Changes in Expression Profiles Revealed by Transcriptomic Analysis in Peripheral Blood Mononuclear Cells of Alzheimer’s Disease Patients

Abstract

Alzheimer’s disease (AD) is an age-related neurodegenerative pathology associated with accumulation of DNA damage. Inflammation and cell cycle alterations seem to be implicated in the pathogenesis of AD, although the molecular mechanisms have not been thoroughly elucidated to date. The aim of the present study was to evaluate whether peripheral blood mononuclear cells (PBMCs) of AD patients display alterations in gene expression profiles, focusing on finding markers that might improve the diagnosis of AD. Blood samples were collected from 22 AD patients and 13 healthy individuals to perform genome-wide mRNA expression. We found 593 differentially expressed genes in AD compared to controls, from which 428 were upregulated, and 165 were downregulated. By performing a gene set enrichment analysis, we observed pathways involved in inflammation, DNA damage response, cell cycle, and neuronal processes. Moreover, functional annotation analyses indicated that differentially expressed genes are strongly related to pathways associated with the cell cycle and the immune system. The results were compared with those of an independent study on hippocampus samples, and a number of genes in common between both studies were identified as potential peripheral biomarkers for AD, including DUSP1, FOS, SLC7A2, RGS1, GFAP, CCL2, ANGPTL4, and SSPN. Taken together, our results demonstrate that PBMCs of AD patients do present alterations in gene expression profiles, and these results are comparable to those previously reported in the literature for AD neurons, supporting the hypothesis that blood peripheral mononuclear cells express molecular changes that occur in the neurons of AD patients.

Keywords

Alzheimer’s disease gene expression profiles neurodegenerative disorders peripheral blood cells

INTRODUCTION

Alzheimer’s disease (AD) is an age-related chronic neurodegenerative disorder that has an immense impact on the healthcare system, considering the overall increasing age of the population worldwide [1]. The pathophysiology of AD is complex, and many hypotheses have been formulated to explain the basis and development of this disease. Several alterations in AD patients have been reported, such as protein deposition, neuronal cytoskeleton modifications, neuronal and white matter loss, mitochondrial dysfunction, ion calcium dysregulation, inflammation, cell cycle dysfunction, and others [2 –7]. Thus, AD characteristics suggest that many pathways are dysregulated in the pathology [8], and attempts have been made to define which factors are involved as causal, and which might be involved as a consequence of AD. However, despite extensive studies in recent decades, the pathogenesis of AD remains to be elucidated.

Moreover, an accurate diagnosis of AD is complicated and depends on clinical and neuropsychological examinations, and requires the exclusion of other types of dementia [9, 10]. Thus, AD diagnosis is generally late and inaccurate. However, it has been proposed that early diagnosis and early treatment are crucial to achieving a better response and delaying the progress of the disease [11].

Imaging techniques and cerebrospinal fluid (CSF) biomarkers have improved diagnostics. However, those methods are either costly or invasive, posing a significant limitation for clinical applications [12]. The use of blood cells may offer possibilities to improve AD diagnosis in clinical practice. The easier sampling of blood compared to CSF and the availability of sensitive methods may allow detection of AD alterations at early phases of the disease.

The use of blood cells for AD diagnostics has already been studied by Booij et al. [12] and Rye et al. [13]. The first work performed a whole genome screening using an oligonucleotide microarray analysis on blood samples from a large cohort of AD patients and healthy cognitive controls. The authors accurately identified and validated gene expression signatures of AD patients at early stages of the disease. Moreover, Rye et al. [13] selected some genes that were already pointed by Booij et al. [12] and created a 96-gene expression array test to detect early stages of AD, and they obtained 80% clinical accuracy in AD diagnostic. Although these results are promising, further studies are needed to confirm those findings in other populations worldwide.

In the present study, we aimed to compare gene expression profiles of peripheral blood mononuclear cells (PBMCs) from Brazilian AD patients and age-matched controls to identify differentially expressed genes that might be relevant to the etiology and/or progression of the disease, thus providing information to validate PBMCs as a reliable tool for AD diagnostics. As a whole, our results showed that PBMCs present alterations in biological pathways that correlate with functional modifications observed in AD patients, supporting the idea that blood cells are an important source of information regarding the changes that occur in AD brains, reinforcing AD as a systemic disease. Furthermore, we found eight genes (DUSP1, FOS, SLC7A2, RGS1, GFAP, CCL2, ANGPTL4, and SSPN) that were upregulated in both blood cells and hippocampus regions, and they are candidates to be further investigated in the search for biomarkers that may characterize AD.

MATERIALS AND METHODS

Subjects

The patients were selected and recruited from the Geriatrics section of the Clinical Hospital at Ribeirão Preto Medical School, University of São Paulo, Brazil. Blood samples were collected from a total of 22 AD patients and 13 age-matched healthy controls. All AD patients were diagnosed based on the criteria established by the DSM-IV (Diagnostic and Statistical Manual of Mental Disorders 4th Edition) and NINCDS-ADRDA (National Institute of Neurological and Communicative Disorders and Stroke and The Alzheimer’s Disease Related Disorders Association). It should be noted that neuropathological exams were not performed, and patients were classified as probable sporadic AD once their symptoms appeared later than 65 years of age (late onset). The AD group was divided into two subgroups according to their Clinical Dementia Rating (CDR), which is the cognitive impairment scale most often used to determine the clinical phase of the dementia. In this study, 15 individuals were classified as CDR1, or mild AD, and 7 as CDR3, or severe AD. Age-matched controls were also recruited while undergoing regular follow-up at the Clinical Hospital. Control individuals were also submitted to all cognitive evaluations and were considered to present normal memory and cognitive capacity. AD and control individuals underwent a series of clinical exams to exclude other diseases. Smokers and individuals with other dementia subtypes were also excluded. Only patients with low Hachinski ischemic scores were included. Characteristics presented by patients are shown in Table 1. The local Ethical Committee from the Clinical Hospital - Ribeirão Preto Medical School, USP, approved the study and informed consents were obtained from all subjects enrolled in the study.

