Neurophysiology of the “Celiac Brain”: Disentangling Gut-Brain Connections

Celiac disease (CD) can be considered a complex multi-organ disorder with highly variable extra-intestinal, including neurological, involvement. Cerebellar ataxia, peripheral neuropathy, seizures, headache, cognitive impairment, and neuropsychiatric diseases are complications frequently reported. These manifestations may be present at the onset of the typical disease or become clinically evident during its course. However, CD subjects with subclinical neurological involvement have also been described, as well as patients with clear central and/or peripheral nervous system and intestinal histopathological disease features in the absence of typical CD manifestations. Based on these considerations, a sensitive and specific diagnostic method that is able to detect early disease process, progression, and complications is desirable. In this context, neurophysiological techniques play a crucial role in the non-invasive assessment of central nervous system (CNS) excitability and conductivity. Moreover, some of these tools are known for their valuable role in early diagnosis and follow-up of several neurological diseases or systemic disorders, such as CD with nervous system involvement, even at the subclinical level. This review provides an up-to-date summary of the neurophysiological basis of CD using electroencephalography (EEG), multimodal evoked potentials, and transcranial magnetic stimulation (TMS). The evidence examined here seems to converge on an overall profile of “hyperexcitable celiac brain,” which partially recovers after institution of a gluten-free diet (GFD). The main translational correlate is that in case of subclinical neurological involvement or overt unexplained symptoms, neurophysiology could contribute to the diagnosis, assessment, and monitoring of a potentially underlying CD.


INTRODUCTION
Celiac disease (CD) is an autoimmune disorder triggered by the ingestion of gluten that, in genetically predisposed individuals, leads to damage of the small intestine and consequent malabsorption. Most patients (95% of them) are carriers of the DQ2 or DQ8 haplotype of the histocompatibility complex class II human leukocyte antigen (Lebwohl et al., 2017). Tissue transglutaminase (tTG) is the main auto-antigen (Alaedini and Green, 2008), whereas gut histopathology shows variable degrees of small bowel mucosal villi atrophy. CD affects 0.3-1.5% of the general population (approximately 1 of 120-300 people in Europe and America) (Morello et al., 2003;Bingley et al., 2004). To date, the only established therapy is a lifetime dietary gluten restriction, which is usually followed by relief of several clinical manifestations, normalization of histological and serological markers, as well as decreased risk of associated malignant and non-malignant complications (Holmes, 1996(Holmes, , 2002. Clinically, although diarrhea and other gastro-intestinal symptoms can commonly be observed at disease onset or in its early phases in both pediatric patients and young adults, they are not as frequent as in the past (Campagna et al., 2017). Many adults (>50%) exhibit significant extra-intestinal involvement even without typical CD manifestations (Cooke and Smith, 1966;Hadjivassiliou et al., 2002aHadjivassiliou et al., , 2010Hadjivassiliou et al., , 2014Bushara, 2005;Uygur-Bayramicli and Ozel, 2011;Castillo et al., 2015). Therefore, CD is considered a complex systemic disorder with multifactorial pathogenesis that should be investigated from genetic, biological, and environmental perspectives.
Currently, little is known about the neurophysiology of central nervous system (CNS) damage in CD. This review aims to summarize the electrophysiological evidence on CNS functioning and pathology, including the clinical and instrumental response to a gluten-free diet (GFD).

