|Home | About | Journals | Submit | Contact Us | Français|
In and of itself, late-onset Alzheimer disease (AD) can be a devastating illness. However, a sub-group of AD patients develop psychosis as the disease progresses. These patients have an added burden of greater cognitive impairment, higher rates of institutionalization, and higher mortality than AD patients without psychosis. While the etiopathogenesis of psychosis in AD (AD+P) is not known, mounting evidence accrued over the past ten years indicates that AD+P represents a distinct phenotype with a genetic basis. Elucidating the genetic mechanism of AD+P is crucial if better pharmaceutical treatments are to be developed for these patients. The goal of this review is to summarize what is currently known regarding the genetic basis of psychosis in AD. Specific attention is given to familial aggregation and heritability, linkage to chromosomal loci, and associations of candidate genes such as APOE and the monoamine neurotransmitter system.
Alzheimer disease (AD) is a neurodegenerative illness that can be broken down into two sub-groups, early- and late-onset. Early-onset AD accounts for less than 5% of all AD cases. In many early-onset AD cases, an autosomal dominant mode of transmission is seen. Causative mutations in the genes presenilin 1 (PSEN1) and presenilin 2 (PSEN2) have been described, while both causal mutations and copy number variations within the amyloid precursor protein (APP), have been reported [1,2,3,4,5]. Late-onset AD is also a highly heritable, albeit complex, phenotype. To date only the apolipoprotein E (APOE) gene, and specifically, the є4 allele, has been accepted as a genetic risk factor [6,7], though substantial data now also supports an association of late-onset AD with genetic variation in SORL [8,9,10,11]. Many other genes and chromosomal regions have been identified as possibly being involved in late-onset AD, but because of the genetic heterogeneity of the disease, these have not been confirmed across studies.
A subgroup of patients with late-onset AD also develops psychosis during progression of the disease (AD with psychosis, AD+P). Psychotic symptoms in AD are typically defined by the presence of delusions and hallucinations (but not by the presence of disorganization of thought and behavior in the absence of delusions or hallucinations). Reported common delusions are of persecution, infidelity, abandonment, and the belief that deceased individuals are still alive . In addition, individuals with AD+P have frequent misidentification delusions, such as the believe that there are phantom boarders in the house, ones’ spouse is an imposter, or ones’ home is not really one’s own . Hallucinations can occur in any modality, but are typically visual [12,13,14]. Recently, Ropacki and Jeste  comprehensively reviewed the literature on psychosis in AD from 1990 to 2003. They reviewed 55 studies comprised of 9,749 subjects. The median prevalence of AD+P was 41% (range=12.2-74.1%). The median prevalence of delusions was 36% (range=9.3-63%), with delusions of theft the most frequently reported type of delusion. Other psychotic symptoms, predominantly misidentification delusions, were the next most frequent psychotic symptom, with a median prevalence of 25% (range=3.6-38.9%). The median prevalence of hallucinations was 18% (range=4-41%) with visual hallucinations (median=18.7%) more common than auditory hallucinations (median= 9.2%). Patients experienced both hallucinations and delusions with a median prevalence of 13% (range=7.8-20.8%).
Greater cognitive impairment was the most consistent correlate of AD+P compared to AD without psychosis (AD−P). Ropacki and Jeste found in 20 of the 30 studies that assessed this relationship that greater cognitive impairment, typically as determined by the mini mental state examination (MMSE) , was associated with the presence of psychosis. Additionally, nine of nine studies found a significant association between a greater rate of cognitive decline and the presence of AD+P. Studies conducted more recently have continued to support the relationship between greater cognitive impairment and AD+P [16,17]. In contrast, only inconsistent associations have been detected between AD+P and age, age at onset of AD, illness duration, gender, education, and family history of dementia or psychiatric illness. AD+P may be associated with African American ethnicity, though it has only been examined in a limited number of studies to date.
Psychotic behaviors in AD patients have a tremendous impact on the patient, family, and caregiver. Oftentimes, these patients also have other psychiatric and behavioral disturbances, the most frequent and troublesome of which are agitation, and verbal and physical aggression[19,20]. Overall, AD+P leads to greater distress for family and caregivers, greater rates of institutionalization[22,23,24,25], and worse general health for the patient, with increased mortality compared to patients with AD−P.
