Anno accademico 2020-2021
UNIVERSITÀ DEGLI STUDI DI GENOVA
SCUOLA DI SCIENZE MEDICHE E FARMACEUTICHE
CORSO DI LAUREA IN MEDICINA E CHIRURGIA
A clinical and metabolomic
fingerprint of potential dysbiosis in patients with epilepsy
Prof. Pasquale Striano Correlatore:
Dott.ssa Antonella Riva
1. Introduction 4
1.1 The microbiota-gut brain (MGB) axis 5
1.2 Communication pathways 7
1.3 Metabolites in the GI tract 11
1.4 Dysbiosis due to alteration of the gut microbiota 16
1.5 Prebiotics, probiotics and Ketogenic diet 18
1.6 Urinary metabolites in epileptic patients 22
2. Aim of the study 35
3. Patients and Methods 36 3.1 Patient selection 36
3.2 Statistical analysis 37
3.3 Chemicals 37
3.4 Mass spectrometry analysis of urine samples 38
3.5 Untargeted metabolomics analysis 40
3.6 Univariate analysis 40
3.7 Principal component analysis (PCA) 41
3.8 Partial least squares discriminant analysis 41
3.9 Pathway analysis 42
4. Results 44
4.1 Clinical data of the studied population 44
4.2 Statistical analysis 45
4.3 Metabolomics Analysis between IE and Epi+ Subgroups 47
4.4 Metabolomics Analysis between drug-R and drug-S epilepsy subgroups 53
5. Discussion and Conclusions 57
6. References 65
7. Supplementary materials 93
8. Acknowledgements 100
Introduction: Patients with neurodevelopmental disorders, including epilepsy, may show abnormal intestinal functioning ranging from mild bloating to severe constipation. Our study aimed to identify a clinical and metabolomic fingerprint of the gut dysbiosis in patients with epilepsy.
Methods: We recruited patients with epilepsy of broad aetiologies and age-matched, neurotypical controls. Gastrointestinal (GI) function was assessed using the Bristol Stool Test (BST) chart and the validated Rome IV Diagnostic Questionnaire.
Untargeted metabolomic analysis was performed on urine samples.
Results: 148 non-related individuals (mean age, 9.4±3.9 years) were enrolled. The epilepsy group included 84 patients with a mean age of 9.3±4.5 years. Thirty patients showed Isolated Epilepsy (IE), while 54 had Epilepsy plus neuropsychiatric comorbidities (Epi+). BST scores were significantly abnormal in epilepsy patients as compared to controls (p=.0026). Comparison of BST scores between patients with or without GI symptoms showed a value of p=.0001. Metabolomic analysis of urine samples revealed specific metabolic profiles associated with the epilepsy subtype and involving specific pathways including those of ABC transporters, and metabolism of amino acids, butyrate, vitamin B6, lysine, caffeine and nicotinamide adenine dinucleotide (NAD+).
Conclusions: This study supports the implementation of a clinical and metabolomic fingerprint of gut dysbiosis in patients with epilepsy, and eventually provides a basis for the optimization of patients’ treatment.
In the last decades, tremendous effort has been made to characterize the bidirectional communication between the central nervous system (CNS), the enteric nervous system, and the gastrointestinal (GI) tract. More recently, a series of preclinical studies have suggested a pivotal role of the gut microbiota in these gut-brain interplay. In rodents raised in a germ-free environment, the gut microbiota appears to influence the development of emotional behavior, stress and pain modulation, as well as the brain neurotransmitter systems through endocrine and neurocrine pathways and, also, that the brain can alter microbial composition and behavior via the autonomic nervous system (1). Future studies will focus on understanding the mechanisms underlying the microbiota-gut brain (MGB) axis and attempt to elucidate microbial-based interventions and therapeutic strategies for neuropsychiatric disorders. (2) Currently, there are multiple interventions acting on the gut microbiota which could be used in a wide range of neurodevelopmental disorders including epilepsy. Prebiotics, probiotics, diet, symbiotic or fecal microbiota transplantation (FMT) can dramatically reduce the use of more invasive treatments such as epilepsy surgery or vagus nerve stimulation (VNS). These “alternative” therapeutic strategies could be particularly interesting in those patients who do not respond to common anti-seizure medications (ASMs), which are up to 25-30% and defined as having drug-resistant epilepsy (DRE).
1.1 The microbiota-gut brain (MGB) axis
The brain and the GI tract are intimately connected to form a bidirectional neuro- humoral communication system, known as the gut-brain axis. Research on the MGB axis were traditionally focused on the psychological status affecting the function of the GI tract. However, recent evidence showed that the gut microbiota communicates with the brain via the MGB axis to modulate brain development and behavioral phenotypes. To elucidate the role of gut microbiota in the gut-brain axis, precise identification of the composition of microbes is an essential step. (3)
The intestinal microbiota is the collection of all the microorganisms in the GI tract and consists of more than 100 billion (10 times the number of cells in the human body) of 1000 different species of microorganisms. Currently, four major microbial phyla are known to represent over 90% of the bacterial component of gut microbiota:
Firmicutes, Bacteroides, Proteobacteria and Actinobacteria. Most of the “good”
bacteria harboring the human gut microbiota are represented by Firmicutes and Cytophaga-Flavobacterium-Bacteroides (CFB). Firmicutes are sub-grouped in Clostridium coccoides (Clostridium cluster XIVa) and Clostridium leptum (Clostridium cluster IV); whereas CFB group is represented mainly by Bacteroides phyla with a great number of Prevotella and Porphyromonas. Moreover, our gut- microbiota includes viruses, especially phages, Eukarya, as Fungi, Blastocystis, Amoebozoa, and Archaea. (4)
The microbiota composition is subject to shaping by host and environmental selective pressures. To protect from injury and maintain homeostasis, the GI tract limits
exposure of the host immune system to the microbiota by recruitment of a multifactorial and dynamic intestinal barrier. The barrier comprises several integrated components including physical (the epithelial and mucus layers), biochemical (enzymes and antimicrobial proteins) and immunological (IgA and epithelia- associated immune cells) factors (5).
Figure 1. Immune mechanisms that limit bacteria– epithelial cell interactions. Several immune mechanisms
work in concert to limit contact between the dense luminal microbial community and the intestinal epithelial cell surface. Goblet cells secrete mucin glycoproteins that assemble into a thick, stratified mucus layer.
Bacteria are abundant in the outer mucus layer, whereas the inner layer is resistant to bacterial penetration.
Epithelial cells (such as enterocytes, Paneth cells and goblet cells) secrete antimicrobial proteins that further help to eliminate bacteria that penetrate the mucus layer. Plasma cells secrete IgA that is
transcytosed across the epithelial cell layer and secreted from the apical surface of epithelial cells, limiting numbers of mucosa-associated bacteria (5).
An individual microbe's longevity is determined by whether it is contributing to the range of essential functions on which host fitness relies. It is proposed that organisms who do not contribute beneficial functions are controlled by, and may occasionally be purged during, for example, transferal of the microbiota to a new host (6,7).
Some of the most important roles of these microbes are to help to maintain the integrity of the mucosal barrier, to provide nutrients such as vitamins or to protect against pathogens. In addition, the interaction between commensal microbiota and the mucosal immune system is crucial for proper immune function.
