Locus coeruleus at asymptomatic early and middle Braak stages of neurofibrillary tangle pathology

The present study analyses molecular characteristics of the locus coeruleus (LC) and projections to the amygdala and hippocampus at asymptomatic early and middle Braak stages of neurofibrillary tangle (NFT) pathology.


Introduction
Alzheimer's disease (AD), the most common cause of cognitive impairment in the elderly, is a progressive neurodegenerative disorder lasting for decades which is characterized morphologically by accumulation of bamyloid in the neuropil forming fibrillar diffuse deposits and senile plaques, b-amyloid deposits in blood vessels leading to amyloid angiopathy and intraneuronal accumulation of hyperphosphorylated and truncated 3Rtau and 4Rtau as the major component of neurofibrillary tangles (NFTs), neuropil threads and dystrophic neurites of senile plaques [1][2][3].
These morphological alterations have a particular pattern of distribution which extends to most of the brain along with disease progression [4][5][6][7] paralleling the progressive appearance of clinical symptoms [8][9][10]. Clinical and pathological changes do not progress in the same way in all individuals; about 85% of the population aged 65 has brain lesions consistent with early changes of AD, yet only about 5% have reached thresholds leading to severe cognitive impairment and dementia (clinical stages of the neurodegenerative process) [11,12]. However, 25% of the population at the age of 85 suffers from dementia of Alzheimer type [12].
The locus coeruleus (LC) is a pontine nucleus containing the largest group of noradrenergic neurons in the central nervous system [13,14]. While the rostral portion of the LC innervates mostly forebrain structures such as the hippocampus and cerebral cortex, the caudal portion innervates mainly hindbrain regions [15][16][17][18]. The LC is the only noradrenergic nucleus innervating the cerebral cortex [19], whereas the amygdala also receives projections from the lateral brainstem tegmentum [20]. The LC receives input from various regions of the brain [21]. All these neuronal networks display the LC as a hub nucleus implicated in superior functions such as arousal, attention, wake-sleep cycles, emotional states, cognition, memory and learning, regulation of blood flow, motor coordination, neuroinflammation, neuronal survival and neurogenesis [14,[22][23][24][25][26][27][28].
Noradrenaline released from the LC acts at specific synapses through a 1 , a 2 or b adrenoceptors [29,30]. Activation of a 1 adrenoceptors generally leads to excitation, and there is some evidence that b adrenoceptors are also excitatory. In contrast, activation of a 2 adrenoceptors leads to inhibition, including inhibition of the noradrenergic neurons themselves through autoreceptors.
The LC is the main noradrenergic nucleus affected in AD [31]. Involvement of the LC at advanced stages of AD has been recognized for decades [14,[32][33][34][35][36][37]. Considering a percentage of neurone loss of about 70% at terminal stages of the disease [33,36], it is conceivable that many functions regulated by the LC are severely dampened in advanced AD. However, the LC, together with raphe nuclei, is one the first regions of the brain showing NFTs in the context of AD-related pathology [38][39][40][41][42].
Based on this scenario, the present study was focused on the LC at the first asymptomatic Braak stages of NFT pathology in an attempt to (i) define characteristics of tau phosphorylation, configuration and truncation, associated expression of tau kinases, proteins involved in energy metabolism, oxidative stress and responses, tyrosine hydroxylase (TH) immunoreactivity and local synapses in the LC; (ii) assess the impact of these alterations of the LC on the expression of noradrenergic receptors and TH in the hippocampus and amygdala; and (iii) unveil altered gene expression to identify altered pathways in LC associated with early stages of NFT pathology using whole-transcription arrays.

