Selective loss of NMDA receptor NR1 subunit ... - Wiley Online Library

Journal of Neurochemistry, 2004, 89, 240–247

doi:10.1111/j.1471-4159.2003.02330.x

Selective loss of NMDA receptor NR1 subunit isoforms in Alzheimer’s disease Matthew R. Hynd, Heather L. Scott and Peter R. Dodd Department of Biochemistry, University of Queensland, Australia

Abstract Previous work had shown that the ratio of NMDA receptor NR1 subunit mRNA transcripts containing an N-terminal splice cassette to those that do not is markedly lower in regions of the Alzheimer’s disease (AD) brain that are susceptible to pathological damage, compared with spared regions in the same cases or homotropic regions in controls. To elucidate the origins of this difference in proportionate expression, we measured the absolute levels of each of the eight NR1 transcripts by quantitative internally standardized RT-PCR assay. Expression of transcripts with the cassette was strongly attenuated in susceptible regions of Alzheimer’s brain, whereas expression of non-cassette transcripts differed little

from that in controls. The expression of other NR1 splice variants was not associated with pathology relevant to disease status, although some combinations of splice cassettes were well maintained in AD cases. The population profile of NR1 transcripts in occipital cortex differed from the profiles in other brain regions studied. Western analysis confirmed that the expression of protein isoforms containing the N-terminal peptide was very low in susceptible areas of the Alzheimer’s brain. Cells that express NR1 subunits with the N-terminal cassette may be selectively vulnerable to toxicity in AD. Keywords: Alzheimer’s disease, excitotoxicity, glutamate, N-methyl-D-aspartate, polyamine. J. Neurochem. (2004) 89, 240–247.

NMDA receptors (NMDARs) are present on most CNS neurones (Petralia et al. 1994) and mediate Ca2+ influx in response to glutamate (Constantine-Paton 1990; Bliss and Collingridge 1993; Seeburg 1993; Hollmann and Heinemann 1994; Constantine-Paton and Cline 1998). Their excessive activation has been implicated in hypoxia-ischemia, epilepsy, and chronic neurodegenerative disorders such as Huntington’s and Alzheimer’s diseases (Choi 1988; Young 1993; Olney 1994; Chen et al. 1999). The impairments in memory and cognition in Alzheimer’s disease (AD) can be correlated to the neuropathological features of the disease, including neuronal loss and plaque and tangle formation in the cortex and hippocampus (Hyman et al. 1986; Albin and Greenamyre 1992; Bobinski et al. 1998). Neuronal loss is generally restricted to the pyramidal cells in layers III and IV. Damage to glutamate-innervated neurones is also observed (Albin and Greenamyre 1992). Axons, terminal boutons, glia, and endothelial and ependymal cells are relatively spared (Choi 1992). NMDARs are heteromeric assemblies comprising a ubiquitous NR1 subunit and two or more NR2 subunits (NR2 A–D) (Monyer et al. 1992; Ishii et al. 1993; Dingledine et al. 1999). An NR3 subunit has also been cloned (Sucher et al. 1995; Das et al. 1998). The function and modulation of

NMDAR channels is spatially and temporally regulated (Bockers et al. 1994; Rigby et al. 1996). While the NR1 subunit is expressed ubiquitously, NR2 subunits show regional and developmental variations (Watanabe et al. 1992, 1993; Akazawa et al. 1994; Monyer et al. 1994). The NR1 subunit undergoes alternate splicing of three exons (4, 20 and 21) to give eight functionally distinct variants (Sugihara et al. 1992; Durand et al. 1993; Hollmann et al. 1993). Exon 4 encodes a 21-amino acid amino-terminal cassette (designated ‘N1’), while exons 20 and 21 encode two consecutive carboxy-terminal 37- (‘C1’) and 38-amino acid (‘C2’) cassettes (Zimmer et al. 1995). NMDARs are regulated by a variety of agents including glycine, Mg2+, Zn2+ and H+ ions, and polyamines. Spermine sensitivity is

240

Received October 20, 2003; revised manuscript received December 4, 2003; accepted December 4, 2003. Address correspondence and reprint requests to Dr P. R. Dodd, Department of Biochemistry, University of Queensland, Brisbane 4072, Australia. E-mail: [email protected] Abbreviations used: AD, Alzheimer’s disease; LBD, Lewy body disease; NR1, NR2, NR3, subunits of the NMDA receptor; NR11XX, NR10XX, NR1 subunit with or without the N-terminal 63 bp/21-amino acid splice insert; PMI, post-mortem interval; TIS, time in storage.

