Introduction

Paired-like homeobox 2b (PHOX2B, also known as PMX2B or NBPhox) is a homeodomain transcription factor, and is known to determine noradrenergic phenotype (Pattyn et al. 2000) and play a role in the development of cranial motor nerves, including the oculomotor (nIII) and trochlear (nIV) nerves (Pattyn et al. 1997) controlling ocular alignment and movement. As a transcription factor, PHOX2B regulates the expression of tyrosine hydroxylase (TH) and dopamine β-hydroxylase (DBH) genes. TH catalyzes the conversion of L-tyrosine to L-dihydroxyphenylalanine (L-DOPA), a precursor of dopamine, and DBH catalyzes the conversion of dopamine to noradrenaline. The protein structure of PHOX2B is characterized by two homopolymeric stretches of alanine residues: one consisting of nine alanines located downstream of the homeodomain; the other comprising 20 alanines (Ala20) on the C-terminal side (Fig. 1). Our prior genomic screening of PHOX2B identified frequent length variations in the Ala20 stretch in the general population, representing an unusual phenomenon compared with other polyalanine-containing transcription factors (Toyota et al. 2004). Variations included −3Ala, −5Ala, −7Ala, −13Ala and +2Ala. These alterations in alanine length resulted in decreased transcriptional ability of the protein and represented the only functional polymorphisms found in the gene. In accordance with the known function of PHOX2B and the functional consequences of these variations, associations between the polymorphisms and general schizophrenia were detected, particularly for schizophrenia manifesting with strabismus (ocular misalignment) (Toyota et al. 2004). That study also raised a possibility of interactions between PHOX2B and other schizophrenia-precipitating factors (genes) for increased risk of the combined phenotype (Toyota et al. 2004).

Fig. 1
figure 1

Schematic representation of the PHOX2B (NM_003924) (above) and ASCL1 (NM_004316) genes (below). Exons are boxed, and initiation and stop codons and protein domains are indicated

Full size image

Human achaete-scute homologue 1 (HASH1; ASCL1 in HUGO nomenclature), a human orthologue of mouse Mash1, is a basic helix–loop–helix (bHLH) transcription factor that is known to co-regulate differentiation of the autonomic system along with PHOX2B (Pattyn et al. 2000). Cross-regulation by the Phox2 and Mash1 genes, and the importance of the HASH1-PHOX pathway in the development of neurons in the noradrenergic lineage have been demonstrated in both mice (Pattyn et al. 1999, 2000), and a human disease mechanism (De Pontual et al. 2003). We therefore speculated that PHOX2B and ASCL1 may affect predispositions to broad catecholamine-related diseases both separately and in combination. The present study examined genetic associations between PHOX2B and ASCL1 and schizophrenia, bipolar disorder and Parkinson’s disease (PD).

Materials and methods

Study subjects

Subjects included 715 schizophrenic patients (394 men, mean age 48.3±12.3 years; 321 women, mean age 50.7±13.3 years), 249 bipolar disorder patients (118 men, mean age 52.6±13.2 years; 131 women, mean age 55.8±12.9 years), 100 PD patients (32 men, mean age 67.3±7.8 years; 68 women, mean age 67.8±7.0 years) and 801 healthy controls (369 men, mean age 40.9±11.4 years; 432 women, mean age 41.3±13.7 years). Compared with the prior study (Toyota et al. 2004), the number of schizophrenia patients was increased by 369 and the number of controls was increased by 260, but these newly added subjects were not screened for strabismus. All subjects were recruited from a geographic area located in central Japan. Diagnosis of schizophrenia and bipolar disorder was based on the Diagnostic and statistical manual of mental disorders (American Psychiatric Association 1994). PD was diagnosed according to the standardized criteria. All PD patients underwent brain computed tomography examination to exclude organic abnormalities. Control subjects were recruited from hospital staff and company employees who were documented as free of psychoses or any kind of neurodegenerative disorder. None of the current subjects displayed mental retardation or congenital central hypoventilation syndrome (De Pontual et al. 2003). This study was approved by the Ethics Committees of RIKEN, Hamamatsu University and Juntendo University, and all subjects provided written informed consent to participate.

