Human Molecular Genetics Advance Access originally published online on December 22, 2004
Human Molecular Genetics 2005 14(3):447-460; doi:10.1093/hmg/ddi041
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Human Molecular Genetics, Vol. 14, No. 3 © Oxford University Press 2005; all rights reserved
Elevated amyloid ß protein (Aß42) and late onset Alzheimer's disease are associated with single nucleotide polymorphisms in the urokinase-type plasminogen activator gene
1Department of Neuroscience and 2Department of Neurology, Mayo Clinic Jacksonville, 4500 San Pablo Road, Jacksonville, FL 32224, USA, 3Center for Genomics and Bioinformatics, Karolinska Institute, Stockholm, Sweden, 4Department of Physiology and 5Sanders-Brown Center on Aging, University of Kentucky, Lexington, KY, USA and 6Department of Health Sciences Research and 7Department of Neurology, Mayo Clinic Rochester, 200 First Street SW, Rochester, MN 55905, USA
* To whom correspondence should be addressed. Email: younkin.steven{at}mayo.edu
Received June 11, 2004; Revised October 8, 2004; Accepted December 10, 2004
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Plasma amyloid ß protein (Aß42) levels and late onset Alzheimer's disease (LOAD) have been linked to the same region on chromosome 10q. The PLAU gene within this region encodes urokinase-type plasminogen activator, which converts plasminogen to plasmin. Aß aggregates induce PLAU expression thereby increasing plasmin, which degrades both aggregated and non-aggregated forms of Aß. We evaluated single nucleotide polymorphisms (SNPs) in PLAU for association with Aß42 and LOAD. PLAU SNP compound genotypes composed of haplotype pairs showed significant association with AD in three independent case–control series. PLAU SNP haplotypes associated significantly with plasma Aß42 in 10 extended LOAD families. One of the SNPs analyzed was a missense C/T polymorphism in exon 6 of PLAU (PLAU_1=rs2227564), which causes a proline to leucine change (P141L). We analyzed PLAU_1 for association with AD in six case–control series and 24 extended LOAD families. The CT and TT PLAU_1 genotypes showed association (P=0.05) with an overall estimated odds ratio of 1.2 (1.0–1.5). The CT and TT genotypes of PLAU_1 were also associated with significant age-dependent elevation of plasma Aß42 in 24 extended LOAD families (P=0.0006). In knockout mice lacking the PLAU gene, plasma—but not brain—Aß42 as well as Aß40 was significantly elevated, also in an age-dependent manner. The PLAU_1 associations were independent of the associations we found among plasma Aß42, LOAD and variants in the IDE or VR22 region. These results provide strong evidence that PLAU or a nearby gene is involved in the development of LOAD. PLAU_1 is a plausible pathogenic mutation that could act by increasing Aß42, but additional biological experiments are required to show this definitively.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
In previous studies, we and others showed that plasma amyloid ß protein (Aß42) (1
) and late onset Alzheimer's disease (LOAD) (2) are linked to a locus at 81 cM on chromosome 10. Together these two independent studies provide strong evidence for a novel Alzheimer's disease (AD) risk gene at this locus that acts on Aß42. The urokinase-type plasminogen activator gene (PLAU) is located on chromosome 10q21–22 at 
95 cM within the 1–LOD support interval of the linkage peak at 81 cM. Urokinase-type plasminogen activator is a serine protease that converts plasminogen to plasmin (3). PLAU expression is induced by Aß aggregates, and plasmin degrades both aggregated and non-aggregated Aß in vitro in a possible negative feedback loop. Thus, PLAU could have an important role in controlling Aß aggregation, which is widely believed to play an important role in AD pathogenesis. Aß is the major proteinaceous component of senile plaques, one of the main pathologic hallmarks of AD (4). Aß composed of 42(43) amino acids (Aß42) is present in all senile plaques, and Aß40 is also found in most cases (5). Aß42 was shown to be elevated in serum, brain and cell culture studies performed in patients and animal models of familial AD due to known autosomal dominant mutations (6–12). Some of these studies have also shown elevations in Aß40, though not as consistently and to the same degree as Aß42. Thus, it is postulated that although both Aß42 and Aß40 associate with one of the pathogenic hallmarks of AD, owing to its consistent and more prominent presence along with the biochemical properties of greater fibrillogenicity and aggregation potential, Aß42 is a more pathogenic biomarker of AD. Given its genomic location and potential functional relevance to AD, we tested PLAU as a candidate gene for LOAD. PLAU is a relatively small, 5970 bp gene found on the NT_008583 chromosome 10 contig and also defined by the AF377330 sequence (www.ncbi.nlm.nih.gov). We genotyped a total of seven single nucleotide polymorphisms (SNPs) spanning the whole length of this gene from the 72nd to the 5885th bp. We found evidence for significant linkage disequilibrium (LD) among these variants, allowing for complex genotype analysis. In this article, we initially depict the results of the association analyses between plasma Aß42 and PLAU haplotypes in 10 LOAD families and between AD and PLAU compound genotypes forming the ‘haplotype pairs’ in three case–control series. After establishing significant association between PLAU variants and two AD phenotypes, plasma Aß42 levels and AD-affected status, we focus on a single missense mutation in a larger data set of 24 LOAD families and six case–control series.
The missense mutation is a variant in exon 6 (rs2227564=PLAU_1) that causes a proline to leucine change (P141L) within the kringle domain of PLAU at the junction between two ß-pleated sheets (13,14). The minor T variant of PLAU_1 does not appear to affect PLAU activity, but Yoshimoto et al. (14) have reported that the P141-PLAU zymogen binds fibrin aggregates less efficiently than the L141-PLAU zymogen, suggesting the possibility of altered extracellular PLAU localization or stability.