Table 1

Characteristics of AD and controls samples utilized in microarrays

Sample^a	Age (y)	Gender	CDR^b	MMSE^c	Education (y)	Ethnic Background (according to self-report)
PDA06	83	M	3	3	–	white
PDA09	91	F	1	14	–	white
PDA11	80	F	1	23	4	white
PDA12	73	M	1	17		white
PDA13	76	F	3	9	8	white
PDA26	84	F	1	17	4	white
PDA30	77	F	1	–	3	white
PDA33	91	M	1	17	2	white
PDA35	65	F	1	–	4	white
PDA36	73	M	1	20	2	white
PDA37	82	F	1	–	8	white
PDA39	77	M	1	23	2	white
PDA40	87	F	1	19	2	white
PDA41	71	F	3	6	2	white
PDA42	76	M	1	21	4	black
PDA43	77	M	3	5	4	black
PDA45	74	F	3	6	5	white
PDA46	79	F	3	9	2	brown
PDA47	79	F	1	14	2	white
PDA48	79	F	3	10	1	brown
PDA50	89	M	1	18	2	white
PDA51	83	F	1	19	–	white
CDA02	69	F	–	–	–	white
CDA07	69	F	–	27	12	white
CDA08	76	F	–	29	8	white
CDA10	78	F	–	28	3	white
CDA18	83	M	–	–	–	white
CDA19	73	F	–	–	–	white
CDA20	72	M	–	29	3	white
CDA21	83	F	–	–	–	white
CDA22	85	F	–	28	4	white
CDA23	84	F	–	27	12	white
CDA24	84	M	–	28	11	white
CDA25	70	F	–	28	3	white
CDA26	79	F	–	27	11	white

^aPDA, patients with DA and CDA: age-matched healthy controls. ^bCDR, Clinical Dementia Rating (1-mild, 2-moderate, 3-severe). ^cMMSE (Mini-Mental State Examination): exam extensively used to estimate the severity of the cognitive impairment in dementia.

Blood collection and isolation of peripheral blood mononuclear cells

Peripheral blood from all patients was collected by venipuncture in vacuum blood tubes containing EDTA (0.054 mL/tube) (BD Vacutainer). The PBMCs were then separated by density gradient using Ficoll-Hypaque (Sigma, St. Louis, MO), following the manufacturer’s instructions.

RNA extraction

Total RNA isolation was performed using the Trizol reagent (Invitrogen, Carlsbad, CA), according to the manufacturer’s protocol. Samples were quantified and evaluated for purity using a NanoDrop ND-1000 (NanoDrop Products, Wilmington, DE). Sample integrity was also evaluated by electrophoresis in a nanofluidic device, the Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA). RNA samples used for microarray analysis were suitable for the experiments only when the RNA integrity number (RIN) was higher than 7.0.

Microarrays

PBMCs expression profiles were carried out using the Human GE 4×44K V2 Microarray kit (G4845A) and the QuickAmp Labeling (monocolor system-Cy3) kit (Agilent Technologies, Santa Clara, CA), according to the manufacturer’s protocol. The microarray slides cover all transcripts already described in humans. Each slide contains 4 arrays comprising 44,000 oligonucleotides representing 27,958 targets and positive controls. Briefly, RNA samples were initially submitted to cDNA synthesis by reverse transcription, followed by generation of fluorescent cRNA. cRNA was purified, and samples were then prepared and added to the slides for the hybridization process. The slides were incubated for 17 h in a hybridization oven and were subsequently washed and scanned in a high-resolution scanner Surescan (Agilent Technologies). Data from the samples included in this study are available at the public database ArrayExpress (EMBL-EBI) (https://www.ebi.ac.uk/arrayexpress/) under accession number E-MTAB-6094.

Normalization and statistical analysis

The raw data were obtained using the Feature Extraction software v.10.7 (Agilent Technologies) and analyzed in the R environment v. 3.3.0 (https://www.r-project.org/). The median signals (gMedianSignal and gBGMedianSignal) were used. Following background subtraction and log2 scale transformation, normalization was performed using the quantile method with the aroma. light package (http://www.bioconductor.org/packages/release/bioc/html/aroma.light.html). Differentially expressed mRNAs were obtained by Rank product analysis using the RankProd package [14], considering p-values and false positive predictions (pfp) ≤0.05. Moreover, we compared the data obtained in this work with the data provided in the Supplementary material of Miller et al., in which they compared hippocampus regions of AD patients and age-related controls [15].

Functional annotation analysis

Genes considered differentially expressed were submitted to functional enrichment analysis using DAVID (Database for Annotation Visualization and Integrated Discovery, http://david.abcc.ncifcrf.gov/) [16], a tool that identifies biological processes associated with differentially expressed genes by integrating a range of databases. Two types of analyses were carried out, considering biological functions according to Gene Ontology, BBID, Biocarta, and KEGG databases. Then, we performed an analysis in which the OMIM and Genetic Association databases were used to identify diseases related to the differentially expressed genes obtained in our work. Only pathways that reached a false discovery rate (FDR) < 25% and p < 0.05 were selected.

Gene set enrichment analysis (GSEA)

The normalized data was used as input into GSEA (Broad Institute). This tool is suitable to evaluate expression variations in gene sets that could be relevant to identify altered biological processes in the comparison between AD and the control group. GSEA considers minimal variations in gene expression levels that are significant when a gene set is considered. The computational method developed and described by Subramanian et al. [17] was applied to define whether a gene set that was a priori defined according to standard databases (KEGG, REACTOME, and BIOCARTA) shows statistical significance for the comparison between two biological states (AD versus controls). The analysis report comprised values for each gene set as an enrichment score, a normalized enrichment score, a false discovery rate, and a nominal p value. Then, this study considered pathways enriched according to the scores obtained for p-values < 0.05 and FDR <25%.

RESULTS

Whole genome RNA microarray analysis was performed to compare profiles of transcriptional gene expression displayed by PBMCs of AD patients compared to age-matched controls. The results generated by bioinformatics and statistical analyses indicated a list of 593 differentially expressed genes: 428 upregulated and 165 downregulated in the AD group compared to its control group (Supplementary Table 1). These lists were submitted to the DAVID tool for identifying pathways associated with those genes; the functional enrichment analysis applied to the set of 428 upregulated genes in AD indicated enriched pathways related to inflammation, immunological system, cell cycle and proliferation (Supplementary Figure 1). For the list of 165 downregulated genes, we found enriched pathways related to gas transport, hemoglobin metabolism and cell signaling (Supplementary Figure 1). Moreover, by performing GSEA on the same data set, we found 70 pathways that were enriched (Table 2 and Supplementary Table 2), such as cell cycle control and immune system.