1.2 Communication pathways
This bidirectional communication network includes the central nervous system (CNS), both brain and spinal cord, the autonomic nervous system (ANS), the enteric nervous system (ENS) and the hypothalamic-pituitary-adrenal (HPA) axis. The autonomic system, with the sympathetic and parasympathetic limbs, drives both afferent signals, arising from the lumen and transmitted through enteric, spinal and vagal pathways to CNS, and efferent signals from CNS to the intestinal wall. The HPA axis is considered the core stress efferent axis that coordinates the adaptive responses of the organism to stressors of any kind (8). It is a part of the limbic system, a crucial zone of the brain predominantly involved in memory and emotional responses. Environmental stress, as well as elevated systemic pro-inflammatory
cytokines, activates this system that, through secretion of the corticotropin-releasing factor (CRF) from the hypothalamus, stimulates adrenocorticotropic hormone (ACTH) secretion from the pituitary gland that, in turn, leads to cortisol release from the adrenal glands. Cortisol is a major stress hormone that affects many human organs, including the brain. Thus, both neural and hormonal lines of communication combine to allow the brain to influence the activities of intestinal functional effector cells, such as immune cells, epithelial cells, enteric neurons, smooth muscle cells, interstitial cells of Cajal and enterochromaffin cells. These same cells, on the other hand, are under the influence of the gut microbiota (9) whose contributing role in brain-gut reciprocal communications has recently been assessed.
Figure 2. Interactions between the gut microbiota and food, the GI tract, the vagus nerve, the
immune system, the CNS and the endocrine system. Image taken from Carmen de Caro et al. “Can we seize the gut-microbiota to treat epilepsy?”. Elsevier. 2019
Dysbiosis also occurs in functional gastrointestinal disorders (FGID) that are highly associated with mood disorders and are linked to a disruption of the MGB axis (10- 12). Data have been provided that both brain-gut and gut-brain dysfunctions occur, the former being dominant particularly in irritable bowel syndrome (IBS) (13). The disruption occurring in the MGB axis determines changes in intestinal motility and secretion, causes visceral hypersensitivity and leads to cellular alterations of the entero-endocrine and immune system. Microbiota may interplay with multiple of these different pathophysiological IBS targets (14) and its role is supported by varying lines of evidence: the presence in IBS patients of alterations in microbiota composition with defects both in its stability and diversity, the development of post- infectious IBS, the possible coexistence with small intestinal bacterial overgrowth and the efficacious treatment of certain probiotics and non-systemic antibiotics (15- 17). Furthermore, the visceral hypersensitivity phenotype, characteristic of IBS, can be transferred via microbiota of IBS patients to previously germ-free rats (18). The concomitant dysregulation of both the MGB axis and the gut microbiota in the pathogenesis of IBS has led to the proposal of considering this FGID as a disorder of the microbiome and the MGB axis (19).
The impact of the microbiota on the axis has been further supported by studies finalized to the manipulation of the gut microbiota using probiotics and/or antibiotics.
These studies also confirm that microbiota affects anxiety and the HPA system by influencing brain neurochemistry (20). Chronic treatment with Lactobacillus rhamnosus JB-1 induced region-dependent alterations in GABA mRNA in the brain.
In comparison to mice with a controlled diet, GABAB1b increased in cortical cingulate and prelimbic regions while concomitantly decreased in the hippocampus, amygdala, and locus coeruleus. In turn, GABAAα2 mRNA expression was reduced in the prefrontal cortex and amygdala but increased in the hippocampus. The probiotics, in parallel, reduced stress-induced release of cortisol, anxiety and depression-related behavior (21). Similarly, transient alteration of microbiota composition, obtained by administration of oral antimicrobials (neomycin, bacitracin, and pimaricin) in specific-pathogen-free mice, increased exploratory behavior and hippocampal expression of the brain derived neurotrophic factor (BDNF) (22). Furthermore, changing the microbiota composition with the probiotics association VSL#3 (which contains Streptococcus thermophilus BT01, Bifidobacterium BB02, Bifidobacterium longum BL03, Bifidobacterium infantis BI04, Lactobacillus acidophilus BA05, Lactobacillus plantarum BP06, Lactobacillus paracasei BP07, Lactobacillus delbrueckii subsp. bulgaricus BD08) leads to an increase in BDNF expression, attenuation of age-related alterations in the hippocampus (23), and reversion of neonatal maternal separation-induced visceral hypersensitivity in a rat model of IBS (24). In this latter model of stress, a change in the expression of subsets of genes involved in pain transmission and inflammation has also been described, which was reset by the early life administration of probiotics.
Evidence indicates that the microbiota communication with the brain involves the vagus nerve, which transmits information from the luminal environment to CNS.
Neurochemical and behavioral effects were not present in vagotomized mice, identifying the vagus as the major modulatory constitutive communication pathway between microbiota and the brain (21). In a model of chronic colitis associated with anxiety-like behavior, the anxiolytic effect obtained with treatment with Bifidobacterium longum was absent in mice that were vagotomized before the induction of colitis (25).
The microbiota may interact with the MGB axis through different mechanisms, the principal one likely being modulation of the intestinal barrier, whose perturbation can influence all the underlying compartments. Probiotic species-specific central effects are indeed associated with restoration of tight-junction integrity and the protection of intestinal barrier, as recently reported in an animal model of water avoidance stress.
1.3 Metabolites in the GI tract
Colonic bacteria express carbohydrate-active enzymes, which provide them with the ability to ferment complex carbohydrates generating metabolites such as short-chain fatty acids (SCFAs) (26). Three predominant SCFAs, propionate, butyrate and acetate, are typically found in a proportion of 1:1:3 in the GI tract (27). These SCFAs are rapidly absorbed by epithelial cells in the GI tract where they are involved in the regulation of cellular processes such as gene expression, chemotaxis, differentiation, proliferation and apoptosis (28). Acetate is produced mostly by gut anaerobes,
whereas propionate and butyrate are produced by different subsets of gut bacteria following distinct molecular pathways (29). Butyrate is produced from carbohydrates via glycolysis and acetoacetyl-CoA, whereas two pathways, the succinate or propanediol pathway, are known for the formation of propionate, depending on the nature of the sugar (29). In the human gut, propionate is mainly produced by Bacteroidetes, whereas the production of butyrate is dominated by Firmicutes (29,30,31). Akkermansia muciniphila is a key propionate producer specialized in mucin degradation (32). Propionate is primarily absorbed by the liver, whilst acetate is released into peripheral tissues (33).
The role of SCFAs on human metabolism has recently been reviewed (31,34).
Butyrate is known for its anti-inflammatory and anticancer activities (31,34). It is a particularly important energy source for colonocytes (28). A decreasing gradient of butyrate from lumen to crypt is suggested to control intestinal epithelial turnover and homeostasis by promoting colonocyte proliferation at the bottom of crypts, whilst increasing apoptosis and exfoliation of cells closer to the lumen (35). Butyrate can attenuate bacterial translocation and enhance gut barrier function by affecting tight- junction assembly and mucin synthesis (31). Moreover, SCFAs appear to regulate hepatic lipid and glucose homeostasis via complementary mechanisms. In the liver, propionate can activate gluconeogenesis, whilst acetate and butyrate are lipogenic.
