A total of 90 patients were included in this study. Among them, 69 met all three diagnostic criteria for Glut1DS, while 17 fulfilled criteria (1) and (3). Four patients met criteria (1) and (2), with all exclusion criteria satisfied, and carried SLC2A1 variants classified as uncertain significance according to the ACMG guidelines.

Detailed clinical information is summarized in Tables S1 and Table 1. The cohort included one mother–child pair and one pair of monozygotic twins, representing 88 families. The male-to-female ratio was 1.2:1, with 49 males (54%) and 41 females (46%). Sixty-three patients (70%) presented with the classic type of Glut1DS, comprising 53 early onset and 10 late-onset patients. The remaining 27 patients (30%) exhibited the non-classic type of the disease. Seizures as the initial symptom presented in 41 patients (46%), developmental delay in 36 (40%), persistent or episodic movement disorders in seven (8%), episodic eye–head movements in five (6%), and episodic psychiatric or behavioral abnormalities in one (1%).

During the course of the disease, seizures occurred in 69 patients (77%), with onset ages ranging from 1 to 144 months (median: 30 months). Single and multiple seizure types were observed in 44 and 25 patients, respectively, including generalized tonic–clonic seizures in 18 , myoclonic seizures in 11 , absence seizures in nine , atonic seizures in five , generalized tonic seizures in four , myoclonic–atonic seizures in three , and eyelid myoclonia in two . Additionally, focal seizures were reported in 47 patients and status epilepticus in 5 .

Movement disorders were observed in 58 patients (64%, 58/90), with onset ages ranging from 1 to 14 years (median 2 years). The most common movement disorder was ataxia (39 patients), followed by ataxia-spastic gait (17), dystonia (nine ), and PED (eight ). Among these, 21 patients experienced persistent movement disorders with periodic exacerbations, while 20 had episodic movement disorders including PED in 8, ataxia in 12, and dystonia in 3, lasting 20 minutes to two hours. The most common trigger for PED was prolonged walking, whereas fasting was the most frequent trigger for ataxia and dystonia, followed by physical exertion, infections, and emotional excitement. Seventeen patients exhibited persistent movement disorders including ataxia, spastic ataxia, and dystonia, without episodic exacerbations. Progression of movement disorders was observed in 25 patients who were not on a ketogenic diet or a pre-ketogenic diet, with worsening between 2 and 18 years of age (median: 4 years). By October 2024, four patients without a ketogenic diet exhibited resolution of movement disorders, with resolution ages ranging from 10 to 20 years (median: 19 years).

Episodic eye–head movements were noted in 17 patients (19%, 17/90), characterized by rapid multidirectional eye movements with concurrent head movement in the same direction. These episodes began between 2 months and 1.5 years of age (median 4 months), with 71% (12/17) occurring within the first 6 months of life. Symptoms resolved between 1 and 4 years of age (median 2.5 years). Among these, 15 patients (88%) also experienced seizures, with 10 developing eye–head movements prior to seizures and five afterward.

Episodic hypersomnolence was observed in 26 patients (29%), episodic psychiatric or behavioral abnormalities in 17 (19%), sleep disturbances in 13 (14%), migraines in eight (9%), periodic vomiting and attention-deficit hyperactivity disorder (ADHD)-like symptoms in seven (8%) each, strabismus in three (3%), and stroke-like episodes in three. Anxiety, depression, type II diabetes, and autism spectrum disorder were each noted in a single case.

Head circumference measurements were taken from 74 patients between 0.2 and 19 years of age (median: 4 years). Among them, 47% (35/74) presented with normal head circumferences, 26% (19/74) displayed microcephaly, and 27% (20/74) exhibited mildly reduced head circumferences.

Eleven families with more than one case of Glut1DS, except a family with a pair of twins, were identified, accounting for 12.5% of the cohort. These families included 12 children and 12 adults, with one family comprising four members carrying SLC2A1 variants. The inheritance pattern in these families was consistent with autosomal dominance, with SLC2A1 variants inherited from the father in eight cases and from the mother in four cases. The identified variants included nine missense forms, one nonsense variant, and one frameshift mutation caused by a small insertion. Notably, three families carried the most prevalent variant, c.997C > T (p.R333W). Among the affected children, eight were male and four were female, with age of onset ranging from 3 to 120 months (median: 6 months). Of these cases, four presented with the early onset classic form, two with the late-onset classic form, and six with the non-classic form. The ages of the parents harboring SLC2A1 variants ranged from 29 to 43 years (median: 37 years). Seven parents exhibited non-classic symptoms, including PED in four cases, migraines in two cases, and poor memory in one case. Five parents were asymptomatic, reflecting an estimated penetrance rate of 79%.