Table 2

Enriched pathways obtained in the comparison of AD and age-matched controls after performing the analysis of transcript expression by using the Gene Set Enrichment Analysis (GSEA) tool

Pathway description	Number of genes	p-value	FDR q-value	Database
Cell cycle mitotic	285	0.000	0.009	REACTOME
Mitotic M G1 phases	150	0.000	0.006	REACTOME
Mitotic prometaphase	91	0.000	0.000	REACTOME
Cell cycle pathway	22	0.002	0.023	BIOCARTA
M phase of mitotic cell cycle	80	0.000	0.001	GENE ONTOLOGY
Mitosis	77	0.000	0.001	GENE ONTOLOGY
M phase	105	0.000	0.001	GENE ONTOLOGY
Regulation of cytokine biosynthetic process	38	0.000	0.017	GENE ONTOLOGY
Viral genome replication	21	0.000	0.018	GENE ONTOLOGY
Positive regulation of cytokine biosynthetic process	25	0.000	0.019	GENE ONTOLOGY
Cell cycle process	182	0.000	0.020	GENE ONTOLOGY
Calcium ion transport	24	0.000	0.021	GENE ONTOLOGY
Cell cycle phase	160	0.000	0.021	GENE ONTOLOGY
Response to bacterium	31	0.000	0.022	GENE ONTOLOGY
Regulation of mitosis	40	0.002	0.023	GENE ONTOLOGY
Positive regulation of translation	34	0.000	0.024	GENE ONTOLOGY
Chromosome segregation	32	0.002	0.026	GENE ONTOLOGY
Cytokine production	72	0.000	0.030	GENE ONTOLOGY
Cytokine biosynthetic process	41	0.000	0.031	GENE ONTOLOGY
Defense response to bacterium	25	0.002	0.036	GENE ONTOLOGY
Di Tri valent inorganic cation transport	29	0.005	0.037	GENE ONTOLOGY

p < 0.05 and FDR < 5%.

Considering the heterogeneity of the AD group composed of patients with mild (CDR1) and severe (CDR3) AD, the results were analyzed aiming to detect differences in gene expression levels according to the disease stages. A comparison of the 15 CDR1 patient subgroup (n = 15) with the control group (n = 13) generated a list of 449 differentially expressed genes (Supplementary Table 3). When the CDR3 (n = 7) and control groups were compared we found 251 differentially expressed genes (p < 0.05) (Supplementary Table 4). Of the genes differentially expressed between either CDR1 or CDR3 AD subgroups and the control group, 116 were common to both AD subgroups (Fig. 1).

Fig. 1.

Amount of differentially expressed genes selected in mild AD (CDR1) and severe AD (CDR3) compared to control group. Venn diagram shows the number of differentially expressed genes in each group, and common genes between the two comparisons: CDR1 versus age-matched controls and CDR3 versus age-matched controls.

Furthermore, 418 differentially expressed genes were found comparing CDR1 and CDR3, of which 332 genes were upregulated and 86 were downregulated in CDR1 relatively to CDR3 (Supplementary Table 5). For the CDR1 group, in general, the upregulated genes were related to cell proliferation, immune system, and neurological functions, while the functions of downregulated genes are linked to cell signaling and hemoglobin metabolism. Considering the upregulated genes in the CDR3 group compared to the CDR1 group, the biological classes were mostly related to neurological function and inflammation, while the downregulated genes are involved in pathways such as ion channels regulation, gas transport and olfactory transduction (Supplementary Figure 2).

In addition, functional enrichment analyses applied to the upregulated genes selected for AD subgroups (CDR1 compared to CDR3) showed pathways linked to neurological process (synaptic and nerve impulse regulation) and inflammation (Supplementary Figure 3). On the other hand, functions attributed to downregulated genes were associated with ionic regulation, cytoskeleton, gas transport, cell signaling and olfactory transduction (Supplementary Figure 3). GSEA was also applied to compare CDR1 versus CDR3, and several pathways were selected as enriched, some of which were also related to cell cycle regulation and neuronal function (Table 3).

Table 3

Pathways obtained from GSEA analysis for the comparison between CDR1 and CDR3. p < 0.05 and FDR < 25%

