A modestly expanded CGG nucleotide repeats (55–200) in the 5′ UTR of the fragile X mental retardation gene, FMR1 (Kogan et al., 2008; Tassone et al., 2012; Battistella et al., 2013), is the hallmark of premutation carriers. Originally, premutation carriers were thought to be free of phenotypic traits; however, the findings regarding fragile X-associated primary ovarian insufficiency [FXPOI; (Cronister et al., 1991)], followed by the discovery of fragile X-associated tremor/ataxia syndrome (FXTAS; OMIM:300623) identified in adult carriers (Hagerman et al., 2001) discredited this notion. Premutation carriers may also suffer from psychological problems, visuo-spatial deficits, and immune dysregulation (Tassone et al., 2012; Winarni et al., 2012; Wong et al., 2012; Battistella et al., 2013; Hagerman and Hagerman, 2013), while affected children are often diagnosed with ADHD, autism, anxiety, and other psychopathologies (Farzin et al., 2006; Chonchaiya et al., 2013). At the cellular level, fibroblasts from premutation carriers (humans or animal models of the premutation) are generally accompanied by high FMR1 gene expression, normal or lower levels of its translation product FMRP (fragile X mental retardation protein) and mitochondrial dysfunction (Ross-Inta et al., 2010; Napoli et al., 2011).

It is currently unknown which carriers of the premutation (over 1.5 million women and men in the United States) will develop FXTAS, as clinical diagnosis fails to identify carriers before significant neurological symptoms are evident. Therefore there is an immediate need for early detection and effective drugs for the cure or the prevention of FXTAS.

An understanding of the molecular characteristics underlying disease processes is a prerequisite for the development of early detection biomarkers to provide high value therapeutics. Metabolic profiling is one of the most important techniques, particularly focused on the detection of biomarkers for diagnosis of diseases (German et al., 2005) including metabolic disorders (Wang et al., 2005), motor neuron disease (Rozen et al., 2005), Parkinson’s disease (Scherzer et al., 2007), and Alzheimer’s disease (Barba et al., 2008). Metabolomics, an omics-level analysis of all metabolites, has become one of the main choices for obtaining an overall view of in vivo disease states, which has been successfully employed to identify metabolites associated not only with disease, but also with toxicity and aging (Lawton et al., 2008; Boudonck et al., 2009a,b; Sreekumar et al., 2009). Ante-mortem studies generally target body fluids, such as cerebrospinal fluid (CSF), urine, saliva or plasma/serum, as they are thought to reflect changes in tissues. Thus, these non-invasively available biofluids can be used as resources for first screening, in early detection of any disease.

To our knowledge, no analysis comparing biomarkers for differentiating premutation carriers (and their endophenotypes) from controls has been reported to date. Here a global plasma profiling has been developed for the discovery of novel plasma biomarkers applied to the premutation with and without FXTAS. Plasma metabolomics was evaluated in premutation carriers and age- and sex- matched controls with the aim of providing early biomarkers with the potential of uncovering metabolic perturbations in pathways (Goodacre et al., 2004; Werner et al., 2008) associated with the presence of the FMR1 premutation.

To the best of our knowledge, this is the first study in which an untargeted serum metabolomic profiling approach combined with sequential metabolite ratio analysis has been applied to discriminate plasma biomarkers in plasma of premutation and FXTAS-carriers. Our results demonstrate that a panel of four core serum metabolites (PEA, oleamide, aconitate, and isocitrate) can be used for sensitive and specific diagnosis of the premutation with and without FXTAS, and one of these ratios (namely, oleamide/isocitrate) as a biomarker of FXTAS. The findings of this study are promising because, despite the relatively small sample size (n = 16 and 23, respectively, for control and premutation individuals), they were obtained from a clinically well-characterized cohort of subjects with the FMR1 premutation representing the wider clinical spectrum of this genetic disorder. Although it could be argued that plasma metabolites may not be predictive of those in CNS because the blood-brain barrier serves to limit the passage of metabolites from and to the periphery, significant correlations have been reported for a number of key metabolites between plasma and CSF in subjects with HIV-1 infection [among them lactate, glutamine, and citrate (Maher et al., 2011)]. Furthermore, several reports have applied metabolomics of non-invasive samples to study complex neurological disorders (Griffin and Salek, 2007) including schizophrenia and Parkinson’s disease (Dumas and Davidovic, 2013). Supporting the notion that plasma metabolites can be used as surrogates of brain function, blood plasma metabolic profiles (including lactate) consistent with metabolic syndrome and increased inflammation have been reported in elderly people with mild cognitive impairment, a transitional state with considerably increased risk for Alzheimer’s disease (Tukiainen et al., 2008), and high CSF citrate levels have been found associated with depression but not to manic features (Mellerup and Rafaelsen, 1981). If these correlations were also applicable to our study, the higher plasma lactate (Figure 1C) and citrate (Figures 1A,C) levels observed in carriers could be linked to their higher incidence of cognitive impairment (Grigsby et al., 2006, 2008; Brega et al., 2008) and depression (Bacalman et al., 2006).

