In recent years, the number of patients diagnosed with autism spectrum disorders (ASD) has increased with current studies reporting a prevalence of 1% [1]. ASD shows extreme clinical heterogeneity; however, the diagnosis of ASD according to the Diagnostic and Statistical Manual of Mental Disorders (5th edition) is based on deficits in two areas—social communication and restricted, repetitive behaviour or interests. The patient must have deficits in both areas, and symptoms must be present from early childhood [2]. The genetic architecture of ASD is very diverse consisting of a variety of genetic alterations, such as chromosomal abnormalities, copy number variations, rare single nucleotide variants (SNVs), common polymorphic variations, and epigenetic modifications; however, only 6–15% of children with ASD have well-defined genetic syndromes [3]. Because of the development of high-throughput sequencing methods, many highly penetrant genetic causes of ASD have been identified, but the underlying genetic background of 70% of cases remains unexplained [4].

Mitochondrial disease (MD) is presently one of the most recognized metabolic diseases caused by the failure of both nuclear and/or mitochondrial DNA (mtDNA). The prevalence of mtDNA mutations responsible for MD is 1 in 5000, whereas that of nuclear mutations is 2.9 per 100,000 cases [5]. Although MD frequently results in a spectrum of disorders with multisystemic presentations, neurological symptoms are common because tissues with high-energy demands, such as neural tissue, are often the most strongly affected by mitochondrial dysfunction. Even though the diagnosis of MD is increasing and becoming more frequent, the exact genetic background in many cases remains unconfirmed. Mitochondrial dysfunction can be caused by either primary MD or secondary mitochondrial damage [6]. Primary MD is because of genetic defects in mtDNA or a defect in a nuclear gene that is important for mitochondrial function. These mutations usually affect proteins involved in reactions of oxidative phosphorylation (OXPHOS). However, many disorders show similar effects in terms of mitochondrial dysfunction, but are elicited by mutations in other genes not related directly to normal mitochondrial function [7]. In other cases environmental factors, associated disorders or ageing are resulting in secondary alterations.

Several authors have proposed that mitochondrial dysfunction may be one of the most common medical conditions associated with autism [8, 9]. Lombard et al. [10] proposed that ASD may be a condition with abnormal mitochondrial function. Clinical and biochemical studies have uncovered an emerging link between mitochondrial dysfunction and neurodevelopmental disorders, including intellectual disability [11], childhood epilepsy, and ASD [9]. Furthermore, mitochondrial dysfunction has been associated with some forms of syndromic ASD [8, 11]. In many of these studies, biochemical changes, such as elevated levels of creatine kinase, lactate, pyruvate, carnitine, ammonia, and alanine were detected in the serum of patients with ASD [11–14]. In other studies, altered respiratory chain enzyme activities [15] or decreased expression of OXPHOS genes were detected in autistic brain [16], findings which indicate abnormal or altered mitochondrial function.

Damage to the OXPHOS system was found in individuals with ASD by Napoli et al. [17] and reviewed by Valenti et al. [11]. Oliveira et al. [14] found that 7% (7/100) of children with ASD, who were clinically indistinguishable from other affected children with ASD, exhibited a mitochondrial respiratory chain disorder. Weissman et al. [18] proposed that defective mitochondrial OXPHOS may be an additional underlying pathogenic mechanism in a subset of individuals with autism.

Despite evidence of altered mitochondrial function in some individuals with ASD, it is not known whether mitochondrial dysfunction is a cause or an effect of ASD. Although a mitochondrial subgroup in ASD could be identified [19], findings from review articles, such as those of Palmieri and Persico [19] and Rossignol and Frye [9], found that even in this subgroup the causative genetic factor could be identified in a proportion of cases (23%). In cases of non-syndromic ASD, mitochondrial dysfunction without mtDNA alterations has been frequently observed [8, 9]. In a systematic review and meta-analysis, Rossignol reported that MD was present in 5% of children with ASD [9], and in this ASD/MD subgroup, mtDNA abnormalities were found in 23% of patients [9]. These findings demonstrate that primary MD may be present in a subgroup of children with ASD.

