Acta Neuropathologica

, Volume 112, Issue 4, pp 439–449

No alteration in tau exon 10 alternative splicing in tangle-bearing neurons of the Alzheimer’s disease brain

Authors

    • Harvard Medical SchoolMassachusetts General Hospital
    • Department of Public Health/Molecular GeriatricsUppsala University
  • Karunya Ramasamy
    • Harvard Medical SchoolMassachusetts General Hospital
  • Ippolita Cantuti-Castelvetri
    • Harvard Medical SchoolMassachusetts General Hospital
  • Lena Skoglund
    • Department of Public Health/Molecular GeriatricsUppsala University
  • Toshifumi Matsui
    • Harvard Medical SchoolMassachusetts General Hospital
  • Jennifer Orne
    • Harvard Medical SchoolMassachusetts General Hospital
  • Hasimoto Kowa
    • Harvard Medical SchoolMassachusetts General Hospital
  • Susan Raju
    • Harvard Medical SchoolMassachusetts General Hospital
  • Charles R. Vanderburg
    • Harvard Medical SchoolMassachusetts General Hospital
  • Jean C. Augustinack
    • Harvard Medical SchoolMassachusetts General Hospital
  • Rohan de Silva
    • Reta Lila Weston Institute of Neurological StudiesUniversity College London
  • Andrew J. Lees
    • Reta Lila Weston Institute of Neurological StudiesUniversity College London
  • Lars Lannfelt
    • Department of Public Health/Molecular GeriatricsUppsala University
  • John H. Growdon
    • Harvard Medical SchoolMassachusetts General Hospital
  • Matthew P. Frosch
    • Harvard Medical SchoolMassachusetts General Hospital
  • David G. Standaert
    • Harvard Medical SchoolMassachusetts General Hospital
  • Michael C. Irizarry
    • Harvard Medical SchoolMassachusetts General Hospital
  • Bradley T. Hyman
    • Harvard Medical SchoolMassachusetts General Hospital
Original Paper

DOI: 10.1007/s00401-006-0095-3

Cite this article as:
Ingelsson, M., Ramasamy, K., Cantuti-Castelvetri, I. et al. Acta Neuropathol (2006) 112: 439. doi:10.1007/s00401-006-0095-3

Abstract

Defective splicing of tau mRNA, promoting a shift between tau isoforms with (4R tau) and without (3R tau) exon 10, is believed to be a pathological consequence of certain tau mutations causing frontotemporal dementia. By assessing protein and mRNA levels of 4R tau and 3R tau in 27 AD and 20 control temporal cortex, we investigated whether altered tau splicing is a feature also in Alzheimer’s disease (AD). However, apart from an expected increase of sarcosyl-insoluble tau in AD, there were no significant differences between the groups. Next, by laser-capture microscopy and quantitative PCR, we separately analyzed CA1 hippocampal neurons with and without neurofibrillary pathology from six of the AD and seven of the control brains. No statistically significant differences in 4R tau/3R tau mRNA were found between the different subgroups. Moreover, we confirmed the absence of significant ratio differences in a second data set with laser-captured entorhinal cortex neurons from four AD and four control brains. Finally, the 4R tau/3R tau ratio in CA1 neurons was roughly half of the ratio in temporal cortex, indicating region-specific differences in tau mRNA splicing. In conclusion, this study indicated region-specific and possibly cell-type-specific tau splicing but did not lend any support to overt changes in alternative splicing of tau exon 10 being an underlying factor in AD pathogenesis.

Keywords

BrainNeurodegenerationAlzheimer’s diseaseTauAlternative splicing

Introduction

Neurofibrillary tangles (NFTs) and amyloid-β (Aβ) plaques are the two neuropathological hallmarks of the Alzheimer’s disease (AD) brain. The main constituent of the NFT is the microtubule-associated protein tau, a critical molecule for cellular cytoarchitecture and proper axonal transport. In AD, tau is phosphorylated at certain epitopes [12] in a sequential order [2] and increased levels of phosphorylated tau have been demonstrated in the AD brain and cerebrospinal fluid [14, 28]. Moreover, it has been proposed that an excess of tau in AD neurons could impair axonal trafficking [23].

The tau gene undergoes alternative splicing of exons 2, 3 and 10, which results in six different tau isoforms [9]. Binding between tau and tubulin is mediated through the C-terminal region of tau, and the inclusion or exclusion of tau exon 10 gives rise to isoforms with three (3R tau) or four (4R tau) microtubule-binding repeat regions, respectively. For certain neurodegenerative disorders with tau pathology, the dysregulation of 4R tau and 3R tau expression is believed to be a pathogenic feature. For example, brains from patients with progressive supranuclear palsy (PSP) generally display a predominance of 4R tau pathology [4]. Moreover, many of the tau mutations causing FTD with parkinsonism (FTDP17) have in vitro been shown to cause a dysregulation of tau exon 10 splicing [13, 15, 16] (and reviewed in [18]). Moreover, a recent study using quantitative PCR (qPCR) techniques demonstrated a two- to sixfold increase in 4R tau/3R tau mRNA in FTDP17 brains [6].

