Genetic association of the low-density lipoprotein receptor-related protein gene (LRP), and apolipoprotein E receptor, with late-onset Alzheimer's disease

Alzheimer's disease (AD) is an age-related neurodegenerative disorder of the CNS characterized by progressive deterioration of mental function. The onset of AD can be influenced by various environmental and genetic factors. Mutations in the amyloid precursor protein (APP), presenilin-1 (PS-1), and presenilin-2 (PS-2) have revealed tight genetic linkage with early-onset familial cases of AD.^1-3 However, the major genetic risk factor for late-onset AD is the ϵ4 allele of the gene coding for apolipoprotein E (apoE).^4,5 Given that the ϵ4 allele is neither necessary nor sufficient for the pathogenesis of AD, other genetic factors may participate independently or in concert with APOE to determine an overall risk for the expression of AD.

Although the role of APOE as a major susceptibility gene for the development of AD is well established, the molecular mechanisms that underlie this process are unclear. Because apoE binds to amyloid β (Aβ), the major constituent of AD amyloid, apoE might serve as a pathologic molecular chaperone, either by promoting Aβ fibrillogenesis^4,6 or by a process involving the sequestration and clearance of soluble Aβ from the extracellular matrix.⁷ Alternatively, apoE may mediate the remodeling of membrane cytoarchitecture and neuronal connections by its ability to bind and transport cholesterol-rich lipids into cell via the interaction with its endocytosis receptors.^8-10 The major apoE receptor in the brain is the low-density lipoprotein receptor-related protein (LRP), selectively found in neurons and reactive astrocytes.^11,12 Moreover, LRP is a multifunctional cell-surface endocytosis receptor,¹³ implicated in the modulation of growth factor responses and catabolism of proteases and protease/protease inhibitor complexes (i.e., activated α-macroglobulin, plasminogen activators). Recently, Kounnas et al.¹⁴ showed that LRP also mediates the endocytosis and degradation of the APP, another central molecule implicated in the pathogenesis of AD. Because the physiologic function of apoE and APP involves the binding to their receptors, we reasoned that the presence of a biologically significant variation in the gene coding for LRP, if any, might influence a pathogenic process underlying the expression of AD. Indeed, LRP and many of its ligands are present in senile plaques,^12,15 suggesting that ligand binding and/or internalization of LRP may be altered in AD. In this case-control study, we present genetic evidence that LRP may be associated with late-onset AD and modifies the pathogenic mechanism.

Methods. Subjects and neuropathologic evaluation. A total of 158 late-onset AD cases (>65 years at onset of disease) and 102 cognitively intact control subjects (age >65 years at clinical/postmortem examination) were evaluate in this case-control study. All subjects were unrelated white Americans of European descent, obtained from the Alzheimer Disease Research Center (ADRC) at the University of California, San Diego. Of the 158 AD cases, 85 subjects (45 women and 40 men) were family history positive (FHP-AD: ≥1 first-degree relatives with senile dementia) and 72 subjects (38 women and 34 men) were family history negative (FHN-AD: no first-degree relatives with senile dementia). In this context, FHN-AD likely represents a purely sporadic form of AD. Because the FHP-AD cases are more likely to reveal genetic components of the disease than those with negative family history, this variable was considered in genotype comparisons. All available control and late-onset AD subjects obtained from the ADRC were included in this study, with the exception of AD cases lacking family history information.

The onset ages of AD were determined by the recognition of obvious memory impairments by spouses and relatives. The mean age of onset among FHP-AD was 72.1 ± 4.6 years (range 66 to 86) and among FHN-AD was 74.4 ± 5.1 years (range 66 to 87). All AD subjects were diagnosed as probable AD by NINCDS/ADRDA criteria among which 74 subjects were confirmed at autopsy by the presence of adequate numbers of senile plaques. Of the 102 healthy individuals (54 women and 48 men) who were clinically free of any signs of neurologic problems, 25 subjects had postmortem confirmation for the lack of neuropathologic lesions. The average age of postmortem subjects was 80.0± 8.1 (range 67 to 98) and among living control subjects was 76.2± 5.9 (range 66 to 95). The clinical and neuropathologic examiners did not have knowledge of APOE or LRP genotypes of the subjects. Neuritic plaques and neurofibrillary tangles were identified in thioflavin S-stained sections of the midfrontal cortex as previously described.^16,17 Neuritic plaques were counted under a 10× objective and a 10× ocular lens (field size 1.6 mm²). Tangles were counted under a 40× objective and a 10× ocular lens (field size 0.16 mm²). For neuritic plaque number comparisons across LRP genotypes, 17 cases were FHP, 37 cases were FHN, and 36 cases were of unknown family history status, which reflected the availability of postmortem samples with neuritic plaque counts. For neurofibrillary tangle comparisons across LRP genotypes, an additional eight cases with unknown family history content were available.

