Autism spectrum disorder (ASD) can be present in children with inborn errors of metabolism, the latter having a prevalence of 1 in every 800 live births [1
]. Less than 5% of ASD cases can be attributed to routinely-screened-for inborn errors of metabolism [2
]. Creatine deficiency syndrome (CDS) is an inherited metabolic disorder that is not included in routine newborn screening panels or in standard ASD diagnostic work up. The presence of autistic symptoms in patients with CDS suggests that genes in the creatine metabolic pathway may play a part in the neurobiology of ASD and may represent a treatable cause of ASD [4
]. We recently ascertained the prevalence of CDS in children with non-syndromic autism in a large, prospective, multicentre study by screening for creatine metabolites in urine and sequencing GAMT
genes (glycine amidinotransferase, guanidinoacetate methyltransferase and solute carrier family 6 member 8) for pathogenic mutations (443 children with ASD). The estimated prevalence of CDS was less than 7 in 1000 and there was no obvious correlation between pathogenic CDS gene mutations and ASD, Therefore, the chances of a child with ASD having a CDS were very low [5
In the study presented here, we investigate the association of autism with non-CDS disease causing variants in the three creatine metabolism genes. Whereas non-synonymous nucleotide changes resulting in amino acid changes are likely to result in pathogenic mutations, synonymous changes that do not alter the amino acid are often attributable to polymorphisms. It has been shown that single nucleotide polymorphisms (SNPs) can explain up to 40% of the variance in liability to ASD [6
]; thus common genetic polymorphisms can be a major risk factor for autism.
The clinical presentation of CDS is characterized by developmental delays and cognitive decline, intellectual disability, impaired speech and language, autistic behaviour, epileptic seizure activity, treatment-refractory epilepsy, and extrapyramidal movement disorders [8
]. Significant behavioural, social and communication challenges seen in ASD can be caused by neurodevelopmental disabilities of complex aetiology. The symptoms in patients with CDS and those with ASD can overlap.
In patients with CDS, the impaired synthesis or transport of creatine leads to depletion of creatine/phosphocreatine in the brain. The creatine–phosphocreatine system plays an essential role in cellular energy homeostasis in the central nervous system. This system acts as a temporal and spatial energy buffer, energy transducer and general regulator of cellular energetics [12
]. Creatine has been shown to act as a neuroprotective agent [13
]; it protects rat hippocampal neurons against glutamate and amyloid beta peptide toxicity [14
]; it seems to function as a potent natural survival and neuroprotective factor for γ-aminobutyric acid (GABA)-ergic neurons in a model for Huntington disease [15
] and of dopaminergic neurons in a model for Parkinson’s disease [16
]. In two studies, creatine supplementation was convincingly shown to improve concentration [17
], as well as short-term memory and learning [18
] in healthy human subjects. Due to these protective functions of creatine the deficiency of creatine could play a part in the neurobiology of ASD.
CDS includes three separate deficiencies: arginine:glycine amidinotransferase (AGAT), guanidinoacetate methyltransferase (GAMT), and creatine transporter (CrT) [4
]. AGAT and GAMT deficiencies are autosomal recessive inherited biosynthesis defects of creatine. The AGAT gene (NCBI Gene ID 2628, official nomenclature: GATM
) located on chromosome 15q15.3 is 16.8 Kb in size and contains nine exons, which encode 424 amino acids. The GAMT
gene (Gene ID 2593) located on chromosome 19p13.3 is 4.46 Kb in size and has six exons encoding 237 amino acids. Dysfunction of the CrT leads to impaired intra-cellular creatine uptake and has X-linked inheritance. The CrT gene (Gene ID 6535, SLC6A8
) located on Xq28 spans 8.4 Kb and consists of 13 exons encoding 635 amino acids [4
]. To date, there are 6, 29, and 88 pathogenic mutations described throughout the GATM
, and SLC6A8
genes, respectively [20
]. The common denominator of all these diseases is the virtually complete absence of creatine and phosphocreatine in the brain that can be assessed with in vivo magnetic resonance spectroscopy (MRS). The diagnosis can be made in spot urine determining the concentration of creatine and guanidinoacetate normalized for creatinine, in 24-h urine measuring the excretion of creatine, guanidinoacetate, and creatinine, and by measuring the concentration of these metabolites in blood [4
]. Among CDS, GAMT deficiency, while presenting as a spectrum from mild to severe, has the most severe phenotype that includes severe intellectual deficits. Treatment-refractory epilepsy and extrapyramidal movement disorders are exclusively seen in GAMT patients [4
]. The few AGAT-deficient patients described so far seem to have a less severe presentation with developmental delays and impaired communication and social contact. Mild seizures have been reported in only some of the cases [9
]. Males with X-linked CrT deficiency present with impairments of speech and language and may have a seizure disorder. They usually show behavioural abnormalities and develop intellectual disability. While most of the female carriers have a normal appearance, some may have mild learning difficulties. Treatment with supplementation of creatine can fully (AGAT) or partly (GAMT) replenish brain creatine and attenuate the clinical course in AGAT and GAMT-deficient patients. In contrast, creatine supplementation has no effect on brain creatine and the clinical course in male CrT patients [4
]. There is an even gender distribution in GAMT [10
] and likely in AGAT deficiencies, whereas predominantly males are affected with the CrT defect. The latter is an X-linked condition and female carriers are mostly asymptomatic [8
Because of the overlap of symptoms between CDS and ASD and because of the potential role of the creatine-phosphocreatine homeostasis in autism, we hypothesize that comparing the ASD population to the general population may reveal an altered frequency of genetic variants in the three CDS genes (GATM, GAMT and SLC6A8) that are associated with autism.
