13 January 2004. This report summarizes some of the genetic findings for late-onset Alzheimer's disease presented last November at the 33rd Annual Meeting of the Society for Neuroscience in New Orleans. It covers recent progress in identifying the genes responsible for the two well-established AD linkage peaks on chromosomes 9 and 10, as well as recent studies of several other candidate genes.
UBQLN1: Several genome screens for late-onset AD loci have agreed upon genetic linkage on chromosome 9. It is fair to say that after chromosome 10, chromosome 9 is one of the best confirmed locations for a novel AD gene (for review, see Tanzi and Bertram, 2001; Myers and Goate, 2001). The gene, UBQLN1, encoding the presenilin-interacting protein ubiquilin (Mah et al., 2000), maps close to the apex of the linkage peak on chromosome 9q22 (e.g., see Blacker et al., 2003). At the conference, Lars Bertram and Rashmi Menon from our group at Massachusetts General Hospital in Charlestown reported the results of analyses of five single nucleotide polymorphisms (SNPs) in the UBQLN1 gene. All five belong to the same haplotype block spanning about 50 kb; more importantly, two of the five SNPs were significantly associated with AD in the NIMH AD family sample (n = 437 families; p-values = 0.018 and 0.046). The 27 best UBQLN1-associated and 9q22-linked AD families from the NIMH sample had an average onset age of 71.7±7.3 years (range: 59-93); 20 of these carried at least one copy of the ApoE ε4 allele, while 7 families did not carry any ApoE ε4. Bertram went on to report that eliminating the 27 families that exhibit the strongest association with UBQLN1 (n = 27, i.e., about six percent of the full NIMH sample) reduced the linkage evidence on chromosome 9q22 by about 50 percent; this corroborates the association of UBQLN1 with AD. Moreover, linkage analysis with these 27 families alone enhanced the linkage peak on 9q22 and shifted it in the direction of UBQLN1.
Kristina Mullin of our group reported further support for the candidacy of UBQLN1 as a novel AD gene. She obtained independent confirmation of association of the same two UBQLN1 SNPs with AD (and the same risk alleles from the NIMH sample) in a second independent set of families, the Consortium on Alzheimer’s Genetics (CAG) sample. CAG has enrolled 314 families, with current genotyping being done on 155 families consisting mainly of discordant sibships. Collectively, these data strongly suggest that UBQLN1 is the one gene responsible for the well-established AD linkage peak on chromosome 9q22. At this time, it remains unclear whether this region harbors additional AD loci.
APBA1: Menon and Bertram presented evidence for single-locus association with one out of the seven SNPs analyzed in the gene encoding X11-α [MINT1] in the NIMH AD family sample. However, conditional logistic regression analysis did not reveal a significant increase in risk for this variant. Further, the best APBA1-associated families (n = 37) did not account for a significant proportion of the 9q22 linkage signal. Finally, the association was not confirmed in the CAG sample, suggesting that genetic variants in APBA1 most likely do not make a major contribution to the overall risk for AD.
ABCA1: Michele Parkinson et al. from our group reported that, contrary to previously published findings (Wollmer et al., 2003), analyses performed on 7 SNPs in ABCA1 showed no association with disease risk or onset age in the NIMH families (see also Sun et al. 2003).
In December 2000, three groups simultaneously reported the first evidence for significant genetic linkage of AD to the long arm of chromosome 10. Efforts to isolate the gene responsible have been encumbered by the fact that the linkage region implicated by the three papers spans more than 40 megabases. This has raised the possibility that there may be two different AD genes on chromosome 10, one in the more proximal region, reported by the Younkin and Goate groups (Ertekin-Taner et al., 2000 and Myers et al., 2000, respectively), and the other located more distally, reported by our group (Bertram et al., 2000).
Chromosome 10 distal linkage region
IDE: In our initial report of linkage of AD to chromosome 10, we had focused on the gene for insulin-degrading enzyme (IDE) as a candidate because of IDE’s capacity to degrade Aβ. We presented evidence for significant genetic linkage and association of AD with IDE. In New Orleans, Bertram presented the results of our analyses of 14 IDE SNPs in the large NIMH AD family sample, which comprises 437 uniformly ascertained and evaluated AD families. Bertram showed the IDE gene to reside next to the KIF11 gene in a region of approximately 200,000 base pairs, which was found to contain three haplotype blocks. The middle haplotype block encompassing the 5’UTR and promoter region of IDE was shown to contain a 4-SNP haplotype that revealed significant family-based association of IDE to late-onset AD (p = 0.0009). The association of AD with this haplotype block accounts for roughly half of the observed genetic linkage of IDE to AD in the NIMH family sample.