LRP exon 3 polymorphism and APOE genotyping. The exon 3 polymorphism was identified by the method of polymerase chain reaction single-strand conformation polymorphism (PCR-SSCP).¹⁸ A set of primers flanking exon 3 of the LRP gene was designed and used to amplify this region from total genomic DNA, which yielded a PCR product of 157 bp in size (LEX-3F, 5′-TGCCCTCAGGTCCACAGA-3′, and LEX-3R, 5′-AAGTCCTTACCTCGGCAG). Fifty nanograms of genomic DNA was mixed with a final concentration of 1× BRL PCR buffer, 1.5 mM MgCl₂, 250 nM each primer, 200 µM dNTPs, 1 µCi α-[³²P] dCTP, and 1 unit of Taq DNA polymerase (BRL) in a total volume of 50µl. The PCR cycling method was as follows: 94 °C 3 minutes, 60°C 30 seconds, 72 °C 30 seconds, followed by 29 cycles of 94 °C 30 seconds, 60 °C 30 seconds, 72 °C 30 seconds, and a final extension step at 72 °C for 5 minutes. An equal volume of amplified product and stop solution (USB) was mixed, denatured at 85 °C for 5 minutes, loaded(5µl) onto an 8% native acrylamide gel (19:1 sequencing format) containing 0.5× TBE buffer, and run for 5 hours at 30 W (0.5× TBE running buffer) in the cold room. The gel was dried under vacuum at 80 °C and subjected to autoradiography. When the polymorphic band was identified, the corresponding position on the dried gel was cut out, eluted in 30µL of 1× BRL PCR buffer at 50 °C for 1 hour, and 10 µL of the eluted product was used to reamplify with PCR using identical parameters as before except with the exclusion of radioisotope. The amplified product was gel purified and used as the template for the ΔTaq cycle sequencing protocol (USB). This revealed that the polymorphic position in the SSCP gel represents base 766 of the LRP cDNA, cytosine to thymine substitution, which does not alter the predicted amino acid sequence (gene bank NID, g34338). Because the T polymorphism did not conveniently create or destroy any restriction enzyme sites, the PCR-SSCP method as described above was subsequently used to genotype all subjects for the LRP exon 3 polymorphism. In all SSCP gels, a positive control for the T polymorphism was included. In a subset of over 60 DNA samples, the genotyping was repeated. In every repeated case, the genotypes were identical to the first run, demonstrating the reliability of this SSCP protocol. APOE genotyping was performed by PCR, followed by Hha I restriction digestion, electrophoreses in a 6% native acrylamide gel, and visualization by ethidium bromide staining.¹⁹

Statistical evaluation. For initial comparisons, Pearson'sχ² test for independence was used to determine potential differences in the distribution of LRP genotypes and alleles between populations. Two-sided Fisher's exact test and odds ratio (OR) was subsequently used to test for association between groups and to estimate the magnitude of the association, respectively. The Bonferroni correction was applied for three-group comparisons with significance value set at p = 0.0167. A χ² test for linear trend was used to assess LRP allele distributions across ordered onset age categories among AD subjects. Two sample t tests were carried out to compare the ages at onset of AD and numbers of neuritic plaques and neurofibrillary tangles across LRP genotypes.