Human brain samples
Human brain samples were obtained from the Institute of Neuropathology Brain Bank (HUB-ICO-IDIBELL biobank) following the guidelines of the Spanish legislation on this matter (Real Decreto Autorizaci on y Funcionamiento de Biobancos, Ministry of Science and Innovation: 1716/2011) and the approval of the local ethics committee of the Bellvitge University Hospital-IDIBELL.
No individuals in the present series had neurological complaints, all of them were categorized as cognitively well preserved at the time of admission to the hospital; depression was not reflected in the clinical records of any case. All of them were admitted and died in the hospital from various non-neurological diseases. NFT pathology was an incidental observation during the post mortem neuropathological study. However, it is worth stressing that clinical histories were not directed to exhaustively analyse 'minor' neurological and neuropsychological changes such as recent modifications in mood, wake-sleep cycle disturbances, impairment of attention and learning capacity. Therefore, these symptoms might have been present without being recorded in some individuals.
Neuropathological diagnosis was made following the Braak and Braak staging for NFTs adapted to paraffin sections [4,43], Thal phases for b-amyloid score [5] and CERAD stages of senile plaques [44]. Cases with combined pathologies were excluded from the present study. Only cases with concomitant mild small vascular disease were accepted. Cases with evident long agonic state, hypoxia or seizures were excluded.
Two series of cases were used. For morphological and immunohistochemical studies, categorization of cases was as follows: Braak stage I of NFT pathology: n = 21, 10 men, 11 women, age 66.7 AE 6.68 years; stage II: n = 19, 12 men, 7 women, age 70.3 AE 7.83 years; stage III: n = 15, 5 men, 10 women, age 77.6 AE 6.83; and stage IV, n = 12, 8 men, 4 women, age 77.75 AE 8.01 years. The post mortem delay was 7.4 AE 5.1 h (between 2 h and 21 h). The purpose of the first series was to analyse changes linked to tau phosphorylation in LC; control cases without NFT pathology were not necessary.
The second series of cases was used for biochemical studies including whole-transcriptome arrays and quantitative real-time polymerase chain reaction (RT-qPCR) validation in LC, and determination of protein levels of tyrosine hydroxylase (TH) and a 2A adrenergic receptor (a 2A -AR) with western blotting in the hippocampus and amygdala. Each of these studies was carried out in all cases. Categorization of cases was as follows: middle-aged individuals with no clinical symptoms and without neuropathological lesions (no bamyloid deposits, NFT stage 0) called MA: n = 10, 6 men, 4 women, age 51.3 AE 5.9; stage I: n = 8, 7 men, 1 women, age 66.2 AE 6.45 years; stage II: n = 4, 1 men, 3 women, age 65.5 AE 9.60 years; stage III: n = 8, 6 men, 2 women, age 76.7 AE 9.20; and stage IV, n = 5, 4 men, 1 women, age 83.8 AE 5.93 years. The post mortem delay was 6.42 AE 3.64 h (between 2 h and 14 h) ( Table S1). The purpose of the second series was to assess changes in LC at early and middle stages of NFT pathology compared with MA individuals with no NFT pathology used as controls. Thal phases varied from 0 to 1 and neuritic plaque score CERAD from 0 to 1 in cases with NFT pathology. b-amyloid deposits in LC were not found in any cases.

Immunohistochemistry
Dewaxed sections, 4-lm thick, of the LC were processed for immunohistochemistry. The sections were boiled in citrate buffer (20 min) to retrieve tau antigenicity. Endogenous peroxidases were blocked by incubation in 10% methanol-1% H 2 O 2 solution (15 min) followed by 3% normal horse serum solution. The sections were then incubated at 4°C overnight with one of the primary antibodies listed in Table S2. Following incubation with the primary antibody, the sections were incubated with EnVision + system peroxidase (Dako, Agilent technologies, Santa Clara, California, USA) for 30 min at room temperature. The peroxidase reaction was visualized with diaminobenzidine and H 2 O 2 . Control of the immunostaining included omission of the primary antibody; no signal was obtained following incubation with only the secondary antibody.

Double-labelling immunofluorescence and confocal microscopy
Dewaxed sections, 4-lm thick, were stained with a saturated solution of Sudan black B (Merck, Barcelona, Spain) for 15 min to block the autofluorescence of lipofuscin granules present in cell bodies, and then rinsed in 70% ethanol and washed in distilled water. The sections were incubated at 4°C overnight with combinations of primary antibodies. After washing, the sections were incubated with Alexa488 or Alexa546 (1:400; Molecular Probes, Eugene, Oregon, USA) fluorescence secondary antibodies against the corresponding host species. Nuclei were stained with DRAQ5 TM (1:2000; Biostatus, Shepshed, UK). After washing, the sections were mounted in Immuno-Fluore mounting medium (ICN Biomedicals, Santa Clara, California, USA), sealed, and dried overnight. Sections were examined with a Leica TCS-SL confocal microscope (Leica, Barcelona, Spain).