2004 International Society for Neurochemistry, J. Neurochem. (2004) 89, 240–247

NMDA splice variants in Alzheimer’s disease 241

controlled by the presence of the amino-terminal insert (Durand et al. 1993). Subunits lacking the N1 cassette are sensitive to physiological pH (Traynelis et al. 1995). Deletion of the cassette potentiates the response to polyamines at saturating glycine concentrations (Durand et al. 1993) and by low-micromolar Zn2+ and Ni2+ concentrations (Hollmann et al. 1993). Both N-terminal variants are inhibited by high-micromolar Zn2+ concentrations and potentiated by polyamines at low glycine concentrations (Zhang et al. 1994). Polyamine potentiation is dose-dependent and involves an increase in both peak channel open probability and peak channel current duration (Araneda et al. 1993; Chen et al. 1999). In AD cases, mean levels of spermidine, the biosynthetic precursor to spermine, are markedly higher in pathologically susceptible brain regions, and lower in pathologically spared regions, compared with controls (Morrison and Kish 1995). In a previous study we measured the ratio of NR1 mRNA transcripts that contain the N1 cassette, designated ‘NR11XX’, to those that do not, designated ‘NR10XX’ (Durand et al. 1993), by using PCR primers that spanned the spliced region (Hynd et al. 2001). We showed that the NR11XX : NR10XX ratio was markedly reduced in those regions of the AD brain most susceptible to pathological damage. In contrast, the ratio was about 1 : 1 throughout the brains of normal controls and in relatively spared areas of AD brain (Hynd et al. 2001). However, that approach could not distinguish whether a scarcity of NR11XX transcripts or a superabundance of NR10XX transcripts had led to the altered ratio in vulnerable areas, nor could it ascertain the expression of the NR1 subunit C-terminal cassettes. We therefore quantified NR1 mRNA using a competitive reverse-transcription polymerase chain reaction (cRT-PCR) assay that allowed the quantification of all eight NR1 splice variants (Hynd et al. 2003). NMDAR NR1 protein expression was also analysed by immunoblotting using an antibody that recognized NR1 subunit isoform proteins which contain the N-terminal peptide insert. We have adopted the terminology ‘susceptible’ rather than ‘affected’ because quantitative morphological data on neuronal loss was unavailable on a case-by-case basis. The data suggest that cells expressing NR1 subunit transcripts and proteins that contain the N1 cassette are selectively lacking in pathologically vulnerable areas of AD brain.

Materials and methods Human brain tissue and case groups Autopsies were performed by authorized pathologists, with full macroscopic and microscopic examinations, for confirmation of disease diagnosis or other clinical imperative. Ethical clearance is current for each hospital from which material was acquired; informed written consent was obtained from the next of kin in all

cases. At autopsy, 1–5 cm3 pieces from either cerebral hemisphere were dissected, slowly frozen in 0.32 M sucrose, and stored at )70C until use (Dodd et al. 1986). Brain regions were taken from cingulate gyrus, temporal cortex and hippocampus, which show marked AD pathology, and from motor and occipital cortices, which are relatively spared (Brun and Englund 1981). At least two pathologically spared and two susceptible areas were studied for each case when possible; limitations on tissue availability meant that it was not possible to include every area from every subject. To confirm a diagnosis of AD, a combined CERAD/Khachaturian schema was employed, with the presence of neocortical neurofibrillary tangles considered obligatory (Khachaturian 1985). Subject groups were defined as definite AD (n ¼ 10), Lewy body disease (LBD) (n ¼ 4), combined LBD/AD (n ¼ 2), and normal controls (n ¼ 10). Full prescription medication information was available from medical records for all subjects. One each of the control (diazepam), AD (carbamazepine) and LBD (phenytoin) subjects had received anti-epileptic medication. Two of the AD cases had each been prescribed one antipsychotic drug (thioridazine and olanzapine, respectively) while one of the AD/LBD subjects was on combination thioridazine, chlorpromazine and haloperidol anti-psychotic therapy. No other neuroactive medication was noted. RNA extraction Tissue samples were homogenized in guanidinium isothiocyanate to isolate total RNA according to the method of Chomczynski and Sacchi (1987). Concentration and purity of RNA was determined by UV absorbance at 260 and 280 nm. RNA quality was assessed by gel electrophoresis. For all samples, the pattern obtained was indicative of undegraded RNA. NMDA receptor nomenclature We have adopted the nomenclature proposed by Durand et al. (1993). NR1 splice variants are denoted by the absence (subscript 0) or presence (subscript 1) of the three alternatively spliced exons in the 5¢ to 3¢ order. The subscript X is used to indicate that an exon may be either present or absent. Competitive RT-PCR Absolute mRNA expression values were quantified by internally standardized competitive RT-PCR in susceptible (hippocampus, cingulate cortex and superior temporal cortex) and spared (occipital and primary motor cortex) areas as described previously (Hynd et al. 2003). Quantification of each hNR1 splice variant was performed using methodology in which a known amount of synthetic ribonucleic acid competitor (internal standard) is co-amplified against total RNA. Four sets of primer pairs were needed to assay all eight variants. The PCR products were subjected to restriction enzyme digest and polyacrylamide gel electrophoresis (PAGE) to separate the various subunit and standard amplimers, and the band intensities were determined by PhosphorImager analysis (Molecular Dynamics, Sunnyvale, CA, USA). An image is shown in Fig. 1; it may be seen that when the amounts and numbers of splice variants in the sample are higher, the intensity of the internal standard band is correspondingly reduced. The amounts of the various transcripts in each sample are calculated from a standard curve based on the ratios of band intensities (unknown : standard).