Whether or not there is a dysregulation of tau exon 10 splicing in AD has been uncertain. Preparations of paired helical tau filaments (PHF tau) from the AD brain consist of excessively phosphorylated species of all six tau isoforms in seemingly equal amounts [8], and previous studies have failed to find any changes in the 4R tau/3R tau ratio in AD [5, 10]. However, by immunohistochemistry with tau isoform-specific antibodies, some studies have suggested that NFT-bearing neurons of the AD brain are predominantly 3R tau-positive [7, 19, 22], whereas in a recent study no differences in tau isoform composition could be seen during early stages of tangle pathology [26].

In the present study, we have examined a neuropathological cohort of AD and control brains in order to investigate whether there is an alteration in overall tau protein and mRNA levels in the AD brain. Moreover, we examined whether dysregulation of tau exon 10 splicing, as for other tauopathies, may be a feature also of the AD brain. Apart from determining total levels of tau protein and tau mRNA, we measured 4R tau and 3R tau mRNA in preparations of temporal neocortex as well as in laser-microdissected tangle- and non-tangle-bearing neurons from the hippocampal CA1 subregion of AD and control brains. Finally, we reasoned that possible alterations in tau splicing could be related to other pathological brain features and, therefore, assessed NFT densities as well as mRNA levels of glial fibrillary acidic protein (GFAP) and synaptophysin (as markers for astrocytosis and synaptic integrity, respectively).

Materials and methods

Human brain samples

Fresh frozen brain tissue was obtained from the brain banks at Massachusetts General Hospital, Harvard University and University of Maryland. Of the 27 AD cases, 52% were males and 48% females. The average age at death was 80.6 (± 1.4) years with 15.5 (± 2.5) h post-mortem interval (PMI). Of the 20 control subjects, 45% were males and 55% females, with an average age at death of 82.6 (± 2.8) years and with 22.2 (± 3.4) h PMI. All AD subjects had been evaluated at Massachusetts General Hospital/Memory Disorders Unit and met both the clinical (NINCDS-ADRDA) [24] and the neuropathological (CERAD, NIA/Reagan) [17, 25] diagnostic criteria for AD. Control brains did not have any history or neuropathological signs of a neurodegenerative disorder.

All tissue dissections for RNA analyses were carried out on coronal sections of fresh frozen brain tissue under stringent, RNAse-free conditions. For the homogenate samples, 50–100 mg of tissue from the anterior part of the temporal neocortex was dissected from each case. In addition, pieces of tissue containing the entire hippocampal formation were dissected from six of the AD and seven of the control brains. The hippocampal tissues were cut as 8 μm sections in the cryostat, mounted on microscopy slides (Gold Seal Rite-On Micro Slides, Portsmouth, NH) and stored in −80°C until further processing.

ELISA

For the tau protein ELISA, frozen tissues were homogenized with a mechanical homogenizer in 10 μl/mg v/ww (volume/wet weight) tris buffer (Tris 50 mM, pH 7.4, 400 mM NaCl, 2 mM EDTA) containing 0.1% Triton X-100, protease inhibitors (Complete, Roche, Indianapolis, IN, USA) and 2% protease-free bovine serum albumin (BSA). The soluble fraction was extracted after centrifugation at 15,000 rpm at 4°C (5 min).

Tau protein levels were measured with the Innotest hTAU Antigen ELISA [27], utilizing AT120 as capturing antibody and HT7/BT2 as reporting antibodies, according to the manufacturer’s instructions (Innogenetics, Ghent, Belgium). ELISA was performed with 25 μl of 100,000× dilution (100 ml/mg ww) of the soluble fraction in duplicates. Finally, samples were analyzed with an automatized plate reader and the average concentrations were calculated. The concentration of tau is presented as ng/g ww tissue.