Quantification of morphological studies
Quantification was carried out in the rostral part of the LC within the dorsolateral pontine tegmentum at the level of the lower segment of the cerebral aqueduct of Sylvius just before the IV ventricle [45]. Neurons with NFTs as revealed with the antibody AT8 were counted in all cases of the morphological series (n = 67), using three nonconsecutive sections per case separated by 50 microns at a magnification of 9 200; the total number of neurons was estimated in the same sections, which were slightly counterstained with haematoxylin. Unbiased stereological estimation of the total number of neurons [46] was not possible in the present series as tissue blocks did not contain the whole LC. However, this limitation had no substantial implications in the purposes at hand, as discussed below.
About 5360 neurons were counted in total. Results were expressed as the percentage of neurons with NFTs in relation to the total number of LC neurons at stages I, II, III and IV of NFT pathology. Mean values for each group were compared with one-way analysis of variance (ANOVA) followed by post hoc Tukey's multiple comparison test, and differences were considered statistically significant at P < 0.05, P < 0.01 and P < 0.001.
The estimation of colocalization of two proteins labelled with specific antibodies and examined with the

Total membrane preparation
Frozen samples of the hippocampus and amygdala were homogenized in ice-cold 10 mM Tris HCl, pH 7.4, 1 mM EDTA, 300 mM KCl buffer containing a protease inhibitor cocktail (Roche Molecular Systems, Basel, Switzerland) using a Polytron for three periods of 10 s each. The homogenate was centrifuged for 10 min at 1000g. The resulting supernatant was centrifuged for 30 min at 12 000g. The membranes were dispersed in 50 mM Tris HCl (pH 7.4) and 10 mM MgCl 2 , washed and resuspended in the same medium as described previously [47]. Protein concentration was determined using the BCA (Bicinchoninic Acid) assay kit (Thermo Fisher Scientific, Inc., Rockford, IL, USA) and 80 lg of protein were used for Immunoblot. Plasmids and transfection used to verify the characteristics of a 2A -AR expressed in HEK-293T cells are shown in Data S2.

Whole-transcriptome arrays
Synthesis and hybridization of cRNA RNA from frozen LC was extracted following the instructions of the supplier (RNeasy Mini Kit; Qiagen â GmbH, Hilden, GE). RNA integrity number (RIN) and 28S/18S ratios were determined with the Agilent Bioanalyzer (Agilent Technologies, Inc., Santa Clara, CA, USA) to assess RNA quality; RNA concentration was evaluated using a NanoDrop TM Spectrophotometer (Thermo Fisher Scientific, Carlsbad, CA, USA). RIN (RNA integrity number) values are presented in Table S1. Suitability of such cases for mRNA expression studies was tested by RT-qPCR using different probes. The outliers were detected using the GraphPad software QuickCalcs (P < 0.05).
Details of microarray hybridization are shown in Data S3. The method of RNA sample preparation was based on the original T7 in vitro transcription technology known as the Eberwine or RT-IVT method [48].
Microarray data analysis Microarray data quality control, normalization and filtering were performed using Bioconductor packages in R programming environment for genes [49]. Processing of raw data for the noncoding regions was performed using Affymetrix Transcription Array Console software (Santa Clara, CA, USA). Cofactor neuritic plaques and gender (M or W) were considered in the analysis as well as the experimental batch (1 or 2). Genes were first selected based on their experimental values using a test for differential expression between two classes (Student's ttest) or a clustering method for coexpressed genes (ANOVA one-way test). Genes were considered differentially expressed if they showed an absolute fold change >1.2 in combination with an unadjusted P ≤ 0.05, as differentially expressed genes obtained from the microarray were lower than currently selected fold change values between 2.0 and 2.5 [50].
Microarray data postanalysis: gene ontology and ingenuity pathway analysis Once a list of differentially expressed genes was identified from microarrays, the selection of putative candidates was analysed for biological significance using a gene enrichment analysis against gene ontology (GO) through bioconductor packages in R programming using P < 0.05 as the cut-off point to determine whether GO database was significantly enriched. The core analysis function included in ingenuity pathway analysis (IPA) (Ingenuity System, Inc., CA, USA) was used to interpret the data in the context of biological processes, pathways and networks.