Mutation screening of ASCL1

ASCL1 is located on human chromosome 12q22-q23 (Renault et al. 1995) and comprises two exons, with the first exon including both the initiation and stop codons (Fig. 1). The protein-coding region contains a polyalanine stretch comprising 13 alanines, and a polyglutamine tract of 12 glutamine residues (Gln12), in addition to the bHLH. The two exons and their flanking genomic stretches were screened using polymerase chain reaction (PCR) amplification and subsequent direct sequencing of genomic DNA from 24 randomly chosen patients. Sequencing was performed using a DYEnamic ET terminator cycle sequencing kit (Amersham, Piscataway, N.J., USA). Information on primer sequences and PCR conditions employed in this study is available on request. Screening detected the insertion of three CAG repeats (coding glutamine) into the polyglutamine stretch. This was the only non-synonymous polymorphism identified, and we therefore focused on this Gln12 length polymorphism in subsequent analyses.

Genotyping

Genotyping of Ala20 length variations in the PHOX2B was performed according to the methods described elsewhere (Toyota et al. 2004). To genotype Gln12 polyglutamine length variations in ASCL1, template DNA was amplified using fluorescently labeled forward (5′-AGCTCTGCCAAGATGGAGAG; 3′ end at nt c.26) and reverse (5′- gtttcttTTGCTTGGGCGCTGACTTGT; 3′ end at nt c.236) primers. The underlined tail sequence was added because Taq DNA polymerase catalyzes the non-templated addition of adenosine to the 3′ end of PCR products to varying degrees. This phenomenon is primer-specific and represents a potential source of genotyping error. Placing the gtttctt sequence at the 5′ end of reverse primers produces nearly 100% adenylation of the 3′ end of the forward strand, facilitating accurate genotyping (Brownstein et al. 1996; Itokawa et al. 2003). PCR products were run on an ABI 3700 genetic analyzer (Applied Biosystems, Foster City, Calif., USA), and the resulting data were analyzed using GeneScan software (Applied Biosystems). Genotypes were confirmed by subcloning the amplicons into a TA vector (Invitrogen, Carlsbad, Calif., USA) and sequencing. Primers were designed to produce a 249-bp DNA fragment for the wild-type allele (Gln12), but GeneScan analysis yielded a band approximately 14 bp shorter than expected (Fig. 2a), with occasional inconsistent genotype results compared with those obtained by subcloning, which could not be resolved by applying the constant 14-bp difference to GeneScan results. This phenomenon was attributed to the secondary DNA structure generated by abundant GCs in the PCR products (Toyota et al. 2004). When 7-deaza-2′-deoxyguanosine triphosphate (c7 dGTP) was added to the PCR reaction mixture (c7 dGTP:dGTP=1:3) to breakdown hydrogen bonds in the GC-rich templates, GeneScan peaks were broadened and two adjacent peaks merged (Fig. 2b). We replaced all dGTP in the PCR reaction mixture with c7 dGTP, and obtained sharp and correctly sized bands, enabling accurate genotyping (Fig. 2c).

Fig. 2a–c
figure 2

GeneScan migration patterns of ASCL1 Gln12 length polymorphisms. DNA fragments with Gln12 or Gln13 genotypes were run after PCR under varying concentrations of c7 dGTP. Exact sizes of the Gln12 and Gln13 alleles were 249 and 252 bp, respectively. a The c7 dGTP was not added to the PCR mixture. Note that displayed allele sizes were 14 bp shorter than actual sizes. b Addition of c7 dGTP to 25% resulted in fusion of the two peaks. c When all dGTP in the PCR reaction mixture was replaced with c7dGTP, peaks appeared at expected sizes with good separation of the two adjacent alleles

Full size image

Statistical analysis

Associations of either PHOX2B or ASCL1 polymorphisms with each neuropsychiatric disorder were evaluated using the Monte–Carlo method implemented in the CLUMP program (T1–T4 modes; number of simulations set to 10,000; random number seed, 100) (Sham and Curtis 1995) or Fisher’s exact test when appropriate. Rare alleles or genotypes showing frequencies of <1% in both comparison groups were removed from the analysis. Hardy–Weinberg equilibrium was evaluated using Arlequin software (http://lgb.unige.ch/arlequin/) (Schneider and Excofier 2000). Logistic regression analysis in the SPSS Regression Models software (SPSS Japan, Tokyo, Japan) was performed to test the joint effects of the two genes. Letting P represent the probability of an individual being a case rather than a control, we modeled P as

$$ {\text{log}}\,{\text{it}}(P) = \beta _{0} + {\sum\limits_{i = 1}^4 {\beta _{i} x_{i} } } + {\sum\limits_{i = 1}^2 {{\sum\limits_{j = 3}^4 {\beta _{{ij}} x_{i} x_{j} } }} } $$

where x1, x2, x3 and x4 represent covariants depending on the genotypes of the individual, β0 is the intercept, and βi and βij are coefficients to be estimated. When applied to the formula, genotypes were dichotomized into two groups: wild-type (w); and mutant (m). Following the approach of Cordell and Clayton (2002) for the possible genotypes of w/w, m/w and m/m, we coded −1, 0 and 1, respectively, to represent the additive effects of allele m and −0.5, 0.5, −0.5, respectively, to represent the dominant effect of allele m over allele w.