Some studies have reported mtDNA deletions in individuals with ASD [12, 20–22]. Single mtDNA deletions have a role in different paediatric and adult onset primary MDs such as Kearns–Sayre syndrome, Pearson syndrome, and progressive ophthalmoplegia externa [23]. Multiple mtDNA deletions occur mostly because of pathogenic mutations in genes responsible for intergenomic communication; however, they are often related to ageing or harmful environmental factors as well because mtDNA has a poor DNA repair system [6, 24].

The aim of the present study was to investigate the presence of the most common pathogenic mtDNA alterations in patients with ASD and to elucidate the etiology of these mtDNA alterations by analysing their co-occurrence with both known ASD-associated genes and genes responsible for mtDNA maintenance and by comparison the targeted NGS data (ASD associated genes and genes responsible for mtDNA maintenance) of cases with and without mtdel-ASD, patients with primary mitochondrial disorders and healthy controls.

Sixty children with ASD (6 females and 54 males, median age = 7 years, IQR = 7.25) were investigated. In our analysed cohort of 60 patients, 29 patients were sporadic (from simplex families) and their family histories were negative for other neurodevelopmental disorders and major psychiatric or neurological disorders. The distribution of the relevant symptoms for mitochondrial disorders in mtdel-ASD and idiopathic ASD group are shown in Fig. 1. There were some differences between ASD cases with and without mtDNA deletion regarding the clinical phenotype. Developmental regression, muscle hypotonia, and additional neurological signs were most common in the mtdel-ASD cases. Multisystemic abnormality appeared also more frequently. Referring to seizures no major differences has been observed between these two groups. Family history was not available in two cases because these children were not living with their biological parents. A positive family history was found in 48% of cases, of which 20 cases (33.3%) had a family history for psychiatric disorders (bipolar disorder, depression, and schizophrenia). In four cases (6.7%), visual and hearing impairments, ataxia, complex endocrine disorders, or a combination of these factors was noted. The co-occurrence of these symptoms is an indicator for MD according to mitochondrial disease criteria (MDC) [36]. In five cases (8%), we found a positive family history for psychiatric disorders and some MDC-related symptoms were also noted.

Minor physical anomalies were identified in 44 children. All children were diagnosed with ASD based on ADI-R, and in most cases, with ADOS as well. We found that serum lactate levels and/or the lactate:pyruvate ratio supported the presence of mitochondrial dysfunction in four patients.

In this study, we provide for the first time a comprehensive genetic analysis of patients with ASD that investigates co-occurrence of the most frequent mtDNA alterations, intergenomic communication disturbances (51 genes), and 101 genes previously associated with ASD. We found co-occurrence of mtDNA deletions with ASD-associated genetic alterations, which supports the previous observation that mitochondrial alterations are frequently associated with ASD. In one patient with ASD (P3), we found a mtDNA deletion with CHARGE syndrome caused by a de novo mutation in CHD7. These genetic alterations in P3 were also accompanied by a TSC2 mutation of uncertain significance. Autistic symptoms are present in approximately 30% of patients with CHARGE syndrome [39], and lactic acidosis is a rare alteration. Based on the phenotype, we conclude that the driving genetic alteration in this patient is the CHD7 mutation, and the mitochondrial gene defect may not be the true causative factor in the etiology of the disease; however, CHD7 function is strongly ATP-dependent [41]. In addition, the associated heterozygous TSC2 mutation is likely a modifying gene. CHD7 is a member of the chromo-domain helicase DNA-binding (CHD) protein family and plays a role in transcription regulation through chromatin remodelling. Mutations in TSC2 are known to cause one syndromic form of ASD; however, our patient did not develop the classic symptoms of tuberous sclerosis until recently. TSC2 mutations may induce activation of mTORC1 leading to increased mtDNA expression and mitochondrial density. mTOR is a Ser/Thr kinase that forms complexes with numerous protein partners to regulate cell growth, mitochondrial membrane potential, and ATP synthetic capacity [42].

