If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Department of Pediatrics, Osaka University Graduate School of Medicine, 2-2 Yamadaoka, Suita, Osaka 565-0871, JapanDepartment of Child Development, United Graduate School of Child Development, Osaka University, 2-2 Yamadaoka, Suita, Osaka 565-0871, Japan
Department of Pediatrics, Osaka University Graduate School of Medicine, 2-2 Yamadaoka, Suita, Osaka 565-0871, JapanDepartment of Child Development, United Graduate School of Child Development, Osaka University, 2-2 Yamadaoka, Suita, Osaka 565-0871, Japan
Department of Clinical Research, National Epilepsy Center, NHO Shizuoka Institute of Epilepsy and Neurological Disorders, 886 Urushiyama, Aoi, Shizuoka, Shizuoka 420-8688, Japan
Department of Pediatrics, School of Medicine, Iwate Medical University, 2-1-1 Idaidori, Yahaba, Shiwa, Iwate 028-3695, JapanEpilepsy Clinic Bethel Satellite Sendai-Station, Comfort Hotel Sendai-Higashiguchi #1F, 205-5 Nakakecho, Miyagino, Sendai, Miyagi 983-0864, Japan
Department of Pediatrics, Aichi Medical University, 1-1, Yazakokarimata, Nagakute, Aichi 480-1195, JapanDepartment of Human Genetics, Yokohama City University Graduate School of Medicine, 3-9 Fukuura, Kanazawa, Yokohama, Kanagawa 236-0004, Japan
Department of Human Genetics, Yokohama City University Graduate School of Medicine, 3-9 Fukuura, Kanazawa, Yokohama, Kanagawa 236-0004, JapanClinical Genetics Department, Yokohama City University Hospital, 3-9 Fukuura, Kanazawa, Yokohama, Kanagawa 236-0004, Japan
Department of Biomedical Statistics, Graduate School of Medicine and Integrated Frontier Research for Medical Science Division, Institute for Open and Transdisciplinary Research Initiatives (OTRI), Osaka University, 2-2 Yamadaoka, Suita, Osaka 565-0871, Japan
Glucose transporter 1 deficiency syndrome severity is predictable.
•
Cerebrospinal fluid glucose correlates with the severity.
•
Cerebrospinal fluid lactate is another predictor of severity.
•
Parameters for prediction differ between older children and those <1 year of age.
•
Innate severity should be considered while evaluating ketogenic diet efficacy.
Abstract
Objective
In glucose transporter 1 deficiency syndrome (Glut1DS), cerebrospinal fluid glucose (CSFG) and CSFG to blood glucose ratio (CBGR) show significant differences among groups classified by phenotype or genotype. The purpose of this study was to investigate the association between these biochemical parameters and Glut1DS severity.
Methods
The medical records of 45 patients who visited Osaka University Hospital between March 2004 and December 2021 were retrospectively examined. Neurological status was determined using the developmental quotient (DQ), assessed using the Kyoto Scale of Psychological Development 2001, and the Scale for the Assessment and Rating of Ataxia (SARA). CSF parameters included CSFG, CBGR, and CSF lactate (CSFL).
Results
CSF was collected from 41 patients, and DQ and SARA were assessed in 24 and 27 patients, respectively. Simple regression analysis showed moderate associations between neurological status and biochemical parameters. CSFG resulted in a higher R2 than CBGR in these analyses. CSF parameters acquired during the first year of life were not comparable to those acquired later. CSFL was measured in 16 patients (DQ and SARA in 11 and 14 patients, respectively). Although simple regression analysis also showed moderate associations between neurological status and CSFG and CSFL, the multiple regression analysis for DQ and SARA resulted in strong associations through the use of a combination of CSFG and CSFL as explanatory variables.
Conclusion
The severity of Glut1DS can be predicted from CSF parameters. Glucose and lactate are independent contributors to the developmental and neurological status in Glut1DS.
Glucose is an indispensable energy source in the brain, and its delivery depends on glucose transporter type 1 (GLUT1), which is the sole glucose transporter expressed on the endothelial cells of blood vessels in the brain. Genetic alterations in GLUT1 (SLC2A1) can result in GLUT1 deficiency syndrome (Glut1DS), which is characterized by a variety of symptoms, including developmental delay, refractory epilepsy, spasticity, ataxia, exertion-induced dyskinesia, fatigability, and other neurological symptoms caused by neuronal energy deficits. Notably, patients with Glut1DS also have decreased glucose levels (CSFG) in the cerebrospinal fluid (CSF) and lower CSFG to blood glucose ratios (CBGRs).
After the first case report of a functional deficit in GLUT1 in 1991 [
]. The classic phenotype is the most common and severe form of the disease and is characterized by nearly the entire symptom constellation associated with Glut1DS. Milder phenotypes present as a diverse combination of symptoms of varying severity.
Previous studies have reported an association between higher CBGRs and the mild phenotypes of Glut1DS [
Based on these correlation studies, the severity of Glut1DS is considered to be proportional to the CSF parameters. We investigated this hypothesis retrospectively, examining the potential predictive factors by regression analysis.
2. Methods
2.1 Standard protocol approvals, registrations, and patient consents
This study was approved by the Osaka University Clinical Research Review Committee (approval number 15332) and the Osaka University Research Ethics Committee (approval number 798) in accordance with the Helsinki Declaration. Written informed consent for publication was obtained from the patients, parents, or caregivers, and no participant received any monetary compensation.