Treatment of psychosis in AD patients has been suboptimal due to the limited efficacy of the classes of drugs available and their high toxicity in this age group. Haloperidol is the most studied of the conventional antipsychotics, and it has demonstrated mild to moderate efficacy relative to placebo in AD patients with psychosis and/or agitated behaviors. However, it also causes serious side-effects in these patients, namely parkinsonism, tardive dyskinesia, and akathisia. More recent studies have examined the efficacy of second generation, or atypical antipsychotics, such as risperidone, olanzapine, and aripiprazole[31,32]. These medications have similar efficacy to conventional antipsychotics in reducing psychotic symptoms, with a lower likelihood of inducing motor side effects. However, they have been associated with an increased risk of cerebral vascular adverse events such as stroke, and increased all-cause mortality[34,35,36,37]. Furthermore, there is no current data to suggest that any of these treatments effectively mitigate against the greater cognitive and functional burden associated with AD+P.
Alzheimer disease can be a devastating illness. It is clear that patients with AD+P have an added burden. Current treatment for AD+P is not adequate. It is, therefore, important to determine the etiopathogenesis of psychosis in AD, so that better pharmaceutical treatments can be developed for this sub-group of patients. One approach results from the observation that AD patients who exhibit psychosis represent a distinct phenotype with a genetic basis. During the past 10 years, substantial information has begun to accrue regarding the genetics of AD+P. The goal of this review is to summarize the current state of knowledge regarding the genetic basis of psychosis in AD, with specific attention to familial aggregation and heritability, linkage to chromosomal loci, and associations of candidate genes.
Familial aggregation refers to the occurrence of a trait in more members of a family than can be accounted for by chance. It is important to remember that even though a trait may aggregate within families, its appearance may still be due to environmental influences that are shared by family members. In contrast, heritability is an estimate of the proportion of risk for a trait which is attributable to genetic variation.
One of the first studies that examined familial aggregation of AD+P was done by Tunstall et al. Looking at sibling pairs diagnosed with AD, Tunstall found the frequency of psychosis in these patients to be 0.41 (with delusions and hallucinations showing frequencies of 0.36 and 0.30, respectively). The pair-wise concordance for psychosis (i.e. frequency of pairs in which both siblings were positive for psychosis) was 0.21, a value that is modestly higher than expected by chance alone, 0.17, and the excess pair-wise concordance was 0.04 (Table 1).
Sweet et al studied a larger cohort of probands and their siblings for familial aggregation of AD+P, examining a total of 371 probands and their 461 siblings from the National Institute of Mental Health Alzheimer Disease (NIMH AD) Genetics Initiative (Table 1). All probands and siblings were diagnosed with AD, and 75.5% of the probands were positive for psychosis. They found a significant association between psychosis in the proband and the occurrence of AD+P in siblings. The Odds Ratio (OR; 95% C.I.) for AD+P in siblings of AD+P probands was 2.4 (1.46-4.0). Similar results were obtained in supplementary analyses covarying for sibling age and age-of-onset and for the presence of extrapyramidal symptoms. When the same model was used to examine a more restrictive definition of AD+P requiring multiple psychotic symptoms to be present over time, the familial aggregation was strengthened, with an OR (relative to single or no psychotic symptoms) of 3.18 (2.17–4.66).
In a follow-up to this study, they used statistical modeling to estimate the heritability of psychosis among 826 of the individuals from their prior report. Though the heritability of any occurrence of a psychotic symptom was modest (29.5%, p=0.04), the heritability of the more restrictive definition of psychosis requiring multiple symptoms was estimated at 60.8%, p=0.004. The OR for at least one psychotic symptom was 2.37 (1.45-3.87), and the OR (relative to no psychotic symptoms) increased to 5.42 (2.62-10.43) for multiple psychotic symptoms (Table 1).
Hollingworth et al undertook a study to replicate and expand the work of Sweet et al by incorporating data from the NIMH AD Genetics Initiative’s pedigrees and families recruited from the United Kingdom (UK). They found a significant association between proband psychosis status and the occurrence of AD+P in family members in both the US and UK samples (Table 1). The estimated OR in the NIMH sample for the development of AD+P in siblings of probands with AD+P was 3.2 (2.05-4.99). Similar findings were obtained in the UK sample with an OR of 4.17 (1.67-10.44). For the combined sample of NIMH and UK families, the OR was 3.38 (2.27-5.05).
