The global incidence of epilepsy is about 60 per 100,000 inhabitants per year, and up to a third of patients may suffer from refractory epilepsy [1, 2]. An alternative treatment to antiseizure medications (ASM) are the ketogenic dietary therapies (KDT), which have been used as a treatment for refractory epilepsy in children since the 1920s [3]. The most commonly-used KDT are Classic KD (CKD), KD (ketogenic diet) with medium chain triglycerides (MCT, MCT-KD) and the modified Atkins diet (MAD) [4]. CKD is a high-fat diet (about 90% of the energy intake) with mostly long chain triglycerides (LCT), very low carbohydrate content and adequate amount of proteins [5]. The most commonly recommended ratios are 3:1 or 4:1, which means that for every 3 or 4 g of fat, 1 g of protein and carbohydrates is administered. The MCT-KD was proposed in 1971 by Huttenlocher and co-workers [6]. In this diet, part of the fat is supplied in the form of MCT oil. MCT are quickly metabolized, ketosis is achieved faster, and ratios 1.2:1 can be used. The MAD is less restrictive than the previous ones because, although carbohydrates are limited to 10 g/day, a free intake of proteins and lipids is allowed [7]. In clinical practice it is also common to supplement the MAD with preparations containing CKD components to increase ketosis [8]. Regardless of the type of KDT, they all seem to be neuroprotective and there is no clear evidence for different efficiency between CKD and MCT-KD [9] or MAD [10]. However, the molecular pathways involved in these antiseizure mechanisms have not yet been established.

Recent research highlights the role of ketone bodies and polyunsaturated fatty acids in the modulation of the KD-induced neuroprotection [4, 11, 12, 13]. The neuroprotective mechanisms include an increase in the cellular mitochondrial function and biogenesis [14], a decrease in oxidative stress [15], changes in potassium channels [15, 16], an increase in the brain-derived neurotrophic factor (BDNF) [17] and in the purinergic and GABAergic neurotransmission [18], attenuation of neuroinflammation, or even changes in gut microbiota [13, 19].

Epigenetic processes are related to changes in neuronal activity and epileptogenesis. Molecular mediators of epigenetics include non-coding RNAs, methylation of DNA and histone modifications [20]. Therapies for epilepsy, including KDT, may bear an influence on epigenetic processes [21].

Transcriptomic analysis allows the study of changes in gene expression in different human diseases as well as the effect of potential interventions with clinical relevance [22]. A massive analysis of gene expression allows the detection of broad coordinated trends which cannot be discerned by more targeted assays.

Our major aim was to determine the changes in gene expression induced by a 6-month treatment with KDT in children with epilepsy to shed light on the potential mechanisms of this type of diets. This is, as far as we are aware, the first transcriptomic study in epileptic children who have followed KDT.

All methods were performed in accordance with the relevant guidelines [23] and regulations. All experimental procedures were approved by the Committee on Ethics in Clinical Research and an informed consent was signed before obtaining the samples.

Eight children with refractory epilepsy were treated with a KDT (Table 1 and Supplementary Table 1). The age at seizure onset was very variable and seizure type was as follows: 3 focal onset, 2 generalized tonic-clonic, 2 tonic, 1 multiple. Six children had seizures daily (1–50 seizures/day) and 2 of them had seizures monthly (1–2 seizures/month). The median number of ASM used before the onset of KDT was 4 (range 2–8) and no patient was treated with epilepsy surgery prior to KDT. The average age of the patients at the beginning of the diet was 19 months, while an average of 7 months elapsed from the onset of the seizures to the onset of the treatment.

Six patients followed a CKD and two of them a MAD. ASM were prescribed concomitantly with KDT in all cases. The nutritional routes were: nasogastric tube (one patient), gastrostomy tube (one patient) and oral feeding (six patients). The onset of KDT was carried out in hospital admission in 6 patients who started CKD, without fasting or caloric restriction, and the fat content of the diet was progressively increased in all cases. Effectiveness of KDT in seizure reduction is in Table 1. Only one patient did not reduce the number of seizures with KDT (Patient E). However, she achieved a faster recovery from the seizures and her cognitive status subjectively and significantly improved during the treatment. Four patients had early secondary effects. Throughout the follow-up, these side effects occurred in some patients, most of them were considered mild: hypercholesterolemia, hypertriglyceridemia, constipation, hypercalciuria, hypocitraturia, proteinuria, acidosis, and elevation of plasma gamma-glutamyl transferase (GGT) activity. The duration of KDT ranged from 9 to 24 months. Three patients discontinued the diet throughout the follow-up (one due to side effects and two because the diet became ineffective in the long-term).