242 M. R. Hynd et al.

Results

Fig. 1 Representative PhosphorImage of a competitive RT-PCR assay of NR1 splice variant transcripts. Lanes 1–4, control subject; lanes 5–8, subject with AD-only pathology. Each lane represents a separate RT-PCR reaction performed using sequence-specific primers with 2 lg of total RNA from cingulate cortex co-amplified with 1 · 107 molecules of internal standard as previously described (Hynd et al. 2003). Band designations: 1 A, 5 A, NR1111; 1B, 5B, NR1101; 1C, 5C, corresponding NR1 internal standard; 2 A, 6 A, NR1011; 2B, 6B, NR1001; 2C, 6C, corresponding NR1 internal standard; 3 A, 7 A, NR1110; 3B, 7B, NR1100; 3C, 7C, corresponding NR1 internal standard; 4 A, 8 A, NR1010; 4B, 8B, NR1000; 4C, 8C, corresponding NR1 internal standard. A marked lack of the bands containing the N1 cassette (5 A, 5B, 7 A, 7B) is apparent in the AD case.

Immunoblotting Tissue homogenates were used for semi-quantitative immunoblotting. Slow-frozen tissue samples were homogenized in lysis buffer (0.32 M sucrose) including complete protease inhibitor cocktail (Boehringer-Mannheim, Mannheim, Germany). Duplicate homogenate samples (10 lg) were mixed with an equivalent amount of sample buffer [2% sodium dodecyl sulfate (SDS)/5% 2-mercaptoethanol/10% glycerol/0.1% Coomassie blue G in 63.5 mM Tris-HCl, pH 6.8] and electrophoresed on 5% SDSpolyacrylamide gels. Proteins were transferred to Hybond nitrocellulose membranes (Amersham Biosciences, Arlington Heights, IL, USA) and blocked with 5% non-fat dry milk in phosphate-buffered saline (PBS) containing 0.05% Tween-20 for 1 h at ambient temperature. Immunoblots were washed in PBS and probed with antibodies to the NMDA receptor NR1 subunit N1 cassette (0.5 mg/mL, rat polyclonal, 1 : 4000, rabbit immunoaffinity, Chemicon International, Temecula, CA, USA). Following incubation with appropriate horseradish-conjugated secondary antibody, immunoreactive bands were visualized on X-ray film using the ECL-Plus western Blotting detection system (Amersham Biosciences). Linearity of the immunoblot assay was confirmed using a standard curve (2, 5 and 10 lg) prepared from total brain homogenates. The concentrations of primary and secondary antibodies, along with the exposure time to the X-ray film, were optimized to provide optical density values that fell within the linear range. Densitometric analysis of immunoblots was performed using ImageQuant software v5.0 (Molecular Dynamics). Protein concentrations were expressed as relative absorbance values. Data analysis All statistics were calculated using the Systat program (Systat Inc., Evanston, IL, USA). Post-hoc testing on ANOVA effects was carried out by the Tukey hsd method with the Bonferroni correction for multiple tests (Winer et al. 1991).