The studies of familial aggregation and heritability of psychosis in AD, summarized above, strongly implicate genetic variation as contributing to AD+P. An important question is what chromosomal loci might be linked to the development of psychosis in AD, and what are the implicated genes? To answer these questions, genetic linkage studies have been done to identify loci that predispose individuals to AD+P.
Bacanu et al reported the first linkage study in AD+P, evaluating 65 families from the National Institute of Mental Health Alzheimer Disease (NIMH AD) Genetics Initiative with two or more members having the restrictive definition of AD+P. Furthermore, they partitioned the families into a subset of 42 families in which two or more individuals were carriers of apolipoprotein (APOE) є4 alleles (AD+P+є4). When looking at this subset, they found evidence of significant linkage on chromosome 2p near marker D2S1356 (64.3 cM). In addition, they found two suggestive linkages on chromosome 6 near marker D6S1021 (112.2 cM), and chromosome 21 at marker D21S1440 (36.8cM). When analyzing the entire group of 65 families, they found a suggestive linkage on chromosome 6, corresponding to the same region that was found in the AD+P+є4 families.
Hollingworth et al examined linkage using the combined NIMH and UK cohort described earlier. In the NIMH sample, they found significant linkage on chromosome 15 between markers D15S655 (85.7 Mb) and D15S652 (90.3 Mb), with a psychosis covariate maximum LOD score at 86 cM. Significant linkage was also found on chromosome 7 with its nearest marker at D7S2204 (78.0 Mb) and a psychosis covariate maximum LOD score at 91 cM. They also found regions of increased linkage on chromosomes 6 and 21 that were close to those previously reported by Bacanu et al. However, the main effect at both loci appeared to be due to inclusion of APOE genotype as a covariate, with a smaller effect of psychosis.
Neuregulin -1(NRG1) is a gene on chromosome 8p that has shown both linkage and association with psychosis due to schizophrenia. Go et al examined 437 families from the NIMH AD Genetics Initiative to determine association and linkage of NRG1 with AD and AD+P. Linkage analysis of chromosome 8 in AD yielded a maximum Zlr score of 2.0, p<0.05, in the 8p11-p12 region containing NRG1. Of the 437 families, 65 families had a proband that was positive for multiple psychotic symptoms as was at least one other sibling. When this subset of families was analyzed, linkage in the 8p11-p12 region increased to a Zlr of 4.2, p=0.000016.
Go et al, followed up on this linkage signal, analyzing the transmission of four SNPs within NRG1 in the AD+P families. The SNPS analyzed included three from the 5′ region of NRG1 strongly implicated in schizophrenia (SNP8NRG24190, SNP8NRG221533, and SNP8NRG243177). An additional exonic SNP, rs3924999 was also tested. In single SNP analyses, rs3924999 demonstrated evidence of significant linkage and association with AD+P. Haplotype analyses indicated that a three SNP haplotype defined by SNP8NRG221533, SNP8NRG243177, and rs3924999) tended to be over-transmitted to siblings with AD+P (p=0.076).
Whereas the studies by Bacanu, Hollingworth and Go showed linkage of chromosomes 2p, 15 and 7, and 8p, respectively, with psychosis in AD, Avramopoulos et al found that chromosome 14q is linked to AD patients without hallucinations. They genotyped 348 individuals from 148 families of the NIMH AD Genetics Initiative and used the presence or absence of hallucinations or delusions as covariates. The majority of these patients had late-onset AD, but Avramopoulos also included some individuals with ages of onset between 50 and 65. They found a linkage to a locus on chromosome 14q24.3 (LOD = 3.91; genome-wide p=0.052), close to the PSEN1 locus related to the absence of hallucinations. Screening of PSEN1 for mutations did not detect any coding region or splice site mutations. They also found some support for the findings of Bacanu et al, with evidence of elevated linkage to the presence of delusions on chromosome 2p.
An alternate approach to assessing the genetic basis of AD+P is to examine whether specific genetic variations are more frequent in AD+P cases than in unrelated AD−P samples. A number of such studies have been reported in AD+P, predominantly focused on APOE and genetic variation within components of monoamine neurotransmitter systems.
The gene for APOE is located on chromosome 19 and is well documented as being a risk factor for the development of late-onset AD[7,47]. However, its association with development of psychosis is less clear. To date, at least 22 studies have examined the relationship between the є4 allele and the presence of psychosis in AD, with nine reporting that є4 is significantly associated with AD+P (See Table 2). These discrepancies may be due to differences in sample sizes (with both false positive and false negative findings more likely in small samples), patient populations, and diagnostic criteria. As a whole, however APOE does not appear to contribute to the risk of psychosis in AD.