CSF glucose levels were measured in 73 patients (81%), ranging from 1.0 to 2.6 mmol/L (median: 1.9 mmol/L; IQR: 1.7–2.0 mmol/L). Among these, 66 patients (90%) exhibited CSF glucose levels below 2.2 mmol/L. CSF-to-blood glucose ratios were available for 71 patients (79%), ranging from 0.20 to 0.63 (median: 0.37; IQR: 0.33–0.42). Notably, 46 patients (65%) and 62 patients (87%) had ratios below 0.4 and 0.45, respectively. Interestingly, one patient with large fragment deletions, including an exon 1 deletion, and presenting with mild clinical symptoms such as predominantly paroxysmal movement disorder, had a CSF glucose level of 2.6 mmol/L and a CSF-to-blood glucose ratio of 0.63. The patient was consistent with the diagnosis of this disease. CSF lactate levels, measured in 40 patients (44%), ranged from 0.2 to 1.5 mmol/L (median: 0.9 mmol/L; IQR: 0.8–1.0 mmol/L). CSF cell counts were normal in all patients.

In patients with the early onset classic phenotype (n = 41), CSF glucose levels ranged from 1.0 to 2.2 mmol/L (median: 1.8 mmol/L; IQR: 1.5–1.9 mmol/L), with CSF-to-blood glucose ratios (n = 40) ranging from 0.22 to 0.53 (median: 0.35; IQR: 0.30–0.38). Lactate levels (n = 23) ranged from 0.2 to 1.5 mmol/L (median: 0.9 mmol/L; IQR: 0.8–1.1 mmol/L). Cases with the late-onset classic phenotype (n = 9) demonstrated CSF glucose levels ranging from 1.7 to 2.3 mmol/L (median: 2.1 mmol/L; IQR: 1.8–2.2 mmol/L) and ratios from 0.37 to 0.46 (median: 0.43; IQR: 0.41–0.44). Lactate levels in these cases (n = 5) ranged from 0.6 to 1.1 mmol/L (median: 0.9 mmol/L; IQR: 0.77–1.0 mmol/L). Patients with the non-classic phenotype (n = 23) exhibited CSF glucose levels between 1.6 and 2.6 mmol/L (median: 1.9 mmol/L; IQR: 1.8–2.1 mmol/L), with ratios from 0.28 to 0.63 (median: 0.39; IQR: 0.35–0.42). Lactate levels in this group (n = 12) ranged from 0.3 to 1.4 mmol/L (median: 0.9 mmol/L; IQR: 0.78–1.1 mmol/L). Significant differences in CSF glucose levels (P = 0.003) and CSF-to-blood glucose ratios (P = 0.001) were observed among the three groups (Table 1).

Sixty-nine unique heterozygous SLC2A1 variants were identified among 90 patients, comprising 39 previously reported and 30 novel variants (Table S2). Of the unreported variants, 23 were classified as pathogenic, three as likely pathogenic, and four as variants of uncertain significance (VUS). All four VUS variants, c.412G > C (p.Gly138Arg), c.431 T > G (p.Val144Gly), c.875A > G (p.Tyr292Cys), and c.1243A > G (p.Asn415Asp), were located at highly conserved amino acid positions. Bioinformatics tools, such as MutationTaster, PolyPhen-2, SIFT, and CADD, predicted these variants to be deleterious. Structural modeling using pMOL software indicated that these amino acid substitutions disrupted protein stability by altering intramolecular interactions, confirming pathogenicity. Clinical presentations of these four patients, as well as CSF glucose levels below 2.2 mmol/L (1.97, 1.86, 2.08, and 1.88 mmol/L) and CSF-to-blood glucose ratios under 0.45 (0.42, 0.4, 0.43, and 0.38), satisfied the clinical diagnostic criteria for Glut1DS.

Among the 69 variants, 46 (51%) were missense mutations, 19 (21%) were frameshift variants (10 deletions, six insertions, and three combined deletion–insertions), nine (10%) were single/multiple exon or whole-gene deletions, seven (8%) were nonsense mutations, four (4%) were splicing variants, three (3%) were in-frame deletions/insertions, and two (2%) affected start codons (Fig. 1). The most common variant, c.997C > T (p.Arg333Trp), was identified in eight probands (9%), followed by c.988C > T (p.Arg330*), c.274C > T (p.Arg92Trp), and c.376C > T (p.Arg126Cys), which were found in six, three, and three probands, respectively. The variants c.102 T > A (p.Asn34Lys), c.398G > A (p.Cys133Tyr), c.457C > T (p.Arg153Cys), c.2 T > A (p.Met1?), and c.680-1G > A were each found in two probands. The remaining 60 variants were unique to individual probands. There were no statistically significant differences in the distribution of SLC2A1 gene variant types among the early onset classic, late-onset classic, and atypical groups (Table 1).

