Journal of the Neurological Sciences 322 (2012) 184–186
Contents lists available at SciVerse ScienceDirect
Journal of the Neurological Sciences journal homepage: www.elsevier.com/locate/jns
Claudin expression profile separates Alzheimer's disease cases from normal aging and from vascular dementia cases Stefan Spulber a, Nenad Bogdanovic b, Mihaela Oana Romanitan c, Ovidiu A. Bajenaru c, Bogdan O. Popescu c, d,⁎ a
Department of Neuroscience, Karolinska Institutet, Stockholm, Sweden Division of Experimental Geriatrics, Alzheimer's Disease Research Center, Department of NVS, Karolinska University Hospital, Karolinska Institutet, Stockholm, Sweden Department of Neurology, University Hospital Bucharest, ‘Carol Davila’ University of Medicine and Pharmacy, Bucharest, Romania d Laboratory of Molecular Medicine, ‘Victor Babeş’ National Institute of Pathology, Bucharest, Romania b c
a r t i c l e
i n f o
Available online 3 June 2012 Keywords: Alzheimer's disease Tight junction proteins Claudin Multivariate analysis PLS-DA
a b s t r a c t We have reported earlier that tight-junction proteins are detectable by standard immunohistochemistry in the brain parenchyma, namely in the cell bodies of neurons, astrocytes, and oligodendrocytes. Here we show, by projection to latent structures — discriminate analysis (PLS-DA), that the immunohistochemical detection profile of tight junction proteins clearly distinguishes the AD cases from healthy aging controls and from the cases of dementia with a predominantly vascular pathology underlying the symptoms (vascular dementia, VaD; cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy, CADASIL; and cerebral amyloid angiopathy, CAA). Our findings might be valuable in the perspective of developing biomarkers for AD. © 2012 Elsevier B.V. All rights reserved.
1. Introduction In order to function properly, the neurovascular unit needs a complex regulation and a strict separation between the brain tissue and the circulatory space. The blood–brain barrier (BBB) is a tightly controlled, two-way diffusion barrier, composed of capillary endothelial cells, basal lamina, astrocytes and pericytes, which has such separating properties. The endothelial cell layer limits the paracellular passage of ions and larger molecules due to peculiar tight junctions (TJs), composed of transmembrane proteins such as occludin, claudins (CLs), and junction adhesion molecules [1]. Only two types of brain cells possess TJs or TJ-like structures: vascular endothelial cells and oligodendrocytes [2]. However, occludin and different CL species are expressed in brain cells without TJs, such as neurons and astrocytes, and their expression increase in Alzheimer's disease (AD) and vascular dementia (VaD), suggesting that they might have roles in cell stress response as well [3,4]. Moreover, disruption of the BBB, associated with a change in TJs properties, is extensively documented in both vascular and degenerative brain disorders [5]. The presence of perivascular pathology in addition to the classical hallmarks of AD, together with the inherent difficulties to distinguish AD from vascular dementia based exclusively on the clinical presentation led to the hypothesis that the vascular pathology plays an ⁎ Corresponding author at: ‘Victor Babeş’ National Institute of Pathology, Laboratory of Molecular Medicine, 99‐101, Spl. Independenței, sector 5, 050096 Bucharest, Romania. Tel.: + 40 744 353433. E-mail address:
[email protected] (B.O. Popescu). 0022-510X/$ – see front matter © 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.jns.2012.05.031
important role in the pathogenesis and the evolution of AD. Here we show that the immunohistochemical detection profile of tight junction proteins clearly distinguishes the AD cases from healthy aging controls and from the cases of dementia with a predominantly vascular pathology underlying the symptoms (VaD; cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy, CADASIL; and cerebral amyloid angiopathy, CAA). 2. Materials and methods We have reported earlier that tight-junction proteins are detectable by standard immunohistochemistry in the brain parenchyma, namely in the cell bodies of neurons, astrocytes, and oligodendrocytes. The staining and counting of immunoreactive cells is described elsewhere [4]. As a further development of the study, we report here the results of using the same data in a multivariate projection-based statistical model that provides a novel view on the relationships among variables as well as among the disease cases included in the analyses. Thus, we have used the expression profile of CLs in neurons, astrocytes and oligodendrocytes as input for data for projection to latent structures — discriminate analysis (PLS-DA) (for PLS-DA principles and application see [6–9]). Multivariate projection techniques have been introduced by Pearson [10] as a powerful method to describe in a synthetic manner the information contained in complex data tables. Thus, principal component analysis (PCA) quantifies the amount of information contained in a data table and provides information about the contribution of each variable to the global variance of the dataset, as well as about which variables have similar contributions (i.e. are
S. Spulber et al. / Journal of the Neurological Sciences 322 (2012) 184–186
correlated to one another). A further development of the model takes advantage of group belonging for each sample and rotates the PCA model in order to maximize the separation among classes (PLS-DA). Therefore, PLS-DA is a supervised projection-based statistical model that estimates (1) the degree to which the spread in the input dataset account for the separation among groups; and (2) which variables are the most important for explaining the separation.