Generation of a 4R tau antibody

4R tau-specific sera was raised in rabbit against the synthetic oligopeptide INKKLDLSNVQSK-C, corresponding to a mid-region of tau exon 10. Antisera were purified by affinity chromatography (UltraLink™ Iodoacetyl, Pierce, Rockford, IL, USA) and purified antibodies against 4R tau (4RT) were tested for specificity on Western blot against a mix of all six recombinant tau isoforms. Only the three 4R tau isoforms, but none of the three 3R tau isoforms, were recognized by 4RT (Fig. 1a).
https://static-content.springer.com/image/art%3A10.1007%2Fs00401-006-0095-3/MediaObjects/401_2006_95_Fig1_HTML.gif
Fig. 1

Demonstration of specificity for tau antibodies and PCR primers. a Western blot on a mix of recombinant tau isoforms, representing all six low-molecular isoforms. Probing with the pan-tau ab Tau 1 visualized all isoforms, whereas the 4RT antibody only recognized the 4R tau isoforms. b Quantitative PCR on 0.05 amol of 4R tau and 3R tau vector cDNA, respectively, using primer sets against 4R tau and 3R tau, demonstrating comparable primer efficiencies between the sets (as indicated by detectable amplification curves at the same PCR cycle numbers). In addition, primer specificities are exceeding 97% for both 4R tau and 3R tau primer sets

Western blot

Sarcosyl-insoluble and sarcosyl-soluble tau protein were extracted, as previously described [8], from temporal neocortex of four of the AD and four of the control brains. Briefly, 50 mg of the tissue was homogenized in 500 μl buffer containing 10 mM Tris–HCl, 0.8 M NaCl, 1 mM EGTA, 10% sucrose and supplemented with protease inhibitors. The homogenate was centrifuged at 20,000g at 4°C (20 min). The supernatant was incubated with 1% N-lauroylsarcosinate at RT (1 h) and centrifuged at 100,000g at 4°C (1 h). The resulting supernatant containing soluble tau was collected and the pellet containing insoluble tau was resuspended in 50 mM Tris–HCl. Samples were resolved together with a tau protein ladder of all six isoforms (rPeptide, Bogart, GA, USA) on a 10% SDS-PAGE gel, transferred to a nitrocellulose membrane and probed with the RD3 [7] (Upstate, Waltham, MA, USA) and 4RT antibodies. Finally, bands were visualized using horseradish peroxidase-conjugated anti-mouse and anti-rabbit antibodies (Pierce, Rockford, IL, USA) followed by chemiluminescence detection (ECL, Amersham, Piscataway, NJ, USA).

Tissue staining for laser-capturing microscopy

The cryostat-sectioned tissues were stained following a protocol, in which molecular biology grade water (Sigma-Aldrich, St Louis, MO, USA) was used for all aqueous solutions. For visualization of tangles, each hippocampal tissue section was lightly fixed in 70% EtOH (40 s). Next, the sections were incubated in 95% EtOH (2 min) and 70% EtOH (2 min), followed by a quick rinse in H2O before staining with 0.1% thioflavine S (thio S) (5 min). After brief washes in 80% EtOH and H2O, the sections were incubated in a toluidine blue solution (Arcturus, Mountain View, CA, USA) (1 min), followed by dehydration in increasingly concentrated ethanols and xylenes before microdissection. All incubations were carried out at room temperature.

Laser-capturing microscopy

Hippocampi from six of the AD and seven of the control brains were included. The 8 μm-thin sections were subjected to laser-capturing microscopy (Arcturus). A light source with normal and UV light combined was used to simultaneously visualize thio S and toluidine blue, thus ensuring that separate neuronal groups could be captured (Fig. 2). From the AD brains, 600 toluidine blue positive/thio S negative non-tangle-bearing neurons and 600 toluidine blue positive/thio S positive tangle-bearing neurons from the CA 1 hippocampal subfield were captured onto separate polyethylene caps (HS Cap, Arcturus). From the control brains, 600 toluidine blue positive CA1 neurons were captured onto a separate cap.
https://static-content.springer.com/image/art%3A10.1007%2Fs00401-006-0095-3/MediaObjects/401_2006_95_Fig2_HTML.jpg
Fig. 2

CA1 tissue section (8 μm) stained with toluidine blue/thioflavine S. Neurons with (orange arrows) and without (green arrows) neurofibrillary pathology were visualized with combined brightfield and UV light before laser capturing

As a second data set, we also captured clusters of tangle-bearing or non-tangle-bearing stellate neurons in layer II of entorhinal cortex (EC) from four AD and four control brains. In total, 5 mm2 of tissue were captured for each sample.

RNA extraction

From the whole tissues, RNA was extracted with the Trizol reagent according to the manufacturer’s instruction (Invitrogen, Carlsbad, CA, USA). In brief, the tissues were homogenized in Trizol in RNAse-free vials, followed by centrifugation at 15,000g and subsequent incubations in chloroform and 2-propanol before obtaining an RNA-pellet which, after washing in EtOH, was resuspended in 20 μl of H2O and immediately placed on dry ice.