Validation of microarray data
Retro-transcription polymerase chain reaction Retrotranscription reaction of RNA samples selected on the basis of their RIN values was carried out with the high-capacity cDNA archive kit (Applied Biosystems, Foster City, CA, USA) following the guidelines provided by the manufacturer and using Gene Amp 9700 PCR System thermocycler (Applied Biosystems). One RNA was processed in parallel in the absence of reverse transcriptase to rule out DNA contamination.
Quantitative real-time polymerase chain reaction RT-qPCR assays were conducted in duplicate on cDNA samples obtained from the retro-transcription reaction diluted 1:10 in 384-well optical plates (Kisker Biotech, Steinfurt, GE) utilizing the ABI Prism 7900 HT Sequence Detection System (Applied Biosystems). TaqMan probes (Thermo Fisher Scientific) used in the study are listed in Table S3. Selection of house-keeping genes was made on the basis of their preservation levels in human post mortem brain [51,52]. DDCT values were obtained from the DCT of each sample minus the mean DCT of the population of control samples (calibrator samples). The fold change was determined using the equation 2 ÀDDCT . Mean fold change values in each group were analysed with the ANOVA one-way test using the Statgraphics Statistical Analysis and Data Visualization Software version 5.1.

Statistical analysis
Results were analysed statistically with SPSS 19.0 (SPSS, Inc., USA) software and GraphPad PRISM (GraphPad Software, Inc.) software. Data were presented as mean AE standard error of the mean (SEM). Pearson's correlation method was used to assess a possible linear association between two continuous variables in the studied samples. Data were compared with two-tailed unpaired Student's t-test or one-way analysis of variance (ANOVA) followed by post hoc Tukey's multiple comparison test when necessary. Differences between groups were considered statistically significant at *P < 0.05, **P < 0.01 and ***P < 0.001. Tendencies were considered at #P < 0.1.

Results
General pathology of the LC: tau phosphorylation and tau kinases The number of neurons with hyperphosphorylated tau deposition was first analysed in three nonconsecutive sections of the rostral part of the LC at the level of the lower segment of the cerebral aqueduct of Sylvius in every case stained with the antibody AT8 which recognizes tau phosphorylation sites Ser202/Thr205, and expressed as the number of positive cells in relation to the total number of neurons per section (about 5360 LC neurons counted in total). The number of AT8immunoreactive neurons increased with age and stage progression from about 1% at stage I of NFT pathology to about 15% at stage IV ( Fig 1A). Abnormal tau deposition was positive with tau-100 and 3Rtau and 4Rtau antibodies and with phosphospecific anti-tau antibodies P-tauThr181, P-tauSer262, P-tauSer422 and double-phosphorylated at Ser396/Ser404 (antibody PHF1). Most of them were also stained with conformation-specific antibodies Alz50 with tau modifications at amino acids 5-15, but only a minority of neurons were labelled with the antibody tau-C3 that recognizes truncated tau at aspartic acid 421 (Fig. 1B). Double-labelling immunofluorescence and confocal microscopy disclosed that 100% of P-tauThr181-positive neurons were stained with AT8 antibodies and about 80% of P-tauThr181-positive neurons with Alz50 antibodies (total number of P-tauThr181-positive neurons counted: 236). Double-labelling immunofluorescence revealed that only about 5% of P-tauThr181positive neurons were stained with tau-C3 antibody (total number of P-tauThr181-positive neurons counted: 128). Double-labelling immunofluorescence identified that only about 5% of AT8-positive neurons were stained with anti-ubiquitin antibodies in the LC at stages III and IV (total number of AT8-positive neurons counted: 214) (Fig. 1C). Immunoreactivity to tau-C3 and ubiquitin in neurons was restricted to those neurons bearing hyperphosphorylated tau (100% colocalization).