SCFAs also play a role in regulating the immune system and inflammatory response (31). They influence the production of cytokines, for example, stimulating the production of IL-18, an interleukin involved in maintaining and repairing epithelial
integrity (28). Butyrate and propionate are histone deacetylase inhibitors that epigenetically regulate gene expression (31,34). SCFAs have also been shown to modulate appetite regulation and energy intake via receptor-mediated mechanisms (36). Propionate has beneficial effects in humans acting on β-cell function (37) and attenuating reward-based eating behavior via striatal pathways (38). Microbial metabolites other than SCFAs have been reported to have an impact on intestinal barrier functions, epithelium proliferation and the immune system (39).
The gut microbiota is also crucial to the de novo synthesis of essential vitamins which the host is incapable of producing (40). Lactic acid bacteria are key organisms in the production of vitamin B12, which cannot be synthesized by either animals, plants or fungi (40,41). Bifidobacteria are the main producers of folate, a vitamin involved in vital host metabolic processes including DNA synthesis and repair (42). Further vitamins, which the gut microbiota have been shown to synthesize in humans, include vitamin K, riboflavin, biotin, nicotinic acid, panthotenic acid, pyridoxine and thiamine (43). Colonic bacteria can also metabolize bile acids that are not reabsorbed for biotransformation to secondary bile acids (44). All these factors will influence host health. For example, an alteration of the co-metabolism of bile acids, branched fatty acids, choline, vitamins (i.e., niacin), purines and phenolic compounds has been associated with the development of metabolic diseases such as obesity and type 2 diabetes (45).
There are many lines of evidence in support of a role for the gut microbiota in influencing epithelial homeostasis (46). Germ-free mice exhibit impaired epithelial
cell turnover which is reversible upon colonization with microbiota (47). A role has been demonstrated for bacteria in promoting cell renewal and wound healing, for example, in the case of Lactobacilli rhamnosus GG (48).
The gut microbiota is also important for the development of both the intestinal mucosal and systemic immune system as demonstrated by the deficiency in several immune cell types and lymphoid structures exhibited by germ-free (GF) animals. A major immune deficiency exhibited by germ-free animals is the lack of expansion of CD4+ T-cell populations. This deficiency can be completely reversed by the treatment of GF mice with polysaccharide A from the capsule of Bacteroides fragilis (49). This process is mainly performed via the pattern recognition receptors (PRRs) of epithelial cells, such as Toll-like or Nod-like receptors, which can recognize the molecular effectors that are produced by intestinal microbes. These effectors mediate processes that can ameliorate certain inflammatory gut disorders, discriminate between beneficial and pathogenic bacteria or increase the number of immune cells or PRRs (50). SFB, a class of anaerobic and clostridia-related spore- forming commensals present in the mammalian GI tract, actively interacts with the immune system (51). Unlike other commensal bacteria, SFBs are closely associated with the epithelial lining of the mammalian GI tract membrane, which stimulates epithelial cells to release serum amyloid A1 (39). Colonization with SFB may also direct post-natal maturation of the gut mucosal lymphoid tissue, trigger a potent and broad IgA response, stimulate the T-cell compartment and up-regulate intestinal innate defense mediators, suggesting immune-stimulatory capacities of SFB (34).
A. muciniphila has been correlated with protection against several inflammatory diseases (52,53,54-58), suggesting that this strain possesses anti-inflammatory properties although the underlying mechanisms have not been completely elucidated (59). Individuals with Chron disease (CD) display mucosal dysbiosis characterized by reduced diversity of core microbiota and lower abundance of F. prausnitzii (60). F.
prausnitzii monitoring may therefore serve as a biomarker to assist in gut disease diagnostics (61). Recently, an anti-inflammatory protein from F. prausnitzii was shown to inhibit the NF-κB pathway in intestinal epithelial cells and prevent colitis in an animal model (62).
The physical presence of the microbiota in the GI tract also influences pathogen colonization by, for example, competing for attachment sites or nutrient sources, and by producing antimicrobial substances (63). Antibiotics have a profound impact on the microbiota that alters the nutritional landscape of the gut and leads to the expansion of pathogenic populations (64). For example, Salmonella Typhimurium and Clostridium difficile utilize fucose and sialic acid. The latter increases after antibiotic treatments and could hence favor the proliferation of these bacteria within the gut (65). Enterohaemorrhagic E. coli has also been shown to access fucose or sialic acid liberated by the gut microbiota from mucins (66). Dietary fiber deficiency, together with a fiber-deprived, mucus-eroding microbiota, promotes greater epithelial access and lethal colitis by the mucosal pathogen Citrobacter rodentium in mice (67).
The gut microbiota, via its structural components and metabolites, also stimulates the host to produce various antimicrobial compounds. These include AMPs such as
cathelicidins, C-type lectins and (pro)defensins by the host Paneth cells via a PRR- mediated mechanism (68). The other mechanism by which the gut microbiota can limit pathogen overgrowth is by inducing mucosal SIgA (69). Induction of SIgAs directed against gut commensal bacteria occurs via an M-cell-mediated sampling mechanism (70). SIgAs are then anchored in the outer layer of colonic mucus through combined interactions with mucins and gut bacteria, thus providing immune protection against pathogens whilst maintaining a mutually beneficial relationship with commensals (71). PRR–MAMP (pattern recognition receptor–microbe- associated molecular patterns) crosstalk results in activation of several signaling pathways that are essential for promoting mucosal barrier function and production of AMPs, mucins and IgA, contributing to host protection against invading pathogens and preventing the overgrowth of the commensals themselves (72).
1.4 Dysbiosis due to alteration of the microbiota
Gut dysbiosis can have a multitude of consequences on the central nervous system (CNS). Microbes can stimulate the release of small molecules, such as cytokines, and produce metabolites serving as neuromodulators, such as SCFAs, γ-aminobutyric acid (GABA), and serotonin precursors (73-75). Reduced levels of SCFAs can increase blood-brain barrier (BBB) permeability, thereby promoting neuroinflammation (76). Similarly, gut dysbiosis can decrease the production of claudins, which seal and render the intestinal lining impermeable to certain
compounds. With increased permeability, there can be leakage of microorganisms, their metabolites (e.g., inflammatory cytokines), and toxins (77).
Studies have suggested that the gut microbiome impacts neurological development and function (78). The gut microbiome, or colonization with good bacteria A.
mucinophilia and Parabacteroides, provided protective benefits against seizures by altering brain neurotransmitter levels, including GABA and glutamate in the hippocampus. Bacterial dysbiosis can alter GABA, which is the major inhibitory neurotransmitter in the brain, and reduced levels are known to exacerbate seizures (79,80)
Lum et al. (81) reported that dysbiosis of the gut microbiota and frequency of epileptic events are often correlated, suggesting that drugs can interact directly with gut microbes that modify their metabolism and thereby impact drug efficacy and toxicity. The gut microbiome can influence the bioavailability of anti-seizure medications (ASMs), brain inflammation, and excitatory-inhibitory balance, representing a potential therapeutic target for treating refractory epilepsy (81).