Results

Table 1 shows the results of association analyses between PHOX2B Ala20 length polymorphisms and the three disease categories. We detected six different genotypes, and distributions of genotypes in each group were all in Hardy–Weinberg equilibrium. None of the modes T1–T4 on CLUMP analysis displayed significant associations for any disease groups. The number of different alleles observed in this study was the same as in our previous study (Toyota et al. 2004), although much larger cohorts were examined here. Again, no allelic associations were detected for any of the three neuropsychiatric disorders.

Table 1 Genotypic and allelic distributions of the PHOX2B Ala20 repeat polymorphism
Full size table

Tables 2 and 3 show the results of genotypic and allelic analyses of ASCL1 Gln12 stretch polymorphisms, respectively. Analysis of the 1,866 subjects yielded 13 different length variations in the Gln12 homopolyer repeat region of ASCL1. These polymorphisms were not genotypically associated with schizophrenia or bipolar disorder, but displayed associations with PD (P<0.05 in T2, T3 and T4) (Table 2). Allelic analysis demonstrated that the allele containing 12 glutamine repeats, the most common of these alleles, was more frequent in PD than in the control group (2×2 Fisher’s exact test, two-sided, P=0.015; odds ratio=1.68, 95% CI=1.10–2.54), while the allele containing 15 glutamine repeats, as the second most common allele, exhibited an opposite distribution pattern (P=0.011; odds ratio=0.57, 95% CI=0.36–0.89) (Table 3). These results suggest that the ASCL1 allele harboring 15 glutamine repeats may play a protective role against PD manifestation.

Table 2 Genotypic distribution of the ASCL1 Gln12 repeat polymorphism
Full size table
Table 3 Allelic distribution of the ASCL1 Gln12 repeat polymorphism
Full size table

Logistic regression analysis was then performed to test the joint effect of the two genes on PD. The Ala20 allele of PHOX2B and the Gln12 allele of ASCL1 were classified as w, with the remaining alleles as m. As a result, only the effect of ASCL1 dominant × PHOX2B additive was found to be significant (P=0.008), among the effects of all possible interaction modes (Table 4).

Table 4 Logistic regression analysis of effects of PHOX2B and ASCL1 genes on Parkinson’s disease
Full size table

Discussion

PHOX2B/ASCL1 and psychiatric disorders

We have previously reported genotypic associations between Ala20 polymorphisms in PHOX2B and overall schizophrenia (P=0.012), with a more prominent association for schizophrenia with strabismus (P=0.004) (Toyota et al. 2004). However, the present study did not detect this association in a larger case-control panel with a 2.2-fold increase in the schizophrenia population and a 1.6-fold increase in control samples. This discrepancy may be partly due to the fact that prior control samples had undergone ocular examinations, and only those subjects who did not suffer from strabismus were chosen, while the present study used control samples without determining the presence of ocular misalignment. The newly added schizophrenic samples in this study were also not screened for ocular misalignment. While the genetic contributions of PHOX2B Ala20 variations to general schizophrenia are more likely to be very weak or even negligible, even by considering genetic interactions with ASCL1 (data not shown), these contributions may be evident only in a subset of schizophrenia (i.e., schizophrenia with strabismus). As might be expected according to this hypothesis, no association was apparent between PHOX2B and schizophrenia without strabismus (P=0.076) in our previous study (Toyota et al. 2004). We also tested here ASCL1 as a singleton or PHOX2B-ASCL1 epigenetic interaction (data not shown) for altered risk of another major psychosis, bipolar disorder, but no significant signals were detected. As a whole, the current results do not support these genetic mechanisms in the manifestation of functional psychoses.