After measurement of the RNA concentration with spectrophotometry, all samples were equally rediluted to 1 μg/μl. DNAse treatment of the samples was performed (R Q1 RNAse-free DNAse, Promega, Madison, WI, USA), after which the RNA concentration of each sample was determined again (to compensate for differing extraction yields).

Next, the RNA was reprecipitated by subjecting the samples to phenol:chloroform:IAA (pH 6.6/7.9, Ambion, Austin, TX, USA) (100% of starting volume) and 3 M NaAc (Ambion) (10% of starting volume). After vortexing briefly, the samples were kept on ice (15 min), followed by centrifugation at 15,000g (15 min) and transferring of the supernatant to a new tube. After adding 2.5 × 100% EtOH and 1 μl glycogen (1:10, Ambion), the samples were vortexed again and kept on ice (30 min) before centrifugation at 15,000g (30 min). Finally, the resulting pellet was washed with 75% EtOH, recentrifuged and dissolved in H2O to a final concentration of 500 ng/μl. All samples were controlled for integrity of 18S and 28S ribosomal RNA by microcapillary electrophoresis (RNA 6000 Nano Assay, Agilent Technologies, Palo Alto, CA, USA) (not shown) and samples showing signs of degradation were at this point excluded from the study.

RNA extraction from the laser-dissected samples was carried out according to the manufacturer’s specifications (PicoPure RNA Isolation Kit, Arcturus). In brief, this solid-phase extraction procedure is based upon a principle in which the RNA from the cell homogenate gets attached to a column-filter and, after DNAse treatment (RNAse-free DNAse set, Qiagen, Valencia, CA, USA), is eluted in 30 μl of elution buffer.

Reverse transcriptase reaction

Reverse transcription (RT) was carried out on all RNA preparations (Superscript II, Invitrogen). For the temporal RNA samples, 2 μg of DNAse-treated RNA was mixed with random hexamers and 10 nmol of dNTPs. After denaturation at 65°C (5 min), the following reagents were added to each sample: 100 nmol of MgCl2, 200 nmol of dithiothreitol (DTT), 40 U of RNAse Out Inhibitor and 2.0 μl of 10× RT-PCR buffer. After incubation in room temperature (2 min), 200 U of Superscript II reverse transcriptase (Invitrogen) was added, followed by incubation at 25°C (10 min) and 42°C (50 min). Finally, the reaction was deactivated at 70°C (15 min).

For the LCM-samples, we used a modified RT protocol, in which the entire 30 μl yield from the RNA-extraction was used in the reaction. All other PCR-ingredients were scaled up proportionally. To reduce the final cDNA sample volume, the phenol:chloroform:IAA-based precipitation protocol described above was used.

Quantitative PCR

Primers against 4R tau, 3R tau, glial fibrillary acidic protein (GFAP), synaptophysin and the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were developed (Table 1) [21]. The specificities of the tau primers were controlled by testing against a known concentration of 4R tau and 3R tau cDNA, respectively, and both primer pairs were found to have a specificity exceeding 97% (Fig. 1b). In addition, the resulting 4R tau and 3R tau amplicons were subjected to microcapillary electrophoresis and sequence analysis, which confirmed the accuracy of the PCR (not shown).
Table 1

Sequences and annealing temperatures of the oligoprimers used in this study

Gene

Forward primera

Reverse primera

Annealing temperature (°C)

4R tau

GAAGCTGGATCTTAGCAACG

GACGTGTTTGATATTATCCT

58

3R tau

AGGCGGGAAGGTGCAAATAG

TCCTGGTTTATGATGGATGTT

58

GFAPb

GATCAACTCACCGCCAACAGC

CTCCTCCTCCAGCGACTCAATCT

58

Synaptophysin

AGGGAACACATGCAAGGAG

CCTTAAACACGAACCACAGG

58

GAPDH

GGTCTCCTCTGACTTCAACA

GTGAGGGTCTCTCTCTTCCT

55

aAll primers are given in the 5′–3′ direction

bPreviously published [21]

For the temporal neocortical samples, duplicates of 50 μl reactions containing 25 μl SYBRgreen Mastermix (Applied Biotechnology, Foster City, CA, USA), 200 nM of each primer and 30 ng of cDNA were used in the qPCR. After initial denaturation at 95°C (6 min), 50 cycles at 95°C (30 s), the primer-specific annealing temperature (Table 1) (30 s) and 72°C (45 s) were performed before a final incubation at 95°C (1 min). The same protocol was used for the LCM-samples, with cDNA from 80 cells used for each reaction.