Considering that a significant number of metabolites identified as potential biomarkers in this study are of mitochondrial origin or involved in mitochondrial metabolism (citrate, aconitate, isocitrate) and given the roles of mitochondria at providing ATP to aerobic tissues—such as brain—and on de novo synthesis of Krebs cycle-associated neurotransmitters [i.e., Glu, Asp, and GABA (Butterworth and Heroux, 1989; Laughlin et al., 1998; Attwell and Laughlin, 2001; Navarro et al., 2008; Napoli et al., 2014)], it is tempting to propose that these metabolites are pointing at an underlying mitochondrial dysfunction. Consistent with this view, we have reported deficits in bioenergetics and/or mitochondrial morphology and dynamics in fibroblasts (Ross-Inta et al., 2010; Napoli et al., 2011, 2016c) from carriers as well as in brains (Napoli et al., 2016a) and ovaries (Conca Dioguardi et al., 2016) from a murine model of the premutation. Methods used for assessing mitochondrial dysfunction are associated with significant drawbacks that limit their routine clinical use. Some of them are invasive (requiring a skeletal muscle biopsy), whereas non-invasive ones require specialized exercise physiology evaluation or NMR studies, which may not be widely available, and not practical for monitoring disease progression or therapeutic responses for subjects with limited mobility. Levels of key plasma metabolites (lactate, alanine, and pyruvate) are often suggestive of respiratory chain diseases or mitochondrial dysfunction (Robinson et al., 1986; Gillis and Kaye, 2002), but simply lack sensitivity and specificity [61% sensitivity based on (Scaglia et al., 2004)]. In this study, by applying a global metabolite profiling utilizing a relatively non-invasive method, a combination of metabolites was identified that have the potential to provide a sensitive and specific assessment of the mitochondria status or “fitness” in carriers, not influenced by SSRI treatments. In this context, it is important to point out that beneficial effects on mitochondrial function have been inferred from various studies based on the fact that anti-depressant treatments normalize decreased cerebral blood flow and glucose metabolism in the prefrontal cortex of depressed patients (Buchsbaum et al., 1997; Kennedy et al., 2001; Dimatelis et al., 2013). However, in our study, the same mitochondrial metabolites with and without SSRIs treatments were identified (citrate, aconitate, isocitrate) with the exception of lactate, which was noted in the SSRI-free group, possibly suggesting a marginal improvement provided by this therapy or that carriers do not respond positively to this pharmacological intervention.

Within these identified biomarkers, the plasma levels of PEA and oleamide were lower than controls, whereas those of two Krebs’ cycle intermediates, aconitate, and isocitrate were higher (Table 2). To understand the biological meaning of these biomarkers and their ratios, we focused on these metabolites and the pathophysiology behind them.

Phenylethylamine- Decarboxylation of Phe in dopaminergic neurons of the nigrostriatal system (Barroso and Rodriguez, 1996) result in the generation of PEA, the latter being a precursor to the neurotransmitter phenylethanolamine. However, some studies have indicated that PEA per se may function as a neuromodulator or neurotransmitter, stimulating the release of dopamine from the cytoplasmic pool and behaving as a dopamine receptor agonist (Raiteri et al., 1977; Barroso and Rodriguez, 1996). High levels of PEA have been found in the urine (but not in serum or CSF) of schizophrenics (Potkin et al., 1980; O’Reilly and Davis, 1994). Low levels—as those observed in carriers—have been reported in the CSF of subjects affected with Rett’s syndrome (Satoi et al., 2000) and in urine from Parkinson’s disease affected individuals (Zhou et al., 1997), children with attention deficit hyperactivity disorder (Kusaga et al., 2002) and subjects affected with depression (Sabelli et al., 1983). Some of these clinical features are shared by carriers of the premutation (Parkinsonism, ADHD, depression, autism; (Bacalman et al., 2006; Farzin et al., 2006; Berry-Kravis et al., 2007). Interestingly, the concentration of PEA or its metabolite phenylacetic acid have been proposed as useful biochemical markers in psychiatric and behavioral research (Davis and Boulton, 1994). This may also be true in neurodegenerative diseases, as a significant negative correlation between CSF concentration of PEA and severity of Parkinson’s disease (Hoehn and Yahr stage) has been reported (Zhou et al., 1997).