Serotonin (5-hydroxytryptamine, 5-HT) is a central nervous system neurotransmitter important in regulating mood, memory, learning, sleep, and appetite. Altered serotonin neurotransmission has also been suggested to contribute to various mood disorders such as depression, obsessive-compulsive disorder, schizophrenia, and with psychosis seen in patients with AD[68,69]. In addition, postmortem and biopsy studies of AD patients’ brains have shown a decrease in the levels of 5-HT, of 5-HT receptors (HTR), and of the 5-HT transporter (SLC6A4)[70,71,72]. A number of investigations, therefore, have looked for an association of genetic variation in serotonin receptors HTR2A and HTR2C, and SLC6A4 with AD+P.
The association of polymorphic variation of the HTR2A gene with AD+P has focused on a single SNP (rs6313), which results in a T(102)C substitution. Both alleles code for a serine in codon 34, and therefore, do not alter the amino acid sequence of the HTR2A receptor. However, this polymorphism has been associated with changes in HTR2A expression. In human postmortem studies, mRNA expression of the rs6313 (C) allele in the temporal cortex of heterozygotes was significantly lower than mRNA expression of the rs6313 (T) allele. Similarly, total levels of HTR2A mRNA and protein in individuals with the CC genotype, as well as the combined CC + TC genotype groups, were lower than in individuals with the TT genotype.
The association of the HTR2A T(102)C genotype with psychosis in AD has been studied by several groups (Table 3). Nacmias et al and Rocchi et al found a significant association between allele frequencies and AD+P, with C being the risk allele. Other groups looked specifically for association with delusions and hallucinations with conflicting results. For example, Holmes et al found a significant association of the C allele with hallucinations, whereas Assal et al did not find an association for allele frequency with hallucinations. Furthermore, Assal and Lam found that T alleles increased risk for delusions, but Holmes didn’t find an association with either allele. Wilkosz et al not only determined that there wasn’t a significant association with the presence of psychotic symptoms, but they also found that HTR2A T(102)C genotype was not associated with the time to onset of psychosis. Others, such as Craig et al and Pritchard et al did not find association with either delusions or hallucinations. Though the above data may be consistent with a modest effect of HTR2A genotype on AD+P risk, the opposing risk alleles found in the different studies would suggest that rs6313 is unlikely to be causative.
Relatively fewer studies have examined the association of AD+P with HTR2C. These studies have all evaluated a single non-synonymous SNP in HTR2C, which results in a Cys23Ser substitution, the function of which is not established. Holmes et al found a significant association of allele frequency with visual hallucinations, but not with auditory hallucinations or delusions, whereas Assal et al, and Pritchard et al, did not find a significant association of allele frequency or genotype with delusions or hallucinations (Table 3).
SLC6A4 gene expression is regulated by several functional polymorphisms. One is within the promoter region of SLC6A4 and is named the 5-HTTLPR. The long allele (L) variant of 5-HTTLPR occurs from a 44 base pair insertion and results in greater transcriptional activity, and therefore, serotonin reuptake, than does the short allele (S) variant[85,86]. Several studies have been conducted to test for an association of the 5-HTTLPR polymorphism with AD+P and have yielded conflicting results (Table 3). Sweet et al found that the LL genotype and the L allele frequency were associated with psychosis, but several subsequent studies did not confirm these findings[76,87,88,89]. Furthermore, Borroni et al found that S allele carriers had increased risk for the development of psychosis.
Another functional polymorphism that affects SLC6A4 gene expression is a variable number tandem repeat polymorphism (5-HTTVNTR) located in the second intron containing 9, 10, or 12 copies of a 17 base-pair sequence,. Pritchard et al and Ueki et al examined if this polymorphism is associated with AD+P. Pritchard found that the 10-repeat allele was significantly associated with psychosis, but Ueki did not find an association of genotype with delusions or hallucinations. This difference could be due to the smaller sample size of Ueki’s study (Table3).
Overall, the serotonergic system is a complex one, and its association with AD+P is still unclear. Small sample sizes, non-uniformity in patient populations and diagnostic criteria across studies, and investigation of only limited genetic variation within individual genes makes a definite conclusion regarding these associations difficult.