Bertram also presented a single SNP in the IDE promoter region (IDE-U4) that was associated with AD in the NIMH sample. This association was confirmed (with the same risk allele) in the CAG sample. Collectively, these data provide additional evidence for the candidacy of IDE as a late-onset AD locus on the long arm of chromosome 10.
In his presentation, Bertram also mentioned that five other labs have observed association of various IDE SNPs or haplotypes with increased risk for AD. Along these lines, also in New Orleans, Nilufer Ertekin-Taner, Steve Younkin, and colleagues at the Mayo Clinic in Jacksonville, Florida, presented evidence that IDE SNPs (from Anthony Brookes’s group at the Karolinska Institute in Stockholm; Prince et al., 2003) were associated with both increased risk for AD and increased plasma Aβ42 levels. This result concurs with the recent publication of positive association of AD with other SNPs in IDE in the Mayo samples by Steve Edland et al. (Edland et al., 2003; see also Edland section in ARF related news story). In addition, at the 53rd Annual Meeting of the American Society of Human Genetics held in Los Angeles just prior to the Neuroscience meeting, Alison Goate and colleagues at Washington University in St. Louis, Missouri, reported genetic association of AD with two independent SNPs in IDE; this is in contrast to her group’s previous report of negative association (see ARF related news story).
Thus, since our original report of IDE linkage and association, converging evidence from our own lab and several others now strongly suggests that IDE is at least one of the AD genes residing on the long arm of chromosome 10. It should be emphasized, however, that it remains unknown which, if any, of the AD-associated IDE SNPs may be pathogenic for AD, as opposed to being in linkage disequilibrium with unidentified pathogenic DNA variants or mutations. Thus, further studies are required.
Malcolm Leissring, Dennis Selkoe, and colleagues at Brigham and Women’s Hospital in Boston presented their analyses of parallel sets of transgenic mice overexpressing IDE or neprilysin (NEP) in the same genetic background, using a CaM kinase II promoter, crossed with APPSwe/Ind transgenic mice. Soluble and insoluble Aβ40 and Aβ42 brain levels were reported to be decreased by more than half in IDE/APP and NEP/APP transgenic mice, relative to age-matched six- to 10-month-old APP transgenic controls. Plaque burden was also found to be decreased in the double-transgenic mice, as was the rate of premature death prior to eight months of age, which is normally elevated in the APP transgenic mice. Interestingly, overexpression of IDE by only twofold over physiological levels still significantly reduced β-amyloid burden in these animals. Frank LaFerla and colleagues at University of California, Irvine, presented evidence that in regions of the human brain that are vulnerable to AD pathology, e.g., hippocampus, and also in muscle fibers sensitive to inclusion body myositis pathology (IBM is a human model for AD-like amyloidosis), steady-state levels of IDE and/or NEP diminish as a function of age. In contrast, muscle and brain regions not associated with significant Aβ accumulation exhibited an age-dependent increase in these catabolic enzymes. The authors suggest that differences in the steady-state levels of these enzymes with aging may account for susceptibility to Aβ-related pathology (see also ARF conference story).
GSTO1/GSTO2: In 2002,Yi-Ju Li, Peggy Pericak-Vance, and colleagues reported linkage of a locus affecting age-at-onset (but not “risk”) for AD and PD on the distal long arm of chromosome 10 (see ARF related news story) in a region telomeric of IDE. Based on expression profiling in AD brains, the same group (Li et al., 2003) reported four genes (Stearoyl-CoA desaturase; NADH ubiquinone oxidoreductase 1 β complex 8; protease, serine 11; and glutathione S-transferase, omega-1, GSTO1) to differ significantly in their expression between AD and control brain samples, and to reside in the AD/PD age-at-onset linkage region on chromosome 10. This group later reported genetic association for two members of the GST omega family (GSTO1 and GSTO2), both of which play roles in the inflammation process (see also ARF related news story).
One of the main findings in this recent study was association of age-at-onset of AD and PD with an amino acid substitution (Ala140Asp) in the GSTO1 gene. At the Neuroscience meeting, however, Bertram presented the absence of association of this same polymorphism with either “age-at-onset” or “risk” for AD in the NIMH sample. Clearly, more work in independent samples is necessary to elucidate whether these genes play a role in AD and PD pathogenesis.