A dysbiosis of the gut microbiome may contribute to disease development (82). It has been found that the “normal” microbiota consists of microorganisms that are both pro and anti-inflammatory. Hence, maintaining a good balance of the microbial community is closely associated with a healthy immune system and overall health (83). There exists a complex communication system between the gut and brain that maintains gastrointestinal homeostasis as well as cognitive functions (84). In the past decade, significant evidence on the impact of gut microbiota regulation in
neurological diseases such as Alzheimer’s disease (AD) (85) Parkinson’s disease (86) and neuropsychiatric disorders (87) have come to light. Recently, the idea of prolonged inflammation playing a role in epilepsy has been introduced (88) and it has been suggested that the gut microbiota can alter this inflammatory and immune response (83,89), thus suggesting the potential of gut-microbiota/microbiome regulation in treating epilepsy. Moreover, ketogenic diet and vagus nerve stimulation may be alternative therapies to refractory epilepsy, and therefore it was hypothesized that the regulation of the intestinal microbiota diversity may provide similar relief.
1.5 Prebiotics, probiotics and the Ketogenic diet
According to the definition of the World Health Organization (WHO), probiotics (so- called lactic acid bacteria) are "Live microorganisms that, when administered/injected in adequate quantities, confer a health benefit on the host" (90). Probiotics are more specific than prebiotics because there is less or no control over which specific bacteria will metabolize and consequently proliferate using the administered prebiotics. Indeed, the intake of beneficial bacteria can regulate the gut microbiota composition, promoting the establishment of a favorable microbial state which in turn may promote the maintenance of beneficial microorganisms (91,92). Probiotics are mainly composed of two genera of bacteria: the lactic acid-producing bacteria Lactobacillus and Bifidobacterium. Some animal studies have demonstrated the beneficial effect of probiotics in treating many diseases such as AD, autistic spectrum disorder (ASD), epilepsy and improving cognitive outcomes, but only a few clinical
studies in humans on the efficacy of probiotics in neuropsychiatric conditions have been conducted. For example, Lactobacillus rhamnosus, helventicus and fermentum, have been shown to restore stress-induced memory impairment in mice and spatial memory impairment caused by microbiota alterations by ampicillin administration. A randomized placebo-controlled clinical trial showed that probiotic supplementation with Lactobacillus rhamnosus could reduce the risk of developing neuropsychiatric disorder in infancy in 75 infants with ASD.
A prebiotic is "a substrate that is selectively utilized by host microorganisms and confers a health benefit" as proposed by the International Scientific Association for Probiotics and Prebiotics (ISAPP). These compounds include fermentable soluble fibers, non-digestible oligosaccharides (NDOs) and human milk oligosaccharides (HMOs) (93). Since fibers’ bonds cannot be broken down by digestive enzymes, soluble fermentable fibers go over relatively intact into the large intestine. Here, they are fermented by commensal bacteria to produce large quantities of acetate, propionate, and butyrate (94); notable among these is alpha-lactalbumin (ALAC), a breast milk protein, the most abundant in colostrum, which can stimulate innate immunity. Although prebiotic therapies could potentially be beneficial due to their enhancement of Lactobacilli and Bifidobacteria, few studies have been published on the beneficial effects of prebiotics on the MGB axis in both animals and humans., The ingestion of L. rhamnosus (JB-1) could regulate stress-induced behavior and alter GABA mRNA expression in mice (95, 96). In 2008, Desbonnet et al. (97) proved that the administration of Bifidobacterium infantis could reduce dopamine and
serotonin metabolites in the frontal cortex of rats, although without any discernible change in rats’ behavior. Later, in 2010, Desbonnet and colleagues demonstrated that the treatment with B. infantis could normalize the immune response, reinstate the basal noradrenaline concentration, and reverse the behavioral deficits (98).
Altogether these results seem to reveal that prebiotics and probiotics could be useful treatment options for neuropsychiatric disorders, but several studies, possibly randomized clinical trials, are mandatory to understand the mechanisms involved and the real efficacy, bearing in mind that correlation does not imply causality.
Moreover, the low-carbohydrate, high-fat ketogenic diet (KD) is an effective treatment for refractory epilepsy, a condition affecting more than one-third of epileptic individuals and defined by a failure to respond to at least two existing anticonvulsant medications (99). The KD is a modality of treatment used since the 1920s as a treatment for intractable epilepsy. It has been proposed as a dietary treatment that would produce similar benefits to fasting, which is already recorded in the Hippocratic collection. The KD has a high-fat content (90%) and low protein and carbohydrate. Evidence shows that KD and its variants are a good alternative for non- surgical pharmacoresistant patients with epilepsy of any age, taking into account that the type of diet should be designed individually, and that less-restrictive and more- palatable diets are usually better options for adults and adolescents (100).
However, despite its value for treating epilepsy and its increasing application to other disorders, including ASD, AD, metabolic syndrome and cancer (101), the use of the
KD remains low due to difficulties with implementation, dietary compliance and adverse side effects. Many studies have proposed roles for ketone bodies, GABA modulation, and mitochondrial anaplerosis in mediating the neurological effects of the KD but exactly how it confers beneficial effects on brain activity and behavior, remains unclear. The gut microbiota is a key intermediary between diet and host physiology; the species composition and function of the gut microbiota is critically shaped by diet, and nutrients made available to the host depend on microbial metabolism (102). Diet-induced changes in the gut microbiota are reproducible and persistent and, as such, have lasting impacts on the host. The gut microbiota modulates several metabolic and neurological pathways in the host that could be relevant to KD-mediated seizure protection. The KD alters the composition of the gut microbiota in mice and its ketosis is associated with altered gut microbiota in humans (103, 104). The microbiota is increasingly associated with changes in neurotransmission, including neurotransmitter signaling, synaptic protein expression, long-term potentiation, and myelination, as well as a variety of complex host behaviors, including stress-induced, social, and cognitive behaviors (105).
1.6 Urinary metabolites in epileptic patients
Seizures are often the first manifestation of central nervous system dysfunction and are common in many inborn errors of metabolism, especially in neonates, infants, and children. A high index of suspicion is required to diagnose inborn errors of metabolism as the cause of seizures (106).
The diagnosis of a genetic defect or an inborn error of metabolism often results in requests for a vast array of biochemical and molecular tests leading to an expensive workup. However, a specific diagnosis of metabolic disorders in epileptic patients may provide the possibility of specific treatments that can improve seizures. In a few metabolic diseases, epilepsy responds to specific treatments based on diet or supplementation of cofactors (vitamin-responsive epilepsies or the ketogenic diet for glucose transporter-1 deficiency), but for most of them specific treatment is unfortunately not available, and conventional antiepileptic drugs must be used, often with no satisfactory success (107).
All living organisms depend on primary and secondary membrane transport for the supply of external nutrients and the removal or sequestration of toxic compounds.