The possible association of dopamine receptors and AD+P is of interest because many antipsychotic agents used to treat psychosis target these receptors. Dopamine receptors are coded by multiple genes. A limited number of studies have been done looking at the association of the polymorphisms of these genes and AD+P (Table 4). Of these, studies of the dopamine-3 receptor (DRD3) have been of greatest interest because of numerous studies suggesting a small, but significant, association of this gene with the risk for schizophrenic psychosis[94,95]. Three studies have been done looking at the association of a single exonic SNP, coding for a serine to glycine substitution in the DRD3 gene, and AD+P with conflicting results[96,97,98]. Sweet et al found the Ser/Ser and Gly/Gly genotype to be significantly associated with psychosis, Holmes et al found the Gly/Gly genotype to be protective against delusions, but did not find any genotype to be associated with hallucinations, and Craig et al did not find an association of any polymorphism with hallucinations or delusions (Table 4).
Sweet et al also tested the association of a single SNP in the dopamine-1 receptor (DRD1) gene with psychosis, whereas Holmes et al tested its association separately with hallucinations and delusions. Sweet found a significant association for the B2/B2 genotype with psychosis whereas Holmes found a significant association for the B1/B2 genotype with hallucinations, with a trend for association of B2/B2, but did not find any genotype to be associated with delusions. Only one study was done to test the association of the dopamine-2 receptor (DRD2) gene with AD+P. Sweet et al found that Cys311 frequency was not significantly associated with psychosis (Table 4). As with the DRD2 gene, Sweet et al were the only ones to test the association of the dopamine-4 receptor (DRD4) gene with AD+P. They found that neither the long nor short allele of the DRD4 exon III repeat sequence polymorphism was significantly associated with AD+P (Table 4).
Pritchard et al investigated the association of the dopamine transporter gene (SLC6A3) with AD+P (Table 4). SLC6A3 has also recently been implicated by association with psychosis in schizophrenia, including both direct effects and interactions with other dopamine system genes. Specifically, Pritchard et al looked at the variable number tandem repeat (VNTR) polymorphism in the 3′-untranslated region (UTR) of SLC6A3. They did not find a significant association with psychosis, hallucinations, or delusions.
The cathechol-O-methyltransferase (COMT) gene is located on chromosome 22q11.2 and codes for an enzyme that inactivates dopamine along with other catecholamine neurotransmitters. Since COMT is responsible for enzymatically degrading dopamine, a mutation in the gene could lead to altered levels of dopamine, and therefore, psychosis. Deletions within 22q11.2 are associated with velocardiofacial syndrome, a developmental abnormality in which 30% of individuals develop schizophrenic psychosis[102,103,104,105]. Furthermore, although numerous positive and negative studies of the association of a single coding SNP, rs4680, which results in a valine to methionine substitution, have been reported in schizophrenia, emerging understanding of the genetic regulation of COMT suggests that COMT haplotype is a risk for psychosis.
Sweet et al examined the possible association of COMT haplotypes in subjects suffering from AD with and without psychosis (Table 5). They genotyped the same SNPs analyzed for association with schizophrenia by Shifman et al (rs737865, rs165599, and rs4680) and an additional SNP (ERE6) located near an estrogen responsive element within the P2 promoter region, evaluating single locus effects and the multiple loci haplotypes. Because of possible effects of the ERE6 locus, they also stratified analyses by gender. In female subjects, single locus analyses revealed that the risk for psychosis increased significantly with the number of G alleles present at the rs4680 locus. Furthermore, there was a trend toward association with rs737865. None of the SNPs were associated with psychosis in men. Haplotype analysis in females revealed an additive affect of risk alleles at rs4680 and ERE6 (or rs737865). There was a significant linear relationship between the number of ERE6 C alleles and rs4680 G alleles and psychosis, with each locus contributing equally to the risk (rs4680: slope = 0.7905, p=0.0007; ERE6: slope = 0.6720, p=0.0032). In males, an interaction between rs4680, ERE6, and rs165599, was close to significant (p=0.08) for psychosis.