185
Table 1 Description of models.
Number of groups Number of principal components extracted Proportion of global variance accounted for Overall significance (Q2)
Model 1
Model 2
5 4 71.91% 0.31
3 4 91.29% 0.68
3. Results We sequentially tested several PLS-DA models, based on the clinical and pathological features of the diseases included in the analysis. Thus, for the first model (Fig. 1A–C), the cases were assigned to one of 5 distinct groups (Control, AD, CADASIL, VaD, CAA; see also Table 1). In this model, the CL's expression profile projected the cases in three main clusters (one for controls, one for AD cases, and one for the other three types of pathology included) mapped in distinct quadrants. We then tested whether reassigning the cases to different groups improves the model. The predictive value of the model displayed the largest improvement as compared to the original 5-group model when we pooled the VaD, CADASIL, and CAA in one single group (dementias with vascular etiology, VaDs) (Fig. 1D–F). Thus, the proportion of global variance accounted for by the multivariate model increased from 71.91% to 91.29%, and the overall significance of the model (Q2) improved from 0.31 to 0.68. Another notable model improved compared to the original 5-group model was obtained by assigning the AD and
A
CAA cases to one group, controls to the second group, and the rest of the vascular dementias to the third group (proportion of global variance accounted for by the model: 83.47%; Q2 =0.58). Yet, the number of CAA cases available in our study was rather low and cannot support further statistical modeling investigation. All other models tested that assigned the cases to other groups resulted in a decrease of the explanatory power of the model (data not shown). 4. Discussion It has been speculated that the pathology of AD has an important vascular component contributing to the progression of the disease (reviewed in [11,12]). However, we found that the expression profile of TJ proteins clearly separates AD from controls and from other vascular brain diseases included in this study. In contrast, CAA, CADASIL, and VaD appear to have very similar CL expression profiles as compared to AD cases and healthy controls. As an exception, the expression profile of
D
4
4 2
0
Ctrl AD VaD CAD CAA
-2 -4
-6
-4
-2
0
2
4
t2
t2
2
0 Ctrl AD VaDs
-2 -4 -6
6
-4
-2
2
4
6
E
B .8
.8 O_CL2
A_CL11 .4 N_CL11 O_CL2 A_CL2 0 N_CL5 N_CL2 A_CL5 -.4 O_CL11 O_CL5 -.8
p2
p2
.4 A_CL11 A_CL2 N_CL11 0 N_CL5 N_CL2 -.4 O_CL11 A_CL5 O_CL5 -.8 -.8
-.4
0
.4
.8
-.8
-.4
0
p1
.4
.8
p1
C
A_CL5
N_CL5
A_CL2
O_CL2
O_CL5
A_CL11
0
N_CL2
.2
N_CL11
A_CL2
O_CL5
A_CL5
N_CL11
N_CL5
N_CL2
O_CL11
0
O_CL2
.2
.4
O_CL11
.4
Variable importance
F
A_CL11
Variable importance
0
t1
t1
Fig. 1. PLS-DA analysis of the expression of claudins in neurons, astrocytes, and oligodendrocytes. The first model tested (A–C) had the cases assigned to 5 groups according to the pathology. The score plot for the first model (A) maps the controls clearly separated from the pathological samples, and shows clustering of the cases with defined vascular pathology. The loading plot for the first model (B) describes the weight of each variable in the principal components plotted in (A). (C), the importance of each variable to the whole model. The second model tested (D–F) had all cases assigned to 3 groups of pathology. The score plot for the second model (D) shows that pooling vascular dementia, CADASIL and congophylic angiopathy in one group (VaDs) improves the separation between AD, controls, and other types of vascular pathology. Note that the loading plot for the second model (E) is only subtly different from the first model. (F), variable importance for each variable in the second model. (C, F), VIP values above 0.22 (dashed line) indicate significant influence. Note that CL11 has the largest influence in the second model. Abbreviations: Ctrl — healthy controls; AD — Alzheimer's disease; VaD — vascular dementia; CAD — CADASIL; CAA — cerebral amyloid angiopathy; CL — claudin; N/A/O — neurons/astrocytes/oligodendrocytes.