The mutation distribution map of the SLC2A1 gene in Glut1DS patients is as follows: the black vertical line represents the SLC2A1 gene, with exons denoted by boxes. Missense and nonsense mutations are displayed to the left of each box, while frameshift mutations, in-frame insertions/deletions, and splice-site mutations are shown to the right. Numbers in parentheses indicate the number of patients harboring each specific mutation. Cases of multi-exon/single-exon deletions or whole-gene deletions (not shown) include eight instances of novel mutations. Asterisks highlight mutation sites that have not been previously reported in the literature

Exon 4 was the most frequently affected, accounting for 26% of the variants, followed by exon 8, which accounted for 19%. The distribution of variants across the GLUT1 protein revealed that 44% (34/77) were in intracellular regions, 31% (24/77) were in transmembrane regions, and 25% (19/77) were in extracellular regions (Fig. 2).

Distribution of SLC2A1 gene mutation sites on the GLUT1 protein. The map illustrates 57 mutation sites identified in 77 patients with Glut1 deficiency syndrome (Glut1DS). Mutations involving multi-exon/single-exon deletions, whole-gene deletions (nine cases), and intron mutations (four cases) are not displayed. Mutation sites marked in red indicate novel mutations unreported previously. Among the mutations, 44% (34/77) are located in intracellular regions, 31% (24/77) in transmembrane regions, and 25% (19/77) in extracellular regions

Among the 69 patients with seizures, 59 (87%) received antiepileptic drug (AED) therapy prior to initiating the ketogenic diet. Of these, 32 patients (53%) used one AED, 18 (30%) used two, six (10%) used three, and four (7%) utilized between four and seven AEDs. The most commonly used AEDs included valproate (34 cases) and levetiracetam (31 cases), followed by oxcarbazepine (13 cases), lacosamide (six cases), lamotrigine and phenobarbital (five cases each), topiramate (four cases), clobazam (three cases), clonazepam and perampanel (two cases each), and gabapentin (one case). Treatment efficacy details are summarized in Fig. 3. In general, AEDs showed some effects in controlling seizures, although the effect was limited.

Efficacy of antiepileptic drugs (monotherapy/combination) in children with Glut1DS. VPA valproate, LEV levetiracetam, OXC oxcarbazepine, LCM lacosamide, LTG lamotrigine, PB phenobarbitone, CLB clobazam, TPM topiramate, CZP clonazepam, PER perampanel, GAB gabapentin

A total of 79 patients underwent ketogenic diet therapy, initiated at ages ranging from 0.25 to 14 years (median: 3.8 years). The duration of therapy varied between 0 and 14 years (median: 4.2 years). At the start of the ketogenic diet, 54 patients were receiving AED therapy concurrently. As of the last follow-up in October 2024, 73 patients remained on the diet, while six had discontinued it. AEDs had been successfully discontinued in 74% (40/54) of patients. Several types of ketogenic diets were used in our cohort, ranked by the number of patients as follows: classic ketogenic diet, medium-chain triglyceride (MCT) diet, and modified Atkins diet.

Among 57 patients with seizures, 47 (82%) achieved seizure freedom, including 40 on the ketogenic diet alone, five on the diet combined with AEDs, and two with an unknown regimen. Four patients experienced > 50% seizure reduction (one on the diet alone, three with AEDs), one experienced < 50% seizure reduction (with AEDs), and five showed no change in seizure frequency (all on a combined regimen). The timing of seizure resolution varied, with 28 patients achieving seizure freedom more than 1 month before ketogenic diet initiation, 15 within one week, one within two weeks, and three within 1 month. Among the five patients with partial seizure control, one improved within a week, three within 2 weeks, and one within a month of starting the diet. In addition to seizure control, 18 of the 47 patients with movement disorders experienced resolution of their symptoms, while 26 showed improvement, and three exhibited no changes. The improvement occurred between one week and six months after initiating the ketogenic diet. Among the 71 patients (including two who discontinued the diet) who adhered to the diet for more than six months, 48 demonstrated cognitive and language improvements, while 23 showed no significant changes.

Six patients discontinued the ketogenic diet after durations ranging from 0.3 to 72 months. Reasons for discontinuation included poor compliance (three patients), severe metabolic acidosis with associated diarrhea and refusal to eat (one ), lack of efficacy accompanied by abdominal pain, vomiting, and increased seizure frequency (one ), and lack of efficacy with increased seizure frequency (one ).

Adverse reactions at the initiation phase of the ketogenic diet were reported in 23 patients, who experienced one to four side effects each. The most common initial side effects were diarrhea or vomiting, followed by metabolic acidosis, hypoglycemia, constipation, electrolyte disturbances, abdominal pain, and liver dysfunction. All adverse events were manageable. During the maintenance phase, 32 patients reported side effects, with the most common being elevated uric acid levels, followed by hyperlipidemia, constipation, hair loss, weight loss or stunted growth, urinary tract stones, intermittent diarrhea, and intermittent abdominal pain. Despite these side effects, the ketogenic diet was well tolerated overall, with most adverse reactions being transient and treatable.