Possible association between common variants of the phenylalanine hydroxylase (PAH) gene and memory performance in healthy adults

Accumulating evidence suggests that genetic factors influence human complex behavior, including neurocognition throughout the life span [1, 2]. Studies of single genes that cause monogenic diseases with cognitive impairments will advance understanding of genetic and molecular mechanisms underlying individual differences in neurocognitive performance [3]. Phenylketonuria [PKU; McKusick OMIM 261600], an inborn error of phenylalanine metabolism, causes intellectual disability (i.e., developmental disorder with general cognitive disabilities), seizures, microcephaly, motor disorder, skin rashes and other symptoms, including psychiatric conditions such as depression [4‐6]. About 98% of PKU is caused by mutations in phenylalanine hydroxylase (PAH) on chromosome 12q. PAH encodes phenylalanine hydroxylase [EC 1.14.16.1] that converts phenylalanine to tyrosine which is a precursor of the neurotransmitter dopamine and other catecholamines [5, 6]. More than 500 mutations in PAH were registered as responsible for PKU (http://​www.​pahdb.​mcgill.​ca/​) [6].

Interestingly, previous studies [7, 8] reported that polymorphisms in PAH confer susceptibility to schizophrenia, a disease characterized by pervasive neurocognitive deficiencies including distinct memory impairment [9], as well as with psychiatric symptoms.

Based on these observations, we hypothesized that common genetic variations in PAH might be associated with neurocognitive function, memory performance in particular. The aim of the present study was to examine the possible association of single nucleotide polymorphisms (SNPs) of PAH with memory performance in healthy adults. To our knowledge, there has been no report that examined such an association.

There were no significant group differences in age or smoking status between the genotype groups (e.g., rs2037639 AA/AG/GG) (Additional file 1: Table S1). With respect to education years, a nominally significant group difference between the genotype groups was observed for rs2037639 (Additional file 1: Table S1). Therefore in subsequent analyses, sex and education years were controlled for. Characteristics of the 6 tag SNPs are shown in Table 1. No significant deviation from the HWE was found for the selected 6 markers. Table 2 shows mean scores of the 5 indices of WMS-R and nominal p values obtained by ANCOVA, in which memory index was a dependent variable, genotypes and sex were independent variables and education years was a covariate. We found 5 out of 30 tests (6 SNPs × 5 indices of WMS-R) were nominally significant without correction for multiple testing. Interestingly, all 5 tests were associations with verbal memory index. Even after Bonferroni correction for multiple testing, the association between rs2037639 and verbal memory index remained significant (p = 0.0013 <0.0017 = 0.05/30). In addition, we compared indices of WMS-R in only nonsmokers to avoid the possible confounding effects of smoking. ANCOVA controlling for sex and education years showed that the association between rs2037639 and age-corrected verbal memory index of WMS-R was significant for multiple testing (p = 0.00024 < 0.0017, F = 8.5). Therefore, we do not think that the association between rs2037639 and age-corrected indices of WMS-R was attributable to smoking status. In subsequent analyses, we focused on the effect of rs2037639 on verbal memory. Figure 1 shows mean verbal memory scores depending on genotypes of rs2037639. There was a significant difference in mean verbal memory score across the three genotype groups (F = 6.7, df = 2, p = 0.0013) (AA genotype: 112.0 ± 12.6, n = 264; AG: 111.3 ± 14.2, n = 254; GG: 106.0 ± 16.0, n = 75). In paired comparisons, subjects with GG genotype showed a significantly lower mean verbal memory score, compared with subjects with AG (F = 5.4, df = 1, p = 0.021) and AA genotype (F = 13.4, df = 1, p = 0.00029). No significant difference was observed in mean verbal memory score between individuals with AA and AG genotype (data not shown). When the AG and AA genotype groups were combined, ANCOVA controlling for sex and education years showed that the GG genotype group was significantly lower in mean verbal memory score than the AA/AG group (106.0 ± 16.0 vs. 111.7 ± 13.4; F = 10.9, df = 1, p = 0.00099).

Figure 2 represents LD coefficients (D’) and constructed two haplotype blocks, block 1 (rs1722387 and rs3817446) and block 2 (rs2037639 and rs10860936), for the 6 SNPs of PAH obtained by the Haploview software. Haplotype analyses for the constructed two haplotype blocks (blocks 1 and 2) using the PLINK software revealed that haplotype block 2, which contained two markers (rs2037639-rs10860936), was significantly associated with verbal memory index (permutation global p = 0.0091, Table 3).