An initial study by Borroni et al looked only at the association of rs4680 with delusions and hallucinations in AD. They found that carriers of the G allele experienced significantly more hallucinations than non-carriers. Borroni et al expanded their analyses of COMT polymorphisms by evaluating the association of four SNPs (rs737865, rs737864, intron 1 C2754delC, and rs4680) with AD+P. An association of the G allele of rs4680 was confirmed for AD+P (Table 5). In addition, haplotype analysis revealed a significant association of haplotypes (A-C-C-G, p=0.044 and G-C-delC-G, p=0.006) for psychosis in AD. The A-C-C-G haplotype further increased the risk by 25% compared to the effect of the rs4680 G allele alone. The G-C-delC-G haplotype further increased the risk by 75% compared to the rs4680 G allele.
Studies with dopamine receptor variation and psychosis in AD have been limited and conflicting. Likewise, studies of COMT have also been limited, although there does appear to be a complex association of COMT haplotypes, including the effects of rs4680, with psychosis. Further work which more thoroughly evaluates the genetic variation within COMT for association with psychosis, and in particular considers the effects of haplotypes known to impact other phenotypes  is needed.
The α7 nicotinic acetylcholine receptor is encoded by CHRNA7 on chromosome 15 and has been found via linkage and association studies to be involved in schizophrenia[111,112]. An initial study by Carson et al looked at whether this gene is associated with psychosis in AD. Analyzing SNPs in this gene in a group of 409 probable AD patients of Northern Ireland descent, they found a significant association with a single SNP (rs6494223) and delusions in AD (p=0.017) with the T allele being the risk factor (OR=1.63, CI=1.22-2.17). This finding has yet to be replicated in other AD populations.
A polymorphism of the interleukin 1β gene promoter was studied in a population of 424 patients diagnosed with possible/probable AD in the United Kingdom. Craig et al found that the CC genotype frequency was significantly higher in patients with delusions (χ2=2.69, p=0.002), with hallucinations (χ2=6.27, p=0.043), and with both delusions and hallucinations (χ2=9.9, p=0.007). The frequency of the C allele was also significantly higher in patients with delusions (χ2=4.86, OR=1.49, CI=1.02-1.94, p=0.028), with hallucinations (χ2=5.95, OR=1.6, CI=1.08-2.39, p=0.014), and with both (χ2=3.91, OR=1.62, CI=0.98-2.70, p=0.048). Once again, independent confirmation of these findings is pending.
If AD+P is heritable, what model may guide future genetic studies of AD+P? Several pathways by which genetic factors may lead to AD+P are presented in Figure 1. Pathway A1 represents disease modifying genes that lead to AD+P in individuals who develop AD itself due to other genetic and environmental influences. This pathway is consistent with certain findings in AD+P. For example AD+P is rare as a prodrome of AD, and occurs with increasing frequency as individuals traverse the early and middle disease stages. Also, as reviewed above, AD+P does not appear to be associated with the one identified genetic risk factor for the development of late-onset AD itself, APOE. Similarly, disease modifying genes may alter the course of other neurodegenerative illnesses to yield psychosis (Pathway A3). There is evidence to support this model in Huntington disease, as psychosis aggregates within families with Huntington disease, but the occurrence of psychosis does not appear to be due to the expansion of the disease causing triplicate repeats in the Huntington disease gene (HD) itself . Some data also exist to support this model in families carrying identified disease-causing mutations in microtubule-associated protein tau (MATP) which results in a frontotemporal dementia. Despite carrying the same disease alleles, families in which some members present with early-onset schizophreniform psychosis have been described, presumably due to an interaction of the disease causing mutation with their specific genetic (and environmental) backgrounds . Finally, as reviewed above, several genes which show some evidence of association with AD+P (NRG1, COMT) have been suggested as putative risk genes for schizophrenic psychosis, possibly indicating they may modify neurodevelopment to yield psychosis risk (Pathway A2).
Alternatively, in pathway B, genes that increase the liability to onset of AD would increase risk for AD+P. Currently, there is little evidence by which to accept or reject this model. For the one gene known to influence risk of late-onset AD, APOE, pathway B does not appear to apply. Another gene with substantial (but not conclusive) evidence of association with late-onset AD, SORL1, has not been studied for association with AD+P. Finally, despite evidence that rates of psychiatric illness, including schizophrenia, may be higher among family members of individuals with early-onset versus late-onset AD , studies of genes known to contribute to early-onset AD for association with psychosis (e.g. APP, PSEN1, PSEN2) have not been reported.
While it is evident that psychosis in AD is a distinct phenotype with a genetic basis, the inconsistent findings of genetic association studies of AD+P indicate that many challenges still face researchers attempting to elucidate the genetics behind this syndrome. These challenges group into two main domains: reducing variability in the phenotypic definition and reducing variability due to the approaches used for genetic analysis. We briefly address these key issues below.
