brains. J Cell Mol Med 2010; 14: 1088–100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Romanitan MO, Popescu BO, Winblad B, Bajenaru OA, Bogdanovic N. Occludin is overexpressed in Alzheimer’s disease and vascular dementia. J Cell Mol Med 2007; 11: 569–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Serrano-Pozo A, Frosch MP, Masliah E, Hyman BT. Neuropathological alterations in Alzheimer disease. Cold Spring Harb Perspect Med 2011; 1: a006189. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Shinohara M, Fujioka S, Murray ME, Wojtas A, Baker M, Rovelet-Lecrux A, et al. Regional distribution of synaptic markers and APP correlate with distinct clinicopathological features in sporadic and familial Alzheimer’s disease. Brain 2014; 137 (Pt 5): 1533–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Shinohara M, Koga S, Konno T, Nix J, Shinohara M, Aoki N, et al. Distinct spatiotemporal accumulation of N-truncated and full-length amyloid-beta42 in Alzheimer’s disease. Brain 2017; 140: 3301–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Shinohara M, Murray ME, Frank RD, DeTure M, Yamazaki Y, Tachibana M, et al. Impact of sex and APOE4 on cerebral amyloid angiopathy in Alzheimer’s disease. Acta Neuropathol 2016; 132: 225–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Shinohara M, Petersen RC, Dickson DW, Bu G. Brain regional correlation of amyloid-beta with synapses and apolipoprotein E in non-demented individuals: potential mechanisms underlying regional vulnerability to amyloid-beta accumulation. Acta Neuropathol 2013; 125: 535–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Storck SE, Meister S, Nahrath J, Meissner JN, Schubert N, Di Spiezio A, et al. Endothelial LRP1 transports amyloid-beta(1–42) across the blood-brain barrier. J Clin Invest 2016; 126: 123–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Terry RD, Masliah E, Salmon DP, Butters N, DeTeresa R, Hill R, et al. Physical basis of cognitive alterations in Alzheimer’s disease: synapse loss is the major correlate of cognitive impairment. Ann Neurol 1991; 30: 572–80. [DOI] [PubMed] [Google Scholar]
  53. Thomas T, Miners S, Love S. Post-mortem assessment of hypoperfusion of cerebral cortex in Alzheimer’s disease and vascular dementia. Brain 2015; 138 (Pt 4): 1059–69. [DOI] [PubMed] [Google Scholar]
  54. Toledo JB, Arnold SE, Raible K, Brettschneider J, Xie SX, Grossman M, et al. Contribution of cerebrovascular disease in autopsy confirmed neurodegenerative disease cases in the National Alzheimer’s Coordinating Centre. Brain 2013; 136 (Pt 9): 2697–706. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Vemuri P, Lesnick TG, Przybelski SA, Knopman DS, Preboske GM, Kantarci K, et al. Vascular and amyloid pathologies are independent predictors of cognitive decline in normal elderly. Brain 2015; 138 (Pt 3): 761–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Vlassenko AG, Vaishnavi SN, Couture L, Sacco D, Shannon BJ, Mach RH, et al. Spatial correlation between brain aerobic glycolysis and amyloid-beta (Abeta) deposition. Proc Natl Acad Sci USA 2010; 107: 17763–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Walchli T, Wacker A, Frei K, Regli L, Schwab ME, Hoerstrup SP, et al. Wiring the vascular network with neural cues: a CNS perspective. Neuron 2015; 87: 271–96. [DOI] [PubMed] [Google Scholar]
  58. Wang M, Roussos P, McKenzie A, Zhou X, Kajiwara Y, Brennand KJ, et al. Integrative network analysis of nineteen brain regions identifies molecular signatures and networks underlying selective regional vulnerability to Alzheimer’s disease. Genome Med 2016; 8: 104. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Winkler EA, Nishida Y, Sagare AP, Rege SV, Bell RD, Perlmutter D, et al. GLUT1 reductions exacerbate Alzheimer’s disease vasculo-neuronal dysfunction and degeneration. Nat Neurosci 2015; 18: 521–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Yamazaki Y, Baker DJ, Tachibana M, Liu CC, van Deursen JM, Brott TG, et al. Vascular cell senescence contributes to blood-brain barrier breakdown. Stroke 2016; 47: 1068–77. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Yamazaki Y, Kanekiyo T. Blood-brain barrier dysfunction and the pathogenesis of Alzheimer’s disease. Int J Mol Sci 2017; 18: 1965. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Yarchoan M, Xie SX, Kling MA, Toledo JB, Wolk DA, Lee EB, et al. Cerebrovascular atherosclerosis correlates with Alzheimer pathology in neurodegenerative dementias. Brain 2012; 135 (Pt 12): 3749–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Ye BS, Seo SW, Kim GH, Noh Y, Cho H, Yoon CW, et al. Amyloid burden, cerebrovascular disease, brain atrophy, and cognition in cognitively impaired patients. Alzheimers Dement 2015; 11: 494–503.e3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  64. Zenaro E, Piacentino G, Constantin G. The blood-brain barrier in Alzheimer’s disease. Neurobiol Dis 2017; 107: 41–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Zhao Z, Nelson AR, Betsholtz C, Zlokovic BV. Establishment and dysfunction of the blood-brain barrier. Cell 2015; 163: 1064–78. [DOI] [PMC free article] [PubMed] [Google Scholar]
  66. Zlokovic BV. Neurovascular mechanisms of Alzheimer’s neurodegeneration. Trends Neurosci 2005; 28: 202–8. [DOI] [PubMed] [Google Scholar]
  67. Zlokovic BV. The blood-brain barrier in health and chronic neurodegenerative disorders. Neuron 2008; 57: 178–201. [DOI] [PubMed] [Google Scholar]
  68. Zlokovic BV. Neurovascular pathways to neurodegeneration in Alzheimer’s disease and other disorders. Nat Rev Neurosci 2011; 12: 723–38. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Figures
Click here for additional data file. (725.8KB, pdf)

Data Availability Statement

The data that support the findings of this study are available from the corresponding author, upon reasonable request.

Articles from Brain are provided here courtesy of Oxford University Press

RESOURCES