Energy metabolism and oxidative stress damage and responses
Antibodies against voltage-dependent anion channel (VDAC) immunostaining were used as a marker of mitochondrial membranes. Double-labelling immunofluorescence and confocal microscopy showed that a minority of LC neurons bearing hyperphosphorylated tau had decreased VDAC immunoreactivity ( Fig. 3A, a-c). Double-labelling immunofluorescence with tau-C3 antibody and anti-VDAC disclosed decreased VDAC immunoreactivity only in the subpopulation of neurons stained with tau-C3 antibody (Fig. 3A, d-f). A decrease in superoxide dismutase 1 (SOD1) immunoreactivity was found in LC neurons with hyperphosphorylated tau deposition when compared with neurons without tau in the same tissue section (Fig. 3B, a-c). This was accompanied by increased expression of neuroketal (NKT) restricted to LC neurons containing hyperphosphorylated tau as revealed by double-labelling immunofluorescence and confocal microscopy (Fig. 3B, d-f). Quantitative studies at stages III and IV identified increased NKT immunoreactivity and decreased SOD1 immunoreactivity in practically the totality of LC neurons stained with anti-P-tauThr181 antibodies (total P-tauThr181-immunoreactive neurons counted: 196).
Finally, double-labelling immunofluorescence with antibodies against specific antihaemoglobin and AT8 showed reduced haemoglobin a and haemoglobin b/c immunoreactivity in all LC neurons containing hyperphosphorylated tau in contrast with preservation of immunoreactivity in neighbouring LC neurons lacking abnormal hyperphosphorylated tau deposits (total AT8immunoreactive neurons assessed: 232) (Fig. 3C, a-f). Importantly, erythropoietin receptor immunoreactivity was preserved in LC neurons with hyperphosphorylated tau deposition (Fig. 3C, g-i).

Tyrosine hydoxylase and synapses
No modifications in TH immunoreactivity were observed in LC neurons bearing phospho-tau at stages I-IV of NFT pathology (Fig. 4A). Reduced synaptophysin immunoreactive perineuronal dots representing synaptic contacts with the neuronal cytoplasm were observed in LC neurons stained with tau-C3 antibodies (Fig. 4B). Quantitative studies (total neurons counted without NFTs: 126; total neurons with NFTs counted: 28) showed a decrease of between 40% and 60% of synaptophysin immunoreactive dots on the cell surface of tau-C3-immunoreactive LC neurons when compared with LC neurons without tau-C3 in the same tissue section.

Glial responses
The number of astrocytes was not increased in LC at stages I-IV (data not shown). In contrast, a moderate increase in the number of microglial cells, as revealed with Iba-1 antibody, was found at stages III and IV of NFT pathology. Some microglial cells adopted a reactive morphology with short, thick varicose branches (Fig. 5A).
Using the same cases employed in whole-transcriptome arrays, RT-qPCR showed significantly increased gene expression of AIF1 (which encodes Iba-1), CD68, PTGS2 (which encodes COX2), IL1b, IL6, and TNF-a in LC at stages III and IV of NFT pathology. In contrast, GFAP and IL10 mRNA expression levels were not modified in LC at the same stages (Fig. 5B).  correspond to nonglycosylated and glycosylated forms of the receptor respectively [53]. As expected, the anti-a 2A -AR did not detect any protein bands in cells transiently transfected with the empty vector (Figure S1, lane: mock), thus indicating that the antibody was specific. a 2A -AR expression was significantly increased in the hippocampal membrane fraction at stage I, but it was significantly decreased at stage IV (Fig. 6A). Interestingly, TH expression declined with age/disease progression, with significantly lower levels at stages III and IV in the same samples (Fig. 6A). The expression of a 2A -AR in the amygdala was also significantly increased at stage I, whereas TH expression was significantly reduced at stages I-IV (Fig. 6B).

Microarray analysis
Pearson's correlation method showed no association among post mortem delay, RIN values and gender variables. The only association was age with NFT stages. After data normalization and filter application, no outliers were detected, and the cofactors were not relevant to the analysis and therefore all samples were included. A total of 6870 gene set sequences were detected across all samples. Major differences in gene profile were identified in LC between stage I and stage IV of NFT pathology, as graphically illustrated in the heat map representation of significantly regulated transcripts at a P < 0.05 (Fig. 7). A complete list of deregulated genes is found in Table S4. and XPNPEP1. AD cases analysed are the same as those used in whole-transcriptome arrays. Mean fold change values for each group were compared with one-way analysis of variance (ANOVA) followed by post hoc Tukey's multiple comparison test and differences were considered statistically significant at *P < 0.05, **P < 0.01, ***P < 0.001 compared with MA; $P < 0.05, $$P < 0.01, $$$P < 0001 compared with stage I; &P < 0.05, &&P < 0.01, &&&P < 0.001 compared with stage II; and~P < 0.05,~~P < 0.01,~~~P < 0.001 compared with stage III. Tendencies were considered at #P < 0.1. Moderate activation of inflammatory responses is manifested by increased AIF1, CD68, PTGS2, ILb, IL6 and TNF-a with disease progression. GFAP and IL10 mRNA expression levels are not modified at these stages.