PATHWAYS	DATABASE
Enriched in CDR1 (in relation to control)
Response to bacterium	GENE ONTOLOGY
Defense response to bacterium	GENE ONTOLOGY
M phase	GENE ONTOLOGY
M phase of mitotic cell cycle	GENE ONTOLOGY
Mitosis	GENE ONTOLOGY
Regulation of Mitosis	GENE ONTOLOGY
Sulfus compound biosynthetic process	GENE ONTOLOGY
Response to other organism	GENE ONTOLOGY
Mitotic cell cycle checkpoint	GENE ONTOLOGY
Regulation of heart contraction	GENE ONTOLOGY
Cell cycle process	GENE ONTOLOGY
Response to biotic stimulus	GENE ONTOLOGY
Cell cycle phase	GENE ONTOLOGY
Cell cycle mitotic	REACTOME
Cell cycle	REACTOME
DNA replication	REACTOME
Mitotic M-M/G1 phases	REACTOME
Mitotic prometaphase	REACTOME
G2 M checkpoints	REACTOME
Enriched in CDR1 (in relation to CDR3)
Intracellular receptor-mediated signaling pathway	GENE ONTOLOGY
Neuron development	GENE ONTOLOGY
Feeding behavior	GENE ONTOLOGY
Defense response to bacterium	GENE ONTOLOGY
Response to bacterium	GENE ONTOLOGY
Steroid hormone receptor signaling pathway	GENE ONTOLOGY
Generation of neurons	GENE ONTOLOGY
G protein signaling coupled to cyclic nucleotide second messenger	GENE ONTOLOGY
G protein signaling adenylate cyclase activating pathway	GENE ONTOLOGY
Neurite development	GENE ONTOLOGY
Neurogenesis	GENE ONTOLOGY
Neuron differentiation	GENE ONTOLOGY
Cellular morphogenesis during differentiation	GENE ONTOLOGY
cAMP Mediated Signaling	GENE ONTOLOGY
Cyclic nucleotide-mediated signaling	GENE ONTOLOGY
G protein signaling coupled to cAMP nucleotide second messenger	GENE ONTOLOGY
Axon guidance	GENE ONTOLOGY
Axonogenesis	GENE ONTOLOGY
Regulation of protein stability	GENE ONTOLOGY
Carbohydrate transport	GENE ONTOLOGY
Central nervous system development	GENE ONTOLOGY
Regulation of heart contraction	GENE ONTOLOGY
Adenylate cyclase activation	GENE ONTOLOGY
Potassium ion transport	GENE ONTOLOGY
PPAR signaling pathway	GENE ONTOLOGY
Ribosome	KEGG
Terpenoid backbone biosynthesis	KEGG
Enriched in CDR3 (in relation to control)
Erythrocyte Differentiation Pathway	BIOCARTA
Cytokines and Inflammatory Response	BIOCARTA
Cell cycle pathway	BIOCARTA
Cytokine pathway	BIOCARTA
IL12 Pathway	BIOCARTA
NFkB activation by Nontypeable Hemophilus influenzae	BIOCARTA
Neuropeptides VIP and PACAP inhibit the apoptosis of activated T cells	BIOCARTA
Viral genome replication	GENE ONTOLOGY
Viral genome replication	GENE ONTOLOGY
M phase of mitotic cell cycle	GENE ONTOLOGY
Mitosis	GENE ONTOLOGY
M phase	GENE ONTOLOGY
Viral reproductive process	GENE ONTOLOGY
Viral infectious cycle	GENE ONTOLOGY
Cell cycle process	GENE ONTOLOGY
Regulation of mitosis	GENE ONTOLOGY
Cell cycle phase	GENE ONTOLOGY
Viral reproduction	GENE ONTOLOGY
Mitotic cell cycle	GENE ONTOLOGY
Positive regulation of cytokine production	GENE ONTOLOGY
Cell cycle checkpoint	GENE ONTOLOGY
Positive regulation of cell cycle	GENE ONTOLOGY
Chromosome segregation	GENE ONTOLOGY
Inflammatory response	GENE ONTOLOGY
Meiotic cell cycle	KEGG
Graft versus host disease	KEGG
NOD-like receptor signaling pathway	KEGG
Hematopoietic cell lineage	KEGG
Cytokine-cytokine receptor interaction	KEGG
Systemic lupus erythematosus	KEGG
Cell cycle	KEGG
Chemokine signaling pathway	KEGG
Antigen processing and presentation	KEGG
Toll-like receptor signaling pathway	KEGG
Homologous recombination	KEGG
Deposition of new CENPA containing nucleosomes at the centromeres	REACTOME
RNA POLI promoter opening	REACTOME
Meiotic recombination	REACTOME
Packing of telomere ends	REACTOME
Meiosis	REACTOME
RNA POLI transcription	REACTOME
Chromosome maintenance	REACTOME
Meiotic synapsis	REACTOME
Telomere maintenance	REACTOME
Cell cycle	REACTOME
Mitotic prometaphase	REACTOME
Chemokine receptors bind chemokines	REACTOME
Amyloids	REACTOME
RNA POLI, RNA POLIII and mitochondrial transcription	REACTOME
Cell cycle mitotic	REACTOME
DNA replication	REACTOME
Mitotic M-M/G1 phases	REACTOME
G2/M checkpoints	REACTOME
IL1 signaling	REACTOME
Mitotic G2-G2/M phases	REACTOME
Homologous recombination repair of replication-independent double-strand breaks	REACTOME
Interferon alpha/beta signaling	REACTOME
Cell cycle checkpoints	REACTOME
Double strand break repair	REACTOME
PERK regulated gene expression	REACTOME
Activation of ATR in response to replication stress	REACTOME
Gaba receptor activation	REACTOME
Enriched in CDR3 (in relation to CDR1)
IL17 pathway	BIOCARTA
Local acute inflammatory response pathway	BIOCARTA
Cytokine pathway	BIOCARTA
Cytokines and Inflammatory Response	BIOCARTA
Erythrocyte Differentiation Pathway	BIOCARTA
Regulation of hematopoiesis by cytokines	BIOCARTA
RNA POLI promoter opening	REACTOME
Chemokine receptors bind chemokines	REACTOME
Packing of telomere ends	REACTOME
Deposition of new CENPA containing nucleosomes at the centromeres	REACTOME
Meiotic recombination	REACTOME
RNA POLI transcription	REACTOME
PERK regulated gene expression	REACTOME
Activation of genes by ATF4	REACTOME
Meiotic synapsis	REACTOME
Meiosis	REACTOME
Telomere maintenance	REACTOME

Considering that neurodegeneration is the most prominent and harmful characteristic of AD and the potential of PBMCs to manifest important molecular changes that might occur in the brain of patients, we looked for common expressed genes between brain and blood. Then, we compared our results in PBMCs with previously published data from Miller et al. [15], in which the authors compared the results obtained for the hippocampus of 17 individuals with AD versus 16 age-matched controls. The authors used Illumina HumanHT-12 v3 Expression Beadchips (Illumina, San Diego, CA, USA) and performed the analysis for two hippocampus regions, CA1 (the most affected region in AD) and CA3 (the least affected region in AD).

Comparing the results obtained in the present work with the list of 1,028 genes differentially expressed reported by Miller et al. [15] in region CA1 of AD patient hippocampi, we found 28 common genes; from these, 14 upregulated and 3 downregulated genes were found in both works (Fig. 2). In addition, regarding Miller’s results obtained for hippocampi region CA3 (603 differentially expressed genes in AD compared to the control), the comparison with our results provided 14 genes that were common in both studies (10 upregulated, 2 downregulated genes and 2 whose regulation was in the opposite direction) (Fig. 3).

Fig. 2.

Differentially expressed genes in PBMCs and hippocampus CA1 region [14] of AD individuals, compared to age-matched controls. Venn diagram is showing differentially expressed genes that are common or exclusive when the results of this study in PBMCs were compared with Miller’s study in CA1 region of the hippocampus [14]. The number of common genes in both studies is represented, as well as the gene names (inside the box). A) Representation of the number of upregulated genes and B) downregulated genes.

Fig. 3.

Differentially expressed genes in PBMCs and hippocampus CA3 region [14] of AD individuals compared to age-matched controls. Venn diagram showing differentially expressed genes that are common or exclusive when the results of this study in PBMCs were compared with Miller’s [14] study in CA3 region of the hippocampus. The amount of upregulated and downregulated genes in both studies is represented, as well as the names of genes (inside the box). A) Representation of the number of upregulated genes; B) downregulated genes.

Therefore, considering the abovementioned results from comparisons between hippocampi and PBMCs, we can highlight the following genes: FOS, SSPN, DUSP1, RGS1, ANGPTL4, and CCL2, which were found to be upregulated in CA1, CA3, and PBMCs.