Previously, Finckh et al. (15) have shown significant association between the major C/C genotype of rs2227564 and LOAD in their series from Germany, Switzerland and Italy. In this article, we show evidence of significant association between LOAD and the minor allele of this variant and discuss potential reasons for the discrepancy. We find evidence of significant association with PLAU variants in both extended LOAD families and independent case–control series using both plasma Aß42 and LOAD-affected status as the phenotypes. In addition, we provide functional data from PLAU knock out (KO) mice, supporting the role of this gene in Aß metabolism. Our results are depicted subsequently.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Identification of SNPs, haplotypes and compound genotypes in PLAU
We identified six SNPs by sequencing PLAU exons in LOAD family members. An additional SNP in exon 2 was identified using the public SNP databases. The seven PLAU SNPs, their locations and NCBI SNP names are summarized in Table 1. These SNPs, which spanned 5.8 kb in the PLAU gene, were analyzed in the 10 LOAD families used to link plasma Aß42 to chromosome 10 (1
). In addition, they were genotyped in three independent case–control series collected at Mayo Clinic Rochester (MCR), Mayo Clinic Jacksonville (MCJ) and University of Kentucky (UKy). The SNP allele frequencies, deviations from Hardy–Weinberg equilibrium (HWE) and allelic association results are summarized in Table 2. Given that the two exon 11 SNPs are in complete LD, the results for only one of these exon 11 SNPs are summarized in Table 2. In summary, the rare exon 2 SNP showed Hardy–Weinberg disequilibrium in the MCJ and MCR series in both the case and the control groups. Apart from this, there was only one instance where an SNP deviated significantly from the HWE in the control group (exon 6 SNP=PLAU_1 in MCJ controls).
|
|
From the PLAU SNP genotypes of the family members, PLAU SNP haplotypes were identified using the haplotyping algorithm within the computer program Simwalk2 (17
,18). Three major PLAU SNP haplotypes were identified, as were three minor haplotypes. Collectively, the six PLAU SNP haplotypes accounted for >95% of all of the haplotypes in these families. The frequency of each PLAU SNP haplotype in the 10 AD families is summarized in Table 1. Because there was profound LD (D'>0.95 for all SNPs and d2>0.4 for all SNPs except exon 8 SNP), only five of the seven SNPs were required to construct all six haplotypes.
Table 3 summarizes the LD parameters, Lewontin's standardized disequilibrium coefficient (19) D' and d2, estimated from the founders in the 10 extended LOAD families. The profound LD between the SNPs can easily be appreciated in this table. Exon 11 deletion mutant was in complete LD with the other exon 11 SNP. Exon 2 SNP was too rare to yield meaningful results. Thus, these two SNPs were omitted from the LD analysis. LD was also estimated in the MCR, MCJ and UKy series within the cases and controls and yielded similar results. As each individual carries two haplotypes, we termed the genotype formed by these two haplotypes as ‘compound genotypes’. As described subsequently, each compound genotype essentially unambiguously defines a haplotype pair due to profound LD between PLAU SNPs and limited haplotype diversity. We next tested for association between plasma Aß42 levels and PLAU haplotypes.
|
Association of PLAU variants with plasma Aß42 levels in 10 LOAD families
To test for association between plasma Aß42 and PLAU variants in family members, we used the variance components methodology implemented in the computer program SOLAR (20
). Using the within-pedigree identity-by-state model of Hopper and Mathews (21), PLAU SNP haplotypes were initially analyzed in family members between the ages of 20 and 65 years because this was the age group in which plasma Aß42 was linked to the chromosome 10 locus (1). In this age group, there was no significant association between PLAU variants and plasma Aß42 levels (Table 4). Inspection of our data suggested that there might be an association between PLAU SNP haplotypes and plasma Aß42 in older family members, so we performed a second analysis using family members of all ages. Remarkably, this group showed significant association between PLAU SNP haplotypes and plasma Aß42 (P=0.02), with PLAU association accounting for an estimated 12% of the total variance in Aß42.
|
Association of PLAU variants with AD in three case–control series
Our working hypothesis is that any gene with variants influencing Aß42 is likely to have variants influencing risk for AD. To determine whether PLAU has variants that are associated with AD in case–control series, we analyzed the PLAU SNPs in three independent series obtained at MCR, MCJ and UKy. All three series were entirely composed of Caucasian subjects, and in every AD patient the age of onset was
59 years. Using the computer program GOLD (22) to analyze the PLAU SNPs, we determined that there is profound LD among the five PLAU SNPs in the case–control series similar to the LOAD family founders (data not shown). An important consequence of the profound LD among the five PLAU SNPs that we analyzed was that there were relatively few common genotypes such that 96–100% of the subjects were accounted for by only 12 compound genotypes (Table 5). In Table 5, the two haplotypes producing six of the 12 compound genotypes (A, D, F, I, J and K) were unambiguous, because no more than one SNP was heterozygous. Each of the remaining six genotypes is produced by one and only one pair of the six haplotypes identified in the LOAD families (Table 1). The only other pairs that could produce these categories would include at least one haplotype that is so rare it was never identified unambiguously in the case–control series or by Simwalk2 haplotype analysis of the LOAD families. As these combinations will be rare, the PLAU SNP haplotypes in 96–100% of the subjects in our LOAD case–control series were easily identified and essentially unambiguous (Table 5).
|
To evaluate association between AD and the compound genotypes summarized in Table 5, we performed
2-tests on each series using the Monte Carlo approach (T1) implemented in CLUMP to determine significance. In each of the three series examined, these genotypes showed significant association with AD (MCR, P=0.001; MCJ, P=0.03 and UKy, P=0.03). Some compound genotypes tended to have the same trends in two or more series analyzed (e.g. 2/2, 1/4 and 3/5), suggesting the presence of functional variants on the backbones of these genotypes. After we found the association between PLAU ‘compound genotypes’ and three case–control series, MCR, MCJ and UKy, we tested for association between the missense PLAU_1 SNP and the larger data set of six case–control series, the results of which are depicted subsequently. Finally, we went back to test for association between PLAU compound genotypes and the three additional case–control series (Gothenburg, Gothenburg-Postmortem, Scottish) and did not find significant association in these series (data not shown).