cRT-PCR for NR1 splice variants The full range of NR1 splice variants was analysed by quantitative internally standardized RT-PCR as previously described (Hynd et al. 2003). The method involves four sets of primer pairs, each with a unique internal standard. Each internal standard was designed to share 100% homology to its native target except for an engineered 104 bp deletion. Quantification was by PhosphorImager assay of the separated PCR products. Representative examples for a control and a subject with AD-only pathology are shown in Fig. 1. Regression analyses Four case groups were used in this study, with subjects grouped according to the presence or absence of AD pathology. The two categories of subjects distinguished were those possessing AD pathology (AD and AD/LBD cases) and those without AD pathology (controls and cases with LBD changes only). Each case-group was matched as closely as possible for age at death, post-mortem interval (PMI) and the time that the tissue had been in storage (TIS). For subjects without AD pathology, the mean age was 75.9 ± 10.3 years; for subjects with AD pathology, it was 76.7 ± 9.0 years (NS). Mean PMI values were 32.1 ± 14.6 h and 26.1 ± 21.7 h (NS), while mean TIS values were 5.3 ± 2.7 years and 3.4 ± 2.8 years (NS) for the same two groups. Before further analyses were carried out, the possible effects of these factors on expression were checked. In no instance was there a significant regression of the expression of any subunit transcript on age. Moreover, analysis of the homogeneity of slopes (Wilkinson et al. 1992) showed that no such regression differed significantly between groups. Tests of the effects of PMI and TIS gave similar results. Since the subjects were well matched on all three variables anyway, these factors would not have materially affected the statistical analyses reported below. Quantitative analyses The expression levels of the splice variants were grouped by cassette to allow patterns of combinations to be discerned. Although this produced a complex ANOVA, the key features were readily distinguished. In every instance where a term in the ANOVA that involved Diagnosis was significant, it was found that the case groups divided in such a way that the levels in controls and LBD-only subjects were comparable, and the levels in subjects with AD-only and AD/LBD pathology were comparable. That is, the presence of LBD pathology had no significant effect on NR1 splice variant expression. Moreover, in every instance where a term that involved Area was significant, expression levels partitioned such that hippocampus and the cingulate and superior


NMDA splice variants in Alzheimer’s disease 243

temporal cortices (all pathologically affected in AD) gave comparable values, and the occipital and motor cortices (relatively spared) gave comparable values. An example of this will be shown below; such groupings in the data simplified interpretation.

cation prior to death, and none had received acetylcholinesterase-inhibitor therapy. Removal from the analysis of subjects on medication made no significant difference to the mean level of expression of any gene product in any casegroup.

Between-subject effects The Main Effect for Case-group, the Main Effect for Area and the overall Group · Area Interaction were all strongly significant (p < 0.001), whether or not the cases and areas were combined as described above. The Main Effect for Group and the simplified Interaction are shown in Fig. 2. These underlying trends need to be borne in mind when more complex effects are considered. As noted in Methods, very few of the subjects had been prescribed neuroactive medi-

Within-subject effects All terms that distinguished the two N-terminal splice variants (hNR10XX and hNR11XX) were significant in the ANOVA (p < 0.001). Hence, the deepest-level interaction (N1 Cassette Expression · Area · Case-group) is relevant, and is shown in Fig. 3, simplified for diagnosis (AD vs. non-AD). In subjects without AD pathology, the two variants were expressed at approximately equal levels, although there were some minor regional variations (Fig. 3a). Markedly distinctive expression was observed in the susceptible regions in cases with AD pathology (Fig. 3b). NR11XX splice variant expression was significantly attenuated in all three areas, both compared with controls and compared with the two spared areas in the same cases (p < 0.01). We have previously reported that the hNR11XX : hNR10XX mRNA expression ratio is markedly reduced in the three susceptible areas (Hynd et al. 2001). The present data demonstrate that virtually all of this reduction can be ascribed to the selective loss of hNR11XX expression. A feature of the N1 · Area Interaction is noteworthy, though it is not related to disease pathogenesis. Inspection of the data showed that the statistical significance of the interaction derived from the fact that in four of the brain areas, hNR10XX expression was greater than or equal to hNR11XX expression, but that this relationship was reversed in occipital cortex (hNR11XX expression > hNR10XX expression). Expression of the C-terminal cassettes, and the interactions between cassette expression, were also analysed. The Main Effect for C1-cassette expression was significant ( p < 0.001), in essence because transcripts lacking the C1 cassette (hNR1X0X) were about 11% more abundant overall. There was a weak C1-cassette · Area Interaction (p < 0.05) because this tendency was selectively reversed in occipital cortex (hNR1X1X about 4% more abundant than hNR1X0X). No term involving Case-group and C1-cassette expression was significant, i.e. there was no differential effect of disease on the expression of transcripts with or without the C1 cassette. The Main Effect for C2-cassette expression was significant ( p < 0.001) because transcripts bearing the C2 (hNR1XX1) rather than the C2¢ (hNR1XX0) cassette were about 11% more abundant overall. This trend was selectively reversed in occipital cortex (hNR1XX0 expression 18% higher than hNR1XX1) to give a significant C2-cassette · Area Interaction (p < 0.001). The C2-cassette · Diagnosis Interaction was also significant (p < 0.01), in essence because the superabundance of C2-containing transcripts over C2¢-containing transcripts was much stronger in controls + LBDonly subjects (15.5%) than in AD + AD/LBD subjects