Results. Distortion of LRP genotypes between late-onset AD and healthy control subjects. The distribution of LRP exon 3 genotypes among late-onset AD (>65 years at onset) and control subjects (>65 years at clinical or postmortem evaluation) are shown intable 1. The LRP T polymorphism represents a minor allele at nucleotide position 766 of the LRP cDNA, cytosine to thymine substitution, which does not alter the predicted amino acid sequence. In aχ² comparison of LRP C/C versus T position genotypes, significant differences revealed a distortion of LRP genotypes between FHP-AD, FHN-AD, and healthy control subjects (seetable 1) (χ² = 11.91, df = 2, p = 0.0026). This distortion was consistent when assessed by LRP allele frequencies (χ² = 10.85, df = 2, p= 0.0044). By two-sided Fisher's exact test, FHP-AD cases displayed a robust difference from control subjects in the distribution of LRP C/C versus other genotype (p = 0.0008, OR = 3.463, 95% CI, 1.7 to 7.2). By allele counting, similar differences were observed (two-sided Fisher's exact test, p = 0.0010). Although the FHN-AD group showed a same directional shift in the distribution of LRP genotypes from healthy control subjects, statistical significance was not achieved (two-sided Fisher's exact test, p = 0.1372, OR = 1.71, 95% CI, 0.87 to 3.33). Moreover, FHP-AD and FHN-AD did not significantly differ in the distribution of LRP genotypes (two-sided Fisher's exact test, p = 0.1040, OR = 2.028, 95% CI, 0.9 to 4.6). Thus, when all AD cases were pooled, the distribution of LRP C/C versus other genotypes (p = 0.0024, OR = 2.410, 95% CI, 1.4 to 4.2) and alleles (p = 0.0068, OR = 2.009, 95% CI, 1.2 to 3.3) significantly differed from control subjects by two-sided Fisher's exact test. These data indicate that LRP might be genetically associated with either the promotion or suppression of late-onset AD.

Age and APOE effects on the distribution of LRP genotypes and alleles in late-onset AD. It was demonstrated that increasing age and presence of the APOE ϵ4 allele are two major risk factors for the expression of late-onset AD.^4,20 Because the C/C genotype was overrepresented in AD, the C allele might be associated with an AD promoting function. Alternatively, the overrepresentation of the T allele in the control subjects may indicate a protective role associated with the T allele. If a genetic variation in LRP is involved in either the promotion or suppression of AD, we reasoned that the former and latter should correlate with an earlier or delayed age of disease onset, respectively. As shown intable 1, the mean age of onset was significantly different (t = 2.306, p = 0.0236) between individuals with LRP C/C genotype (71.63 ± 4.22) and T positive genotypes (74.83 ± 5.80) among FHP-AD. When all AD cases were pooled, the same effect was observed (C/C genotype, 72.68 ± 4.73; T positive genotypes, 75.03 ± 5.42;t = 2.384, p = 0.0183). However, there was no difference in the current age (at clinical or postmortem evaluation) of normal subjects between C/C and T positive genotypes(t = 0.329, p = 0.7422).

Because the ages at onset of AD significantly differed as a function of different LRP genotypes, this effect was further assessed by stratifying the onset ages into three ordered age categories: ≤70, 71 to 75, and >75 years of age. In a χ² test for linear trend, there was a significant linear effect for increase in the T allele across increasing age categories among FHP-AD (table 2)(χ² = 8.719, df = 1, p = 0.0031). This linear trend was consistent when all AD cases were pooled (χ² = 7.138, df = 1, p = 0.0075), suggesting that the observed difference in the mean age of onset between C/C and T positive genotypes is likely not due to sampling error. Moreover, this phenomenon was specific to AD given that control subjects did not display age-dependent linear changes in the distribution of LRP alleles(χ² = 0.937, p = 0.3329, df = 1; data not shown). Thus, the observed unevenness in the distribution of LRP genotypes between AD and control subjects primarily resulted from AD cases with younger ages of disease onset. Hence, LRP T positive genotypes were severely underrepresented among FHP-AD ≤75 years compared with control subjects by two-sided Fisher's exact test (table 3) (p < 0.0001, OR = 5.878, 95% CI, 2.3 to 14.9). This effect was consistent when all AD cases ≤75 years were pooled (p = 0.0008, OR = 4.229, 95% CI, 1.7 to 10.3). As predicted, AD cases > 75 years showed no trend for differences in LRP genotypes (p = 0.5842) or allele frequencies (p = 0.7537) from the control subjects by two-sided Fisher's exact test (data not shown). These data indicate that LRP might be associated with the expression of AD, either by promoting or delaying the disease onset. The negative AD risk associated with the T allele is reflected in the underrepresentation of T positive genotypes in AD and relatively delayed age of disease onset. Inversely, the overrepresentation of the C/C genotype in AD and relatively earlier age of disease expression potentially reflects an AD risk associated with the C allele.