Association of PLAU_1 SNP (rs2227564) with plasma Aß42 levels in 24 LOAD families
Of the PLAU variants analyzed, rs2227564, which we internally termed PLAU_1, is a relatively common missense SNP, which leads to a proline to leucine change in the kringle domain of PLAU at the junction between two ß-pleated sheets (13,14) and has been shown to have a functional effect on the fibrin binding of PLAU zymogen (14). To determine whether PLAU_1 genotypes are associated with plasma Aß42, we analyzed a larger set of 24 extended LOAD families, which includes the original group of 10 families. To take family relationships into account, these analyses were performed using variance components methodology implemented in the program SOLAR (20).
In the members of these 24 families, the change that occurs in plasma Aß42 with aging is complex. Plasma Aß42, which is elevated in young subjects, first decreases and then increases again with age >50. This is shown in Figure 1, where plasma Aß42 stratified by PLAU_1 genotype is plotted as a function of age. Non-linear analysis, in which we analyzed age, age(2) and PLAU_1 genotype (CC, CT, TT or CT+TT) and the interactions among age, age(2) and PLAU_1 genotype in the variance components framework, showed a significant age-dependent association between plasma Aß42 and PLAU_1 CT and TT genotypes (CTxage, TTxage(2), CT+TTxage(2) and age(2) were all significant at P<0.05) (Table 6). As our main concern was to determine whether there is a significant age-dependent elevation of plasma Aß42 in elderly subjects associated with the CT, TT or pooled CT+TT genotypes, we performed a second analysis in which family members >50 years (Fig. 2) were analyzed using linear regression in the variance components framework. This analysis showed that the CTxage and TTxage covariates had significant effects (P=0.003 and P=0.007, respectively) in the 24 LOAD families (Table 7). When the CT and TT variables were tested as a single CT+TT variable, the CT+TTxage covariate was even more significant (P=0.0006). It is noteworthy that even when we correct for the fact that three age groups were tested prior to the analysis of the 50 age group summarized in Table 7, using the most conservative approach of Bonferroni correction, there is still highly significant age-related PLAU genotype effect in these families (overall corrected P-values are 0.000024 for Model 1 and 0.00006 for Model 2).
|
|
|
|
On the basis of these findings, when we re-analyzed the haplotypic associations in the 10 LOAD families for subjects
50 years of age, the association was even more significant at P=0.005 (Table 4). We subsequently went back to analyze the haplotypes in the 24 families in this age group, which accounted for 22% of the total variance in the plasma Aß42 phenotype and was also significant (P=0.009, data not shown).
Association of PLAU_1 SNP (rs2227564) with AD in six case–control series
We initially analyzed three series from two groups, MCR and MCJ series from Mayo Clinic and UKy series from the University of Kentucky. To analyze the missense PLAU_1 SNP, we obtained three more series from our collaborators at Karolinska Institute: the Gothenburg LOAD-control series, an autopsy-confirmed LOAD series from Gothenburg and a Scottish early onset AD (LOAD) series.
In the MCR series, significant association was observed for the CT (P=0.006) and CT+TT (P=0.008) genotypes with odds ratio (OR) of 1.7 (1.2–2.6) and 1.7 (1.1–2.4), respectively. The TT genotype was enriched in the LOAD patients of the MCR series, but this enrichment did not achieve significance. In the UKy series, the first series where PLAU_1 was analyzed, the TT genotype showed significant (P=0.029) association with LOAD and had an OR of 3.5 (1.1–10.7). Three of the four remaining series (MCJ, Gothenburg and Scottish) showed some non-significant enrichment of the CT and/or TT genotypes (Table 8), but this did not occur in the autopsy-confirmed Gothenburg series. Testing for homogeneity (Breslow–Day) showed no evidence for heterogeneity among the six series (P=0.45), and meta analysis gave a pooled estimate of the OR for the CT+TT genotype of 1.21 (1.0–1.47) with a fixed effects P=0.056 and a random effects P=0.051 (Fig. 3).
|
|
Analysis of Aß in PLAU knockout mice
Our finding that PLAU variants are associated with elevated plasma Aß42 predicts that plasma Aß42 should be elevated in KO mice lacking a functional PLAU gene. To test this, we analyzed plasma Aß42 in 11 PLAU KO mice (16
) and 14 control mice with the same genetic background. For comparison, KO mice lacking a functional neprilysin gene, which is known to be involved in Aß degradation (24), were analyzed in parallel. When compared with controls, 3 to 6-month-old PLAU KO mice showed a highly significant increase in plasma Aß42 (P=0.006 by Mann–Whitney test) (Fig. 4A). Plasma Aß40 also increased (P=0.003) in the PLAU KO mice (Fig. 4B). By 11 months, plasma Aß42 and Aß40 increased and the difference between KO and control mice appeared to increase in the small number of mice analyzed (two PLAU KO versus six controls). Thus, the PLAU gene influences plasma Aß42 and our data suggest that this effect may become more pronounced with aging, consistent with our genetic analyses in LOAD families. Remarkably, brain Aß42 (Fig. 5A) and Aß40 (Fig. 5B) showed no detectable increase in PLAU KO mice when compared with control mice. The implications of this result are presented in the Discussion.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Identification of an LOAD risk locus on chromosome 10 (2
), which also affects plasma Aß (1) levels in extended LOAD families, has led to the search for genes that are good functional candidates in this region. PLAU is both a good functional and a good positional candidate gene that converts plasminogen to plasmin, which in turn degrades Aß. We genotyped seven SNPs within this relatively small gene spanning almost its entire length. We found that these SNPs were in profound LD with each other leading to limited haplotype diversity in the region analyzed, a finding consistent with previous work (25–27). The limited haplotype diversity allowed us to perform association analysis using the compound genotypes that make up the haplotype pair for each subject. The advantage of this approach is that it does not require phase information as each compound genotype forms a group per se. Nonetheless, given the small number of compound genotypes observed in all series, one can identify the underlying haplotype pair for >95% of all subjects essentially unambiguously. The other advantage of the compound genotype approach is that it allows for the analysis of the ‘haplotype pair’ rather than individual haplotypes. Thus, variants that are effective in an additive fashion, but not when encountered alone, will be easier to identify using this approach. We determined significant association between the PLAU compound genotypes with AD in three independent case–control series.