(a)

(b)

Fig. 2 Between-subject effects. (a) Main Effect for Group (p < 0.001). NR1 mRNA expression was averaged across all five areas and all eight splice variants. The two AD groups (AD + AD/LBD; shaded bars) differed significantly from the two non-AD groups (controls + LBD; open bars), but neither set differed within itself. (b) Interaction between Local Vulnerability and Diagnosis (p < 0.001). NR1 expression was averaged across all eight splice variants, across spared (open bars) and susceptible (shaded bars) areas, and across cases with and without AD pathology. Note the marked difference in susceptible areas in subjects with AD pathology. Values are means ± SEM; *significantly different from all others, p < 0.001.


244 M. R. Hynd et al.

Fig. 3 N1 cassette Expression · Area · Diagnosis Interaction. Expression of the NR1 variants was averaged across each of the two N-terminal cassette sets; the three-factor interaction was significant (F4,111 ¼ 25.147; p < 0.0001). (a) Subjects without AD pathology (controls + LBD-only). Post-hoc tests showed that NR11XX expression was significantly different from NR10XX expression in superior temporal cortex only (p < 0.01), while NR10XX expression in superior temporal cortex differed significantly from NR10XX expression in both cingulate and occipital cortex (p < 0.01). (b) Subjects with AD pathology (AD + AD/LBD). Post-hoc tests showed that NR11XX expression differed significantly from NR10XX expression in three pathologically susceptible areas: hippocampus, cingulate cortex and superior temporal cortex. NR11XX expression in these three areas also differed significantly from NR11XX expression in occipital and motor cortex in the same cases (p < 0.01), and from NR11XX expression in the same three areas in subjects without AD pathology (p < 0.01). NR10XX expression in hippocampus differed significantly from NR10XX expression in cingulate and occipital cortex, and NR10XX expression in superior temporal cortex differed significantly from NR10XX expression in occipital cortex. Cing, cingulate cortex; Hippo, hippocampus; S. Temp, superior temporal cortex; Motor, primary motor cortex; Occip, occipital cortex. (c) Summary of expression in pathologically susceptible (susc) and spared regions. Key: *, expression of the splice variants differed (p < 0.01); à, expression differed from that of the same variant in the other region in the same case-group (p < 0.01); , expression differed from that of the same variant in the same region in the other case-group (p < 0.01). Open columns, NR11XX expression; shaded columns, NR10XX expression; means ± SEM.

(a)

(b)

(c)

(4.5%). However, the C2-cassette · Area · Case-group Interaction was not significant, i.e. this disease differential was not associated with the regional distribution of pathology. The N1 · C1-cassette Interaction was significant (p < 0.001) because transcripts containing either or both cassettes were progressively less abundant overall (order: transcripts hNR111X < hNR110X < hNR101X < hNR100X; containing neither cassette were about 24% more abundant than those with both). No deeper-level interaction involving Area or Case-group was significant, i.e. the pattern of N1 + C1 cassette interaction did not vary significantly between areas or diagnostic groupings. Similarly, the N1 · C2-cassette Interaction was significant (p 0.01). The order of abundance was hNR11X0 hNR11X1 hNR10X0 < hNR10X1; in this classification, hNR10X1 was about 25% more abundant than the other three sets. However, no deeper-level interaction was significant, i.e. the pattern of the N1 + C2-cassette interaction did not vary significantly between areas or between case-groups.

The C1 · C2-cassette Interaction was significant (p < 0.001). The order of abundance was hNR1X10