As mentioned earlier, there is substantial variability between studies in the rates of psychosis reported in AD patients . To the extent this reflects inaccurate classification of individuals, it will either reduce power to identify true associations, or possibly inflate the false positive rate of genetic association studies. While variable classification of subjects with regard to psychosis presence may reflect multiple factors, in practice two design variables prove most important. First, the prevalence of psychosis in subjects with AD increases from the earliest stages of cognitive impairment to reach a maximum value by the middle stages of illness . Thus, studies would do best to only consider an individual accurately characterized as without psychosis if they have reached at least this stage of illness. Second, individuals may have brief psychotic symptoms due to a number of factors, including intercurrent medical illness or drug toxicity. The likelihood of detecting one of these events increases with the use of frequent, systematic screening for psychosis symptoms. One way to avoid the reduction in study power introduced by these phenocopies is to use more stringent requirements for the presence of psychosis. For example, we have found that estimates of familial aggregation and heritability are increased if multiple and/or recurrent symptoms of psychosis have been reported .
Several factors contribute to both false-positive and false-negative findings in genetic association studies of candidate genes. Perhaps the most important is small sample size. It seems obvious that too small a sample may cause the failure to detect a true association due to lack of statistical power, what is less obvious is that small sample sizes not uncommonly lead to false positive findings due to random chance alone. Because these findings are not stable, they result in failed replications. While there is no absolute standard, experience with the use of genome-wide association studies in other complex disorders has suggested the correct sample sizes for association tests number in the thousands, not the tens or hundreds.
Another factor which can further contribute to variation in findings in association studies includes problems of population stratification. This occurs when the affected (e.g. AD+P) and unaffected (e.g. AD−P) groups have different allele frequencies due to unidentified differences in genetic ancestry. This problem can be handled in a straightforward way by genotyping a set of ancestry SNPs within all subjects and controlling for ancestry in the analysis of the variants of interest .
The association studies contained within this review have mostly relied on testing a single SNP within genes of interest. Failure to find association may lead to the false-negative conclusion that the gene of interest does not contribute to AD+P risk, a conclusion that could be avoided with more systematic interrogation of genes. Currently it is reasonable to attempt complete coverage of all haplotype blocks (defined by elevated correlation of SNPs within a block) within a gene, using for each block a representative (tag) SNP. Study of single SNPs can also lead to non-replicated positive findings. If the associated SNP is not causal, but only associated due to its correlation with an unassessed genetic variant which does contribute to disease risk, two different situations may arise. Because these correlations can differ across populations, subsequent studies may find either a significant association with the same SNP, but with the opposite allele identified as the risk allele, or may not replicate any association at all.
The choice of genes of interest that have been tested in association studies of AD+P is also problematic. These genes have been chosen based on inference about their potential relationship to psychosis. Such an approach has no power to identify true associations unless the set of genes examined have a causal relationship to AD+P. An alternate approach currently being used with success in other complex phenotypes of aging (e.g. prostate cancer and macular degeneration) is genome-wide association . This approach interrogates SNPs from across the entire genome for association with disease. While not biased by mistaken assumptions of causality, this approach is highly dependent on the presence of large samples, typically many thousands of subjects.
Psychosis occurs in a subset of patients with Alzheimer Disease, in whom it is associated with a more aggressive cognitive deterioration and worse outcomes. There is growing evidence that psychosis in AD aggregates within families and is likely to result, in part, from effects of genetic variation. Association studies with a number of candidate genes have been reported. The most studied has been the APOE gene with most reports failing to find an association with AD+P. However, even in this case no firm conclusion can be made as to whether the APOE є4 allele is associated with development of AD+P as most individual studies have been small, and there is substantial variability in phenotypic characterization and analytic approach between reports. A number of studies have evaluated putative schizophrenia risk genes, with intriguing initial findings for both NRG1 and COMT. These will require confirmation in larger cohorts who are genetically characterized for ethnicity and with more thorough interrogation of the genetic variation present in these genes. Finally, if sufficiently large cohorts can be homogenously collected and characterized, consideration of alternate approaches to identifying genetic variants association with AD+P risk, such as Genome-Wide Association Studies should be considered.
This work was supported by USPHS grant AG 027224 and AG 05133, and there are no potential conflicts.