CH25H-LIPA: In a poster presentation, Parkinson and colleagues from our group reported that four SNPs in the CH25H-LIPA genes located within a 50 kb interval on chromosome 10q23 (near our linkage region in the vicinity of the IDE-gene) revealed a marginally significant effect in the NIMH AD sample, with one variant located in intron 2 of LIPA, and only in ApoE ε4/4-positive families (p-value = 0.034, OR 3.4 [1.2-9.8]). However, no other variants in this region showed evidence for association, and the haplotype analyses performed on this set of polymorphisms yielded results that were inconsistent with the single SNP findings. We are currently investigating whether these findings are specific for the CH25H-LIPA region, or whether they are perhaps the result of linkage disequilibrium effects with the IDE association about 3 Mb further distal. (CH25H encodes cholesterol 25-hydroxylase; LIPA is an atherosclerosis candidate gene.)
Chromosome 10 proximal linkage region
VR22: Recently, Nilufer Ertekin-Taner, Steve Younkin, and colleagues reported evidence that the locus responsible for increasing plasma Aβ42 on the more proximal linkage peak on chromosome 10 was the gene encoding α-T catenin, a binding partner of β-catenin (Ertekin-Taner et al., 2003). Specifically, they reported that two intronic VR22 SNPs (4360 and 4783, in strong linkage disequilibrium with one another) were associated with higher plasma Aβ42 levels in 22 late-onset AD families. In New Orleans, Younkin reviewed these findings and also reported that the association with VR22 accounted for a significant portion of the genetic linkage that his group originally reported for this region (Ertekin-Taner et al., 2000).
Bertram presented the results of our analysis of these and three other SNPs in the VR22 gene in the NIMH AD sample. While the overall results (including for SNP4360) were negative (p = 0.34), it should be noted that a trend toward association with risk for AD (p = 0.08) was observed in the subset of families with onset after 65 years. Thus, it may still be worthwhile to search for potentially pathogenic SNPs/mutations in VR22. While at this point it is unlikely that the tested VR22 SNPs represent pathogenic variants, they may still be in linkage disequilibrium with other unknown pathogenic variants or mutations.
PLAU: The urokinase-type plasminogen activator (PLAU) can degrade Aβ aggregates by generating plasmin (Tucker et al., 2000). The PLAU gene resides on 10q21-22 at about 95 cM within the 1-lod support interval of the proximal linkage peak at about 80 cM. Younkin and colleagues reported on analyses of various SNPs and haplotypes in PLAU, particularly the P141L polymorphism. The CT/TT genotype of P141L was reported to be associated with an age-dependent increase in plasma Aβ42. Interestingly, Younkin added, plasma Aβ is elevated in PLAU KO mice. Younkin went on to review the data of 10 different studies performed in collaboration with Brookes—Matthias Riemenschneider at the Technical University, Munich, and Steve Estus at the University of Kentucky in Lexington. (Estus also reported negative data on several other genes related to the plasmin pathway.) While the overall findings supported an association of PLAU with risk for AD, the results of these studies revealed inconsistent risk alleles across cases and families. Thus, the data suggest that P141L is most likely in linkage disequilibrium with unknown pathogenic variants and mutations that are yet to be identified. Younkin also emphasized the need to subdivide samples genetically when searching for mutations.
Other AD candidate genes
CYP46: Parkinson et al. reported analyses of two noncoding SNPs flanking exon 3 in the cholesterol 24-hydroxylase gene, CYP46. While neither revealed overall association with risk for AD, stratification by ApoE ε4/4-genotype was found to reveal strong but opposite associations with AD in the NIMH sample. Specifically, the minor alleles of both variants (i.e., C-allele in CYP46-intron 2 and T-allele in CYP46-intron 3) exhibited significant overtransmission to affected subjects in the ApoE ε4/4-positive samples (p = 0.0009 and 0.007, respectively), but this did not translate into a significantly elevated risk for AD in these families. On the other hand, the opposite alleles were overtransmitted in the ApoE ε4/4-negative sample (p =0.017 in both cases), and this elevated the odds for AD by about twofold. This latter finding is in agreement with previous studies that have reported association of the major alleles of both of these variants with increased risk for AD (Koelsch et al., 2002; see also ARF related news story).
In considering the candidacy of CYP46 as an AD gene, some caveats should be mentioned. First, Desai et al. found no association in their case-control sample (Desai et al., 2002). Second, in New Orleans, Parkinson and Bertram reported no evidence of association between these two variants and AD in the independent CAG family sample. Third, to date, there is no reported evidence for genetic linkage of AD to the region of chromosome 14q32 where the CYP46 gene is located. Thus, the potential contribution of CYP46 to AD genetics may turn out to be genuine, but with limited effect in the general population.