For the standard curve, we subcloned cDNA amplicons, generated by using the respective qPCR primers (Table 1), in the pcDNA 3.1 vector system (Invitrogen) according to the manufacturer’s instructions. The cDNA generated in this way will be in a circular conformation, but is denatured during an initial denaturation step of the PCR. After verifying the specificity by sequencing, the cDNA clones were used to generate individual standard curves, ranging from 3 to 30,000 molecules, thus allowing for calculation of specific primer efficiencies and respective initial number of mRNA molecules for the different samples. Nevertheless, an exact estimation of molecule numbers cannot be obtained since PCR carried out in a “target-rich environment” (as for the standard curve samples) and in a complex mix of cDNA molecules (as for the brain-derived cDNA samples) will affect primer efficiencies differently. The “true” numbers of molecules is therefore likely to be higher than those given here, but the overall interpretations should not be affected since the same conditions were used to derive all numbers.

Statistical methods

Comparison of protein and mRNA levels was performed by using a t test (Statview, SAS, Cary, NC, USA). A sample size analysis demonstrated that there was 0.80 power at a 0.05 significance level to test a hypothetical twofold difference in the 4R tau/3R tau ratio between AD patients and control subjects. A simple regression analysis was used to correlate the 4R tau/3R tau ratio with NFT densities and mRNA levels of GFAP and synaptophysin (Statview). The protein and mRNA levels of the temporal cortex samples are given as both number of mRNA molecules/μg total mRNA and as number of mRNA molecules/GAPDH molecules. For the LCM-based samples, levels are only given as number of mRNA molecules/GAPDH molecules. All data are presented with ± 1 standard error (SE).

Results

No difference in total soluble tau protein levels or in the ratio of 4R tau/3R tau proteins between AD and control brains

Although it is well established that insoluble highly phosphorylated tau is abundant in AD as compared to control brains, we examined whether there were any changes in the amount of soluble, presumably cytoplasmic, tau in AD neocortex. Levels of tris-soluble tau were measured by ELISA on 14 of the AD and 13 of the control brains. Alzheimer brains had 63.2 ± 5.0 ng of tau /g ww tissue whereas the control brains had 70.4 ± 7.5 ng of tau /g ww tissue (P = 0.43) (Fig. 3a).
https://static-content.springer.com/image/art%3A10.1007%2Fs00401-006-0095-3/MediaObjects/401_2006_95_Fig3_HTML.gif
Fig. 3

Analysis of tau protein in temporal neocortex. a ELISA measurements of total soluble tau protein showed no difference between AD and control temporal neocortex. b Western blot on the sarcosyl-insoluble fraction showed increased levels and a decreased electrophoretic mobility in the AD brains for both 4R tau (upper left panel) and 3R tau (lower left panel) isoforms. The amount of sarcosyl-soluble tau was similar between AD and control brains. No apparent shift in the 4R tau/3R tau ratio was seen for either the AD or the control group

As expected, Western blot analysis revealed markedly elevated amounts of both 4R tau and 3R tau protein in the sarcosyl-insoluble phase of the AD brains (Fig. 3b). However, in accordance with the ELISA data, no obvious quantitative difference could be seen for sarcosyl-soluble tau, when comparing samples from diseased and non-diseased brains. Finally, it could be seen that 4R tau and 3R tau isoforms were roughly equally represented within both the AD and the control group (Fig. 3b).

No significant changes in levels of 4R tau and 3R tau mRNA between AD and control brains

For samples derived from temporal neocortex, we calculated the amount of mRNA in two ways: the amount of target molecules per microgram total RNA, which is advantageous in that it makes no assumptions about (the lack of) alterations in any particular housekeeping gene used as a normalization factor. Moreover, we determined the amount of target molecules as a ratio to GAPDH mRNA, a housekeeping gene frequently used as a normalization factor. In our samples, GAPDH was slightly but non-significantly diminished in AD, and results using either method of normalization were roughly equivalent.

Alzheimer’s disease brains contained 122 ± 25 × 103 4R tau mRNA molecules/μg total RNA, whereas the average level in the control brains was 149 ± 24 × 103 4R tau mRNA molecules/μg total RNA (P = 0.45) (Fig. 4a). When normalized to GAPDH mRNA, the 4R tau mRNA levels were 0.18 ± 0.02 4R tau mRNA molecules/GAPDH mRNA molecule in AD and 0.17 ± 0.02 4R tau mRNA molecules/GAPDH mRNA molecule (P = 0.79) (not shown).
https://static-content.springer.com/image/art%3A10.1007%2Fs00401-006-0095-3/MediaObjects/401_2006_95_Fig4_HTML.gif
Fig. 4

No difference in levels of 4R tau mRNA (a), 3R tau mRNA (b) or 4R tau/3R tau mRNA ratio (c) in AD as compared to control temporal neocortex

The AD brains expressed 111 ± 20 × 103 3R tau mRNA molecules/μg total RNA, whereas for control brains, the levels were 166 ± 23 × 103 3R tau mRNA molecules/μg total RNA (P = 0.07) (Fig. 4b). When normalized to GAPDH mRNA, the 3R tau mRNA levels were 0.20 (± 0.04) 3R tau mRNA molecules/GAPDH mRNA molecule in AD and 0.21 (± 0.03) 3R tau mRNA molecules/GAPDH molecule in control brains (P = 0.79) (not shown).