In Patient 8, we found that the mtDNA deletion was accompanied by a heterozygous pathogenic DHCR7 mutation and a rare variant of uncertain significance in NHS actin remodelling activator (NHS). DHCR7 catalyses cholesterol production from 7-dehydrocholesterol, and defects in this protein cause SLOS. Furthermore, a high 7-dehydrocholesterol level results in mitochondrial dysfunction [43]. However, the significance of a heterozygous mutation in this gene is not known. We hypothesize that in the case presented here the co-occurrence of the DHCR7 heterozygous mutation and CMV infection may play a role in changes of mitochondrial biogenesis and in the pathogenesis of autistic features. The patient was tested for SLOS; both serum cholesterol and 7-dehydrocholesterol levels were normal, which rules out the presence of typical SLOS.

Evidence of mitochondrial dysfunction in ASD was first described 19 years ago [10]. Currently, it is the most common metabolic abnormality known in ASD with a prevalence of 7.2% [14]. In a subgroup of the CHARGE (Childhood Autism Risk from Genes and Environment) study, decreased NADH activity was found in lymphocytes in 8 of 10 cases. In this cohort, only 2 of the 10 patients had mtDNA deletions and 5 patients had altered mtDNA copy numbers [12]. However, the genetic background was not clarified in 79% of the patients with ASD-MD [9]. Therefore, the possibility of secondary damage to mitochondria cannot be excluded. A small pilot study examining 12 patients with ASD described 8 mitochondrial deletions [21], which could be the result of intergenomic communication disturbances, environmental factors, or other gene–gene interactions. As is the case for many other disorders, it is still not clear whether the detected mitochondrial dysfunction in ASD is a primary or secondary event either having a key role in disease pathogenesis or is simply a downstream effect.

In our study, mtDNA deletions were identified in 16.6% of evaluated patients with ASD. During mtDNA hotspot screening and NGS analysis, no concomitant primary MD was detected. To examine whether mtDNA deletion is a primary or secondary event in ASD in our cohort, we used different comparison groups to screen the nDNA background of the mtDNA deletion. Pathogenic or likely pathogenic variants were detected in both mtdel-ASD and MD without ASD cases, all in heterozygous form. A high number of VUS in intergenomic communication genes were detected in the MD without ASD cohort (4/6), and a few rare variants were identified in patients with ASD that lacked mtDNA deletion and in healthy controls (Table 3). Homozygous or compound heterozygous mutations in MGME1 and SUCLG1 have been previously correlated with severe early-onset mitochondrial disorders (OMIM 615084, OMIM 245400) [44]. The importance of the presence of heterozygous mutations is not well understood. It is known in some genes responsible for intergenomic communications heterozygous mutations may result in a less severe phenotype than that found with the homozygous form [45].

The question has also been raised whether patients with MD and ASD symptoms have special characteristics. In a study by Rossignol and Frye [9], a cohort of ASD/MD children were compared to two comparison groups: children with general ASD and children with general MD. In the ASD/MD group, increased lactate and pyruvate levels, seizures, motor delays, and gastrointestinal abnormalities were significantly more prevalent compared to children with general ASD. A more balanced male:female ratio was also detected in the ASD/MD group [9]. Our results confirm the observations by Rossignol and Frye; however, in our ASD cohort with mtDNA deletion, elevated lactate levels and/or an elevated lactate:pyruvate ratio were found in only four cases, whereas most of our ASD patients with mtDNA deletion had symptoms common to MD, such as hypoacusis, muscle weakness, hypotonia, delayed motor development, and movement disorder (Fig. 1). Significant difference between mtdel-ASD and non-mtdel-ASD group was found regarding clinical phenotypes (developmental regression, muscle hypotonia, additional neurological signs and multisystemic alterations were more common in cases mt-delASD). Interestingly, the phenotypes of classic mitochondrial deletion syndromes, such as Pearson syndrome, progressive ophthalmoplegia externa, and Kearns–Sayre syndrome, were not detected in any of our patients. The family histories of mtdel-ASD children in our cohort differed from the family histories of the ASD cohort without a mtDNA deletion, since various psychiatric disorders were common among family members of mtdel-ASD cases both on maternal and paternal side. However none of the parents reached the MDC scoring cut-off value for definitive MD, which could not be independently verified because none of the family members agreed to perform muscle biopsy. MtDNA disorders are usually inherited maternally, however single mtDNA deletions are considered sporadic events with low inheritance risk, whereas multiple mtDNA deletions are the result of primary nuclear defects in genes responsible for mtDNA maintenance or nucleoside metabolism and follow Mendelian inheritance patterns [46].