The observed lower PEA concentrations in carriers than controls may reflect any of these possibilities:

In the context of the premutation, the lower plasma PEA concentrations may reflect either incipient nigrostriatal degeneration in subjects without yet significant signs of Parkinsonism or a reduced modulatory effect of PEA on a postsynaptic receptor in response to dopamine.

Oleamide- Oleamide is a natural occurring fatty amide that accumulates in the CSF during sleep deprivation and induces sleep in animals. In this regard, it is being studied as a potential medical treatment for mood and sleep disorders, and cannabinoid-regulated depression. The mechanism of action of oleamide’s sleep inducing effects is not well understood yet; however, it is likely that this compound interacts with multiple neurotransmitter systems. Oleamide is structurally related to the endogenous cannabinoid anandamide. Both anandamide and oleamide elicit behavioral effects indicative of cannabinoid activity, but only anandamide binds the cannabinoid (CB1) receptor in vitro. Oleamide, anandamide, and myristic amide are degraded to the corresponding fatty acids (oleic acid, arachidonic acid, and myristic acid) by the enzyme fatty-acid amide hydrolase (FAAH) to terminate the signaling functions of these molecules (Wei et al., 2006). This enzyme is most abundant in neocortex, hippocampal formation, amygdala, and cerebellum suggesting that this CNS distribution supports the degradation of neuromodulatory fatty acid amides at their sites of action influencing their effects on sleep, euphoria, and analgesia (Thomas et al., 1997).

Oleamide has been proposed to induce its behavioral effects by serving as a competitive substrate for the brain FAAH and inhibiting the degradation of endogenous anandamide. Disruption of the endocannabinoid pathway induces metabolic imbalances (Di Marzo et al., 2001; Osei-Hyiaman et al., 2005, 2006), outcomes generally considered detrimental to any tissue but, they gain more weight in brain given its high aerobic capacity (Laughlin et al., 1998; Attwell and Laughlin, 2001) and the localization of CB1 receptors in neurons and neuronal mitochondria (Benard et al., 2012). In FAAH^+/+ and FAAH^-/- mice, oleamide induced hypomotility, hypothermia, and ptosis, all of which were enhanced in FAAH^-/- mice, with negligible binding to the CB1 receptor in brain extracts from either genotype (Lichtman et al., 2002). In contrast, anandamide exhibited a 15-fold increase in apparent affinity for the CB1 receptor in brains from FAAH^-/- mice, consistent with its pronounced CB1-dependent behavioral effects in these animals. Other reports indicated that oleamide neither directly activates CB1 receptors nor acts via the proposed “entourage” effect (Mechoulam et al., 1997) to increase concentrations of anandamide through FAAH inhibition, rather the selective effects of oleamide on theta-burst-conditioning may reflect modulation of GABAergic transmission (Lees and Dougalis, 2004). It has been proposed that genetic variations in FAAH can be associated with susceptibility to polysubstance abuse (OMIM: 606581). Indeed, a homozygous T129 variant is strongly associated with drug and alcohol abuse and methamphetamine dependence (Sipe et al., 2002; Filbey et al., 2010; Buhler et al., 2014) whereas increasing amygdala anandamide (and possibly oleamide) enables extinction-driven reductions in fear in mouse and may promote stress-coping in humans (Gunduz-Cinar et al., 2013).

Based on these reports and in the context of the premutation, lower levels of oleamide may be predictive or linked to the increased incidence of substance abuse, including alcohol (Kogan et al., 2008), mood/anxiety issues of carriers of the premutation (Kogan et al., 2008), their increased anxiety as a reflection of lower stress-coping abilities (Bourgeois et al., 2007) and a reflection of their sleep disturbance which is seen in the majority of carriers (Chonchaiya et al., 2010).