We then analyzed a common missense SNP, PLAU_1 (rs2227564) in a larger set of 24 families and six case–control series that include the initial series analyzed. In the extended families, this SNP was associated with an age-related elevation in plasma Aß levels in individuals >50 years of age. Plasma Aß displays a complex characteristic in that it appears to decline in an age-related fashion until 50 years of age and increase thereafter. This characteristic is much more pronounced in subjects who have the CT or TT genotype for PLAU_1. It is also noteworthy that the significance of association with PLAU haplotypes is enhanced when the >50 age group is analyzed separately in the 10 families, as expected (Table 4). The finding of age-related increase in Aß is also evident in the plasma Aß of PLAU KO mice, consistent with the finding in humans. These mice have significantly higher plasma Aß levels when compared with wild-type littermates; however, their brain Aß levels did not show significant difference. This finding could have several explanations. This may suggest that PLAU variants influence AD through an effect on plasma but not brain Aß42. While this is conceivable, it is more likely that PLAU influences both plasma and brain Aß in a way that is not evident in KO mice. Thus, PLAU may have a greater influence on secreted than on cell-associated Aß. As most brain Aß is cell-associated (28), effects on secreted Aß that are easily detected in plasma could be hard to detect in brain. Alternatively, compensatory changes triggered by PLAU KO may maintain brain Aß more effectively than plasma Aß.
We analyzed PLAU_1 in six case–control series using meta-analysis. Lohmueller et al. (29) reported their meta-analyses of published genetic associations with various disease phenotypes. Of the 25 reported associations that they investigated by meta-analysis, eight showed significant association that replicated the original study, with most ORs in the 1.1–2.0 range. On the basis of these findings, the authors recommended that large, collaborative genetic association studies should be encouraged, that all sound studies, positive or negative, should be published and that reports of association should include a meta-analysis of all available data. Our meta-analyses of the PLAU_1 SNP in six case–control series indicate that the PLAU_1 CT+TT is associated with a modest (OR=1.21), but marginally significant (P=0.051–0.056) increase in risk for AD. This result is similar to that of Lohmueller et al. (29), whose meta-analyses identified many other common genetic variants with modest, but significant effects on other disease phenotypes. Two other groups had previously published their findings on this SNP (15,30). Finckh et al. (15) found significant association in three case–control series from Germany, Switzerland and Italy, whereas Myers et al. (30) did not find evidence of association with any of the PLAU SNPs they analyzed including PLAU_1. Interestingly, the former found that in their series, it is the CC genotype that is associated with increased risk for LOAD, whereas in our series, it is the CT and TT genotypes. Furthermore, another group from Germany also found significant association with the CT and TT genotypes (M. Riemenschneider, personal communication). Thus, there clearly is heterogeneity among all these findings, with one group having significant results in the same direction as our series (M. Riemenschneider), another finding significance with the common genotype (15) and a third group not finding any significance. Clearly, all results need to be analyzed in one meta-analysis subsequent to their publication. Nonetheless, as demonstrated by Lohmueller et al. (29), such discrepancies are frequent among studies of common genetic variants in complex diseases, particularly when the effect size is modest. There are many reasons for such discrepancies including small sample size, gene–gene and gene–environment interactions not modelled in studies, intra-sample heterogeneity/population substructure and analysis of the ‘marker’ variant instead of the ‘disease’ variant with varying degrees of LD. While we tried to minimize these factors by gathering large series, age matching the Mayo Clinic series and choosing a variant that is likely to be functional itself, it is still plausible that some of these problems may be contributing to inter-series variations observed for PLAU_1.
Indeed, the association for PLAU_1 is marginally significant with the confidence interval of 1.0–1.47 in the six series analyzed. Nonetheless, this statistical result should be considered in the presence of evidence from the family and animal data on the effect of PLAU in Aß; and with the understanding that there likely exist other variants in PLAU that affect the risk of AD, as suggested by the ‘compound genotype’ data in the case–control series. Furthermore, given the likely presence of important variants in other genes, such as VR22 and IDE, the finding of highly significant P-values for variants of modest effect in complex diseases may need to await the identification of most, if not all, functional mutations in multiple genes and their analyses in multiple large series.
We recently published our findings of significant association in two other genes on chromosome 10 (31,32), IDE and VR22. When we tested the key variants in all three genes, we did not find evidence of LD. Joint analysis of the key variants demonstrates separate, independent effects. Of all of the variants, only those in VR22 account for the linkage on chromosome 10. PLAU variants are different from the other two in their age-related effects that manifest themselves more strongly after age 50. Thus, each of these genes appears to act independently of one another via apparent separate mechanisms. While the unequivocal effects of these genes in the pathogenesis of AD await the identification of all of the functional variants and testing them in large series, these findings provide evidence for the existence of LOAD risk variants that affect Aß, within and/or in the vicinity of these genes.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Subjects
We initially collected 10 extended LOAD families. Four of these families were collected via an LOAD patient who had extremely high plasma Aß42 and/or Aß40 levels (‘extremes’, top 10th percentile of the 545 AD patients in our series). The remaining six families were ascertained via a LOAD proband who had multiple relatives with LOAD, without prior Aß measurements (‘non-extremes’). One of the probands from the ‘non-extreme’ families was determined to have extremely high-plasma Aß levels after the family collection; therefore, this family was grouped with the ‘extreme families’ subsequently. The detailed collection strategies and family profiles for these families are provided elsewhere (1
). In addition to these 10 families, 14 independent extended LOAD pedigrees were subsequently collected. These additional pedigrees were collected via a proband who was a first-degree relative of a LOAD patient. All except one of these probands for the 14 additional pedigrees had elevated plasma Aß levels. The LOAD families studied in this article lacked a clear autosomal dominant mode of segregation for AD. In addition to the extended LOAD families, six different case–control series were analyzed in this study. These series are named as follows: Mayo Clinic Rochester (MCR), Mayo Clinic Jacksonville (MCJ), University of Kentucky (UKy), Gothenburg, Gothenburg-Postmortem and Scottish. MCR samples were collected as part of the Mayo Clinic Alzheimer's disease patient registry, a community-based prospective registry and the Mayo Clinic Rochester memory disorders clinic case–control series. MCJ samples were collected at Mayo Clinic Jacksonville memory disorders clinic. UKy series were collected as part of the University of Kentucky memory disorders clinic and BRAINS program. MCR, MCJ and UKy series consisted of Caucasian subjects from the United States. The Gothenburg and Gothenburg-Postmortem series consisted of Swedish AD cases recruited from a prospective longitudinal study of patients with dementia (The Mölndal prospective dementia study). Swedish controls consisted of healthy volunteers (Gothenburg) and autopsy cases (Gothenburg-Postmortem). Finally, the Scottish samples were composed of EOAD patients from the Lothian Psychiatric case register, and Scottish controls consisting of local church congregation volunteers from the Lothian region. The EOAD materials were selected from non-familial cases, and the most common presenilin-I mutations have been screened for. The few positives found were discarded from the test set.