3.1 Treatment with the KDT changes the mRNA expression profile in epileptic paediatric patients

The peripheral blood mononuclear cells (PBMCs) gene expression profile from the paediatric patients treated with the KDT for at least 6 months, significantly differed from the pre-intervention one. 4630 mRNAs differentially expressed (p 0.05) were found. Three-dimensional unsupervised PCA showed two clustered groups that included the pre- and post-intervention samples (Fig. 1A). The unsupervised hierarchical clustering (heatmap) also showed significant differences in the gene expression from the two groups of samples (Fig. 1B).

Fig. 1.

Transcriptomic analysis of PBMCs before and after a KD intervention in epileptic paediatric patients. (A) Principal components analysis (PCA) displays clear spatial separation in the expression between the two groups of samples identified by unsupervised hierarchical clustering. In the 3-dimensional plot, the three principal components PC#1, #2 and #3 of all samples with over 20,000 well-annotated genes and their respective variations are expressed on the x-, y- and z-axis. The total percentage of PCA mapping variability is 35%. Each data point represents one sample. The ellipsoids highlight the portioning of the different samples. Assignment of samples by colour: pre-intervention (red) and post-intervention (green). (B) Heatmap of pre-intervention (red bar) and post-intervention (green bar) samples and 4630 genes derived from ANOVA. Each line represents a patient and each column a gene. Up-regulated genes are represented in red and down-regulated genes in blue. This shows a complete separation of transcriptomic profile pre- (red) and post- (green) intervention. Images obtained by using The Pathway Studio Software (Ariadne software, https://www.pathwaystudio.com).

According to Pathway Studio, of the 10 biological pathways that show the highest statistically significant treatment-induced change in their gene expression, we can highlight the following ones: synaptic transmission (19 genes involved, p-value 3.40 $\times$ 10${}^{-7}$), potassium ion transmembrane transport (8 genes involved, p-value 6.99 $\times$ 10${}^{-5}$), and nervous system development (16 genes involved, p-value 1.09 $\times$ 10${}^{-4}$). Table 2 shows the top ten modulated biological processes according to Pathway Studio.

Table 2.Biological processes regulated by the mRNA differently expressed in the patients treated with KDT.

	Biological process	p-value	Number of genes
1	Multicellular organismal development	5.61 × 10${}^{-9}$	36
2	Synaptic transmission	3.40 × 10${}^{-7}$	19
3	Protein homooligomerization	5.97 × 10${}^{-6}$	11
4	Cell differentiation	5.00 × 10${}^{-5}$	22
5	Response to denervation involved in regulation of muscle adaptation	5.54 × 10${}^{-5}$	3
6	Potassium ion transmembrane transport	6.99 × 10${}^{-5}$	8
7	Transmembrane transport	7.11 × 10${}^{-5}$	20
8	Protein ubiquitination	8.62 × 10${}^{-5}$	13
9	Nervous system development	1.09 × 10${}^{-4}$	16
10	Small molecule metabolic process	1.12 × 10${}^{-4}$	30
Top ten biological processes whose gene expression change after intervention with KDT ($>$6 months) in epileptic children according to PS analysis.

3.3 Different miRNA expression profile in epileptic paediatric patients before and after the treatment with KDT

A global analysis of miRNA expression (miRNome) indicates that there are 230 miRNAs which are expressed differently before and after KDT. Fig. 2 shows the heatmap of the expression profile of these mRNA regulators. If additionally the miRNAs are restricted with a fold change $>$$|$1.3$|$, 11 relevant miRNAs whose expression differs after KDT in PMBCs are found: miR-3978, miR-6726-3p, miR-130a-3p, miR-4758, miR-6745, miR-532, and miR-185-5p (down-regulated after the KDT); and miR-4538, miR-602, miR-330-5p, and miR-4673 (up-regulated after the KDT). As with mRNA analysis, the miRNA expression profile before and after KDT intervention allowed complete separation of each case (no green bars intertwined with the red ones or vice versa).