NEUROLOGICAL COMPLICATIONS OF CELIAC DISEASE: A BRIEF OVERVIEW
Nowadays, it is widely accepted that the typical disease represents a small proportion of the so-called "CD iceberg, " because 5-6 fold more patients present with atypical or silent forms (Green et al., 2005). Neurological manifestations may either precede or follow the disease, or be present at its onset (Hadjivassiliou et al., 2002a(Hadjivassiliou et al., , 2010(Hadjivassiliou et al., , 2014Briani et al., 2008). Therefore, a sensitive and specific diagnostic method able to recognize early disease process, progression, and complication would be desirable.
To date, both the causative factors and pathophysiological mechanisms of neurological involvement in CD remain a matter of debate. According to the literature, the nervous system may be one of the elective sites of gluten-mediated pathogenesis, including cross-reacting antibodies, immune-complex deposition, direct neurotoxicity, other immune-mediated factors, and deficiency of vitamin and other nutrients secondary to chronic malabsorption (Zelnik et al., 2004;Bushara, 2005;Abenavoli, 2010;Parisi et al., 2015). Recently, studies using single photon emission computed tomography showed regional changes in cerebral perfusion, with regression after institution of a GFD (De Santis et al., 1997;Usai et al., 2004). The authors argued that the cerebral hypoperfusion might be related to intestinal hyperemia from immune-mediated or endothelial damage due to immune-complex deposition likely involving antibodies against gliadin (De Santis et al., 1997). Alternatively, cortical brain hypoperfusion could reflect focal vasculitis secondary to perivascular inflammation (Usai et al., 2004).

Cerebellar Ataxia
"Gluten ataxia" is one of the first recognized symptoms (Cooke and Smith, 1966) and the most frequent neurological disturbance in CD (Hadjivassiliou et al., 2015). Dysarthria, corticalspinal signs, eye and gaze movement disorders, and cerebellar ataxia are representative presentations. Recent studies showed deposits of antibodies against tTG on cerebellar blood vessels, adding support to a blood-brain barrier (BBB) dysfunction in CD (Hadjivassiliou et al., 2006(Hadjivassiliou et al., , 2015. Interestingly, gluten ataxia is not usually related to intestinal manifestations or vitamin deficiency, and improvement with a GFD is possible (Hadjivassiliou et al., 2006).

Peripheral Neuropathy
Peripheral neuropathy is the second most common neurological manifestation of CD (up to half of patients) after cerebellar ataxia, and can appear even before diagnosis (Chin et al., 2003;Chin and Latov, 2005). Studies on the effect of a GFD on peripheral neuropathy are conflicting, with some authors reporting clinical improvement whereas others concluding a lack of relevant response (Luostarinen et al., 2003;Siqueira Neto et al., 2004). A previous study on 32 consecutive adult patients complaining of peripheral neuropathy, autonomic dysfunction, or both, and showing anti-neuronal antibodies found no response despite the adoption of a GFD (Tursi et al., 2006).

Epilepsy
A bidirectional link between epilepsy and CD has been established in several studies, although not all (Vieira et al., 2013), with rates of prevalence from 3.5 to 7.2% (Cooke and Smith, 1966;Zelnik et al., 2004;Bushara, 2005;Uygur-Bayramicli and Ozel, 2011;Hadjivassiliou et al., 2014;Parisi et al., 2015). A large population-based cohort study observed an increased risk of CD in subjects of all ages, including children, even when epilepsy was independently restricted to patients receiving the diagnosis of epilepsy and those with prescriptions of antiepileptic drugs (Ludvigsson et al., 2012). The hypotheses accounting for epilepsy in CD included a gluten-mediated toxicity, an immune-induced cortical damage, the presence of cerebral calcifications, and vitamins/trace elements malabsorption. GFD usually controls seizures refractory to antiepileptic drugs (Hadjivassiliou et al., 2002a;Canales et al., 2006). CD-related progressive ataxia is associated with stimulus-sensitive myoclonus, opsoclonusmyoclonus, and sometimes with seizures (Borg, 2006;Deconinck et al., 2006).

Headache
It has been reported that a GFD results in complaints of less severe headache symptoms by celiac patients (Hadjivassiliou et al., 2001). Accordingly, structural and functional neuroimaging studies were in favor of an association between migraine and CD, with relief after gluten restriction (Hadjivassiliou et al., 2001). However, when headaches in CD patients were compared with those in the general population, a conclusive association was not proven (Nikpour, 2012).