Clinical AD diagnoses in all series were made according to the NINCDS-ADRDA criteria (33). All neuropathologically diagnosed AD patients (Gothenburg-Postmortem) also fulfilled the clinical NINCDS criteria for probable AD and met the neuropathological CERAD criteria for definitive AD. The ADs in all series except the Scottish series had ages of onset 59 years.
This study was approved by the appropriate institutional review boards, and appropriate informed consent was obtained from all participants.
Identification and genotyping of PLAU SNPs
Eleven members were selected from the 10 extended LOAD families that gave the linkage signal on chromosome 10 (1). The genomic DNA from five family members and two probands with high-plasma Aß level and from two members and two married-in spouses with low levels were sequenced. The samples came from four different families, three of which had probands with ‘extreme high’ Aß levels. Intronic primers flanking the PLAU exons were used for PCR amplification. All of the coding regions, 400 bp of the 5' untranslated region (5' UTR) and 1000 bp of the 3'-UTR of PLAU were sequenced. The sequencing was done via a semi-automated fluorescent method using an ABI377 sequencer and associated Factura software packages. The SNP databases at the National Center for Biotechnology Information (NCBI; URL: http://www.ncbi.nlm.nih.gov/SNP/) and University of Washington and the Fred Hutchinson Cancer Research Center (UW-FHCRC; URL: http://pga.mbt.washington.edu) were also screened for SNP identification. The SNPs identified via the sequencing effort were used in the analysis as well as a rare missense SNP in exon 2 identified in the bioinformatics screen. When the seven SNPs, thus identified, were analyzed in haplotype analysis, it was determined that two of them, the missense exon 2 SNP and exon 11 deletion SNP yielded redundant information. Therefore, remainder of the analysis was carried out with five SNPs.
PLAU SNP frequencies estimated in MCJ, MCR, UKy series and 24 LOAD families are summarized in Table 2. Of all the SNPs tested in all these groups, the rare Exon 2 SNP is in Hardy–Weinberg disequilibrium in the control groups in MCJ and MCR. This SNP is not used in the compound genotype analysis. In addition, exon 6 SNP is in Hardy–Weinberg disequilibrium in the MCJ control series.
DNA was extracted from peripheral blood leukocytes using routine methods. The Gothenburg, Gothenburg-Postmortem and Scottish series were genotyped using dynamic allele specific hybridization (34,35). The other series were genotyped on an ABI7900 instrument using TaqMan technology. For the TaqMan assays, software for designing primers and probes implemented within the ABI PRISM 7900 HT sequence detection system was utilized. The NCBI GenBank PLAU gene sequence AF377330 was used. The pedigree structure, phenotypic and genotypic information, was maintained in PEDSYS (36) (http://www.sfbr.org/sfbr/public/software/pedsys/pedsys.html), which also produced output files for the analysis performed with SOLAR, as explained subsequently.
Analysis of LD among the SNPs
We measured LD between the SNPs within the 10 extended LOAD families by using the GOLD program (22). This program determines various pairwise statistical parameters for LD, such as Lewontin's standardized disequilibrium coefficient (19) D' and d2. LD was detected from the founders and married-ins using the expectation–maximization algorithm of Slatkin and Excoffier (37). We also detected LD within the ADs and controls from MCR, MCJ and UKy series using the same program.
Tests of association between PLAU SNP haplotypes and plasma Aß levels in LOAD families
We estimated the haplotypes for the five PLAU SNPs genotyped in the members of the ten LOAD families using GOLD (22) and Simwalk2 (17). There was highly significant LD among all five PLAU SNPs, and no crossover events were observed. In our 10 extended LOAD families, we tested for association between PLAU haplotypes and plasma Aß42 phenotype using the variance components method implemented in the computer program SOLAR (20).
This method, described in detail elsewhere (20), estimates the amount of variance in a quantitative trait owing to residual genetic factors (
g2), shared-family environment (2h) and individual specific, random environmental factors (2e), based on the phenotype covariance between arbitrary relative pairs. The covariance matrix takes the form
 |
is the matrix of kinship coefficients describing polygenic factors, I is the identity matrix describing sporadic environmental factors and H is the household matrix. This is a matrix whose ijth element is equal to 1 if individuals i and j share a specified environment, and 0 otherwise. Our family collection is largely composed of extended pedigrees, where we have measured the plasma Aß of most members from each family in a single batch of ELISAs. To rigorously account for any possible assay batch to batch variations that may affect the variance in plasma Aß, we included a household matrix in the models. As almost all individuals from each family were measured in one batch, the household matrix included an element equal to 1 for those individuals from the same extended family, and 0 otherwise. We realize that by doing so, we may be overly conservative. Our collection is largely weighted towards families with high-Aß segregating within the family. Given that the shared-family effect (household) overlaps with kinship, we may have over corrected and underestimated the shared-genetic effect by including the household term in the model. The scalars g2, h2 and e2 are estimated using maximum likelihood.