Fig. 2.

Heatmap of pre-intervention (red bar) and post-intervention (green bar) samples of 230 miRNA derived from ANOVA. Each line represents a patient and each column a miRNA. Up-regulated miRNAs are represented in red and down-regulated miRNAs in blue. Image obtained by using The Pathway Studio Software (Ariadne software, Elsevier, https://www.pathwaystudio.com).

3.4 Gene and miRNAs expression combined analysis

A combined analysis of the miRNAs and mRNAs that change significantly with KDT intervention was also carried out (see the list of genes and miRNAs in the Supplementary Information). Several of the selected miRNAs regulate genes whose expression change with the intervention (see Fig. 3). All the relevant changes in mRNA and miRNAs (and their physiopathological significance) are summarized in Fig. 4. The results show that 8 genes related to the nervous system modulate their expression after KDT and are known to be regulated by the miRNAs modified during the intervention. These genes are: CDH2 (regulated by miRNA-199b-5p), NEGR1 (regulated by miRNA-875-5p), NEDD4 (regulated by miRNA-9-5p), FGF12 (regulated by miRNA-9-5p and miRNA-93-5p), CNTN2 (regulated by miRNA-185-5p), BSN (regulated by miRNA-185-5p and miRNA-153-3p), CACNA1C (miRNA-323a-3p and miRNA-153-3p), and DOC2A (regulated by miRNA-153-3p).

Fig. 3.

Genes and miRNAs with biological relevance in our study. In all panels, blue indicates genes or miRNAs that are down-regulated and red those that are up-regulated. Each panel shows our selection strategy. (A) From 4630 genes differently expressed after KDT, the 19 involved in the synaptic transmission pathway have been selected to be discussed. Protein encoded by these genes can be found in Table 2. (B) From 230 miRNAs differently expressed after KDT, the 11 which have a fold-change $>$$|$1.3$|$ have been selected to be discussed. (C) From the combine analysis of miRNAs and genes whose expression changes with KDT, those related to the central nervous system (CNS) were selected. (D) List of genes and miRNAs selected through strategy A, B and C, and which finally are discussed.

Fig. 4.

Treatment with KDT ($\mathrm{>}$6 months) in refractory epileptic children change the transcriptomic profile of genes involved in epileptic processes that encode ion channels, neurotransmitter receptors, and synapse structural proteins. The down-regulated genes and miRNAs are shown in blue, whereas those up-regulated are shown in red.

The results show that synaptic transmission is the most relevant biological pathway related to epilepsy, targeted by our intervention with KDT in epileptic children. In general, genes which encode proteins involved in this pathway are down-regulated after the intervention. These genes are basically: ion channels, neurotransmitter receptors, and structural synapse proteins.

Regarding the changes in the expression of ion channels, of the four potassium channels whose expression varies with KDT, three of them have been previously linked to epilepsy: Kir 2.2, whose up-regulation has been described in epilepsy [26]; KCNMA1, a Ca${}^{2+}$-dependent potassium channel of the family known as BK proposed to be responsible of higher neuronal excitability [27]; and KCNK6, whose up-regulation contributes to epileptic seizures [28]. CACNA1C, which encodes an L-type Ca${}^{2+}$ voltage-gated Ca${}^{2+}$ channel (Cav 1.2), is the only ion channel up-regulated after KDT. It has been described mutations in this gene lead to epileptic phenotypes [29], and its dysfunction has been also associated to other neurological disorders such as autism and bipolar disease [30].

The results show a KD-induced down-regulation of several genes encoding neurotransmitters’ receptors in epileptic children: orexin, serotonin, melatonin, glutamatergic and cholinergic receptor. Antagonists of the orexin receptors, HCTR1 and HCTR2, can reduce seizures in animal models [31]. Serotonin receptor 5-HT6 seems to be involved in the regulation of cognition, feeding, and, possibly, affective state and seizures [32]. Blockade of the 5-HT6/mTOR pathway can be used as a target for epileptic treatment [33]. It has been found that the expression of melatonin receptors, MT1 and MT2, is increased in the hippocampus of pilocarpine-induced epileptic rats during the epileptic state, while it decreases in a chronic state [34]. Glutamatergic systems are involved in the control of excitation and inhibition. Higher expression of mGlu2/3 receptors have been found in the hippocampus of patients with temporal lobe epilepsy [35]. Finally, the results also show a decrease in the expression of CHRNB3, a cholinergic receptor, already been described as a possible gene associated with epilepsy [36].