Cognitive Impairment and Dementia
Adult CD patients often complain of mild cognitive symptoms called "brain fog, " which improves when gluten-restriction is started, but re-appears with dietary contamination (Lichtwark et al., 2014;Yelland, 2017). Concentration and attention difficulties, episodic memory deficits, word-retrieval problems, reduced mental acuity, and episodes of confusion or disorientation are the commonly reported features (Lurie et al., 2008). In some severely affected patients, dementia can develop as acalculia, confusion, amnesia, and personality disorders (Collin et al., 1991;Hu et al., 2006;Lurie et al., 2008;Casella et al., 2012). Despite long-term administration of a GFD, patients older than 65 years exhibited worse cognitive performance than age-and sex-matched controls (Casella et al., 2012).
Reactive anxiety that usually ameliorates with a GFD is the predominant form of anxiety disorder in these patients. Depressive disturbances, which affect a relevant number of subjects, may significantly impair quality of life, and are a good predictor of lack of dietary compliance (Zingone et al., 2010). Therefore, screening patients for depression is of pivotal importance both at diagnosis and follow-up in order to advice psychological support and/or pharmacological therapy. Possible causative factors of mood disorders might be tryptophan deficiency secondary to chronic malabsorption (Hallert et al., 1982;Hernanz and Polanco, 1991) or co-morbidity with thyroid disease (Carta et al., 2002). Decreases in levels of serotonin, dopamine, and noradrenaline metabolites in cerebrospinal fluid as well as tryptophan and other monoamine precursors in serum were observed in untreated patients (Hallert et al., 1982;Hernanz and Polanco, 1991).
Clinical improvement was reported only after long-term administration of a GFD (>5 years) (van Hees et al., 2013), highlighting the importance of prolonged alimentary restriction on extra-intestinal CD symptoms as well.

NEUROPHYSIOLOGICAL TECHNIQUES USED TO PROBE CNS INVOLVEMENT IN CELIAC DISEASE Electroencephalography
The spectrum of electroencephalography (EEG) features associated with CD is rather wide, although focal activity in terms of unilateral or bilateral spike or slow waves, mainly localized in the occipital regions, have been reported in most of the wakefulness EEG studies (Magaudda et al., 1993;Labate et al., 2001;Pratesi et al., 2003;Ranua et al., 2005;Lionetti et al., 2010;Licchetta et al., 2011;Aksoy et al., 2016). However, as recommended (Parisi et al., 2014), EEG patterns should not be considered disease-specific.
A recent prospective study among 307 CD children compared with 197 age-and sex-matched controls observed that patients were more prone to epileptiform activities on EEG (spike/sharpwave discharges, especially in the occipital lobes but also in the central-temporal sites and in diffuse distribution). However, early and strict adherence to a GFD effectively decreased these findings (Işıkay et al., 2015a). In addition, a positive correlation between tTG level and epileptiform changes during sleep and awake EEG was found (Işıkay et al., 2015a). The concept that the occipital region is frequently involved in CD seems to be supported by the evidence of occipital calcium deposition, occipital lobe semiology, and EEG findings. The preferential involvement of this lobe may lie on several factors, such as its vulnerability to some metabolic circumstances (e.g., hypoglycemia and hypoxia) and its thinner morphological structure than other cortical regions (Işıkay et al., 2015b). However, the opposite scenario (the occurrence of CD in patients with "posterior" epileptic semiology) is not always true because in a group of 90 pediatric epileptic patients with occipital EEG abnormalities, tTG antibody was positive in only two (Canales et al., 2006). CD-associated epilepsy has also been reported in association with other neurological signs, such as ataxia, tremor, and progressive myoclonus (Javed et al., 2012;Sarrigiannis et al., 2014). In these cases, epilepsy was usually refractory, and EEG demonstrated spike and waves in the right anterior and midtemporal lobes, as well as bilateral slow and sharp waves. Some of these spike waves were present in association with localized jerks of the upper or lower limb, although without periodic complexes (Javed et al., 2012;Sarrigiannis et al., 2014). Finally, the fixationoff sensitivity phenomenon could be observed (Casciato et al., 2015).
In regard to the impact of dietary restriction, a recent study in 19 children with biopsy-proven CD revealed abnormal EEG findings in 48% of them that were no longer evident in most of the patients after 6 months of GFD (Parisi et al., 2015). However, some asymptomatic children and adolescents still manifested hyperexcitability to EEG despite the diet (Parisi et al., 2015).
In summary, CD screening should be performed in patients with cryptogenic and/or refractory epilepsy, or in the presence of unexplained EEG findings. Gluten restriction is usually effective in ameliorating the clinical-instrumental correlates.