Enrichment analysis against GO database and IPA
Statistical analysis revealed the most significant enrichment analysis against GO database (P < 0.05) was between stages I and IV of NFT pathology. Upregulated genes coded for proteins were linked to protein folding, heat shock protein binding and ATP metabolism. Downregulated genes were linked to DNA binding and members of the small nucleolar RNAs family (SNOR).
IPA revealed upregulated pathways corresponding to 'neurological disease', 'hereditary disorder' and 'psychological disorders'. The analysis of molecular and cellular functions identified 'post-translational modification mechanisms', 'protein folding system' and 'cellular compromise components'. Detailed analysis is shown in Data S4 and Figure S2.

Validation of microarray data by qRT-PCR
We selected five genes for validation from the four main altered functional pathways including (i) protein folding mechanisms HSPA1B and DNAJB1; (ii) ATP metabolism ASPA6; (iii) DNA/RNA binding proteins: ID1; and (iv) noncoding small nucleolar RNAs SCARNA2. Stages I, II, III and IV were included in the study.
RT-qPCR in LC samples confirmed significant upregulation of HSPA1B and DNAJB1 at stage IV when compared with stage I; DNAJB1 was also  Table 1). One-way analysis of variance (ANOVA) followed by post hoc Tukey's multiple comparison test were used; differences are considered statistically significant at *P < 0.05 compared with control; $P < 0.05 and $$P < 0.01 compared with stage I. significantly upregulated at stage IV when compared with stage II.
HSPA6 was significantly upregulated at stage IV when compared with stage I. In contrast, ID4 was downregulated at stage III when compared with stage I. A nonsignificant trend of reduction was also observed for ID4 at AD stage IV when compared with stage I.
Finally, SCARNA2 was downregulated in LC at stages II, III and IV when compared with stage I (Fig. 8).