To formulate an association model for PLAU SNP haplotypes, we used the within-pedigree association model of Hopper and Mathews (21). Under this model, the matrix of allelic sharing within pedigrees, 
=(
ij), is defined such that ij=1 if individuals i and j are from the same pedigree and share both alleles of a haplotype identical in state, ij=1/2 if individuals i and j are from the same pedigree and share one allele identical in state and ij=0 otherwise. Under the ‘association’ model, SOLAR estimates the variance in the quantitative trait owing to association (a2), residual genetic factors (g2), same-household effects (h2) and environmental effects (e2). Under the null-model of ‘no association’, a2 is fixed to zero. The difference between the log10 likelihoods of these two models produces a test statistic with the same distribution as the LOD score of genetic linkage analysis. There is one degree of freedom between the two models (association fixed, association tested) and one-sided test is performed.
Tests of association in the LOAD case–control series by CLUMP
In our case–control series, we tested for association between the PLAU compound genotypes and LOAD using CLUMP. The CLUMP program (23) measures the statistical significance of a variety of 2xm contingency tables by simulation. We employed the T1 test within CLUMP, which analyzes the raw 2xm table where m is the total number of categories. In our application of CLUMP, the 2xm table consisted of two diagnostic categories and m genotypic categories, each corresponding to a unique combination of genotypes at the five PLAU SNPs.
Aß analysis
Plasma Aß40 and Aß42 were measured as previously described (11). The relationship between plasma Aß42, age and PLAU_1 genotypes in LOAD family members was initially analyzed by fitting polynomial regression curves separately for all family members with rs2227564 CC, CT, TT or CT+TT genotypes, using 10 log(Aß42) as the dependent variable and age and age(2) as the independent variables. These analyses showed genotype-specific age-related effects on plasma Aß42 levels and indicated that plasma Aß42 increases with aging beginning at age 50 in those individuals with a CT or TT genotype.
We had used the 20–65 year age group in our previous studies where we found linkage between plasma Aß levels and a locus on chromosome 10 (1). As described previously, the 20–65 year age group was initially chosen because of our observation that plasma Aß is elevated in individuals <25 and >65 years old. In addition, plasma Aß appears to have complex likely deposition-related changes in individuals with AD, who are in the >65 age group. In this study, we determined significant age-related effects of PLAU_1 genotypes on plasma Aß levels in those individuals >50 years, depicted subsequent. Thus, all analyses were performed in the ‘all ages’, ‘20- to 65-year-old’ and ‘>50 years’ groups.
Tests of association between the PLAU_1 (rs2227564) SNP and plasma Aß levels in LOAD families
We applied variance components methodology implemented in the software package SOLAR (20) to estimate the effect of covariates on plasma Aß levels, while controlling for the pairwise genetic relationships and shared-family effects among family members in the extended LOAD pedigrees.
We tested for the association of PLAU_1 genotypes on plasma Aß by performing a multivariate regression analysis, while including the kinship, identity and household matrices, described earlier, in the model. We tested two different models. The significance and coefficients for all covariates were tested using the polygenic-screen-all option of SOLAR, which provides a p-value and coefficient by comparing a model that includes the covariate to a model that excludes it. This option causes all covariates to stay in the model regardless of their significance. In Model 1, the covariates tested were CT and TT genotypes, age, CT genotypexage and TT genotypexage. For Model 2, the CT and TT genotypes were pooled into a single CT+TT variable encoded as one for subjects with a CT or TT genotype and zero for those with a CC genotype. All analyses were performed in family members who were 
50 years of age. We selected this subset because we determined an age-related increase in plasma Aß42 levels after age 50 in those individuals with PLAU_1 genotypes CT or TT. All analyses were performed both with all Aß values and a second time by excluding the two extreme ‘outliers’ with plasma Aß levels ±4 SD beyond the mean. The PLAU_1 CTxage and PLAU_1 CT+TTxage were still significant variables in Model 1 and 2, respectively, after excluding the outliers.
To determine the age-related effect of PLAU_1 genotypes and PLAU_1 genotypes on plasma Aß42 levels in LOAD family members of all ages, we fitted two non-linear regression models, while taking into account the same variance components as described earlier. In Model 1 for family members of all ages, age(2), PLAU_1 CTxage(2) and PLAU_1 TTxage(2) were tested as covariates in addition to age, PLAU_1 CT, PLAU_1 TT, PLAU_1 CTxage and PLAU_1 TTxage. In Model 2, age(2) and PLAU_1 CT+TTxage(2) were included as covariates besides age, PLAU_1 CT+TT and PLAU_1 CT+TTxage. The rationale for testing these non-linear models is described in detail in the main text.
Logistic regression analysis to test for association with the PLAU_1 (rs2227564) SNP in case–controls
Logistic regression analysis implemented in SAS version 8 was utilized to test for association between AD and PLAU_1 genotypes. Initially, a model that estimated and tested the effects of PLAU_1 CT and TT genotypes separately was utilized. The results of this analysis suggested similar risky estimates for these genotypes. Given this result and that the PLAU_1 TT genotype is relatively rare, a second model was fitted, where CT and TT genotypes were combined into one variable, CT+TT, coded as 1 for those with CT or TT genotype and 0 for those with CC genotype.
Meta-analysis of PLAU_1 (rs2227564) SNP in 10 case–control series
Using the genotype counts for PLAU_1 CT+TT versus CC genotypes in AD cases and controls from the six series described in this article, we performed meta-analysis implemented in the StatsDirect statistical package. Breslow–Day test was performed to test for heterogeneity among the samples. Pooled OR estimates were calculated using both fixed effects and random effects model as implemented in the Mantel–Haenszel, and DerSimonian–Laird tests, respectively.