The homozygous presence of APOE ϵ4 is extremely robust, displaying near complete AD penetrance before the age of 75.^17,21 This observation is consistent with our data set. Hence, in the absence of APOE ϵ4/ϵ4 genotype, the LRP T positive genotypes were further underrepresented in FHP-AD (≤75 years) compared with control subjects (see table 3) (p< 0.0001, OR = 9.677, 95% CI, 2.8 to 33.2). The same effect was observed when all AD cases (≤75 years without APOE ϵ4/ϵ4) were compared with control subjects (genotype C/C versus T positive: p = 0.0001, OR = 4.088, 95% CI, 1.9 to 8.6). Thus, in the absence of APOE ϵ4/ϵ4, the putative LRP-associated risk becomes more robust.

LRP modifies the severity of AD pathology. The pathologic hallmark of AD is the accumulation of neuritic plaques and neurofibrillary tangles.²² The severity of these pathologic measures is remarkably heterogeneous among postmortem AD cases, suggesting that various environmental and genetic factors can drive or modify the pathogenic mechanism. We and others have shown that AD cases harboring the APOE ϵ4 allele display more β-amyloid containing senile plaques²³ and more neuritic plaques¹⁷ compared with those lacking the ϵ4 allele, with ϵ4 homozygotes showing the strongest effect. Thus, the potential involvement of LRP in the development of plaque and tangle pathology was examined. In the following analysis, all available late-onset pathology-confirmed cases, regardless of family history content, were used for comparisons. In the midfrontal cortex, the cases harboring LRP C/C genotype (n = 66) displayed significantly more neuritic plaques than those with T positive genotypes (n = 24)(table 4) (t = 2.762, p = 0.0069, df = 88). In the absence of APOE ϵ3/ϵ4 and ϵ4/ϵ4 genotypes, the LRP effect on neuritic plaque numbers was stronger (seetable 4) (t = 3.641, p = 0.0008, df = 36), with doubling of average neuritic plaque counts among cases with the C/C genotype compared with those with T positive genotypes. In addition, the C/C carriers (n = 72) showed similar increases in neurofibrillary tangles compared with T positive genotypes (n = 26) in the midfrontal cortex; however, this did not reach statistical significance (seetable 4) (t = 1.730, p = 0.0869, df = 96). In the absence of APOE ϵ3/ϵ4 and ϵ4/ϵ4 genotypes, the difference in tangle numbers between C/C and T positive genotypes was larger, but without statistical significance (see table 4) (t = 2.019, p = 0.0506). These data indicate that LRP may primarily affect neuritic plaque development. Neurofibrillary tangles, on the other hand, do not seem to be directly associated with LRP. Taken together, these data support the notion that the positive or negative AD risk associated with LRP exerts a biologically penetrating effect on the severity of AD pathology. If further augments the idea that molecular processes involving LRP and its ligands, particularly apoE, contribute to the manifestation of late-onset AD.

Discussion. Genetic and biochemical evidence has demonstrated that apoE exerts a crucial role in the pathogenesis of AD,^4,7 although the precise mechanisms that underlie this process are not clear. Because the physiologic function of apoE involves the binding to its receptor,²⁴ we investigated whether LRP, a major brain apoE receptor, might be a significant counterpart in this process. In this study, we detected a significant distortion of LRP genotypes and allele frequencies between late-onset AD and control subjects. This difference was stronger in AD cases with positive family history of senile dementia. In addition, the age of disease onset was significantly altered as a consequence of different LRP genotypes. Hence, the unevenness in the distribution of LRP genotypes between AD and control subjects was more robust among cases with younger age of disease onset.