Moreover, there was no difference in the ratio of 4R tau/3R tau mRNA levels between the two groups; the average 4R tau/3R tau mRNA ratio for the AD brains was 1.18 ± 0.13, whereas for control brains, the ratio was 1.17 ± 0.34 (Fig. 4c).

Finally the total tau mRNA levels were calculated by adding the levels of 4R tau mRNA and 3R tau mRNA. In the AD brains, there were 233 ± 42 × 103 tau mRNA molecules/μg total RNA, whereas the average level in the control brains was 315 ± 40 × 103 tau mRNA molecules/μg total RNA (P = 0.17) (not shown).

Density of neurofibrillary tangles in the AD temporal neocortex does not correlate with mRNA levels of tau

The NFT density of the temporal cortex (superior temporal sulcus) was available for 21 of the AD cases and had been assessed according to previously published stereological principles [20]. The average NFT density for these cases was 2,334 ± 320 tangles/mm3 and when the individual densities were compared with the 4R tau and 3R tau mRNA levels, no correlations were seen. There was also no correlation between the tangle density and the 4R tau/3R tau mRNA ratio or 4R tau and 3R tau mRNA levels alone (not shown).

Increased levels of GFAP mRNA and decreased levels of synaptophysin mRNA in AD temporal neocortex, but no correlation to tau mRNA levels

In the AD brains, there was a fourfold increase in GFAP expression with 24.9 ± 5.0 × 106 GFAP mRNA molecules/μg total RNA as compared to the control brains with 6.1 ± 1.7 × 106 GFAP mRNA molecules/μg total RNA (P = 0.003) (Fig. 5a). The expression of synaptophysin was > 50% lower in AD with 98 ± 24 × 103 synaptophysin mRNA molecules/μg total RNA whereas for the control brains, the levels were 232 ± 32 × 103 synaptophysin mRNA molecules/μg total RNA (P = 0.002) (Fig. 5b). The same qualitative results and statistically significant changes were observed also when the data were expressed as a ratio of target message to GAPDH (not shown). No correlations between GFAP or synaptophysin mRNA levels and mRNA levels of 4R tau or 3R tau could be seen (not shown).
https://static-content.springer.com/image/art%3A10.1007%2Fs00401-006-0095-3/MediaObjects/401_2006_95_Fig5_HTML.gif
Fig. 5

Increased levels of GFAP mRNA in AD temporal neocortex as compared to control brains (a) and lower levels of synaptophysin mRNA in AD than in control temporal neocortex (b); **P < 0.01

No change in the 4R tau/3R tau mRNA ratio in tangle-bearing neurons of the AD brain

In order to assess the overall specificity for our laser-capture procedure, we used 80 captured hippocampal neurons from one of the cases included in the study to estimate the GFAP mRNA levels. The level from the LCM sample was compared with the GFAP level in the temporal homogenate from the same case (normalized to GAPDH). The microdissected sample was found to have approximately 20× lower relative GFAP expression, illustrating that our LCM procedure led to a successful dissection from non-neuronal mRNA in the processed samples.

For the LCM-based samples, we compared the amount of tau mRNA species to the amount of GAPDH mRNA measured in the same set of samples, since measuring the total amount of mRNA in the laser-capture sample would have used up the majority of each sample. Tangle-bearing AD CA1 neurons were then found to express 0.19 ± 0.09 4R tau mRNA molecules/GAPDH mRNA molecule. Non-tangle-bearing neurons expressed 0.04 ± 0.01 4R tau mRNA molecules/GAPDH mRNA molecule whereas control brain neurons expressed 0.05 ± 0.01 4R tau mRNA molecules/GAPDH mRNA molecule. The differences in 4R tau mRNA levels between the separate sample groups were not statistically significant. Also, when analyzing the second data set of EC neurons, there were no differences between the groups (control neurons 0.08 ± 0.01, non-tangle-bearing AD neurons 0.09 ± 0.02, tangle-bearing AD neurons 0.09 ± 0.02 4R tau mRNA molecules/GAPDH mRNA molecule).