Analysis of plasma and brain Aß in PLAU KO mice
Plasma Aß42 and Aß40 were analyzed by sandwich ELISA as previously described (11). Brain Aß42 and Aß40 were analyzed by ELISA following direct extraction into formic acid and neutralization as previously described (28). All KO mice were homozygous for the null allele. All results were normalized to the Aß present in 3- to 6-month-old control (wild-type) mice. Numbers analyzed were as follows: 3- to 6-month-old PLAU KO mice (16) (n=9), control mice (n=6); 11-month-old PLAU KO mice (n=2), control (n=7) mice; 3- to 6-month-old neprilysin KO mice (24) (n=3), control mice (n=3). The neprilysin KO mice were used for additional positive control (24).
To evaluate brain Aß42 and Aß40, we used the same methods that we previously employed to demonstrate that endogenous brain Aß42 increases in transgenic mice expressing presenilin 1 mutations linked to early-onset familial AD (9).
|
ACKNOWLEDGEMENTS |
|---|
We acknowledge the Mayo Clinic and UKy ADRC members for their help in the collection of samples. We are grateful to all of the individuals who participated in this study, without whom this work would not be possible. This study was supported by the NIH grants AG18023 (S.G.Y.), AG06656 (S.G.Y.), AG20903 (N.E.T.), AG21545 (S.E.), 2P50AG05144 (W.M.), a grant from American Federation for Aging Research Grant PD01062 (N.E.T.) and a Robert and Clarice Smith Fellowship (N.E.T.).
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
-
Ertekin-Taner, N., Graff-Radford, N., Younkin, L.H., Eckman, C., Baker, M., Adamson, J., Ronald, J., Blangero, J., Hutton, M. and Younkin, S.G. (2000) Linkage of plasma Aß42 to a quantitative locus on chromosome 10 in late-onset Alzheimer's disease pedigrees. Science, 290, 2303–2304.
[Abstract/Free Full Text] -
Myers, A., Holmans, P., Marshall, H., Kwon, J., Meyer, D., Ramic, D., Shears, S., Booth, J., DeVrieze, F.W., Crook, R. et al. (2000) Susceptibility locus for Alzheimer's disease on chromosome 10. Science, 290, 2304–2305.
[Abstract/Free Full Text] -
Tucker, H.M., Kihiko, M., Caldwell, J.N., Wright, S., Kawarabayashi, T., Price, D., Walker, D., Scheff, S., McGillis, J.P., Rydel, R.E. et al. (2000) The plasmin system is induced by and degrades amyloid-beta aggregates. J. Neurosci., 20, 3937–3946.
[Abstract/Free Full Text] - Glenner, G.G. and Wong, C.W. (1984) Alzheimer's disease and Down's syndrome: sharing of a unique cerebrovascular amyloid fibril protein. Biochem. Biophys. Res. Commun., 122, 1131–1135.[CrossRef][ISI][Medline]
-
Gravina, S.A., Ho, L., Eckman, C.B., Long, K.E., Otvos, L., Younkin, L.H., Suzuki, N. and Younkin, S.G. (1995) Amyloid beta protein (A beta) in Alzheimer's disease brain. Biochemical and immunocytochemical analysis with antibodies specific for forms ending at A beta 40 or A beta 42(43). J. Biol. Chem., 270, 7013–7016.
[Abstract/Free Full Text] - Citron, M., Oltersdorf, T., Haass, C., McConlogue, L., Hung, A.Y., Seubert, P., Vigo-Pelfrey, C., Lieberburg, I. and Selkoe, D.J. (1992) Mutation of the beta-amyloid precursor protein in familial Alzheimer's disease increases beta-protein production. Nature, 360, 672–674.[CrossRef][Medline]
-
Suzuki, N., Cheung, T.T., Cai, X.D., Odaka, A., Otvos, L., Jr, Eckman, C., Golde, T.E. and Younkin, S.G. (1994) An increased percentage of long amyloid beta protein secreted by familial amyloid beta protein precursor (beta APP717) mutants. Science, 264, 1336–1340.
[Abstract/Free Full Text] - Borchelt, D.R., Thinakaran, G., Eckman, C.B., Lee, M.K., Davenport, F., Ratovitsky, T., Prada, C.M., Kim, G., Seekins, S., Yager, D. et al. (1996) Familial Alzheimer's disease-linked presenilin 1 variants elevate Abeta1–42/1–40 ratio in vitro and in vivo. Neuron, 17, 1005–1013.[CrossRef][ISI][Medline]
- Duff, K., Eckman, C., Zehr, C., Yu, X., Prada, C.M., Perez-tur, J., Hutton, M., Buee, L., Harigaya, Y., Yager, D. et al. (1996) Increased amyloid-beta42(43) in brains of mice expressing mutant presenilin1. Nature, 383, 710–713.[CrossRef][Medline]
-
Hsiao, K., Chapman, P., Nilsen, S., Eckman, C., Harigaya, Y., Younkin, S., Yang, F. and Cole, G. (1996) Correlative memory deficits, Abeta elevation, and amyloid plaques in transgenic mice. Science, 274, 99–102.