Evoked Potentials
Few studies, most of which used somatosensory evoked potentials (SEPs), have explored CD with evoked potentials. Di Lazzaro and co-workers reported a patient whose lower limb SEPs presented enlargement of lumbar waves and bilateral lack of cortical responses, suggesting an impaired somatosensory conduction along the spinal dorsal columns; GFD induced complete clinical-instrumental recovery (Di Lazzaro et al., 2010). Another case presented one child with prolonged central conduction time among 27 treated children (Cakir et al., 2007). In patients with cerebellar ataxia associated with subclinical CD responding to a GFD, normal SEPs were reported (Pellecchia et al., 1999). However, in a large cohort of adult ataxic patients, more than half had loss or delayed P40 cortical response, suggesting dorsal column degeneration (Bürk et al., 2001). In regard to SEPs in cortical mycolonus, a previous study in two CDproven subjects with myoclonic ataxic syndrome showed giant and time-locked cortical responses that preceded the myoclonus (Tijssen et al., 2000). The authors speculated that, in spite of the neurophysiological evidence of cerebral cortical involvement, the hyperexcitability was mainly located in the cerebellum, and that the effects on sensory-motor cortex represented a remote influence from cerebellar dysfunction (Tijssen et al., 2000). Conversely, the cortical electrophysiological origin of the myoclonus was argued by other researchers who found their patients responded poorly to a GFD and worsened progressively (Lu et al., 1986;Tison et al., 1989;Bhatia et al., 1995).
Although CD may impact the auditory system (Bürk et al., 2001), brainstem auditory evoked potentials (BAEPs) and vestibular evoked myogenic potentials were reported to be normal (Pawlak-Osińska et al., 2007). More recently, it was found that only 1 of 25 patients had abnormalities in BAEPs in terms of moderate sensorineural hearing loss (Aksoy et al., 2016).
Complications affecting visual pathways may develop in CD and be evidenced by visual evoked potentials (VEPs) (Bürk et al., 2001;Freeman, 2008;Hadjivassiliou et al., 2010). In particular, patients can show abnormalities on VEPs without evident lesions at neuroimaging (Aksoy et al., 2016). A previous case described a slight increase of P100 wave latency bilaterally at pattern-reversal VEPs that reverted back to normal after a GFD (Pellecchia et al., 1999). Given their role in detecting even preclinical pathology in subjects with normal ophthalmological and brain imaging exams, VEPs may provide useful insights in neurologically asymptomatic CD patients (Aksoy et al., 2016).