In our genetic association study, although the subjects were racially controlled (white), they were not controlled for a specific ethnic background. Thus, we used an additional experimental approach to assess the potential association of LRP with the severity of AD pathology. Postmortem AD cases harboring the C/C genotype showed significantly higher numbers of neuritic plaques than those with T positive genotypes. Although a trend was present, the tangle numbers were not significantly altered as a consequence of different LRP genotypes. Because admixture of various ethnic groups among white Americans is a challenging problem to overcome, the correlation of genetic risks to clinical and pathologic indices seems to be critical for evaluation of case-control comparisons. Although various laboratories have reported genetic associations to late-onset AD in genes coding for α₁-antichymotrypsin,²⁵ VLDL receptor,²⁶ and PS-1,²⁷ these findings remain controversial.^28,29 As an internal measure, it may be useful to seek a correlation between these genetic markers and clinical/pathologic status.

One limitation of this study is that the LRP exon 3 polymorphism represents a silent nucleotide substitution, which does not alter a splice site or the predicted amino acid sequence. Thus, a direct causal effect of this polymorphism on LRP biology is an unlikely possibility. Conceivably, the exon 3 polymorphism might alter the secondary structure of the LRP mRNA such that the efficiency of translation or stability of the message is affected, thereby changing the steady-state level of the protein. Future experimental studies will test the plausibility of this idea. However, we prefer to interpret the above data to indicate that this LRP genetic marker is in linkage disequilibrium with a putative nearby AD susceptibility locus. Although we cannot exclude the possibility that the observed disequilibrium may be a result of genetic association to another gene proximal to LRP, studies with the cystic fibrosis gene have shown that among unrelated individuals, association between two genetic markers rapidly subsides within genomic distance of 30 kb.³⁰ The gene coding for LRP, which resides on chromosome 12, covers more than 90 kb of genomic DNA³¹ and thus favors the notion that the putative disequilibrium arises from within the LRP gene. Although the potential association of the LRP T/C polymorphism with a protective or pathogenic function cannot be clearly distinguished from this study, the negative association of the T allele with AD may directly reflect a positive AD risk associated with the C allele. However, additional genetic and experimental studies are necessary to support this conclusion. Because the LRP C allele is common in both AD and control subjects, identification of other genetic markers in close proximity to LRP exon 3 might be useful for future genetic studies. As no other polymorphisms either in the coding or promoter/enhancer regions of the LRP gene have been reported to date, we are currently examining these regions for potentially informative genetic variations.

Although the role of LRP as a multifunctional endocytosis receptor has been well characterized in peripheral tissues,¹³ the functional significance of LRP in the CNS is largely unknown. Our finding that different LRP genotypes affect the numbers of neuritic plaques is consistent with the abundant immunohistochemical localization of LRP in senile plaques.¹⁵ Moreover, many known LRP ligands are also present in senile plaques,¹² suggesting that altered clearance of these molecules (i.e., apoE, apoE-Aβ complexes, etc.) may be relevant in AD. Conceivably, the clearance efficiency of apoE or the apoE-Aβ complex from the extracellular compartment may be modified as a consequence of different LRP genotypes. As LRP is predominantly found in neurons¹¹; however, an alternative possibility is that neuronal homeostasis is affected through a process involving the endocytosis of apoE-enriched lipids via LRP. In the presence of lipids, apoE4 is a dominant inhibitor of neurite outgrowth in primary and immortalized neuronal cultures, whereas apoE3, in the absence of apoE4, is effective in extending neurite outgrowth.^8,10 Moreover, inhibitors of apoE binding to LRP impede the neurite promoting effects of apoE3.⁹ Thus, a qualitative or quantitative change in the expression of LRP may alter the delivery of apoE-enriched lipids into neurons. Recently, Kounnas et al.¹⁴ showed that the Kunitz protease inhibitor (KPI) containing forms of APP are targeted for degradation via LRP-mediated endocytosis. In light of several reports providing evidence that KPI forms of APP are overabundant in AD,^32,33 an alteration in the binding or internalization of APP via LRP may be, in part, responsible for this imbalance. Taken together, our findings indicate that LRP might be genetically associated with the expression of late-onset AD and support the idea that the LRP endocytic pathway may provide a common pathophysiologic link between two apparently unrelated molecules, apoE and APP. Thus, future genetic and experimental studies in LRP may shed insight in the molecular basis of neurodegeneration in late-onset AD.

Acknowledgments

We thank Robert Davignon and Patty Melendrez for their editorial help in the preparation of this manuscript. We also thank Christine Kim and Mary Sundsmo for their technical and moral support. This work is dedicated in the memory of Professor Tsunao Saitoh.