We examined whether common genetic variations of PAH are associated with memory performance in Japanese healthy adults. To our knowledge, this is the first attempt to examine such an association. We found a nominally significant association of verbal memory index of WMS-R with 5 SNPs (rs1722387, rs3817446, rs2037639, rs10860936 and rs11111419) of the 6 tag SNPs in PAH. The association between SNP rs2037639 and verbal memory remained significant after correction for multiple testing. Haplotype-based analysis revealed that a haplotype containing two markers (rs2037639-rs10860936) was significantly associated with verbal memory index. These results suggest that common functional variants in PAH impact verbal memory performance in healthy adults. Alternatively, there might be a possibility that functional variants in other genes nearby, which would be in linkage disequilibrium with rs2037639 or the haplotype, may play a role. Since there were no PKU patients in our sample, the obtained result cannot be ascribed to mutations responsible for PKU. Further, considering the carrier frequency for PKU of about 1/150 that corresponds to the PKU incidence among Japanese of 1/70000 to 1/ 120000 [21, 22], the number of individuals heterozygous for a rare mutation responsible for PKU is estimated to be less than 5 in our sample. Therefore, it is unlikely that the obtained results are attributable largely to such rare mutations.

Of note, the SNP rs2037639 was previously reported to be associated with schizophrenia in a male Bulgarian sample (nominal p = 0.03) [7]. Since verbal memory performance is one of the most disturbed neurocognitive functions in schizophrenia, it is possible that the SNP confer susceptibility to schizophrenia, at least in part, via affecting memory performance, an endophenotype of the disease.

There is evidence that rare PKU mutations in PAH are related with impaired working memory in humans [23, 24] and chemically induced (ethylnitrosourea, ENU) mouse model of PKU (Pah^enu2) [25]. However, to our knowledge, there is no report focusing the relationship between common polymorphisms of PAH and cognition. Since there is no direct evidence of functional effects of the SNPs, the mechanism underlying the relationship is currently unclear. The SNP (rs2037639) may have some functional effects such as regulation of transcription of PAH. Alternatively, the SNP might be in linkage disequilibrium with an unknown functional polymorphism. The functional effect could alter phenylalanine level, which then alters tyrosine level and activation of tyrosine hydroxylase. Finally the PAH genotypes could affect dopamine level in the brain [26, 27], which is related with memory [28]. In addition, neurotransmitter serotonin was also reportedly found to be decreased in PKU [29‐31], possibly being accounted for by inhibitory effect of phenylalanine on the activity of tryptophan hydroxylase [27]. Serotonin also affects dopamine metabolism and serotonin itself is related with human neurocognition.

Previous genetic studies have shown an association of neurocognition with genes encoding enzymes regulating catecholamine metabolism, such as catechol-O-methyltransferase (COMT) and monoamine oxidase (MAO). Krach et al. reported that the functional COMT Val158Met polymorphism links neural activation pattern during episodic memory tasks [32]. In their study, 84 healthy subjects performed a memory encoding and a retrieval task while they underwent fMRI scanning. Bilateral insula and anterior hippocampus activations were increased linearly with the number of Met alleles (Val/Val [low PFC dopamine] < Val/Met < Met/Met [high PFC dopamine]) during memory encoding. Giakoumaki et al. found that tolcapone, the COMT inhibitor, enhanced working memory assessed by the letter–number sequencing tasks in healthy subjects with the Val/Val genotype [33]. On the other hand, Enge et al. found that the MAOA variable number of tandem repeats (VNTR) polymorphism, which is located in its promoter region and controls the transcriptional activity of the gene, was associated with working memory-related performance, indicating that n-back performance measured by reaction time was poorer in individuals with MAO-L (the short allelic variant and lower enzymatic activity) than in individuals with MAO-H (the long allelic variant and higher enzymatic activity) [34]. Since PAH converts phenylalanine to tyrosine which is a precursor of dopamine and other catecholamines, functional PAH variants could affect synthesis of catecholamines.

There are limitations in the study. The present sample size may have been small, which is subject to type II errors. When we performed a post-hoc power analysis to evaluate the statistical power for the association analysis between verbal memory index and the PAH rs2037639 genotypes, the current sample size had a 80% power (β = 0.2) to detect mean difference of 3.3, 4.9, and 5.4 between AA vs. AG, AA vs. GG, and AG vs. GG, respectively, at the α (two-tailed) = 5% level. Further studies with a larger sample size might be required to detect smaller differences. The mechanism by which polymorphisms of PAH influences memory performance is currently unclear, i.e., there is no direct evidence of functional effects of the examined SNPs. Future animal and human studies are therefore warranted to elucidate the mechanism.

We performed an association study between genetic variations of PAH and memory performance assessed with WMS-R in Japanese healthy adults. We obtained evidence suggesting that genetic variation of PAH (SNP rs2037639 and the haplotype containing two markers, rs2037639 and rs10860936) are associated with verbal memory. Considering the carrier frequency for rare mutations responsible for PKU, our data suggest that common functional polymorphisms in PAH impact verbal memory. Alternatively, functional polymorphisms in other genes nearby may play a role in genetic basis for verbal memory.