Autism is a severe disorder of childhood diagnosed on the basis of impaired social interaction and communication, presence of rigid and stereotyped behaviors, and disease onset prior to 3 years of age [1]. Altered prenatal and early-postnatal neurodevelopment plays a pivotal role in autism pathogenesis. Microscopic cytoarchitectonic CNS abnormalities are responsible for the miswiring of neural pathways and for altered information processing, typically in the absence of macroscopic neural malformations or facial dysmorphology [2, 3]. Family and twin studies have conclusively demonstrated prominent genetic contributions to autism, with concordance rates of 82–92% in monozygotic twins vs 1–10% in dizygotic twins, sibling recurrence risk at 2–3% vs an incidence of 1–2/1000 in the general population, and heritability estimates above 90% [4, 5]. The identification of these genetic underpinnings has proven more complex than anticipated, possibly due to interindividual heterogeneity, epistasis, gene-environment interactions and epigenetic events [6]. In an attempt to pin down single-gene contributions, several investigators have begun selecting candidate genes not merely based on their chromosomal localization, but also by linking their cellular function to specific "endophenotypes" (i.e. heritable clinical, biochemical or morphological traits especially frequent among affected individuals and their first-degree relatives), or by using proteomic studies of post-mortem brain tissue as a starting point. Following these approaches, and under the assumption that single genes may not necessarily yield a detectable signal in linkage studies of polygenic disorders such as autism, TPH2 and GLO1 were initially assessed despite a relative lack of support by sib-pair analyses for their chromosomal localizations (ch. 12q21.1 and 6p21.3-p21.2, respectively). These initial studies yielded promising evidence supporting TPH2 and GLO1 roles as autism vulnerability genes [7, 8].

Tryptophan hydroxylase-2 (TPH2) is the recently-discovered rate-limiting enzyme for 5-HT synthesis in the CNS, whereas the "classical" TPH isoform, now termed TPH1, is found only in peripheral tissues, except for the pineal gland [9–11]. Several issues contribute to raise interest in TPH2 as a candidate gene for autistic disorder: first, 5-HT exerts prominent neurotrophic roles during development, in addition to mediating "classical" neurotransmission [12]; secondly, TPH2 gene variants were found associated with childhood onset OCD [13], and 5-HT influences compulsive and/or stereotyped behaviors to the extent that drugs blocking 5-HT reuptake represent their first-line treatment not only in OCD but also in autism [14]; thirdly, brain imaging studies have revealed that neural responsiveness to emotional stimuli in the amydgala is significantly modulated by TPH2 gene variants [15, 16]; finally, one report has provided initial evidence of a possible association between TPH2 gene variants and autism, particularly in patients characterized by more severe repetitive and stereotyped behaviors [7].

Glyoxalase I (GLO1) has been identified using a proteomic approach on post-mortem brain tissues [8]. Gray matter samples from brains of autistic patients display more frequently than unaffected controls a more negatively-charged GLO1 isoform resulting from the single aminoacid substitution A111E, produced by SNP C419A [8]. GLO1 is a cytosolic enzyme responsible for the glutathione-dependent detoxification of α-oxoaldehydes, such as methylglyoxal, spontaneously forming from intermediates of the Embden-Meyerhof glycolytic pathway, such as glyceraldehyde-3-phosphate and dihydroxyacetonephosphate [17]. Interestingly, this initial report also shows reduced GLO1 enzymatic activity in brain extract from autistic patients compared to controls, and accumulation of advanced glycation products resulting from low GLO1 activity [8]. At the genetic level, enhanced frequencies of the A419 allele are found in a small case-control contrast performed on genomic DNA extracted from the same brain tissues of patients and controls [8].

The present study attempts to replicate and extend these findings by using both case-control and family-based association approaches. We also perform quantitative-trait analyses employing some of the best-characterized endophenotypes in autism research, namely cranial circumference, 5-HT blood levels and urinary peptide excretion rates [18, 19]. Indeed, macrocephaly (i.e., cranial circumference > 97^th percentile) has been consistently found in approximately 20% of autistic patients, 5-HT blood levels are elevated in at least 25% of patients, and excessive peptiduria is found in up to 60% of cases depending on ethnicity and country of origin; sizable familiality and heritability is present for all three parameters ([20–23], and Persico et al, manuscript in preparation for familiality of peptiduria).

The present study provides evidence that TPH2 and GLO1 gene variants do not contribute significantly to autism pathogenesis in our sample, while GLO1 could seemingly exert a protective effect in unaffected siblings. Our conclusions are strengthened by the consistency of multiple statistical approaches and by the parallel assessment of some of the most reliable biochemical and morphological endophenotypes described in autism research to this date, namely macrocephaly, hyperserotoninemia and enhanced peptiduria [20–23]. These heritable traits are believed to be more directly linked to the genetic make-up of an individual than complex clinical symptoms, and could thus possibly identify subgroups of autistic patients characterized by relatively homogenous pathogenetic processes. These parameters were thus chosen on this ground, and not following a-priori hypotheses. In particular, 5-HT blood levels do not reflect TPH2 activity, as peripheral 5-HT largely comes from the digestive tract and is thus produced by TPH1 [9–11]. Therefore, the non-significant trend towards an association between TPH2 gene variants and 5-HT blood levels either represents a spurious finding, or necessarily stems from an indirect pathophysiological link which will require further investigation.