Due to the chemical diversity of cellular molecules, it comes as no surprise that a significant part of the proteome is dedicated to the active transport of cargo across the plasma membrane or the membranes of subcellular organelles. Transport against a chemical gradient can be driven by, for example, the free energy change associated with ATP hydrolysis (primary transport) or facilitated by the potential energy of the chemical gradient of another molecule (secondary transport). Primary transporters
include the rotary motor ATPases (F-, A-, and V-ATPases), P-type ATPases and a large family of integral membrane proteins referred to as “ABC” (ATP binding cassette) transporters. ABC transporters are widespread in all forms of life and are characterized by two nucleotide-binding domains (NBD) and two transmembrane domains (TMDs). ATP hydrolysis on the NBD drives conformational changes in the TMD, resulting in alternating access from inside and outside of the cell for unidirectional transport across the lipid bilayer. (108)
The family of transporters was subsequently termed ABC transporters in recognition of the “cassette-like” nature of the ATP-binding subunit (109). Around the same time, biochemical studies on the mammalian multidrug resistance (MDR) export pump P-glycoprotein revealed the presence of the very same motifs in its ATP- binding domain, demonstrating that the family of ABC transporters was represented not only in bacteria but also higher eukaryotes, including mammals. From the current sequence information of microbial genomes, ABC transporters represent the largest protein family identified to date, highlighted by the fact that between 1 and 3% of bacterial and archaeal genomes encode for subunits of ABC transporters (110).
Figure 3. Molecular architecture of ABC transporters. Image taken from Rees, D., Johnson, E. & Lewinson, O. ABC transporters: the power to change. Nat Rev Mol Cell Biol 10, 218–227 (2009) .
There are 48 ABC transporters in humans (111, 112) and many of these have been shown to be responsible for, or involved, in disease states, including cystic fibrosis, Tangier disease, adrenoleukodystrophy, and cancer. These ABC transporters in humans can be subdivided by phylogenetic analysis into seven distinct subfamilies A-
G (111, 112). Mammalian ABC transporters are involved in the cellular export of several groups of molecules, including cholesterol and sterols, lipids, retinoic acid derivatives, bile acid, iron, nucleosides, and peptides. Another prominent group of human ABC transporters is found in the liver, placenta and blood-brain barrier (BBB) where they are involved in the detoxification of hydrophobic organic molecules (113). The group includes P-glycoprotein (ABCB1), one of the best-studied ABC transporters, the MRP (ABCC1) and ABCG2. These transporters, when found highly expressed in the plasma membrane of tumor cells, can result in the failure of chemotherapy by protecting the cancer cells from the cytotoxic drugs used to fight the disease. Much effort has been spent on identifying selective inhibitors for these MDR transporters and while many compounds have been identified that inhibit P- glycoprotein function in, for example, human cell culture, no broadly applicable inhibitor is in use yet, due to significant side effects of the compounds (114).
In recent years, both animal and clinical trials have shown that the expression of ABC transporters are increased in patients with intractable epilepsy; additionally, epileptic seizures can lead to an increase in the number of sites that express ABC transporters.
These findings suggest that ABC transporters play an important role in the drug resistance mechanism of epilepsy. ABC transporters can perform the functions of a drug efflux pump, which can reduce the effective drug concentration at epilepsy lesions by reducing the permeability of the blood-brain barrier to ASMs, thus causing resistance to antiepileptic drugs (115).
Numerous studies have shown that ABC transporters are highly expressed in isolated epilepsy (IE). Rizzi et al. found that the MDR gene level and P-gp expression are increased by 1.8 and 5 times in the hippocampus and entorhinal cortex, respectively, of rats after 3 months of stimulation and kindling (116). Sisodiya et al. collected samples from epileptic patients that had a neuroepithelial tumor with embryonic developmental anomalies, limited cortical dysplasia, and hippocampal sclerosis, and they found that the MRP1 expression was increased in the above intractable epilepsy lesions (117). An MDR1 (ABCB1) cDNA probe was used to measure the MDR1 gene expression. The results showed that the P-gp immune-positive samples and the expression of MDR1 in the temporal lobe vascular endothelial cells of IE patients were higher than those in the control group of temporal lobe vascular endothelial cells from aneurysm resection (118). Not only an increase in the expression level of ABC transporters but also an increase in the number of expression sites is seen after epileptic seizures. Under normal physiological conditions, the MDR1 gene expression of the central nervous system is limited to capillary endothelial cells and the astrocytes around the capillaries. However, after epileptic seizures, the P-gp of the patients with epileptic seizures is also expressed on parenchymal astrocytes, even on the neurons, in addition to its previous expression in brain capillary endothelial cells and astrocytes around the capillaries (119). Normal brain neurons and astrocytes do not express MRP, but in the samples from intractable epilepsy patients after resection (e.g., limited resections for cortical dysplasia and hippocampal sclerosis), MRP can be detected in astrocyte and hypogenetic neuron membranes (118). This expression
may provide a histological explanation for the significant correlation between drug- resistant epilepsy and the high expression of ABC transporters. Increased expression of ABC transporters in the brain will enhance the function of the BBB, and ASMs are the natural substrate of ABC transporters because of their high lipid solubility (120);
thus, a high expression of ABC transporters may limit the passage of ASMs through the BBB, prevent ASMs from entering brain tissue, and reduce the concentration of antiepileptic drugs in epilepsy brain lesions (121, 122). Therefore, even for a blood drug concentration that is within the scope of treatment, the internal and external concentrations of parenchymal cells are not sufficient to provide the treatment effect, thus leading to drug resistance. This theory has been confirmed by considerable research. Studies have analyzed the concentration of phenobarbital (PB) in the serum and cerebrospinal fluid (CSF) of epilepsy patients, and the results showed that the daily PB dose and serum PB concentration are similar between the observation group (decreased seizure frequency ≥ 50%) and the control group (decreased seizure frequency < 50%). However, the CSF PB concentration and the CSF to serum ratio of PB in the observation group were both higher than those in the control group, suggesting that the amount of PB that penetrates the BBB is lower in the control group (123). Studies have confirmed (124) that excessive glutamate is associated with seizures and that the glutamic acid N-methyl-D-aspartate (NMDA) receptor cyclooxygenase-2 (COX-2) prostaglandin E2 receptor P-gp pathway is involved in the regulation of P-gp expression in epilepsy. Recurrent epileptic seizures can induce the upregulation of the expression of a variety of proinflammatory cytokines, such as
IL-1β, IL-6, and TNF-α, in brain tissue; these inflammatory factors can enhance the binding of glutamic acid to the NMDA receptor, which further aggravates seizures (125). NF-kB is a player in the regulation of inflammation. High expression of NF- kB is found in neurons and glia from samples obtained during operations on epilepsy patients. The activation level of peripheral NF-kB in children with epilepsy was higher than that in a healthy control group; these patients were reviewed after effective antiepileptic treatment when the activation level of peripheral NF-kB was significantly decreased (126). In in-depth studies, some scholars found that the NF- kB complex can activate the transcription of the gene associated with the MDR1 promoter. The main mechanism occurs through transcriptional regulation loci on the NF-kB subunit that binds with p65 in the MDR1 gene promoter region. When a seizure induces inflammation, a variety of inflammatory factors can act on MDR1 through NF-kB and P-gp expression is upregulated; thus, the inflammatory response is associated with drug-resistant epilepsy. Therefore, it is possible that the regulation of NF-kB is also a feasible way to reverse drug-resistant epilepsy, and this method also provides a clinical basis for the anti-inflammatory treatment of drug-resistant epilepsy (126).