Krebs’ cycle intermediates: citrate, isocitrate and aconitate- Two intermediates of the TCA cycle, located within the first half of the cycle, namely aconitate and isocitrate, were identified as potential biomarkers along with citrate (Figures 1A,B). Their levels were higher in plasma from carriers than controls (1.4-fold of controls for aconitate and 1.2-fold for both citrate and isocitrate; p < 0.05). Increases in these tricarboxylic acids located within the first half of the cycle may indicate a slower Krebs’ cycle activity. This is based on the fact that increases in succinate, malate, and fumarate (located in the second half of the cycle) have been documented to accompany exercise, effect termed as the TCA cycle expansion (Gibala et al., 1997a,b). The “expansion” is believed to result from a relative excess of glycolysis compared to OXPHOS, which then results in the shunting of pyruvate and Glu to Ala and alpha-ketoglutarate (AKG) via Ala aminotransferase. As a result, AKG is then converted to other TCA cycle intermediates (succinate, malate, and fumarate). Thus, increases in the first half of this cycle are suggestive of lower alpha-ketoglutarate dehydrogenase (AKGDH) activity. This would result in higher AKG concentrations, which via the near-equilibrium of glutamate dehydrogenase, results in the increase of Glu concentrations. The latter is an allosteric inhibitor of glutaminase, and AKG is also an allosteric inhibitor of mitochondrial transport of Gln. Hence, an increase in AKG, due to a decreased activity of AKGDH, should lead to increases in Glu and Gln, and given that Glu is a precursor of GABA, also an increased synthesis of GABA or related metabolites. Two experimental evidences supported this option suggesting an increased flux from AKG to Gln: (i) the ratio of Gln-to-Glu was higher in plasma of premutation than controls (0.86 vs. 0.37; p < 0.05) and (ii) higher 4-hydroxybutyrate (GHB, a GABA derivative) was observed in plasma from carriers (twofold; p < 0.05) suggesting an increased synthesis of GABA from Glu and, as a result, of its product GHB.

In the context of the full and premutation, dysregulation of neurotransmitter systems, including the mGluR1/5 pathway and GABA pathways, has been reported in fragile X syndrome (Hagerman et al., 2010), neurons from a KI mouse model of the premutation (Cao et al., 2012) and in the brain of individuals with FXTAS (Pretto et al., 2014). Although not pathognomonic, the increased plasma levels of three TCA cycle metabolites (citrate, aconitate, and isocitrate) along with a higher plasma lactate (1.5-fold of controls; p = 0.095) have been observed in some mitochondrial diseases (Crippa et al., 2015), usually associated with the formation of analogs of TCA intermediates with the potential of depleting oxaloacetate and slowing down the TCA cycle, especially if not accompanied by the activation of the anaplerotic reaction catalyzed by pyruvate carboxylase. These results are consistent with the reports on mitochondrial dysfunction and abnormal network/distribution in cells from premutation carriers (Ross-Inta et al., 2010; Napoli et al., 2011, 2016c) as well as in KI mouse models of the premutation (Kaplan et al., 2012; Conca Dioguardi et al., 2016; Napoli et al., 2016a).

In the broader context, this and other studies (Fonteh et al., 2007; Griffin and Salek, 2007; Dumas and Davidovic, 2013; Tsuruoka et al., 2013; Napoli et al., 2015) highlight the potential of metabolic phenotyping to (a) identify biomarkers of neurological diseases utilizing non-invasive samples such as plasma and urine, (b) characterize metabolic profiles that contribute to accurate diagnosis in complex neurological conditions (endophenotypes), and (c) assess treatment efficacy and determine metabolic secondary side effects, especially in subjects with multidrug treatments associated with metabolic disorders.

Further studies are warranted to allow the construction of robust models for predicting neurological changes from the less invasive approach of plasma metabolic profiling and verifying the important biological roles of these key metabolites with endophenotypes (e.g., with and without Parkinsonism). If successful, the specific biochemical fingerprints could eventually be used to identifying key pathways for therapy and helping in monitoring disease progression. In this regard, the biomarker oleamide/isocitrate seemed to discriminate carriers with FXTAS from those not affected; however, future research would need to be performed to test whether this ratio can be used to diagnose more accurately stages of FXTAS as well as if it has any predictive value in terms of discriminating carriers that will develop FXTAS later in life.

CG conceptualized and designed the study, wrote the manuscript, and approved the final manuscript as submitted; EN separated the plasma from samples, contributed to the writing of the manuscript, revised and approved the final version as submitted; FT provided the CGG repeat sizing, revised the manuscript and approved the final manuscript as submitted; JH helped with the collection and classification of nutritional supplements and medications for the study group, revised and approved the manuscript as submitted; RH carried out clinical assessment of these subjects, wrote clinical findings, revised the manuscript and approved the final manuscript as submitted.

RH has received funding from Novartis, Roche/Genentech, Alcobra, and Neuren for treatment trials in fragile X syndrome, autism and Down syndrome. She has also consulted with Novartis, Zynerba and Roche/Genentech regarding treatment for fragile X syndrome. All the other authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.