DISCUSSION

Analysis of gene expression profiles carried out for PBMCs of AD patients compared to age-matched control subjects generated a list of 593 differentially expressed genes in AD patients, and functional analyses indicated the involvement of relevant pathways. We found 428 upregulated genes in the AD group, and most gene functions were found associated with cell proliferation, apoptosis, immune response, and learning/memory, among others. Interestingly, some of these pathways are exclusively involved in brain physiology, such as learning and memory.

Previous studies demonstrated that neurons, the principal cells implicated in AD, are able to re-enter the cell cycle following genomic instability. This process can lead to apoptosis, further culminating into neurodegeneration [18 –21]. Additionally, cell cycle dysfunction was also reported in blood cells of AD patients [22 –24]. Alterations in the expression of proteins that participate in cell cycle control in lymphocytes of AD patients, such as TP53, a key protein involved in cell fate after induction of DNA damage and apoptotic signaling, have also been demonstrated [25]. Moreover, an increased expression of CDKN1A, which encodes a cyclin-dependent kinase inhibitor whose role is related to the control of G1/S transition, has been observed [25]. Thus, considering that cell cycle alterations play a role in AD neurodegeneration and that most PBMCs are also postmitotic cells, alterations on cell proliferation observed in this work reinforce the results reported by other researchers [22 –24]. Therefore, cell cycle dysregulation in neurons and PBMCs from AD patients seems to be an important aspect in the pathology, which deserves further investigation.

Pathways related to inflammation were also found to be enriched in our study with AD patients, and this is compatible to alterations in the immune system and inflammatory responses that were already observed in the brain and blood of AD subjects [26, 27]. Furthermore, some authors consider that immune system dysfunction can be the main cause of AD and protein deposition might be a consequence of this altered immune response [27 –30]. Inflammation is an intrinsic condition associated with several alterations observed in AD, as modification of insulin metabolism, mitochondrial dysfunction, oxidative stress, leucocyte migration, apoptosis, and alterations in calcium homeostasis [29, 31]. Therefore, dysregulation of those pathways found in AD, both in brain and blood cells, reinforces the systemic nature of AD and the importance of studying those biological processes in both tissues.

Curiously, learning, and memory pathways were also found upregulated in blood cells of AD patients, despite the exclusive association of these pathways with neurons. In Booij et al. [12], pathways related to neuronal activities were also found to be enriched in blood cells. In our work, differentially expressed genes related to learning and memory, such as EGR1, FOS, EGR2, PTGS2, S100B, PRKAR1B, and PTN, are also implicated in other cellular functions. In general, they are transcription factors or growth factors. Therefore, in PBMCs, these genes can participate in different processes. Thus, alterations in their expression might be systemic and possibly could be related to the symptoms that affect learning and memory.

Moreover, the 145 downregulated genes in AD subjects are correlated with pathways such as gas transport, G-coupled receptor, hemoglobin metabolism, cell signaling, plasma membrane, and, notably, olfactory transduction. Interestingly, some studies showed a relationship between AD and deficiency in mechanisms related to hemoglobin production, culminating in anemia [32], and anemia has been considered a risk factor for AD [33].

Olfactory receptors are members of the G-protein family, and olfactory transduction was considered enriched among downregulated genes in PBMCs of AD, although this pathway has been associated with functions linked to the neuronal system. Interestingly, a decrease in the olfactory capabilities of AD patients has been reported earlier in preclinical phases of the disease [34, 35]. Thus, these findings suggest that the inhibition of the olfactory system could be a feature of AD. Additionally, alterations in pathways related to DNA damage response and DNA repair were also observed in this work in accordance with a decrease in DNA repair capacity previously reported in AD patients [25 , 37].

Delvaux et al. [38] performed a multivariate analysis of transcript expression in peripheral blood leukocytes of AD patients and observed alterations in the expression of gene sets whose functions were related to cell cycle, apoptosis, cellular stress and inflammatory systems. Furthermore, these alterations could be used to distinguish AD patients from patients with Parkinson’s disease and age-related cognitive normal individuals [38].

Thus, the present results may indicate that there is a possible correlation among the alterations in transcript profiles observed in PBMCs and the functional alterations that occur along the development of AD.

Another approach in this work consisted in studying gene expression alterations according to the clinical stage of AD patients: mild AD (CDR1) and severe AD (CDR3), compared to the control group. Considering AD as a systemic disease, it is possible that exclusively expressed genes in CDR1 can represent the earliest alterations related to mechanisms responsible for triggering the disease, while genes altered just in CDR3 can be related to modifications occurring as a consequence of the pathology per se.

The results obtained by functional enrichment analyses indicate that processes such as immune response, neurological functions, and inflammation were induced in both the CDR1 and CDR3 groups when compared to the control group, confirming the importance of these pathways during different phases of AD. Bossers et al. [39] analyzed samples of prefrontal cortex of AD patients and found altered gene functions associated with synaptic plasticity, development of nervous system and plasticity. Those genes were highly expressed at early stage of AD in comparison to the later stage [39].

It is important to highlight that in our enrichment analysis, CDR3 compared to controls showed differential transcript expression for pathways related to DNA repair and telomere maintenance. Despite many studies demonstrating telomere shortening in AD blood cells [40 –42], the same was not observed in brain regions such as the hippocampus [43]. Although the role of telomere stability is not well established yet in AD, our results suggest some involvement of telomere maintenance in PBMCs, which may be linked to the severity of the disease.

Genes playing a role in DNA repair pathways, such as homologous recombination (HR) and nonhomologous end-joining (NHEJ), were also found enriched in CDR3 compared to controls. Moreover, there is evidence of increased oxidized bases in brains and peripheral cells of AD [44 –46], and some studies have associated such alteration with deficiency in DNA repair mechanisms [36 , 47–49].

As a whole, the present results indicate that it is possible to distinguish transcriptional expression profiles between AD stages, and in spite of the small number of patients enrolled in our study, our data represent a proof of concept that supports subsequent investigations towards important signaling pathways compromised in AD.