Another important facet to evaluate is the vitamin B6 and its metabolism. Vitamin B6 (B6) comprises a group of six related compounds: pyridoxal (PL), pyridoxine (PN), pyridoxamine (PM), and their respective 5′-phosphates (PLP, PNP, and PMP). The major forms in animal tissues are PLP and PMP; plant-derived foods contain primarily PN and PNP, sometimes in the form of a glucoside. In humans, the major
excretory form is 4-pyridoxic acid (4-PA). Pyridoxine and its related compounds function as a coenzyme in the metabolism of amino acids, glycogen, and sphingoid bases. PLP is a coenzyme for more than 100 enzymes involved in amino acid metabolism, including aminotransferases, decarboxylases, racemases, and dehydratases (127). An inborn abnormality of PLP-dependent GABA synthesis induced by glutamate decarboxylase (GAD) deficiency was postulated as a cause of epilepsy, and lifelong pyridoxine administration was recommended (128). So, an inborn error of pyridoxine metabolism (accentuated by high pyridoxine requirement during early development) is inherent in epilepsy. Being a starting point for neurotransmitter disorders, such an error may be a key determinant of epileptic diathesis. An impairment of GABA (as well as serotonin and taurine)-mediated inhibition along with an enhancement of glutamate (and aspartate)-mediated excitatory transmission evidently facilitates spreading of ictal activity throughout the brain and thereby generation of seizures (129, 130).
The data obtained in S. Dolina et al. (127) testify that concentrations of compounds formed or metabolized during PLP-dependent TRP degradation, as well as correlations between them, are quantitative urinary biomarkers for the determination of clinical status—from the first seizure attack up to progressively worsening condition. These biomarkers are also indicative for the evaluation of ASM treatment effectiveness and its individual monitoring. The parameters reflecting kynureninase activity turned out to be the most sensitive link of this chain. Once the initial seizure attack has occurred, the drastically increased levels of tryptophan (TRP), kynurenine
(KYN), and toxic 3-hydroxykynurenine (3HOKYN), and the drastically reduced level of indoxyl sulfate (IND) pointed to the disordered PLP-dependent TRP degradation.
Pyridoxine-dependent epilepsy (PDE) is an autosomal recessive epileptic encephalopathy characterized by a therapeutic response to pharmacological dosages of pyridoxine and resistance to conventional antiepileptic treatment (131). Recently, the underlying genetic defect was identified as deficiency of antiquitin (ATQ) (α- aminoadipic semialdehyde dehydrogenase, ALDH7A1), an enzyme that facilitates cerebral lysine catabolism (132). Lysine is catabolized in mammals through the saccharopine and pipecolate pathways; the former is mainly hepatic and renal, and the latter is believed to play a role in the cerebral lysine oxidation. Both pathways lead to the formation of α-aminoadipic semialdehyde (AASA) that is then oxidized to aminoadipate (AAA) by antiquitin (ALDH7A1). Mutations in the ALDH7A1 gene result in the accumulation of AASA and its cyclic form, piperideine-6-carboxylate (P6C), which causes pyridoxine-dependent epilepsy (PDE) (133).
ATQ deficiency results in accumulation of chemical substrates arising from lysine degradation proximal to the deficient enzyme activity including AASA, its cyclic equivalent P6C, and pipecolic acid. Inactivation of PLP via chemical reaction with P6C is a pathophysiological mechanism of pyridoxine dependency. While treatment with pyridoxine compensates chemical PLP inactivation, the accumulation of substrates from lysine degradation is not sufficiently reduced. These potentially neurotoxic compounds could explain for the limited efficacy of pyridoxine, as 75–
80% of patients suffer developmental delay or intellectual disability (IQ < 70) despite
seizure control (134, 135). This article is the first to report on the biochemical and clinical results of dietary lysine restriction as adjuvant therapy for PDE due to ATQ deficiency. The fact that 75–80% of patients suffer global developmental delay or intellectual disability (IQ < 70) despite adequate seizure control by pyridoxine treatment (134, 135, 136) illustrates the need for additional therapeutic strategies in PDE. Extrapolating experience from other neurotoxic inborn errors of metabolism, early intervention with dietary therapy is a beneficial approach to prevent brain damage and optimize neurodevelopment (135, 136).
Caffeine is the most consumed central nervous system (CNS) stimulant worldwide (137). Given its stimulative effects, it is not surprising that people with epilepsy and health providers question whether caffeine is a trigger for seizures (138). Caffeine (1,3,7-trimethylxanthine) has been found to offset fatigue and enhance vigilance, reaction speed, information processing, arousal, and motor activity. The effects of caffeine are likely due to its interaction with various neurotransmitters, most importantly adenosine (139). Adenosine is produced as a by-product of neuronal firing (140). It promotes sleep and reduces cortical excitability through binding to the adenosine receptors (141). As the molecular structure of caffeine is similar to adenosine, caffeine can also bind to the adenosine A1 and A2A receptors and in doing so prevents adenosine from binding, thus acting as an adenosine antagonist.
Caffeine also interacts with GABA by modulating GABA-A receptors (142-144).
The activation of the adenosine A1 receptor inhibits dopamine, a neurotransmitter involved in focus and motivation, and release of glutamate, an important excitatory
neurotransmitter. Caffeine thus increases dopamine and glutamate release and inhibits GABA, resulting in a stimulating net effect (145, 146). Contrary to clinical studies, there is a relatively large number of studies on the interaction between caffeine and anti-seizure medication in animal models. Rats were injected with caffeine as well as one of the following ASMs: carbamazepine, phenytoin, phenobarbital, valproic acid, felbamate, oxcarbazepine, lamotrigine, tiagabine, gabapentin, and topiramate. Single-dose caffeine injections reduced the seizure threshold and increased the amount of phenobarbital, carbamazepine, phenytoin, topiramate, gabapentin, valproic acid, and felbamate needed to protect 50% of the rats against electro convulsions, whereas no changes were seen for oxcarbazepine, lamotrigine, and tiagabine (147-149, 151-153). When caffeine was chronically administered, the amount of phenobarbital, carbamazepine, phenytoin, topiramate, gabapentin, and valproic acid needed to protect 50% of the rats against seizures was also increased, but again unaltered for oxcarbazepine, lamotrigine, and tiagabine (148, 150-153). The interaction effects between caffeine and ASMs may occur on two different levels. First, as the concentrations of phenobarbital, clonazepam, phenobarbital, valproic acid, carbamazepine, gabapentin, topiramate, and ethosuximide were unaffected in the presence of caffeine, caffeine may act as an antagonist of the anticonvulsant properties of these medications (154, 149, 152, 153).
Second, more simple explanation of the pharmacodynamic interaction between caffeine and ASMs is that caffeine increases seizure susceptibility, indirectly
increasing the need for drugs, which makes it analogous to the interaction between any seizure precipitants.