Transcranial Magnetic Stimulation
In 1999, Pellecchia and co-workers first reported motor evoked responses to transcranial magnetic stimulation (TMS) in a CD patient who exhibited reduced amplitude in the rectus femoris muscle that improved with diet; however, motor responses remained undetectable in the tibialis anterior muscle (Pellecchia et al., 1999). A year later, report of delayed motor response in the left tibialis anterior muscle and abnormal cortical inhibition was published in one of three CD patients with cortical myoclonus (Tijssen et al., 2000).
Specific TMS studies before and after GFD were published more recently. TMS is an electrophysiological tool able to noninvasively explore the excitation state of motor cortical areas and conductivity of the pyramidal tract in vivo. Moreover, it is capable of unveiling subclinical central motor involvement in different neurological and psychiatric diseases or systemic disorders affecting the CNS (Bella et al., 2011a(Bella et al., ,b, 2016Pennisi et al., 2011aPennisi et al., ,b, 2015Pennisi et al., , 2017Concerto et al., 2013;Lanza et al., 2013Lanza et al., , 2015aLanza et al., ,b, 2017aCantone et al., 2014Cantone et al., , 2017, also providing prognostic Pennisi et al., 2016) and therapeutic implications Concerto et al., 2015;Bordet et al., 2017). Lastly, the so-called "pharmaco-TMS" can selectively probe the functioning of different central neurotransmission pathways, such as glutamate, gammaaminobutyric-acid (GABA), monoamine, and acetylcholine, by testing their pharmacological agonists or antagonists (Paulus et al., 2008;Ziemann et al., 2015).
The first TMS study investigated 20 de novo CD patients without apparent neurological involvement and 20 age-matched controls (Pennisi et al., 2014). TMS revealed cortical motor disinhibition and hyperfacilitation, which is a profile compatible with dysfunctional GABAergic and glutamatergic transmissions, in patients. The authors hypothesized that an imbalance of excitatory and inhibitory circuits within the motor cortex might be the neurochemical correlate of the cross-interaction between antibodies against gliadin and specific neuronal antigens. An alternative explanation was the deposition of tTG-immunoglobulin leading to an abnormal ion levels across neuronal membrane. Likewise, antibodies synthesized within the CNS and directed against glutamic acid decarboxylase might disrupt the functioning of GABAergic interneurons (Pennisi et al., 2014).
The same cohort of patients underwent re-evaluation after a relatively short time of a GFD (median of 16 months) . Their gastrointestinal symptoms were ameliorated but, unexpectedly, the cortical excitability to TMS further increased. This finding was hypothesized to represent a plastic re-organization of the cerebral cortex triggered by gluten exposure and independent of GFD. On the other hand, diet duration or compliance might not have been enough to induce an adequate remission . A recent cross-sectional TMS study after a much longer GFD (mean period of 8.35 years) showed that a more prolonged period of gluten restriction was required to revert the cortical changes in adult CD patients. Nevertheless, regardless of diet, some specific excitatory features to TMS remained, probably suggesting an intracortical synaptic rearrangement, mostly involving glutamate-mediated interneurons .