To compare our gene expression profiles in PBMCs with studies using brain samples, we compared our results with those obtained by Miller et al. [15] regarding transcriptional profile analysis in the hippocampi of AD patients [14]. Miller et al. showed expression profiles of two hippocampus regions, highly affected (CA1) and less affected (CA3). The authors highlighted the importance of comparing CA1 and CA3 regions to define risk factors and indicate the vulnerable regions to be affected by AD. By comparing differentially expressed genes in CA1 and PBMCs, we found 28 common genes. However, 11 genes (TMEM176A, OR7A5, PRDM16, HLA-DQA1, LINGO2, TYMS, PEG3, NUAK1, GRIK1, LRRN3, and RCAN2) presented different regulation, being upregulated in CA1, but downregulated in PBMCs, or vice-versa. Moreover, for the comparison of PBMCs versus CA3, we found 14 common genes; among them, SELENBP1 and GINS2 presented opposite regulation.

Therefore, the present work emphasizes shared upregulated genes between PBMCs and brain regions (CA1, and CA3), thus supporting the hypothesis that these genes present a great potential as AD biomarkers. We found 8 common upregulated genes: FOS, SLC7A2, RGS1, GFAP, DUSP1, ANGPTL4, SSPN, and CCL2. Among them, CCL2 and ANGPTL4 were already reported as genes associated with AD in brain and blood cells [50 –52]. Although SLC7A2, RGS1, and DUSP1 were not previously related to AD, they have the potential to be investigated as molecular biomarkers of AD, since they were found upregulated in both studies. Regarding gene functions, while SLC7A2 encodes a plasma membrane protein related to amino acid transportation and inflammatory response, RGS1 plays a role in signal transduction and immune response, while DUSP1 encodes a MAP kinase that functions in the cell cycle and calcium ion regulation. Therefore, these three genes present functions related to pathways that were found altered in AD in the present work and are important candidates as biomarkers for AD diagnosis.

FOS, GFAP, and SSPN have already been found to be associated with brain dysfunction in AD [53 –55], but there is no similar information in PBMCs of AD patients. Thus, the present work points to the relevance of the FOS, GFAP, and SSPN genes to be investigated as candidate biomarkers of AD.

Interestingly, Blalock et al. [56] also performed whole-genome transcriptomic analysis in the hippocampus of AD patients and, among the upregulated genes selected by these authors, 8 genes (KCNJ15, GFPT2, GSTM4, IFNG, CYP4F3, RGN, SSPN, and GFAP) are in accordance with the upregulated genes found in the present study. Furthermore, SSPN and GFAP were likewise found by Miller et al. [15]. Similarly, Ravetti et al. [55] also analyzed data from AD hippocampi and found some common genes with those selected in the present study, such as EGR, GNAS, GSTA3, IL1A, LTF, SLC8A2, and SSPN.

Moreover, Ray et al. [57] performed a study in blood plasma and pointed out 18 genes as possible markers of AD, among them, EGF, G-CSF, IL3, and IL1A. In our study, the expression of these four genes was also found altered in PBMCs.

Rye et al. [13] performed extensive prospective studies aiming to find a pattern of expression in blood cells of AD patients, providing a blood test for the early detection of the pathology. Despite some concordance in the enriched pathways, such as stress response and chromatin packaging, their results do not match with the data obtained in our study. The differences in both studies may be due to different populations studied (Brazilian in our study and Norwegian in their case), and differences in the methods used in each work. In this context, more studies are needed to establish blood biomarkers that may be able to characterize AD.

In spite of the limitations of the present study, our results provide some important candidate genes that deserve further investigation in the search for blood biomarkers and can be a proof-of-concept supporting PBMCs as valuable diagnostic tools in AD. Therefore, additional studies in a broad cohort that might include previous stages of the disease (such as mild cognitive impairment) are needed to establish biomarkers of AD present in blood cells that might characterize the disease.

Conclusions

Transcriptomic profiles analyzed in PBMCs of AD patients indicate pathways involved in the disease, such as inflammation, cell cycle regulation, and neuronal process, as well as DNA damage response. Our gene expression data compared to the results obtained in the hippocampus [15] identified 8 common genes (DUSP1, FOS, SLC7A2, RGS1, GFAP, CCL2, ANGPTL4, and SSPN) that represent potential peripheral blood biomarkers of AD.

Finally, our findings reinforce the hypothesis that AD is a systemic pathology and supports the assumption that blood cells can be useful tools to understanding the etiopathogenesis and development of the disease. Moreover, our work indicates potential genes to be investigated for this purpose, on the basis of evidence that PBMCs do manifest changes in gene expression profiles compared to non-AD controls.

Footnotes

ACKNOWLEDGMENTS

The authors would like to thank Dr. Douglas V.P.N. Oliveira for the critical reading of this manuscript and Dr. Silvio Bernardes Filho for his effective participation in blood collection. This research received financial support from Brazilian agencies (FAPESP – Proc. 2012/07879-5; 2013/09352-7, CNPq and CAPES) and University of São Paulo – USP (NAP-ENV Project – Pro-Reitoria de Pesquisa-USP).

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Avramopoulos

(2009) Genetics of Alzheimer’s disease: Recent advances. Genome Med 1, 34.

Caspersen

, Wang

, Yao

, Sosunov

, Chen

, Lustbader

, Xu

, Stern

, McKhann

, Yan

(2005) Mitochondrial Abeta: A potential focal point for neuronal metabolic dysfunction in Alzheimer’s disease. FASEB J19, 2040–2041.

Hirai

, Aliev

, Nunomura

, Fujioka

, Russell

, Atwood

, Johnson

, Kress

, Vinters

, Tabaton

, Shimohama

, Cash

, Siedlak

, Harris

, Jones

, Petersen

, Perry

, Smith

(2001) Mitochondrial abnormalities in Alzheimer’s disease. J Neurosci 21, 3017–3023.

Isaacs

, Senn

, Yuan

, Shine

, Yankner

(2006) Acceleration of amyloid beta-peptide aggregation by physiological concentrations of calcium. J Biol Chem 281, 27916–27923.

Swerdlow

, Khan

(2004) A “mitochondrial cascade hypothesis” for sporadic Alzheimer’s disease. Med Hypotheses 63, 8–20.

Swerdlow

, Burns

, Khan

(2014) The Alzheimer’s disease mitochondrial cascade hypothesis: Progress and perspectives. Biochim Biophys Acta 1842, 1219–1231.

Wyss-Coray

, Mucke

(2002) Inflammation in neurodegenerative disease-a double-edged sword. Neuron 35, 419–432.