Nicotinamide adenine dinucleotide (NAD+) plays key roles in energy metabolism, calcium homeostasis, gene expression, cell signaling and DNA damage repair. NAD+
deficiency is an important pathological factor in several neurological disorders such as brain ischemia, Parkinson's disease, Alzheimer’s disease and epilepsy (155, 156, 157). Cellular NAD+ content is regulated by a balance between NAD+ synthesizing enzymes including NAMPT (nicotinamide phosphoribosyltransferase), NAPRT (nicotinic acid phosphoribosyltransferase) and NMNATs (nicotinamide mononucleotide adenylyltransferases) and NAD+ consuming enzymes such as poly (ADP-ribose) polymerases (PARPs), NADK (NAD+ kinase) and CD38 (158). The most recent study is the Sadeghi et al. (159) work that showed a significant increase in the expression of CD38 in the hippocampus of kindled and pilocarpine treated rats.
In addition, applying a CD38 inhibitor reduced the tonic-colonic seizure severity and seizure duration in the PTZ-kindled rats. However, the role of CD38 in epileptogenesis and its association with NAD+ homeostasis has not yet been elucidated. It has been indicated that NAD+ treatment at early stages can suppress epileptogenesis by reducing neuronal apoptosis in the mice hippocampus (159), emphasizing contribution of NAD+ to epileptogenesis. It was also found a significant decrease in the hippocampal NAD+ levels due to epileptogenesis. Among proteins related to NAD+ homeostasis, CD38 was identified to be a key enzyme involved in the degradation of NAD+. We demonstrated that CD38 protein, mRNA, and activity
increased in hippocampus during epileptogenesis. It is likely possible that a decline in NAD+ and elevation of calcium could have a crucial role in the development of epilepsy, therefore, cADPR signaling proteins appear to be a potential target for epilepsy treatment (160).
2. Aim of the study
We conducted a controlled study to evaluate potential intestinal dysbiosis and species richness using the Bristol Stool Test (BST), comparing epilepsy patients and age- matched healthy controls. Moreover, we compared urinary metabolites and associated pathways between the Isolated Epilepsy (IE) and Epilepsy plus (Epi+) and between drug-resistant (drugR) and drug-sensitive (drugS) epilepsy patients’ subgroups.
3. Patients and Methods
3.1 Patients’ selection
Patients with epilepsy of various aetiologies were recruited at the Department of Paediatric Neurology and Muscular Disease Unit, IRCCS Istituto Giannina Gaslini, between August 2019 and May 2020. Clinical and instrumental data, including brain MRI and EEG findings as well as treatments data and genetic results, were reviewed through our local database.
Stool consistency was assessed through the BST that illustrates stool shapes together with precise descriptions regarding their consistency as an ordinal scale of stool types ranging from the hardest (type 1) to the softest (type 7): types 1 and 2 are considered abnormally hard stools while types 5, 6 and 7 are considered abnormally liquid stools. Types 3 and 4 are generally considered normal stool forms. To simplify BST scores collection, only one score was reported when patient indicated two different stool types: 3-4 scores= 3; 1-2 scores= 1; 5-7 scores= 7.
The occurrence of GI symptoms was assessed referring to the validated Rome IV Diagnostic Questionnaire (the Rome Foundation, Inc.;
https://theromefoundation.org), which investigates the rate and intensity of the following GI symptoms: abdominal pain, constipation, diarrhea, reflux, bloating, dyspepsia, nausea and vomiting. To evaluate urinary metabolites, early morning urine samples of hospitalized epileptic patients were collected and stored at -80 °C soon after.
We did exclude patients with progressive neurological disorders (e.g., autoimmune encephalitis, and CNS tumors). The control group included age-matched neurotypical children, recruited from the Blood Transfusion Centre of our Institute.
Written informed consent was signed by patients or their parents/legal guardians and approved by our local Independent Ethics Committee (IEC).
3.2 Statistical analysis
Epileptic patients were stratified into different subgroups: the Isolated Epilepsy (IE) group, including patients suffering from epilepsy and without intellectual disability or associated neuropsychiatric comorbidities; the Epilepsy plus (Epi+) group defined as epilepsy patients also showing intellectual disability, and/or psychiatric symptoms;
the drug-resistant (drug-R) and the drug-sensitive (drug-S) groups that, according to the International League Against Epilepsy (ILAE), included patients taking ≥ 2 and <
2 anti-seizure medications (ASMs).
Groups were compared using the Fisher’s Exact test. The thresholds of p-value were set at 0.05 (statistical significance) and 0.001 (highly statistical significance).
Ammonium formate, acetonitrile, methanol and formic acid (LC-MS grade) were purchased from Sigma Aldrich Srl (Milan, Italy). Water was purified by reverse osmosis and filtrated through a Milli-Q purification system (Millipore, Milford, MA, USA).
Reversed-phase column ACQUITY C18 BEH 1.7µm 2.1 X 100 mm (Waters S.p.A., Sesto San Giovanni, Milan, Italy) and HILIC column ACQUITY BEH Amide 1.7µm 2.1 X 150 mm (Waters S.p.A., Sesto San Giovanni, Milan, Italy) were used.
3.4 Mass spectrometry analysis of urine samples
Fifty µL urine, extracted adding 150 µL cold (-20°C) methanol, vortex-mixed and centrifuged at 14,000 rpm for 10 min, were used for both hydrophilic interaction liquid chromatography (HILIC) and reversed-phase (RP) chromatography. Samples were analyzed using Vanquish Horizon UHPLC coupled to a Q-Exactive Orbitrap mass spectrometer.
The extracted metabolites were diluted 1:2 with H2O and 5 µl samples were directly injected into RP and HILIC columns. The linear gradient for reversed-phase columns started with 1% B and in 15 minutes increased up to 100% with a flow rate of 250 µl/min, then the columns were normalized for 5 minutes with 1% phase B. The linear gradient for HILIC columns started with 90% B and decrease to 30% B in 15 minutes with a flow rate of 200 µl/min, the columns were then normalized with 90 % phase B for 9 min.
Mass spectrometry (MS) data were acquired in full scan mode in both positive and negative ionization, using 70000 resolution, 1e6 AgC and 100 ms maximum injection time. In the identification phase separately for each polarity, the experiments were done in data‐dependent acquisition mode alternating MS and MS/MS experiments. A
maximum of 5 MS/MS experiments were triggered per MS scan. The intensity threshold was set at 1.6e5 using an isolation window of 1.4 Da. The m/z values of signals already selected for MS/MS were put on an exclusion list for 20 s. 70000 and 17500 resolution, 1e6 and 1e5 AgC, 100 ms and 50 maximum injection time were used for MS1 and MS2 scan respectively. If no further inclusion list entries are identified in a scan event will be selected other masses. Normalized stepped collision energy of 30, 40, 50 was used.
Raw data files were processed by Compound Discoverer™ 3.1 software, including peak detection, peak alignment and peak integration. Raw files were aligned with adaptive curve settings. Unknown compounds were detected with a 5ppm mass tolerance, 3 signals to noise ratio, 30% of relative intensity tolerance for isotope search, and 500,000 minimum peak intensity, and then grouped with 5 ppm mass and 0.2 min retention time tolerances. A procedural blank sample was used for background subtraction and noise removal during the pre-processing step. Peaks with less than a 3-fold increase, compared to blank samples were removed from the list.