186
S. Spulber et al. / Journal of the Neurological Sciences 322 (2012) 184–186
claudins in CAA appears to resemble the profile found in AD cases. This is in agreement with earlier reports on perivascular drainage impairment in AD and CAA as contributing to the accumulation of amyloid pathology [13,14]. Our results suggest that increased CL expression might be a part of the cellular response to stress and different CL species overexpression might be triggered by different pathological molecular operators. The comparison of AD with VaD suggests that vascular pathology may account for part of the changes in CL expression profile, but AD pathology is able to further alter this profile. Interestingly, CL family protein expression is also altered in cancerous cells, and CLs might play a role in carcinogenesis and in the metastatic process [15], data which support their role in cell cycle/apoptosis regulation. The most important factor for cluster differentiation appears to be the expression of CL11 in oligodendrocytes, astrocytes, or neurons. CL11 (oligodendrocyte specific protein) is expressed in autotypic TJs of myelin sheaths in brain and has a role in rapid nerve conduction principally for small diameter myelinated axons, which help information processing, learning and memory [16]. Therefore, AD pathology might trigger overexpression of CL11 in oligodendrocytes as an endogenous defense response. In addition, follicle stimulating hormone (FSH) and luteinizing hormone (LH) stimulate CL11 expression, and their serum levels are significantly increased in AD patients [17]. The gonadotropins might be responsible for the increased CL11 expression in neurons, since their receptors are present in the neuronal populations affected in AD [18]. A recent report demonstrated an upregulation of CL11 in type II astrocytes differentiated from oligodendrocyte precursor cells [19]. These results are in line with our data, supporting alternative functions of CLs, in addition to regulation of TJs. In conclusion, we demonstrated here that AD, VaD and aging Controls have distinct profile of CL expression in different brain cell types. Our finding might be valuable in the perspective of multiple dementia biomarkers development. Conflicts of interest The authors declare no conflict of interest. Acknowledgements This paper is supported by the Sectorial Operational Programme Human Resources Development (SOP HRD), financed from the European Social Fund and by the Romanian Government under the contract number POSDRU/89/1.5/S/64109 and by the Executive Unit for Financing Higher Education, Research, Development and Innovation — Romania
(UEFISCDI) and by Program 4 (Partnerships in Priority Domains), grant nr. 41-013/2007.
References [1] Kniesel U, Wolburg H. Tight junctions of the blood–brain barrier. Cell Mol Neurobiol 2000;20:57–76. [2] Rubin LL. The blood–brain barrier in and out of cell culture. Curr Opin Neurobiol 1991;1:360–3. [3] 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. [4] Romanitan MO, Popescu BO, Spulber S, Băjenaru 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. [5] Popescu BO, Toescu EC, Popescu LM, Bajenaru O, Muresanu DF, Schultzberg M, et al. Blood–brain barrier alterations in ageing and dementia. J Neurol Sci 2009;283: 99–106. [6] Westman E, Spenger C, Wahlund L, Lavebratt C. Carbamazepine treatment recovered low N-acetylaspartate+N-acetylaspartylglutamate (tNAA) levels in the megencephaly mouse BALB/cByJ-Kv1.1(mceph/mceph). Neurobiol Dis 2007;26:221–8. [7] Spulber S, Bartfai T, Winblad B, Schultzberg M. Morphological and behavioral changes induced by transgenic overexpression of interleukin-1ra in the brain. J Neurosci Res 2011;89:142–52. [8] Rüetschi U, Zetterberg H, Podust VN, Gottfries J, Li S, Hviid Simonsen A, et al. Identification of CSF biomarkers for frontotemporal dementia using SELDI-TOF. Exp Neurol 2005;196:273–81. [9] Oberg J, Spenger C, Wang F, Andersson A, Westman E, Skoglund P, et al. Age related changes in brain metabolites observed by 1H MRS in APP/PS1 mice. Neurobiol Aging 2008;29:1423–33. [10] Pearson K. On lines and planes of closest fit to systems of points in space. Philos Mag 1901;2:559–72. [11] van Norden AGW, van Dijk EJ, de Laat KF, Scheltens P, Olderikkert MGM, de Leeuw FE. Dementia: Alzheimer pathology and vascular factors: from mutually exclusive to interaction. Biochim Biophys Acta 2012;1822:340–9. [12] Bell RD, Zlokovic BV. Neurovascular mechanisms and blood–brain barrier disorder in Alzheimer's disease. Acta Neuropathol 2009;118:103–13. [13] Weller RO, Subash M, Preston SD, Mazanti I, Carare RO. Perivascular drainage of amyloid-beta peptides from the brain and its failure in cerebral amyloid angiopathy and Alzheimer's disease. Brain Pathol 2008;18:253–66. [14] Hawkes CA, Härtig W, Kacza J, Schliebs R, Weller RO, Nicoll JA, et al. Perivascular drainage of solutes is impaired in the ageing mouse brain and in the presence of cerebral amyloid angiopathy. Acta Neuropathol 2011;121:431–43. [15] Escudero-Esparza A, Jiang WG, Martin TA. The Claudin family and its role in cancer and metastasis. Front Biosci 2011;16:1069–83. [16] Devaux J, Gow A. Tight junctions potentiate the insulative properties of small CNS myelinated axons. J Cell Biol 2008;183:909–21. [17] Bowen RL, Isley JP, Atkinson RL. An association of elevated serum gonadotropin concentrations and Alzheimer disease? J Neuroendocrinol 2000;12:351–4. [18] Casadesus G, Garrett MR, Webber KM, Hartzler AW, Atwood CS, Perry G, et al. The estrogen myth: potential use of gonadotropin-releasing hormone agonists for the treatment of Alzheimer's disease. Drugs R&D 2006;7:187–93. [19] Hu J, Wang Y, Zhou J, Chen C, Wang F, Li X, et al. Differential gene expression in oligodendrocyte progenitor cells, oligodendrocytes and type II astrocytes. Tohoku J Exp Med 2011;223:161–76.