Zhao

, Bhattacharjee

, Dua

, Alexandrov

, Lukiw

(2015) microRNA-based biomarkers and the diagnosis of Alzheimer’s disease. Front Neurol 6, 162.

McKhann

, Drachman

, Folstein

, Katzman

, Price

, Stadlan

(1984) Clinical diagnosis of Alzheimer’s disease: Report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology 34, 939–944.

10.

Satoh

, Kino

, Niida

(2015) MicroRNA-seq data analysis pipeline to identify blood biomarkers for Alzheimer’s disease from public data. Biomark Insights 10, 21–31.

11.

Galimberti

, Scarpini

(2011) Disease-modifying treatments for Alzheimer’s disease. Ther Adv Neurol Disord 4, 203–216.

12.

Booij

, Lindahl

, Wetterberg

, Skaane

, Saebo

, Feten

, Rye

, Kristiansen

, Hagen

, Jensen

, Bardsen

, Winblad

, Sharma

, Lonneborg

(2011) A gene expression pattern in blood for the early detection of Alzheimer’s disease. J Alzheimers Dis 23, 109–119.

13.

Rye

, Booij

, Grave

, Lindahl

, Kristiansen , Andersen

, Horndalsveen

, Nygaard

, Naik

, Hoprekstad

, Wetterberg

, Nilsson

, Aarsland

, Sharma

, Lonneborg

(2011) A novel blood test for the early detection of Alzheimer’s disease. J Azheimers Dis 23, 1–21.

14.

Hong

, Breitling

, McEntee

, Wittner

, Nemhauser

, Chory

(2006) RankProd: A bioconductor package for detecting differentially expressed genes in meta-analysis. Bioinformatics 22, 2825–2827.

15.

Miller

, Woltjer

, Goodenbour

, Horvath

, Geschwind

(2013) Genes and pathways underlying regional and cell type changes in Alzheimer’s disease. Genome Med 5, 48.

16.

Huang da

, Sherman

, Lempicki

(2009) Bioinfor-matics enrichment tools: Paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res 37, 1–13.

17.

Subramanian

, Tamayo

, Mootha

, Mukherjee

, Ebert

, Gillette

, Paulovich

, Pomeroy

, Golub

, Lander

, Mesirov

(2005) Gene set enrichment analysis: A knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA 102, 15545–15550.

18.

Moh

, Kubiak

, Bajic

, Zhu

, Smith

, Lee

(2011) Cell cycle deregulation in the neurons of Alzheimer’s disease. Results Probl Cell Differ 53, 565–576.

19.

Silva

, Santos

, Farfel

, Grinberg

, Ferretti

, Campos

, Cunha

, Begnami

, Rocha

, Carraro

, de Braganca Pereira

, Jacob-Filho

, Brentani

(2014) Repair of oxidative DNA damage, cell-cycle regulation and neuronal death may influence the clinical manifestation of Alzheimer’s disease. PLoS One 9, e99897.

20.

Yates

, Zafar

, Hubbard

, Nagy

, Durant

, Bicknell

, Wilcock

, Christie

, Esiri

, Smith

, Nagy

(2013) Dysfunction of the mTOR pathway is a risk factor for Alzheimer’s disease. Acta Neuropathol Commun 1, 3.

21.

Zekanowski

, Wojda

(2009) Aneuploidy, chromosomal missegregation, and cell cycle reentry in Alzheimer’s disease. Acta Neurobiol Exp (Wars) 69, 232–253.

22.

Bialopiotrowicz

, Kuzniewska

, Kachamakova-Trojanowska

, Barcikowska

, Kuznicki

, Wojda

(2011) Cell cycle regulation distinguishes lymphocytes from sporadic and familial Alzheimer’s disease patients. Neurobiol Aging 32, 2319.e2313–2326.

23.

Stieler

, Lederer

, Bruckner

, Wolf

, Holzer

, Gertz

, Arendt

(2001) Impairment of mitogenic activation of peripheral blood lymphocytes in Alzheimer’s disease. Neuroreport 12, 3969–3972.

24.

Zhou

, Jia

(2010) P53-mediated G(1)/S checkpoint dysfunction in lymphocytes from Alzheimer’s disease patients. Neurosci Lett 468, 320–325.

25.

Leandro

, Lobo

, Oliveira

, Moriguti

, Sakamoto-Hojo

(2013) Lymphocytes of patients with Alzheimer’s disease display different DNA damage repair kinetics and expression profiles of DNA repair and stress response genes. Int J Mol Sci 14, 12380–12400.

26.

Akiyama

, Barger

, Barnum

, Bradt

, Bauer

, Cole

, Cooper

, Eikelenboom

, Emmerling

, Fiebich

, Finch

, Frautschy

, Griffin

, Hampel

, Hull

, Landreth

, Lue

, Mrak

, Mackenzie

, McGeer

, O'Banion

, Pachter

, Pasinetti

, Plata-Salaman

, Rogers

, Rydel

, Shen

, Streit

, Strohmeyer

, Tooyoma

, Van Muiswinkel

, Veerhuis

, Walker

, Webster

, Wegrzyniak

, Wenk

, Wyss-Coray

(2000) Inflammation and Alzheimer’s disease. Neurobiol Aging 21, 383–421.

27.

Song

, Qian

, Peng

, Han

, Wang

, Bai

, Crack

, Lei

(2016) Perturbation of the transcriptome: Implications of the innate immune system in Alzheimer’s disease. Curr Opin Pharmacol 26, 47–53.

28.

Drachman

(2014) The amyloid hypothesis, time to move on: Amyloid is the downstream result, not cause, of Alzheimer’s disease. Alzheimers Dement 10, 372–380.

29.

Macchi

, Marino-Merlo

, Frezza

, Cuzzocrea

, Mastino

(2014) Inflammation and programmed cell death in Alzheimer’s disease: Comparison of the central nervous system and peripheral blood. Mol Neurobiol 50, 463–472.

30.

McCaulley

, Grush

(2015) Alzheimer’s disease: Exploring the role of inflammation and implications for treatment. Int J Alzheimers Dis 2015, 515248.

31.

Mattson

(2002) Oxidative stress, perturbed calcium homeostasis, and immune dysfunction in Alzheimer’s disease. J Neurovirol 8, 539–550.

32.

Faux

, Rembach

, Wiley

, Ellis

, Ames

, Fowler

, Martins

, Pertile

, Rumble

, Trounson

, Masters

, Bush

(2014) An anemia of Alzheimer’s disease. Mol Psychiatry 19, 1227–1234.