Tangle-bearing AD CA1 neurons were found to express 0.31 ± 0.09 3R tau mRNA molecules/GAPDH mRNA molecule whereas non-tangle-bearing AD neurons expressed 0.13 ± 0.03 3R tau mRNA molecules/GAPDH mRNA molecule. The control brain neurons expressed 0.09 ± 0.01 3R tau mRNA molecules/GAPDH mRNA molecule. The relative expression of 3R tau mRNA in the AD tangle-bearing CA1 neurons was significantly higher than in the control neurons (P = 0.02) (Fig. 6b). However, when analyzing the second data set of EC neurons, no differences were seen (control neurons 0.10 ± 0.03, non-tangle-bearing AD neurons 0.06 ± 0.01 and tangle-bearing AD neurons 0.08 ± 0.03 4R tau mRNA molecules/GAPDH mRNA molecule).
https://static-content.springer.com/image/art%3A10.1007%2Fs00401-006-0095-3/MediaObjects/401_2006_95_Fig6_HTML.gif
Fig. 6

No difference in 4R tau/3R tau mRNA ratio between laser-captured samples of tangle-bearing AD neurons, non-tangle-bearing AD neurons and control brain neurons from a the CA1 hippocampal subregion or from b layer II of the entorhinal cortex

Finally, when calculating the 4R tau/3R tau mRNA ratio, there were no differences between the various groups. For the tangle-bearing CA1 AD neurons, the 4R tau/3R tau mRNA ratio was 0.54 ± 0.13, whereas the non-tangle-bearing neurons displayed a ratio of 0.48 ± 0.22. The 4R tau/3R tau mRNA ratio for the control brain neurons was 0.66 ± 0.12 (Fig. 6a). The corresponding ratio for the laser-captured tangle-bearing AD neurons in EC was found to be 1.44 ± 0.55. Non-tangle-bearing AD neurons displayed a ratio of 1.60 ± 0.45 and neurons from the control brains had a 4R tau/3R tau mRNA ratio of 1.16 ± 0.57 (Fig. 6b).

Overall, the 4R tau/3R tau mRNA ratio for the laser-captured CA1 samples was thus approximately 50% as compared to the respective ratios for temporal neocortex and entorhinal cortex. In order to clarify the observed difference in 4R tau/3R tau between the two types of sample preparations, we also analyzed whole tissue preparations from CA1 of four AD and four C brains. The 4R tau/3R tau ratio for the CA1 whole tissue samples was then found to be 0.69 ± 0.05 for the AD brains and 0.74 ± 0.15 for the control brains, i.e. similar to the numbers obtained from the laser-captured CA1 samples.

Discussion

In the present study, we assessed various aspects of tau metabolism on temporal neocortex from AD and control brains. An established ELISA was adopted to measure levels of soluble tau protein and a sensitive qPCR-based assay was developed to determine mRNA levels of 4R tau, 3R tau, GFAP and synaptophysin. Apart from investigating overall levels of these target molecules, the most important question was to answer whether a shift in the ratio of 4R tau/3R tau mRNA could be observed in AD temporal neocortex. In addition, laser-capturing subsets of neurons from the CA1 hippocampal subregion and from entorhinal cortex enabled us to separately analyze subsets of neurons with and without tangles, thereby addressing the question whether cells that develop tangle pathology differ in tau exon 10 splicing as compared to seemingly healthy neighboring neurons.

Overall levels of soluble tau protein, as determined by ELISA, were not altered in the AD brain. These data were further corroborated by Western blot analysis performed on sarcosyl-soluble extracts from a subset of AD and control brains. However, as could also be shown by Western blot, the levels of phosphorylated 4R tau and 3R tau protein were clearly augmented in the sarcosyl-insoluble fraction of AD brain homogenates. These observations are consistent with a previous ELISA-based study, which found increased levels of phosphorylated tau in AD temporal cortex [14].

We next examined the transcriptional patterns of tau containing or lacking exon 10, corresponding to 4R tau and 3R tau proteins. The results did not support any alteration of overall tau levels in AD as compared to control temporal neocortex. Furthermore, analyzing 4R tau mRNA and 3R tau mRNA separately did not reveal any significant differences between AD and control brains, although there was a trend for a decrease in the absolute amount of 3R tau mRNA in the AD brains. However, when normalizing to GAPDH mRNA, levels of both 4R tau mRNA and 3R tau mRNA were comparable between the groups.