Discussion
Cases in the present series were selected on the basis of the presence of NFTs in the LC in the context of early Braak staging of NFT pathology together with the absence of clinical symptoms related to cognitive impairment and major depression. All these cases can be categorized as preclinical AD stages according to the ABC classification proposed by the NIA/AA [54,55]. However, it can be argued that some cases might not represent AD but rather primary age-related tauopathy (PART) [56,57], because most had no b-amyloid plaques in the brain (particularly those at NFT stages I and II, and some at stage III), and lesions were restricted to tau pathology [58]. This is difficult to discuss at present, as PART (formerly dementia with only tangles when advanced) has also been considered part of the AD spectrum [59]. For this reason, herein we have avoided the term 'AD' and preferred to use the descriptive term of 'NFT pathology categorized as Braak stages'.
Seminal quantitative studies using unbiased methods have shown that neurone loss in the LC occurs at advanced stages of AD [60,61]. This has recently been determined in a more comprehensive study, using the unbiased stereological method of optical fractionators, showing that cell loss in the LC was insignificantly small at preclinical stages of AD [62]. Neurone loss occurs in prodromal AD in cases with mild cognitive impairment, and neurone numbers markedly decrease in the LC along with progression of dementia [62]. The present study has taken advantage of these observations to focus on Braak stages of NFT pathology which do not significantly compromise the total number of neurons in the LC.
Expression of activated kinases involved in tau phosphorylation colocalizing with hyperphosphorylated tau deposits in the LC has the same characteristics as those reported in cerebral cortex in AD [63][64][65][66]. As described in other regions, phosphorylation of tau precedes altered tau conformation; tau truncation occurs only after complex tau phosphorylation, and tau truncation in neurons at aspartic acid 421 is needed to fully develop the cascade of NFT formation, including recruitment of full-length tau [67][68][69][70][71][72][73]. Strong association has been found between tau truncation at aspartic acid 421 and ubiquitin deposition in neurons in AD [74], and tau truncation has been proposed as an inducer of cell death via caspase activation [75].
Decreased VDAC immunoreactivity, suggesting a decrease in the number of mitochondria, was identified only in neurons containing truncated tau as revealed with the antibody tau-C3. Hyperphosphorylated tau and particularly truncated tau produce mitochondrial damage [76][77][78][79], thus accounting for decreased numbers of mitochondria in LC neurons bearing tau-C3.
Hyperphosphorylated tau deposition in LC neurons is accompanied by increased neuroketal adducts, indicating oxidative stress damage [80], and reduced SOD1 immunoreactivity, suggesting reduced oxidative stress responses in the same neurons and arguing in favour of increased oxidative damage in the LC at early stage NFT pathology [81][82][83]. Haemoglobin a and haemoglobin b/c immunoreactivity is also reduced in LC neurons with hyperphosphorylated tau deposition, as previously reported in neurons of the cerebral cortex in AD [84]. It has been suggested that brain haemoglobin subunits may participate in redox homeostasis and mitochondrial activity [85,86]. a-Haemoglobin modulates antioxidant defences and facilitates tyrosine hydroxylase synthesis and activity in peripheral catecholaminergic cells [87].
Microglia in the LC at middle stages of NFT pathology appear ramified with short branches, buds and varicosities described as senescent microglia [88], a feature also observed in association with tau pathology in other brain regions [89]. These morphological changes are accompanied by increased mRNA expression of AIF1 (which encodes Iba-1), PTGS2 (which encodes COX2), IL1b, IL6 and TNF-a.
The timing of these alterations is similar to that seen in P301S transgenic mice used as a model of Cases analysed are the same as those used in whole-transcriptome arrays. Mean fold change values of each group were compared with one-way analysis of variance (ANOVA) followed by post hoc Tukey's multiple comparison test and differences are considered statistically significant at *P < 0.05, **P < 0.01, ***P < 0.001 compared with MA; $P < 0.05, $$P < 0.01, $$$P < 0001 compared with stage I; &P < 0.05, &&P < 0.01, &&&P < 0.001 compared with stage II; and~P < 0.05,~~P < 0.01,~~P < 0.001 compared with stage III. Tendencies are considered at #P < 0.1. frontotemporal lobar dementia linked to mutations in MAPT (FTLD-tau) [90] and differs from biochemical responses of neurons and glial cells in mouse models of b-amyloid deposition [91].
Focal loss of synaptic contacts is found only in neurons bearing truncated tau. This alteration is seldom accompanied by focal synaptophysin immunoreactivity clustering, suggesting altered localization of synaptic molecules in this subpopulation of neurons. Reduced synaptic contacts have also been reported in tau-containing neurons in the hippocampus in AD [92], reinforcing the notion that reduced connectivity of neurons with NFTs may be due, in part, to progressive loss of synaptic inputs, leading to progressive neuronal isolation.
Decreased TH immunoreactivity and increased neurons with NFTs in the LC occur in AD at advanced stages of the disease [14,93]. The present findings show preserved TH immunoreactivity in LC neurons at asymptomatic Braak stages I-IV. Reduced cortical noradrenergic innervation and expression of adrenergic receptors have been reported in symptomatic AD [14,[94][95][96]. Reduced noradrenergic innervation is accompanied by sprouting noradrenergic afferents in the hippocampus [97] and compensatory changes in the LC and hippocampus in advanced stages of AD [98,99]. Interestingly, dopamine beta-hydroxylase activity in plasma is decreased at early stages of AD, suggesting a compensatory mechanism for the loss of noradrenergic neurons [100].
The present findings show increased expression levels of a 2A -ARs in the hippocampus and amygdala at first stages of NFT pathology occurring in parallel with local decrease of TH protein levels. The present observations show that compensatory changes in noradrenergic target nuclei might occur at very early Braak stages of NFT pathology.
The differentially expressed genes obtained after the microarray raw data processing did not reach the high fold change levels commonly established for transcript selection. For this reason, we selected those transcripts that showed an absolute fold change equal to or greater than 1.2 and unadjusted p < 0.05.
Whole-transcriptome arrays have identified deregulated clusters including (i) genes encoding proteins linked with protein folding, chaperone binding and heat shock protein binding which were upregulated at stage IV when compared with stage I; (ii) genes associated with ATP metabolism, including genes associated with ATPase activator activity and ATPase regulator activity which were upregulated at stage IV when compared with stage I; (iii) genes coding for DNA-binding proteins which were downregulated at stage IV when compared with stage I; and (iv) members of the small nucleolar RNAs family (SNOR) downregulated at stage IV when compared with stage I.
Representative elements of these four clusters were validated with RT-qPCR including heat shock protein members HSPA1A/Bis, HSPA6 (Hsp70, member 6) and DNAJB1 (HSP40) [101][102][103][104]. Altered expression of these genes is not unexpected as they are likely linked to the abnormal tau protein and folding in NFTs. ID4 (inhibitor of DNA binding 4) encodes a member of the inhibitor of DNA-binding protein family which acts as helix-loop-helix transcription factors whose activity depends on the protein binding partner [105]. Small nucleolar RNAs (snoRNAs), including SCARNA2 (small Cajal body-specific RNA 2), are noncoding RNAs which regulate post-transcriptional processing of other noncoding RNAs, among other functions such as micro-RNA-dependent gene silencing and alternative splicing [106][107][108]. The involvement of DNA-binding proteins and the snoRNA family in the pathology of LC in AD is relevant, as nuclear and nucleolar RNAs and proteins are emerging as key participants in neurodegeneration [109][110][111], and altered protein synthesis from the nucleolus to the ribosome has recently been reported in AD [112].