Mitochondrial haplogroups were also investigated in association with ASD. Chalkia et al. found that individuals with European haplogroups designated I, J, K, X, T and U (55% of the European population) had significantly higher risks of ASD compared to the most common European haplogroup, HHV. Asian and Native American haplogroups A and M also were at increased risk of ASD [47]. In Hungary it is not rare that a person has ancient European haplotype such as T, K, and U haplotype, and rarely Asian haplotype such as B can occur as well. In some Hungarian patients the mtDNA deletion was coexisting with ancient haplotype [48].

In 90% (9/10) of children from the mtdel-ASD cohort, we found rare SNVs in ASD-associated genes (Table 4). A rare mutation was detected in AUTS2 in which deletions are inherited in an autosomal dominant manner and are associated with neurological symptoms including intellectual disability and developmental delay [49]. In a modest study of 13 cases of ASD associated with AUTS2 alterations, only one patient had a nonsense mutation; all the other patients had a deletion [49]. The significance of the missense mutation identified in our study is uncertain (her mother harbours the mutation as well); however, clinical symptoms of the patient correlate with the phenotype of previously published AUTS2 mutations. This rare AUTS2 variant coexisted with a rare variant in retinoic acid induced gene 1 (RAI1). The gene–gene interaction of these two alterations are hypothesized.

In addition, we found that patients had mutations of uncertain significance in forkhead box P2 (FOXP2), RAI1, phosphodiesterase 10A (PDE10A), katanin catalytic subunit A1 like 2 (KATNAL2), and reelin (RELN). Most of these genes play a role in cell regulation, signal transduction, and various signalling pathways, which could influence mitochondrial function. FOXP2 is an evolutionarily conserved transcription factor that regulates the expression of a variety of genes. Mutations in this gene cause speech-language disorder 1 (OMIM 602081), which is also known as autosomal dominant speech and language disorder with orofacial dyspraxia [50]. RAI1 acts as a transcriptional regulator of chromatin remodelling by interacting with basic transcriptional machinery [51]. RAI1 deletion is associated with Smith–Magenis syndrome, whereas duplications are associated with Potocki–Lupski syndrome [52]. Several heterozygous mutations are also associated with Smith–Magenis syndrome [53]. In our case, we found that the typical symptoms of Smith–Magenis syndrome were not present. Mutations in PDE10A can affect cyclic nucleotide concentrations. This phosphodiesterase selectively catalyzes the hydrolysis of 3′ cyclic phosphate bonds in cAMP and/or cGMP. The phosphodiesterase family of proteins regulates cellular levels, localization, and duration of action of these second messengers by controlling the rate of their degradation. In addition, phosphodiesterases are involved in many signal transduction pathways and are implicated in the pathogenesis of bipolar disorder [54].

Mitochondrial dysfunction may be associated with several forms of syndromic ASD, but is also frequently related to non-syndromic cases [8, 9]. During our comprehensive analysis, we found examples of both but in most cases we did not find the causative genetic mutation that accounts for the mitochondrial dysfunction. In the examined children from the general ASD cohort (without mtDNA deletion), we found several VUS, most of which were identified in genes without previous correlation to mitochondrial dysfunction. Based on our findings, we conclude that the detected mitochondrial DNA deletions in patients with ASD in our cohort are a secondary effect. By investigating the most common mtDNA alterations and the most common nuclear genes responsible for intergenomic communications, we did not identify the clear genetic etiology in most of our cases. Therefore, further investigation and characterization is warranted.

The aim of our study was to gain a better understanding of mitochondrial dysfunction in autism. We found that mtDNA alterations were more common among our cohort of patients with ASD than in control individuals. In addition, we found that the mtDNA deletion was usually not the single genetic alteration identified in ASD, but co-occurred in both syndromic and non-syndromic forms of ASD with either ASD-associated genetic risk factors and/or alterations in genes responsible for intergenomic communication. Our findings indicate a very complex pathophysiology of ASD in which mitochondrial dysfunction is not rare and can be caused by mtDNA deletion, which may be considered as de novo mutations or the consequence of the alterations of the causative culprit genes for autism or genes responsible for mtDNA maintenance.