DISCUSSION
The main translational value of this review is that clinical neurophysiology can contribute to the diagnosis, assessment, and monitoring of CD even in patients with subclinical CNS involvement or unexplained neurological symptoms (Table 1,  Figure 1). In particular, the majority of electrophysiological changes are often subclinical ("celiac iceberg"), and these need to be strictly monitored because of the possibility of progression to clinically visible neurological syndrome in both young and adult patients ("symptomatic celiac disease"). Accordingly, it has been shown that there is an increased risk of neurological complications in atypical or silent CD forms, especially in older patients or those older at diagnosis (Aksoy et al., 2016). It is worth noting that, despite their valuable role, anti-ganglioside antibodies and neuronal antigens are not always specifically linked to the neurological manifestations and their progression        in the course of CD (Kaplan et al., 1988;Gobbi et al., 1992;Hadjivassiliou et al., 1998Hadjivassiliou et al., , 2002bAksoy et al., 2016). From a pure neurophysiological perspective, findings from EEG, SEPs, and TMS seem to converge on an overall profile of "hyperexcitable celiac brain, " albeit this may not be confined to the cerebral cortex. Indeed, an increase in cerebral cortical excitability may arise from enhanced inputs from the cerebellum (Tijssen et al., 2000). In this context, it is important to remember that malabsorption syndrome (with the consequent deficiency of vitamins and other nutrients) probably does not account for these cortical manifestations, given that it takes place in the most severely affected patients whose intestinal mucosa are seriously damaged and do not recover after institution of a GFD (Pennisi et al., 2014).
Regarding humoral autoimmunity to neuronal antigens, deposits of anti-tTG2 and anti-tTG6 antibodies have been found not only in the small intestine but also in different CNS sites (cerebellum, pons, medulla, brain blood vessels) (Hadjivassiliou et al., 2008). Furthermore, a possible BBB lesion, secondary to diffuse infiltration of T-lymphocytes and inflammatory cells within the perivascular cuffing might expose cerebral tissues to antibodies (Hadjivassiliou et al., 2010). The result may be a vicious circle that eventually leads to a prevailing synaptic hyperexcitation and a weaker inhibition at the cortical level (Pennisi et al., 2014). The increased excitability may also be the correlate of a glutamate-induced cortical rearrangement or a dysfunctional control of GABAergic inhibitory interneurons. In particular, because glutamate is of pivotal importance in synaptic plasticity, it can be speculated that immune system dysregulation triggered by gluten ingestion, might result in a long-standing activation of post-synaptic glutamate receptors accounting for the enhanced hyperexcitability .
The neurophysiological-based approach to CD should take into account potential pitfalls and critical aspects related to both the techniques themselves and methodological biases in the studies reviewed here. First, as mentioned, electrophysiological changes are not disease-specific. Second, an association finding does not mean causative relationship. For instance, an association between CD and amyotrophic lateral sclerosis was previously reported in different investigations (Turner et al., 2007(Turner et al., , 2013Brown et al., 2010;Bersano et al., 2015;Gadoth et al., 2015) but not confirmed in a large population-based cohort study (Ludvigsson et al., 2014). Finally, it is mandatory to make a differential diagnosis between hyperexcitability-related seizures and incidental EEG findings in neurologically asymptomatic CD subjects. In the latter case, EEG changes represent a confounding factor and a long follow-up is required (Parisi et al., 2014).
The response of neurological symptoms to a GFD is still controversial. Current knowledge encompasses an initial phase when patients are "gluten-sensitive" and a subsequent stage characterized by "gluten-insensitivity" (Tursi et al., 2006). An older age at diagnosis or a prolonged period of gluten ingestion may account for persistent neurological symptoms after a relatively short period of GFD . Moreover, gluten restriction is not usually effective in patients with refractory CD and in those with an associated autoimmune disease or some neurological complications (Hadjivassiliou et al., 2010;Castillo et al., 2015;Campagna et al., 2017). It is reasonable to conclude that some neurological aspects improve after diet restriction whereas others persist, supporting the concept that the more prolonged the GFD, the more likely clinical and neurophysiological remission may occur. However, given that neurological impairment may develop despite an adequate adherence to a GFD (Luostarinen et al., 2003;Chin and Latov, 2005;Tursi et al., 2006;Bürk et al., 2009), other causative factors have to contribute (McKeon et al., 2014): (a) accidental minimal gluten contamination despite a good dietary compliance (Green and Jabri, 2003); (b) direct gliadin-mediated inflammatory attack; (c) other components that are independent of GFD (Tijssen et al., 2000).
Based on further understanding of the pathogenesis and treatment of CD, neurophysiology-targeted non-dietary therapies are in development (Schuppan et al., 2009). Similarly, modern rehabilitation approach involves organizational measures that promote not only clinical recovery but also a better quality of life (Sabel'nikova et al., 2013) through the support of different medical and non-medical specialists (Usanova et al., 2012).
In conclusion, neurophysiology, together with clinical, serological, and imaging data, can help in disentangling the multifaceted physiopathological and neurobiological mechanisms coupling gut and brain in CD. The eventual identification of neurophysiological markers might be useful in the diagnosis and monitoring of CD, aiming to improve the healthcare of both single subjects and the global community.

AUTHOR CONTRIBUTIONS
All authors provided substantial contributions to the conception, drafting, critical revision for important intellectual content, final approval, and agreement to be accountable for all aspects of the work. In particular, MP and AB conceived and designed the study, MC and GL reviewed the literature and drafted the manuscript, and GP and RB critically reviewed and finalized the paper.