Conclusions
The present study shows that intraneuronal deposition of hyperphosphorylated tau and misfolded tau, and incorporation of truncated tau in the LC at asymptomatic early and middle stages of NFT pathology are accompanied by altered mitochondria, reduced intraneuronal haemoglobin a and b, increased oxidative stress and reduced stress responses. This concurs with microglia proliferation and increased expression of AIF1, CD68, PTGS2, IL1b, IL6 and TNF-a. Whole-transcription analysis has shown novel altered metabolic pathways in the LC at early stages of NFT pathology including upregulation of genes associated with protein folding and chaperone binding, and genes associated with ATP metabolism, and downregulation of genes coding for DNA-binding proteins and members of the small nucleolar RNAs family. Tyrosine hydroxylase (TH) immunoreactivity is preserved in neurons of LC but TH protein levels are decreased in the amygdala and hippocampus. Parallel increased expression levels of a 2A -ARs in the hippocampus and amygdala suggest compensatory activation in the face of decreased adrenergic input. Although subtle changes in mood and wake-sleep cycle disturbances, and discrete impairment of attention and learning capacity, cannot be excluded in this series, present observations unveil novel alterations in the LC and projections in asymptomatic individuals at early and middle stages of NFT pathology.

Supporting information
Additional Supporting Information may be found in the online version of this article at the publisher's web-site: Figure S1. Immunoblot detection of human a 2A -AR expressed in HEK-293T cells. Figure S2. (A) IPA analysis of top altered networks in LC and network-related diseases and biofunctions comparing Braak stage IV with stage I of NFT pathology. (B) Interaction diagram of altered pathways. Table S1. Cases used for whole-transcriptome assays and mRNA validation in the LC. Table S2. Characteristics of the antibodies used for immunohistochemistry (IHC) and immunoblotting (IB). Table S3. Probes used for TaqMan assays; AARS, GUSb, HPRT1 and XPNPEP1 were used as house-keeping genes. Table S4. List of deregulated genes. Data S1. Brain tissue was obtained from the Institute of Neuropathology Brain Bank following the guidelines of Spanish legislation on this matter and the approval of the local ethics committee. Data S2. The cDNA encoding the human a 2A -AR cloned in pcDNA3.1+ was purchased to UMR cDNA Resource Center (University of Missouri, Rolla, MO, USA). Data S3. Samples were assessed by microarray hybridization with GeneChip â Human Transcriptome Array 2.0 (HTA 2.0) and Affymetrix microarray 7000G platform from Affimetrix (Santa Clara, CA, USA). Data S4. Statistical analysis revealed the most significant enrichment analysis against GO database (P < 0.05) was between AD stage I and stage IV.