On the other hand, MAOA is the enzyme of the external mitochondrial membrane that degrades neurotransmitters. Its inhibition has anticonvulsant effects [37]. The KDT down-regulated the MAOA expression.

Some structural synapse proteins have shown also different gene expression after the KDT. DLGAP1 indirectly interacts with a voltage-dependent Ca${}^{2+}$ channel subunit that can modulate AMPA ($\alpha$-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) glutamate receptors and slow their deactivation, thereby increasing neuronal excitability [38]. Our results show that the KDT down-regulates this signalling pathway and therefore decreases excitability. Several studies in animal models indicate that the deletion of the chromosome where the DOC2A gene is encoded [39, 40] causes seizures. The treatment, by inducing an overexpression of this gene may explain the reductions in the number of seizures in our patients. PRKCG encodes the gamma subunit of PRKC, a protein kinase enzyme involved in many metabolic pathways, which has been associated to epilepsy [41]. It has been described that its inhibition has anticonvulsant effects [42]. KDT down-regulate this gene in our study.

In general, the changes we have seen in the gene expression related to synaptic transmission fit with a decrease in neuronal excitability, explaining at least part of the therapeutic effect of KDT in epilepsy.

Changes in miRNAs expression are especially important since they regulate the translation of mRNA and therefore control gene expression. Some of the miRNAs whose expression changed more significantly after KDT are involved in antioxidant pathways. It has been previously shown that glucose-induced oscillations of miR-185-5p impairs the cellular antioxidant response in humans through the deregulation of GPx-1 [43]. KDT reduce the oscillation in glucose levels and consequently, in our study, we have found a down-regulation in the expression of miR-185-5p.

On the other hand, miR-330 has been related to the stability of hippocampal neurons [44]. A protective role of this miRNA has been assigned in neurological disorders as Alzheimer’s disease [45]. miR-330-3p is up-regulated by PUFA enriched diets [46] and our results also show an up-regulation of this miRNA after fat enriched KDT.

The combined analysis of both (miRNAs and mRNA) transcriptomic results shows that miR-330-5p regulates the expression of 14 genes and miR-185-5p regulates the expression of 11 genes, two of them related to the central nervous system (CNS): CNTN2 and BSN. Other miRNAs which regulate CNS related genes are miRNA-199b-5p, miRNA-875-5p, miRNA-9-5p, miRNA-93-5p, miRNA-153-3p, and miRNA-323a-3p. Almost all the CNS-genes regulated by these miRNAS have been previously discussed. There are four additional genes regulated by these miRNAs that deserve to be briefly mentioned: CDH2, NEGR1, NEDD4 and CNTN2. Cadherin 2 and contactin 2, encoded by the CDH2 and the CNTN2 genes, respectively, are membrane proteins whose gene deletion has been linked to epilepsy [47, 48]. The KD-induced down-regulation of miRNA-199-5p (CDH2 expression regulator) and miRNA-185-5p (CNTN2 regulator) should be accompanied with an increased expression of these membrane proteins. However, this was the case only for CDH2. CNTN2 was slightly down-regulated in our patients after the treatment. A paradoxical result regarding the NEGR1 gene that is regulated by miRNA-875-5 was also found. It encodes a neuronal growth regulator related to axon regeneration whose deficiency increases seizures susceptibility [49]. The results show a down-regulation of miRNA-875-5 after KDT that is not accompanied with an up-regulation of NEGR1 expression. Finally, we studied a well-known epilepsy associated gene, i.e., NEDD4, known to be regulated by miRNA-9-5p [50]. Using loss of function studies, it has been found that mice deficient in Nedd4-2 increase their seizure susceptibility [50]. Our results show a down-regulation of miRNA-9-5p by KDT in the epileptic children. However, this was not accompanied with a concomitant increase in NEDD4 expression in the patients. Collectively our data show the complexity of the gene regulation and strengthen the idea that it involves not only miRNAs.