Some differences must be drawn in the interpretation of the data pertaining to TPH2 and GLO1. Our results exclude causal TPH2 roles in our sample, and do not support the hypothesis that TPH2 gene variants may exert a major influence on repetitive and stereotypic behaviors in autistic patients. This conclusion confirms the outcome of a recent study, reporting no association either with autism, or with obsessive-compulsive and self-stimulatory behaviors in autistic patients [34]. Collectively, these findings, although negative, are important due to the relevant roles played by 5-HT during neurodevelopment [12], and because they help us interpret contributions to autism by serotoninergic genes. TPH2 and the 5-HT transporter (5-HTT) play pivotal roles in central serotoninergic neurotransmission. TPH2 determines the rate of 5-HT biosynthesis in the CNS [9–11], whereas the 5-HTT terminates the action of extracellular 5-HT on its receptors through reuptake [35]. During neurodevelopment, 5-HT exerts impressive neurotrophic effects on cell proliferation, differentiation, and migration, programmed cell death, cell-cell coupling, synaptogenesis and cytoskeletal plasticity [12]. Many of these processes appear altered in neuropathological studies of autistic brains [2, 3]. Furthermore, a recent quantitative brain imaging study has revealed increased cortical gray matter volumes in autistic patients carrying 5-HTT gene variants yielding slower extracellular 5-HT clearance rates, resulting in higher extracellular 5-HT levels [36]. This finding parallels results obtained in 5-HTT knockout mice backcrossed into C57BL/6J, displaying increased neuronal cell density and thickness of supragranular and infragranular neocortical layers compared to wild-type mice [37]. In contrast to these neurobiological data, strongly supporting the trophic effects exerted by elevated extracellular 5-HT levels in autism, results of genetic studies on serotoninergic gene variants in autism are mostly negative or mixed at best. Indeed, the present and prior studies jointly indicate that (a) in the vast majority of cases, functional serotoninergic gene variants play a modulatory, but not a causal role in autism; (b) 5-HT synthesis does not contribute to produce increased levels and/or prolonged persistence of extracellular 5-HT, which largely stems from reduced 5-HT uptake and could possibly play a role in eccessive neurotrophism and macrocephaly in autism; (c) 5-HT could modulate not only morphological features such as head circumference, but also repetitive and stereotypic behaviours in autism, as well as in OCD [7, 13, 14, 38]. Unfortunately the approved Italian translation of the Autism Diagnostic Interview-Revised (ADI-R) has become available only recently [25], and this instrument was not available when the vast majority of our sample was recruited. However, the presence/absence of relevant verbal and motor stereotypic behaviors that was assessed here by an experienced clinician at intake largely corresponds to the current presence of "stereotyped utterances and delayed echolalia", "hand and finger mannerisms", "other complex mannerisms or stereotyped body movements", and "midline hand movements", as assessed by ADI-R items n. 33, 77, 78, and 79, respectively [25]. Therefore, despite representing a crude, single time-point measure, it allows us to exclude prominent roles for TPH2 gene variants in these behavioural symptoms. Searches for modulatory effects on repetitive and stereotypic behaviors will require finer, possibly longitudinal measures.

On the other hand, GLO1 allele A419 seemingly exerts a protective effect among unaffected siblings, whose allelic transmission patterns differ significantly from those of their autistic siblings, both using TDT and FBAT. The discrepancy between initial findings by Junaid and Colleagues [8], proposing A419 as an autism vulnerability allele, and our findings may stem from at least two potential sources. First, it is important to notice that the A419 allele frequencies found by Junaid and Colleagues [8] in autistic patients are very similar to those reported here (0.6056 vs 0.5785 and 0.5693 in our Italian and Caucasian-American patients, respectively), whereas their control allele frequencies are much lower than those found in our controls (0.4400 vs 0.5439). It is likely that a small sample size including only 50 controls, encompassing both normal subjects and patients with Batten disease or fragile-X syndrome instead of randomly chosen unaffected individuals, may have diminished the reliability of allele frequency estimates in the general population, skewing case-control statistics [8]. The definition of A419 as a risk allele for autism would thus not seem fully justified already on the basis of the genetic data set presented in the initial study [8]. Secondly, the present study is the first one including also unaffected siblings, which were not assessed previously [8]. The existence of significant differences in allelic transmission rates between our autistic patients and unaffected siblings does point towards the existence of functional GLO1 gene variants, deserving further characterization in order to understand their protective role in families with an autistic proband. These variants may be represented by the A419 allele itself, but could also consist in polymorphisms located in the transcriptional regulatory regions of the GLO1 gene and in linkage disequilibrium with the C419A SNP. This possibility is consistent with the decreased glyoxalase I enzymatic activity found in post-mortem neocortical tissues of autistic patients, as compared to controls [8].

Based on the consistency between case-control and family-based association analyses, we can exclude in our sample a major role for TPH2 gene variants in autism pathogenesis, and in determining the presence of prominent repetitive and stereotypic behaviors assessed at the time of patient recruitment by an experienced clinician. Our study cannot exclude that functional TPH2 gene variants may influence the intensity and course of repetitive and stereotypic behaviors in the longer term. Also GLO1 protein variants encoded by the C419A SNP seemingly do not confer autism vulnerability in this sample, but allele A419 apparently carries a protective effect among unaffected siblings, spurring further interest into the functional correlates of the C419A SNP and of other polymorphisms located in transcriptional regulatory regions of this gene.