Metabolites identified in the processed raw data of mass spectral peaks were searched against both ChemSpider™ chemical structure database and mzCloud spectral library. A customized AMRT database integrated into CD was also used for metabolites identification.
3.5 Untargeted metabolomics Analysis
All annotated and normalized metabolite data obtained from the three metabolite detection systems were as follows: 11,190 for C18-positive, 11,110 for C18-negative, and 4,773 for HILIC-positive. For filtration of the features from each of the detection platforms, relaxed filtering in which features satisfying both “mzVault Best Match”
value of 50, and “mzCloud Best Match” value of 70 were kept and the remaining was discarded. Accordingly, 277, 327, and 174 features passed the filtration process for C18-positive, C18-negative, and HILIC-positive, respectively. All the filtered data coming from three platforms were merged. Since more than half of the features in each data type did not satisfy normal distribution test, the Mann-Whitney U test was carried out between IE and Epi+, or between drugR and drugS epilepsy patients’
subgroups. Among multiple abundance entries of the same metabolite, the one with the minimum p-value was retained, and if there is more than one metabolite with the same minimum p-value, average values of the abundances were calculated. Lastly, contaminants were also filtered out. After filtration and merging steps, a total of 369 metabolite data were obtained for 43 patient samples. For drug-resistant/sensitive group analysis, two patient samples that did not have a relevant record were discarded.
3.6 Univariate Analysis
For normality and homoscedasticity, Shapiro-Wilk and Levene’s test was performed in R, respectively. Since a higher proportion of data was not following normal
distribution, a non-parametric univariate test, the Mann-Whitney U test was performed to find metabolites statistically significantly changing between two groups. For significance, metabolites with p-values equal to or less than 0.05 were used in downstream analysis.
3.7 Principal Component Analysis (PCA)
Principal Component Analysis (PCA) is a dimensional reduction method and used as an unsupervised classification method (161) that can also be used to test classification between two groups of interest, using R packages “factoextra” and “FactoMineR”
(162). Additionally, a multivariate version of ANOVA, PERMANOVA, in which the significance between centroids of clusters is tested, was performed with Euclidian distance using adonis function in R package “vegan” (163).
3.8 Partial least squares discriminant analysis (PLS-DA)
Partial least squares discriminant analysis (PLS-DA), considered as supervised version of the PCA method, is another one of the machine learning algorithms that can be used for classification and feature selection, especially in metabolomics studies (164). PLS-DA is also sensitive to imbalanced data, as most machine learning methods (165). One of the oversampling methods, Synthetic Minority Oversampling Technique (SMOTE) was performed on KNIME Analytics Platform to balance IE/Epi+ or drug-resistant/drug-sensitive data before PLS-DA (166). For PLS-DA, R package “mixOmics” was used by setting the number of components to 5 (167). The
performance of the model was evaluated through 10-fold cross-validation. In addition to the area under curve (AUC) score, two model evaluation metrics, R2Y and Q2 scores were recorded. R2Y describes the variance of class response (Y) captured by the model, and Q2 reflects the quality of the prediction ability of the built model (168). Even though there is no exact threshold for Q2, models with Q2 scores higher than 0.5 are considered to have good prediction ability (169). Additionally, another validation technique, a permutation test was performed with 999 permutations, using R package “RVAideMemoire”. When the model passed the performance tests, Variable Importance in Projection (VIP) scores were calculated from the final validated model to quantify the contribution of metabolites in the separation of the classes (170). Metabolites with VIP scores higher than 1.0 were selected. Lastly, the predictive ability of the model was tested on test dataset, and error rate was recorded.
3.9 Pathway Analysis
For pathway analysis, metabolites with both p-values ≤ 0.05 from Mann-Whitney U test and VIP score > 1 from PLS-DA were selected. Respective KEGG IDs of the metabolites were manually retrieved from KEGG Compound Database, based on match with name and formula in our dataset (171). A list of identified KEGG IDs was submitted to the online pathway enrichment analysis tool, Metabolites Biological Role (MBROLE) 2.0 (172). For analysis, “KEGG Pathways” selected for annotation, and organism-dependent enrichment analysis was carried out by selecting “Homo
sapiens” in options. This tool calculates p-values of the enriched pathways based on hypergeometric test, and for this study enriched pathways with FDR corrected p values ≤ 0.05 were recorded.
4.1 Clinical data of the studied population
148 non-related individuals (61 females) were enrolled. The age range was 1.4-18.8 years (mean= 9.4±3.9 years). The epilepsy group included 84 patients (37 females) with a mean age of 9.3±4.5 years. Likewise, the control group consisted of 64 healthy (24 females), age-matched, individuals (mean= 9.4±3.1 years). The IE subgroup included 30 patients (11 females) with a mean age of 10.7±3.7 years, while the Epi+
subgroup reached up to 54 patients (26 females) and the mean age was 8.6 ± 4.7 years (Supplementary table 1).
In the IE subgroup, 20 (67%) patients had focal epilepsy, while 10 (33%) had generalized epilepsy. Twenty-eight (93%) patients were chronically taking one ASM (50% valproate). In the Epi+ subgroup, patients were affected by developmental and epileptic encephalopathy (DEE). A genetic diagnosis was confirmed in 25/30 (46%) patients. Nine (17%) individuals showed structural cause of epilepsy (X MCD, 1 hypoxic-ischemic neonatal injury, 1 post-herpetic encephalitis). For 17 (31%) patients the etiology was unknown. Intellectual disability (ID) was the most common comorbidity found in 53 (98%) patients. Twenty-four (44%) patients were under ASM monotherapy, twenty-six (48%) patients were treated with ≥ 2 ASMs, and four (7%) were not under treatment at the time of the assessment. Valproate (VPA) and carbamazepine (CBZ) were the most used ASMs, found in 30 (56%) and 15 (28%) patients (Supplementary Table 2 and Table 3).
Finally, the anamnestic interview conducted through the validated Rome IV Diagnostic Questionnaire found 30 (56%) patients in the Epi+ subgroup and 5 (17%) in the IE subgroup who did suffer from GI discomfort. A total of 31 (37%) patients had abnormal BST scores: 5 (17%) patients in the IE group and 26 (48%) patients in Epi+ group, respectively. Only 9 (14%) children in the control group had abnormal BST scores.
4.2 Statistical analysis
We first matched the control group with both IE and Epi+ subgroups resulting in a value of p=.0026, then we compared the control group with the two epilepsy subgroups: comparison between the control group and the IE subgroup resulted in any relevant difference between BST scores with p=.76; conversely, comparison between the control group and the Epi+ subgroup did show a p=.0001. There was no difference between the two epilepsy subgroups (IE vs Epi+) (p=.0047); Figure 4).
Figure 4. BST scores comparison between groups.
Finally, we explored the relation between GI discomfort and BST (Figure 5). GI discomfort was detected in 35 (42%) patients: 25 (71%) had abnormal BST, 10 (29%) had normal BST. GI symptoms were absent in 49 (58%) patients: 6 (13%) patients had abnormal BST scores, 43 (87%) had normal BST scores. Patients with abnormal BST scores reported increased GI discomfort symptoms (p=.0001) (Figure 5).