Ben Wolozin and colleagues at Loyola University in Maywood, Illinois, reported on the distribution of CYP46 in the brain and found signal in neurons in both AD and control brain. In AD brain, an increase in the levels of cholesterol 24-hydroxylase was also reported. Additionally, CYP46 co-localized with neuritic plaques along with microglia and astrocytes. CHO cells overexpressing APP that were either transfected with CYP46 or treated with 24(S)-hydroxycholesterol were reported to lead to decreased PMA-stimulated secretion of APP and decreased APP-CT. The authors proposed the hypothesis that CYP46 levels may be elevated near neuritic plaques to remove cholesterol from degenerating neurites, and that increased production of 24(S)-hydroxycholesterol increases Aβ secretion. (See also Rebeck section of ARF conference story.)
OLR1: The OLR gene encoding the lipoprotein receptor for oxidized proteins resides on chromosome 12, close to the α2-macroglobulin (A2M) gene. In a poster presentation, Parkinson et al. reported the results of testing a 3’UTR SNP (rs1050283) that was previously reported to be associated with AD risk in two studies (Luedecking-Zimmer et al., 2002; Lambert et al., 2003). In the NIMH AD family sample, this same SNP was not found to be associated with AD. This negative result for OLR1 contrasts with the consistently positive findings that have been previously reported in the NIMH sample for multiple polymorphisms in the gene encoding α2-macroglobulin (Saunders et al., 2003), which maps within 1 Mb proximal of OLR1. Parkinson et al. also reported no evidence for significant linkage disequilibrium between variants in the OLR1 and A2M genes, suggesting that the previously reported associations of AD with OLR1 are most likely not due to linkage disequilibrium with A2M. Thus, the DNA variant in OLR1 does not appear to substantially affect AD risk independently of A2M in the NIMH AD family sample.
TNFA: After testing two SNPs in the tumor necrosis α (TNFA) gene in an age- and gender-matched case-control study, E. M. Pfeiffer and colleagues at University of Washington Medical Center in Seattle presented evidence that the TNF-863 polymorphism, which leads to decreased production of TNFa protein, is associated with a reduced risk for developing AD. Interestingly, TNFA resides in a region of chromosome 6 shown in genome screens by multiple groups to contain a novel AD locus. These results provide further support for the hypothesis that genetic variability in the production of TNFa might affect the inflammatory response in AD pathogenesis. Further study is warranted.
MTHFR: Given that epidemiological studies have recently revealed elevated plasma homocysteine levels in AD (see ARF related news story), Yosuke Wakutani, Kazuhiro Nakashima, and colleagues of Tottori University in Japan performed a case-control study in a Japanese population to test three polymorphisms (C677T [Ala222Val], A1298C [Glu429Ala], and A1793G [Arg594Gln]) in the gene encoding methylenetetrahydrofolate reductase (MTHFR) for association with late-onset AD. They found MTHFR to contain four major haplotype alleles, one of which, haplotype-C (677C-1298C-1793G) was less represented in cases than in controls. This protective effect was stronger in patients who lacked the ApoE4 allele. These results suggest that this haplotype of MTHFR may protect against the development of LOAD, perhaps by regulating homocysteine levels (see also ARF Live Discussion.
STH: The Saitohin (STH) gene resides in the intron between exons 9 and 10 of the human tau gene on chromosome 17 (ARF related news story, see also Conrad et al., 2002). In New Orleans, Chris Conrad, Peter Davies, and colleagues at Albert Einstein College of Medicine in the Bronx, New York, reported that the STH is expressed similarly to the tau gene. Moreover, the STH SNP, Q7R, was also reported to be overrepresented in the homozygous state in late-onset AD cases. Based on this association, the potential for a pathological role of this amino acid change in STH was suggested. However, it will also be important, in parallel, to test whether the association of AD with STH might involve possible linkage disequilibrium of Q7R with polymorphisms/mutations in the tau gene, especially those associated with frontotemporal dementia.
DLD: Abe Brown, John Blass, and colleagues of Weill Medical College of Cornell University in White Plains, New York, presented evidence for association of the mitochondrial-ketoglutarate dehydrogenase complex with AD (see also ARF Live Discussion with Blass). Four SNPs in the DLD gene, which resides on chromosome 7, were genotyped in a case-control series of 297 Caucasians from New York City, which included 229 Ashkenazi Jews. Association with AD was reported specifically for the male population independently of ApoE status. Interestingly, this result was noted to be consistent with a previous report of association of AD with markers from this same region of chromosome 7, exclusively in paternal families (Bassett, et al., 2002).—Rudy Tanzi, Massachusetts General Hospital, Charlestown.