PHOX2B/ASCL1 and Parkinson’s disease

PD is a common neurodegenerative disorder, characterized clinically by resting tremor, rigidity and bradykinesia. Neuropathological studies have revealed degeneration of the dopamine-producing substantia nigra and various other regions, including the basal ganglia, brainstem, autonomic nervous system and cerebral cortex (Dekker et al. 2003). Clinically defined PD represents an etiologically heterogeneous group of conditions encompassing a small population of individuals with Mendelian-type inheritance and a larger population of apparently sporadic cases (Hattori et al. 2003). Accumulating evidence has suggested that genetic predispositions exist even for sporadic PD (Marder et al. 1996). Dopamine deficiency is a primary pathomechanism in PD, and genes involved in dopamine neurotransmission, such as those for dopamine transporter, dopamine receptors, tyrosine hydroxylase, catechol-it O-methyltransferase and monoamine oxidase, have been examined in population-based association studies over the past decade. However, few of these genes have been definitively established as conferring susceptibility to sporadic PD (reviewed in Warner and Schapira 2003).

Perturbation of PHOX2B and ASCL1 function has the potential to disturb catecholaminergic neurons, as these genes control the expression of the TH and DBH genes, which encode enzymes for the biosynthesis of dopamine (TH) and noradrenalin (TH and DBH) biosynthesis. Ludecke et al. (1996) reported a female infant who manifested L-dopa responsive Parkinsonism and carried a Leu205Pro mutation in exon 5 of the TH gene, reducing the catalytic ability of TH. The current study identified a positive association between PD and ASCL1 polymorphisms. However, whether these ASCL1 variants result in a predisposition to PD through direct effects on dopamine neurons remains unclear, as ASCL1 expression in the human substantia nigra has not yet been confirmed. In contrast, expression of ASCL1 in developing noradrenergic neurons in the human brainstem (locus coeruleus: LC) has been reported (De Pontual et al. 2003). The LC is known to play an important role in the pathophysiology of PD (reviewed in Gesi et al. 2000). Zarow et al. (2003) found more severe neuronal loss in the LC than in the substantia nigra in a postmortem examination of brains from PD patients. Mavridis et al. (1991) demonstrated that monkeys with LC lesions displayed impaired recovery from Parkinsonism induced using 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP). Other studies have also shown that animals with LC lesions exhibit marked dopamine loss on administration of MPTP or methamphetamine (Bing et al. 1994; Fornai et al. 1997). These data suggest a protective role of the LC against the development of PD. Indeed, Srinivasan and Schmidt (2004) reported that the enhancement of noradrenergic transmission in the LC by β2-adrenoceptor antagonists exerts a prophylactic effect against 6-hydroxydopamine-induced Parkinsonism. The present finding that the ASCL1 allele containing 15 glutamines is less represented in PD than in controls might suggest that the 15-repeat allele could confer protective benefits compared to the most common 12-repeat allele, perhaps allowing the development of a well-functionalized LC that in turn helps to protect the substantia nigra from various insults.

Because of the presumed multigenic nature of complex traits, it would be desirable to analyze several polymorphisms jointly and investigate their effects and possible interactions on disease outcome (Ott 2001). One of the statistical methods that can be used to resolve this problem is logistic regression analysis. When applied to the current data, this analysis indicated that the dominant effect of ASCL1 with the additive effect of PHOX2B was positive. The biological consequences resulting from the interaction between ASCL1 and PHOX2B might thus offer useful insights into the pathogenesis of PD. Further studies elucidating the detailed mechanisms of this interaction are thus warranted.

Polyglutamine length variations in ASCL1

Polyglutamine expansion has been found in various neurodegenerative disorders, including Huntington’s disease, spinocerebellar ataxia types 1, 2, 3 and 7, dentatorubral-pallido-luysian atrophy and spinobulbar muscular atrophy (Lipinski and Yuan 2004). The aggregation or accumulation of proteins with expanded polyglutamine sequences is considered to represent a critical contribution to neurodegeneration in these diseases. Generally these aggregate-forming proteins display more than 30 glutamine repeats, while ASCL1 displays repeats of less than 20 glutamines. None of the Gln12 length variations for ASCL1 detected in this study are thus likely to exert deteriorative effects on neurons. However, the functional consequences evoked by variations of the polyglutamine stretch in ASCL1 are yet to be examined.

In summary, we performed an association study for PHOX2B and ASCL1, genes that are functionally closely related and display imperative roles in the development of neurons in the noradrenergic (dopaminergic) lineage, in three major neuropsychiatric diseases. Significant contributions of ASCL1 and ASCL1-PHOX2B interactions to PD were detected. These results require genetic replication studies in different populations and further biological investigations to clarify the precise mechanisms and effects.