We have performed the transcriptional study in the epileptic paediatric patients in PBMCs to reduce the invasiveness of the study. We are assuming that KDT may have a similar effect on the nervous system cells than that found in blood mononuclear cells. We are aware that changes in gene expression varies from one tissue to another, but we would like to emphasize that the vast majority of transcriptomic human studies, especially in patients, are usually performed in blood cells [51, 52]. In addition, despite the coherence between the different individuals in the study and the fact that our results are consistent with the clinical effects of KDT, the number of patients analyzed is small, and this makes it difficult to generalize our results.

Treatment with KDT for more than 6 months reduces the number of seizures in epileptic paediatric patients and induces significant changes in their gene expression profile. Our study reveals changes at the transcriptomic level in genes encoding ion channels, neurotransmitter receptors, and synapse structural proteins. These findings may explain the effect of the diet on the epileptic crises. Our results contribute to understand the mechanisms of action of KDT and reinforce their clinical importance in the epilepsy treatment.

ASM, antiseizure medications; BDNF, brain-derived neurotrophic factor; BHB, $\beta$-Hydroxybutyrate; BMI, body mass index; BSN, bassoon presynaptic cytomatrix protein; CACNA1C, Ca${}^{2+}$ voltage-gated channel subunit alpha1 C; CDKL5, cyclin dependent kinase-like 5; CHRNB3, Neuronal acetylcholine receptor subunit beta-3; DLGAP1, Disks large-associated protein 1; DOC2A, Double C2-like domain-containing protein alpha; CKD, classic ketogenic diet; FGF12, fibrfoblast growth factor 12; GGT gamma-glutamyl transferase; GNG10, Guanine nucleotide-binding protein G(I)/G(S)/G(O) subunit gamma-10; GRM3, Glutamate Metabotropic Receptor 3; HCRTR1, hypocretin receptor type 1; HCRTR2, hypocretin receptor type 2; KCNA10, potassium voltage-gated channel subfamily A member 10; KCNJ12, Kir2.2. ATP-sensitive inward rectifier potassium channel 12; KCNK6, K2P6.1. potassium channel subfamily K member 6; KCNMA1, KCa1.1. Ca${}^{2+}$-activated potassium channel subunit alpha-1; KD, ketogenic diet; KDT, ketogenic dietary therapies; LCT, long chain triglycerides; MAD, modified Atkins diet; MAOA, monoamine oxidase A; MCT-KD, ketogenic diet with medium chain triglycerides; MTNR1B, Melatonin Receptor 1B; NMDA, N-methyl-D-aspartate; PBMCs, peripheral blood mononuclear cells; PCA, principal components analysis; PRKCG, Protein kinase C gamma; RBFOX1, RNA binding fox-1; SCN1A, sodium voltage-gated channel alpha subunit 1; SCN9A, sodium voltage-gated channel alpha subunit 9; SLC22A2, solute carrier family 22 member 2.

JRH, GOG, ECV, CPG, and JV planned the experiments. ES, AGC and GOG performed the transcriptomic experiments and JG, MCGC and JV supervised the experiment. CPG supervised each clinical step. GOG and JV wrote the first draft and all authors commented criticized, and reviewed the manuscript. All authors accepted the final version of the manuscript.

All experimental procedures were approved by the Committee on Ethics in Clinical Research of the University Children’s Hospital Niño Jesús of Madrid (R-0002/15). All patients or their relatives were fully informed of the aims and scope of the research and signed an informed consent.

We thank Marilyn Noyes for her kind help in reviewing the manuscript.

This work was supported by Sira Carrasco Grant 2015 from SEGHNP, Instituto de Salud Carlos III and FEDER [grant number PIE15/00013], SAF2016-75508 -R from the Spanish Ministry of Education and Science (MEC), CB16/10/00435 (CIBERFES), PROMETEOII2014/056 from Conselleria de Sanitat de la Generalitat Valenciana and EU Funded CM1001 and FRAILOMIC-HEALTH.2012.2.1.1-2, ADVANTAGE-724099 Join Action (HP-JA) 3rd EU Health Program and DIALBFRAIL-LATAM (825546 H2020-SC1-BHC).

The authors declare no conflict of interest.