Three rare, non-synonymous genetic variants were identified in coding regions in GATM
genes in four individuals with ASD. The variants were heterozygous in GATM
, implying possible carrier status for a pathogenic mutation; and hemizygous for SLC6A8
in a single male, suggesting creatine transporter deficiency if the variant is proven pathogenic. All three variants were predicted as being damaging using SIFT, PolyPhen-2 and MutationTaster. No creatine transporter defect had been demonstrated in males with the hemizygous SLC6A8
c.1162G > A variant. This variant was subsequently proven to be benign, based on its presence in two patients with mental retardation and no creatine transporter defect [31
]. The variants in GATM
can only be proven as truly pathogenic if functional studies are performed. No homozygous or compound-heterozygous variants classified as “damaging” were identified in the ASD population, suggesting the prevalence of CDS in autism patients in our sample is zero. We have recently published another study confirming the prevalence to be zero, as no cases of CDS were identified in 443 children with ASD (screening of patients was done using both metabolic and molecular methods) [5
]. Schiff et al. [34
] screened urine from 203 children with non-syndromic autism and found none to be affected with CDS. The only other genetic screen was carried out by Newmeyer et al [35
]. They screened 100 males with ASD for mutations in the SLC6A8
gene, and reported one affected child. However, this male patient had the c.1162G > A variant. He was later further investigated and had a normal urine creatine/creatinine ratio, suggesting the variant was benign.
Statistical Differences between MAFs in CDS Genes in ASD and the General Population
Fisher’s exact test was used to determine the statistical significance of any differences identified in the allele frequencies of the autism cohort compared to the 1000 Genomes database (divided into four population groups).
There were some differences seen in the MAFs calculated for GATM, GAMT and SLC6A8 variants identified in the ASD patient group, but the only statistically significant results were seen in the GATM gene. These included variants in non-coding regions (3′ and 5′UTRs, introns) as well as changes in exons. The Asian autism cohort was compared to the East Asian data in 1000 Genomes initially. Nine variants were identified at a statistically lower frequency in our Asian autism cohort compared to the 1000 Genomes database. The data for these nine variants was then compared to the South Asian data in 1000 Genomes, which further suggested that the MAFs were not significantly different. We do not know the diversity of ethnicities within our Asian cohort and so cannot determine if our population is skewed towards one or another subpopulation. This potentially interesting finding would need further investigation with Asian subpopulations. Two variants in the European autism cohort also presented at a statistically lower allele frequency than in the 1000 Genomes database.
4. Materials and Methods
A prospective group consisting of 71 subjects was enrolled in Toronto, and a retrospective sample of 95 subjects was collected as part of the Canadian Autism Genome Project (families with only one autistic child). The former was also part of the prevalence study of CDS in ASD [5
]. Supplementary Table S1
shows the cohort divided by ethnicity and sex.