33.

Shah

, Buchman

, Wilson

, Leurgans

, Bennett

(2011) Hemoglobin level in older persons and incident Alzheimer disease: Prospective cohort analysis. Neurology 77, 219–226.

34.

Franks

, Chuah

, King

, Vickers

(2015) Connectivity of pathology: The olfactory system as a model for network-driven mechanisms of Alzheimer’s disease patho-genesis. Front Aging Neurosci 7, 234.

35.

Masurkar

, Devanand

(2014) Olfactory dysfunction in the elderly: Basic circuitry and alterations with normal aging and Alzheimer’s disease. Curr Geriatr Rep 3, 91–100.

36.

Migliore

, Fontana

, Trippi

, Colognato

, Coppede

, Tognoni

, Nucciarone

, Siciliano

(2005) Oxidative DNA damage in peripheral leukocytes of mild cognitive impairment and AD patients. Neurobiol Aging 26, 567–573.

37.

Sykora

, Misiak

, Wang

, Ghosh

, Leandro

, Liu

, Tian

, Baptiste

, Cong

, Brenerman

, Fang

, Becker

, Hamilton

, Chigurupati

, Zhang

, Egan

, Croteau

, Wilson III

, Mattson

, Bohr

(2015) DNA polymerase deficiency leads to neurodegeneration and exarcerbates Alzheimer disease phenotypes. Nucleic Acids Res 43, 943–959.

38.

Delvaux

, Mastroeni

, Nolz

, Chow

, Sabbagh

, Caselli

, Reiman

, Marshall

, Coleman

(2017). Multivariate analyses of peripheral blood leukocyte transcripts distinguish Alzheimer’s, Parkinson's, control, and those at risk for developing Alzheimer’s. Neurobiol Aging 58, 225–237.

39.

Bossers

, Wirz

, Meerhoff

, Essing

, van Dongen

, Houba

, Kruse

, Verhaagen

, Swaab

(2010) Concerted changes in transcripts in the prefrontal cortex precede neuropathology in Alzheimer’s disease. Brain 133, 3699–3723.

40.

Boccardi

, Pelini

, Ercolani

, Ruggiero

, Mecocci

(2015) From cellular senescence to Alzheimer’s disease: The role of telomere shortening. Ageing Res Rev 22, 1–8.

41.

Hochstrasser

, Marksteiner

, Humpel

(2012) Telom-ere length is age-dependent and reduced in monocytes of Alzheimer patients. Exp Gerontol 47, 160–163.

42.

Zhang

, Dilley

, Mattson

(2007) DNA damage responses in neural cells: Focus on the telomere. Neuroscience 145, 1439–1448.

43.

Thomas

, NJ

, Fenech

(2008) Telomere length in white blood cells, buccal cells and brain tissue and its variation with ageing and Alzheimer’s disease. Mech Ageing Dev 129, 183–190.

44.

Coppede

, Migliore

(2009) DNA damage and repair in Alzheimer’s disease. Curr Alzheimer Res 6, 36–47.

45.

Kadioglu

, Sardas

, Aslan

, Isik

, Esat Karakaya

(2004) Detection of oxidative DNA damage in lymphocytes of patients with Alzheimer’s disease. Biomarkers 9, 203–209.

46.

Mecocci

, Polidori

, Cherubini

, Ingegni

, Mattioli

, Catani

, Rinaldi

, Cecchetti

, Stahl

, Senin

, Beal

(2002) Lymphocyte oxidative DNA damage and plasma antioxidants in Alzheimer disease. Arch Neurol 59, 794–798.

47.

Leandro

, Sykora

, Bohr

(2015) The impact of base excision DNA repair in age-related neurodegenerative diseases. Mutat Res 776, 31–39.

48.

Shackelford

(2006) DNA end joining activity is reduced in Alzheimer’s disease. Neurobiol Aging 27, 596–605.

49.

Weissman

, Jo

, Sorensen

, de Souza-Pinto

, Markesbery

, Mattson

, Bohr

(2007) Defective DNA base excision repair in brain from individuals with Alzheimer’s disease and amnestic mild cognitive impairment. Nucleic Acids Res 35, 5545–5555.

50.

Bose

, Cho

(2013) Role of chemokine CCL2 and its receptor CCR2 in neurodegenerative diseases. Arch Pharm Res 36, 1039–1050.

51.

Naert

, Rivest

(2013) A deficiency in CCR2+monocytes: The hidden side of Alzheimer’s disease. J Mol Cell Biol 5, 284–293.

52.

Vienberg

, Kleinridders

, Suzuki

, Kahn

(2015) Differential effects of angiopoietin-like 4 in brain and muscle on regulation of lipoprotein lipase activity. Mol Metab 4, 144–150.

53.

Mueller

, Schrag

, Crofton

, Stolte

, Muckenthaler

, Magaki

, Kirsch

(2012) Altered serum iron and copper homeostasis predicts cognitive decline in mild cognitive impairment. J Alzheimers Dis 29, 341–350.

54.

, Wen

, Li

, Zhou

(2015) Upregulation effects of Tanshinone IIA on the expression of NeuN, Nissl body, and IkB and downregulation effects on the expressions of GFAP and NF-kB in the brain tissues of rat models of Alzheimer’s disease. Neuroreport 26, 758–766.

55.

Ravetti

, Rosso

, Berreta

, Moscato

(2010) Uncovering molecular biomarkers that correlate cognitive decline with the changes of hippocampus' gene expression profiles in Alzheimer’s disease. PLoS One 5, e10153.

56.

Blalock

, Geddes

, Chen

, Porter

, Markesbery

, Landfield

(2004) Incipient Alzheimer’s disease: Microarray correlation analyses reveal major transcriptional and tumor suppressor responses. Proc Natl Acad Sci USA 101, 2173–2178.

57.

Ray

, Britschgi

, Herbert

, Takeda-Uchimura

, Boxer

, Blennow

, Friedman

, Galasko

, Jutel

, Kary-das

, Kaye

, Leszek

, Miller

, Minthon

, Quinn

, Rabinovici

, Robinson

, Sabbagh

, So

, Sparks

, Tabaton

, Tinklenberg

, Yesavage

, Tibshi-rani

, Wyss-Coray

(2007) Classification and prediction of clinical Alzheimer’s diagnosis based on plasma signaling proteins. Nat Med 13, 1359–1362.