[Abstract/Free Full Text] - Scheuner, D., Eckman, C., Jensen, M., Song, X., Citron, M., Suzuki, N., Bird, T.D., Hardy, J., Hutton, M., Kukull, W. et al. (1996) Secreted amyloid beta-protein similar to that in the senile plaques of Alzheimer's disease is increased in vivo by the presenilin 1 and 2 and APP mutations linked to familial Alzheimer's disease. Nat. Med., 2, 864–870.[CrossRef][ISI][Medline]
- Citron, M., Westaway, D., Xia, W., Carlson, G., Diehl, T., Levesque, G., Johnson-Wood, K., Lee, M., Seubert, P., Davis, A. et al. (1997) Mutant presenilins of Alzheimer's disease increase production of 42-residue amyloid beta-protein in both transfected cells and transgenic mice. Nat. Med., 3, 67–72.[CrossRef][ISI][Medline]
- Li, X., Bokman, A.M., Llinas, M., Smith, R.A. and Dobson, C.M. (1994) Solution structure of the kringle domain from urokinase-type plasminogen activator. J. Mol. Biol., 235, 1548–1559.[CrossRef][ISI][Medline]
- Yoshimoto, M., Ushiyama, Y., Sakai, M., Tamaki, S., Hara, H., Takahashi, K., Sawasaki, Y. and Hanada, K. (1996) Characterization of single chain urokinase-type plasminogen activator with a novel amino-acid substitution in the kringle structure. Biochim. Biophys. Acta, 1293, 83–89.[CrossRef][Medline]
- Finckh, U., van Hadeln, K., Muller-Thomsen, T., Alberici, A., Binetti, G., Hock, C., Nitsch, R.M., Stoppe, G., Reiss, J. and Gal, A. (2003) Association of late-onset Alzheimer disease with a genotype of PLAU, the gene encoding urokinase-type plasminogen activator on chromosome 10q22.2. Neurogenetics, 4, 213–217.[CrossRef][ISI][Medline]
- Carmeliet, P., Schoonjans, L., Kieckens, L., Ream, B., Degen, J., Bronson, R., De Vos, R., van den Oord, J.J., Collen, D. and Mulligan, R.C. (1994) Physiological consequences of loss of plasminogen activator gene function in mice. Nature, 368, 419–424.[CrossRef][Medline]
- Weeks D.E., Sobel E., O'Connell J.R. and Lange K. (1995) Computer programs for multilocus haplotyping of general pedigrees. Am. J. Hum. Genet., 56, 1506–1507.[ISI][Medline]
- Sobel, E. and Lange, K. (1996) Descent graphs in pedigree analysis: applications to haplotyping, location scores, and marker-sharing statistics. Am. J. Hum. Genet., 58, 1323–1337.[ISI][Medline]
- Lewontin, R.C. and Kojima, K. (1960) The evolutionary dynamics of complex polymorphisms. Evolution, 14, 450–472.
- Almasy, L. and Blangero, J. (1998) Multipoint quantitative-trait linkage analysis in general pedigrees. Am. J. Hum. Genet., 62, 1198–1211.[CrossRef][ISI][Medline]
- Hopper, J.L. and Mathews, J.D. (1982) Extensions to multivariate normal models for pedigree analysis. Ann. Hum. Genet., 46, 373–383.
-
Abecasis, G.R. and Cookson, W.O. (2000) GOLD—graphical overview of linkage disequilibrium. Bioinformatics, 16, 182–183.
[Abstract/Free Full Text] - Sham, P.C. and Curtis, D. (1995) Monte, Carlo tests for associations between disease and alleles at highly polymorphic loci. Ann. Hum. Genet., 59, 97–105.[ISI][Medline]
-
Iwata, N., Tsubuki, S., Takaki, Y., Shirotani, K., Lu, B., Gerard, N.P., Gerard, C., Hama, E., Lee, H.J. and Saido, T.C. (2001) Metabolic regulation of brain Abeta by neprilysin. Science, 292, 1550–1552.
[Abstract/Free Full Text] -
Patil, N., Berno, A.J., Hinds, D.A., Barrett, W.A., Doshi, J.M., Hacker, C.R., Kautzer, C.R., Lee, D.H., Marjoribanks, C., McDonough, D.P. et al. (2001) Blocks of limited haplotype diversity revealed by high-resolution scanning of human chromosome 21. Science, 294, 1719–1723.
[Abstract/Free Full Text] - Daly, M.J., Rioux, J.D., Schaffner, S.F., Hudson, T.J. and Lander, E.S. (2001) High-resolution haplotype structure in the human genome. Nat. Genet., 29, 229–232.[CrossRef][ISI][Medline]
-
Gabriel, S.B., Schaffner, S.F., Nguyen, H., Moore, J.M., Roy, J., Blumenstiel, B., Higgins, J., DeFelice, M., Lochner, A., Faggart, M. et al. (2002) The structure of haplotype blocks in the human genome. Science, 296, 2225–2229.
[Abstract/Free Full Text] -
Kawarabayashi, T., Younkin, L.H., Saido, T.C., Shoji, M., Hsiao, Ashe, K. and Younkin, S.G. (2001) Age-dependent changes in brain, CSF, and plasma amyloid ß protein in the Tg2576 transgenic mouse model of Alzheimer's disease. J. Neurosci., 21, 372–381.
[Abstract/Free Full Text] - Lohmueller, K.E., Pearce, C.L., Pike, M., Lander, E.S. and Hirschhorn, J.N. (2003) Meta-analysis of genetic association studies supports a contribution of common variants to susceptibility to common disease. Nat. Genet., 33, 177–182.[CrossRef][ISI][Medline]
- Myers, A.J., Marshall, H., Holmans, P., Compton, D., Crook, R.J., Mander, A.P., Nowotny, P., Smemo, S., Dunstan, M., Jehu, L. et al. (2004) Variation in the urokinase-plasminogen activator gene does not explain the chromosome 10 linkage signal for late onset AD. Am. J. Med. Genet., 124B, 29–37.[CrossRef]
-
Ertekin-Taner, N., Ronald, J., Asahara, H., Younkin, L., Hella, M., Jain, S., Gnida, E., Younkin, S., Fadale, D., Ohyagi, Y. et al. (2003) Fine mapping of the {alpha}-T catenin gene to a quantitative trait locus on chromosome 10 in late-onset Alzheimer's disease pedigrees. Hum. Mol. Genet., 12, 3133–3143.
[Abstract/Free Full Text] - Ertekin-Taner, N., Allen, M., Fadale, D., Scanlin, L., Younkin, L., Petersen, R.C., Graff-Radford, N. and Younkin, S.G. (2004) Genetic variants in a haplotype block spanning IDE are significantly associated w