Participants received a diagnosis of ASD from clinician experts based on diagnostic and statistical manual of mental disorders, 4th edition (DSM-IV) criteria for ASD (autism, Asperger’s, or pervasive developmental disorder—not otherwise specified (PDD-NOS)) with the diagnosis confirmed by the autism diagnostic observation schedule (ADOS-G) [36
]. An additional group of children with moderate to severe ASD eligible for a publicly-funded Intensive Behavioural Intervention Therapy program for children with ASD was recruited. Children included from this subgroup also underwent assessment by clinician experts, however, rather than the ADOS, their assessment included observation with the Childhood Autism Rating Scale (CARS) [37
] and the DSM-IV checklist. Individuals were excluded if they had Rett syndrome, Childhood disintegrative disorder, severe bilateral visual impairment, or severe bilateral hearing impairment.
4.1. Study Design and Measurement
To investigate the genetic variability in genes involved in creatine metabolism in children with ASD, Sanger-based DNA sequencing chemistry was used (ABI-3730) to sequence the three genes, GATM
, and SLC6A8
(CrT). The entire gene was sequenced including 3′ and 5′UTRs, coding regions and flanking intronic segments. Examination of raw sequence data was completed manually for missense, nonsense, or small insertion/deletion events, by aligning sequences with NCBI reference genes. Genetic variations were compared to sequence data in Alamut (Alamut Visual version 2.6, Interactive Biosoftware, Rouen, France) and three variation databases: the 1000 Genomes phase 3 dataset [38
], Exome Sequencing Project [39
] and the Exome Aggregation Consortium [40
]. Fisher’s exact test was used to calculate “goodness-of-fit” between the allele frequencies in published databases (primarily the 1000 Genomes phase 3 dataset and the autistic patient population, based on ethnicity. If ethnicity was not known, or admixed, then allele frequencies were compared to the combined allele frequency for all ethnicities. Significant variation between allele frequencies for the two populations was noted when p
To reduce the false discovery rate, and help reduce type 1 errors (false positives) we applied the Benjamini–Hochberg procedure. The variants for each ethnic population for each gene (excluding the novel variants) were ranked in ascending p-values. Each individual p-value’s Benjamini–Hochberg critical value was calculated using the formula (i/m)Q (with i = the individual p-value’s rank; m = total number of tests; and Q = the false discovery rate (0.01)). The highest p-value that was also smaller than the critical value was noted, and that value and all values with lower p-values were considered significant. These variants are identified on the Manhattan plots.
4.2. Molecular Genetics
DNA was prepared from blood or lymphoblasts; PCR products were amplified using the primers listed in Supplementary Table S5
, and sequenced by TCAG (Toronto Center of Applied Genomics, Toronto, ON, Canada). GATM
gDNA was amplified using Herculase II Fusion Taq polymerase (Agilent Tech., Mississauga, ON, Canada). GAMT
gDNA was amplified using Hotstar Taq polymerase (Qiagen, Toronto, ON, Canada). In some cases, a second nested amplification was carried out on the first PCR product; some of these primers had a GC clamp at their 5′ end. SLC6A8
was amplified using Herculase II Fusion Taq polymerase, Hotstar Taq polymerase or TaKaRa La Taq polymerase (Takara Bio Inc., Shiga, Japan). Some of these primers had an M13 clamp at their 5’ end (G. Salomons, personal communication).
4.3. Ethics, Consent and Permissions
The study was approved by the Research Ethics Boards of The Hospital of Sick Children, Surrey Place Center, and Holland Bloorview Kids Rehabilitation. Subjects enrolled in the study after parents completed written informed consent.
We hypothesized that genetic variability (polymorphisms) in the three genes associated with CDS could impact the health of children and result in an autistic phenotype. This hypothesis was suggested by the fact that rare variants have been shown to have a cumulative effect, contributing to ASD risk (narrow sense heritability); and that CSD and ASD have overlapping phenotypes suggesting some similar disease mechanisms.
Our findings suggest there could be a lower association of some specific GATM gene variants in Asians and Europeans with autism compared to the 1000 Genomes database (East Asian population and all Europeans), observations that would need to be corroborated in a larger group of autism patients in which ethnic sub-populations are known. Variants in GATM have not been associated with autism previously, nor has the chromosomal region of 15q21.1. There are no genes overlapping GATM that have been associated with autism.
By sequencing the three CSD genes, the findings demonstrate that genetic variability in genes of biosynthesis (AGAT and GAMT) and transport of creatine (CrT) is unlikely to play a part in the pathogenesis of ASD in children. Through this work, we have corroborated results determined in our prior prevalence study in which none of the 443 children with ASD was found to be affected with CDS [5
]. Both studies taken together suggest that, despite the possibility that there is a shared feature that might cause the separate but similar ASD and CDS autistic symptoms, we should look elsewhere for causation of ASD.