A primary goal of the study was to determine whether a shift in the 4R tau/3R tau ratio, as for PSP and FTD, is a pathological feature also in the AD brain. Since many neurons in the FTD brain undergo degeneration without inclusions, and since many more neurons are lost in AD than can be accounted for by NFT numbers [11], we felt it was important to examine the 4R tau/3R tau mRNA ratios in both NFT-containing and non-NFT-containing (but vulnerable) neurons. However, no support could be gained for such a mechanism being of importance in AD, at least not in this well-characterized cohort of brains. Western blot analyses of sarcosyl-insoluble tau from AD and control tissues did not suggest a predominance of either 4R tau or 3R tau in paired helical filaments. Additionally, both for the temporal neocortical samples and for the preparations of laser-capture-microdissected hippocampal CA1/entorhinal cortex neurons, no significant differences in 4R tau/3R tau mRNA between Alzheimer and control brain tissues could be seen. These results are in accordance with a previous study, in which no disease-related differences in levels of 4R tau and 3R tau mRNA were found when examining AD and control neocortical areas [3]. By contrast, a transgenic mouse model expressing human tau on a mouse tau null background that develops substantial numbers of tangle-like inclusions and marked neuronal death was shown to overexpress 3R tau to a much greater extent than 4R tau [1].

The trend for decreased tau mRNA levels in AD temporal cortex probably does not reflect a disease-specific impairment of tau expression but is more likely explained by a relatively smaller contribution of neuronal mRNA in AD tissues, in which astrocytes and other cells of non-neural origin are known to have an increased transcriptional activity.

Morphologically, it is well known that brains from AD patients display a severe astrocytosis/gliosis as well as an impaired synaptic integrity. Similarly, increased levels of GFAP protein and decreased levels of synaptophysin protein were previously reported for this cohort of brains [20]. As expected, we could now demonstrate a corresponding pattern on the mRNA level, with increased GFAP and decreased synaptophysin mRNA levels in AD as compared to control brains, results that also served to confirm that our qPCR assays were appropriately designed. However, no correlation between 4R tau or 3R tau mRNA levels and GFAP or synaptophysin mRNA levels could be found, indicating that alternative tau splicing and neuroinflammatory or synaptic changes may not be intimately related phenomena.

When comparing the different types of RNA preparations, the LCM-based CA1 samples were found to have approximately 50% lower 4R tau/3R tau mRNA ratio than samples derived from temporal cortex. In order to investigate whether this difference was due to region- or cell-type-specific gene regulation, we also analyzed whole tissue preparations from the CA1 region. The 4R tau/3R tau mRNA ratios from whole tissue CA1 and laser-captured neuronal preparations from the same area were fairly similar to each other and different from whole tissue temporal cortex, which mainly indicates regional differences in tau exon 10 splicing. However, the findings may to some extent also reflect cell-type-specific differences, since CA1 predominantly contains pyramidal neurons whereas temporal cortex has various neuron types in its different layers. In addition, the two regions may be different with respect to types and/or numbers of glial cells. The notion that different cell types may express and accumulate different tau isoforms is intriguing. By developing quick immunostaining protocols, it should be possible also to laser-dissect the non-neuronal cells from brain tissues to determine the expressional profile of tau, e.g. in astrocytes.

The impact on tau splicing by the FTDP17 tau mutations clearly suggest that tau isoform imbalance may lead to neuronal instability and eventually cause neurodegeneration. However, the associated pathogenic mechanisms are unknown and it is still unclear whether aberrant tau splicing could be a general pathogenic factor in neurodegenerative disorders. In this study, we observed a subtle shift in tau splicing in the CA1 hippocampal subregion, an early and extensively affected area in the AD brain. As the FTDP17 tau mutations have illustrated, a disturbed tau isoform balance with a shift towards an increase of either 4R tau or 3R tau may render neurons more susceptible to degeneration. Accordingly, we speculate that the regional alteration in tau splicing observed here may contribute to the selective vulnerability for NFT formation of the pyramidal neurons in the CA1 region; further analysis of splice form ratios in additional brain areas would be necessary to test this hypothesis.

In conclusion, analysis of a large number of AD cases using quantitative protein and mRNA measurements, including laser-capture microdissection methods to separately analyze neurofibrillary tangle-containing populations, did not reveal any clear changes between 4R tau and 3R tau isoforms in AD. The study was powered to detect a potential twofold difference between diseased and control brains, analogous to the size of change observed in FTPD17, leading us to conclude that no overt changes in 4R tau/3R tau occur in AD. However, techniques with yet higher sensitivity may reveal more subtle ratio differences, which possibly could be of pathogenic significance also in the AD brain. In addition, splicing of other tau exons may also impact tau biology and it should be of potential interest to investigate coordinated changes in the expression of all six tau isoforms, both in Alzheimer’s disease and other neurodegenerative disorders.

Acknowledgments

This study was supported by NIH (P50 AG005134 and AG08487) and the Rubenstein Foundation. M.I. was sponsored by The Swedish Society of Medicine, The Swedish Alzheimer Foundation and the Swedish Research Council. RdeS was funded by the Reta Lila Weston Trust for Medical Research. We are also grateful to Karlotta Fitch for technical support.

Copyright information

© Springer-Verlag 2006