2.2 Participant recruitment
The medical information of patients with Glut1DS who visited Osaka University Hospital between March 2004 and December 2021 was retrospectively collected. The diagnostic criteria were those of a recent consensus statement [
]. The criteria were: characteristic clinical features, hypoglycorrhachia, and the presence of a pathogenic SLC2A1 variant. Patients were classified into four categories: confirmed, probable, possible, and negative. A confirmed patient satisfied all three criteria. A probable patient was required to have hypoglycorrhachia and one of the other two criteria. A possible patient had either hypoglycorrhachia or the pathogenic SLC2A1 variant.
Patients were classified according to phenotype and genotype to allow comparison of their clinical features with those of a preceding study [
]. The phenotype classification recognized three phenotypes: the early-onset classical phenotype, which shows seizure onset before the age of 2 years; the late-onset classical phenotype, which shows a later seizure onset; and the non-classical phenotype, which shows no epileptic episodes. The confirmed patients were classified into three genotypic groups: A, with a heterozygous deletion involving the entire gene; B, with nonsense, out-of-frame frameshift, or splice-site variants; and C, with in-frame frameshift or missense variants.
2.3 CSF analysis
CSF was acquired by lumbar puncture after >4–6 h of fasting, and blood glucose sampling was performed immediately before CSF sampling. Some patients underwent several lumbar punctures. CSFG and CBGR are normally low in infants ≤6 months of age, and the levels vary widely during the first year. Moreover, CSFL can be relatively high at <4 weeks of age [
]. Therefore, the CSF parameters at diagnosis were adopted as representative values. The CSF parameters of patients diagnosed at <1 year of age were analyzed separately, and their CSF parameters were reevaluated after 1 year of age if their parents or caregivers consented.
2.4 Gene analysis
A variant analysis of all exons and intron-exon boundaries of the SLC2A1 gene was performed. Genomic DNA was extracted from blood (preserved with ethylenediaminetetraacetic acid) using a Maxwell 16 system (Promega, WI, USA). Polymerase chain reaction (PCR) template primer sets (Table 1) were designed to cover the exons and exon-intron boundaries of the SLC2A1 gene. PCR was conducted using the KOD FX (Toyobo, Osaka, Japan) according to the manual. The amplified products were purified with ExoSAP-IT (Thermo Fisher Scientific, MA, USA). The SLC2A1 gene was directly sequenced using 20 ng of purified DNA with the BigDye Terminator v3.1 Cycle Sequencing Kit (Thermo Fisher Scientific, MA, USA). Capillary electrophoresis was performed using the Applied Biosystems 3730DNA analyzer (Thermo Fisher Scientific, MA, USA).
Table 1The primer sets for the direct sequencing of the SLC2A1 gene.
There is a common insertion polymorphism at c.1278 + 31insATTTCTCACC in intron 9. Therefore, another forward primer is necessary to confirm the sequence in exon 10.
PCR, polymerase chain reaction.
It indicates primers identical to the primer used for template preparation.
† There is a common insertion polymorphism at c.1278 + 31insATTTCTCACC in intron 9. Therefore, another forward primer is necessary to confirm the sequence in exon 10.
Variant nomenclature was based on the coding region of the SLC2A1 transcript (SLC2A1: RefSeq NM_006516.3, SLC2A1: NP_006507.2). After detecting a novel variant, we searched the Human Gene Mutation Database [
], which utilizes Polyphen-2, SIFT, PROVEAN, and PANTHER, was applied. The variant was analyzed using Human Splicing Finder 3.1 to assess the pathogenicity of any intronic variant that appeared to create a new splice acceptor site [
For a patient with a variant that may cause an in-frame frame shift at a splice site, cDNA analysis was conducted. RNA was extracted from the patient's and her parents' blood using the Nucleospin RNA Blood kit (Macherey-Nagel, Düren, Germany), and cDNA was synthesized using ReverTra Ace qPCR RT Master Mix with gDNA Remover (Toyobo, Osaka, Japan). After cDNA synthesis, the same kits were used in the DNA sequence confirmation described above. The primers for cDNA template amplification were: forward, 5’-TGGCCGGCGGAATTCAATGC-3′; and reverse, 5’-ACAGCGACACGACAGTGAAG-3′. The forward primer for chain termination was 5’-ACTGGGCAAGTCCTTTGAGA-3′.
If a variant was not found by Sanger sequencing, multiplex ligation-dependent probe amplification (MLPA) was performed to search for single or multiple exon deletions or duplications. When MLPA detected a single-allele deletion, a whole-genome single nucleotide polymorphism (SNP) array assay was performed (Lab Corp Japan, Tokyo, Japan).
2.5 Clinical information
Patient history, clinical findings, and laboratory data were obtained from medical records.
The Kyoto Scale of Psychological Development 2001 (KSPD), which was developed in Japan, was used to assess development [
]. It is administered face-to-face by a psychologist, contains 328 test items, and can be used with all age groups, from infants to adults. It calculates a developmental quotient (DQ) by dividing the developmental age by the chronological age. Therefore, in a typically developing child, the DQ is 100. The KSPD was administered during regular clinical evaluations, and the timing of administrations was not specifically planned for this study. For patients who underwent KSPD testing multiple times, the DQ score obtained closest in time to the lumbar puncture was used.
Patients with Glut1DS often present complex movement disabilities comprising spasticity, ataxia, and others. Spasticity more often influences motor performance than other elements depending on severity. However, existing evaluation systems of spasticity, such as the Ashworth Scale or modified Ashworth Scale, consist of subjective scales of each muscle, and require a certain degree of patient cooperation [
]. Among these neurological manifestations, cerebellar ataxia was the most steadily observed in our patients. There is a well-established and easily applicable scaling system for cerebellar ataxia: the Scale for the Assessment and Rating of Ataxia (SARA). SARA is a validated assessment system for ataxia in the four extremities and trunk and is utilized broadly for clinical and research purposes [
]. Patients with Glut1DS often present complex movement disabilities other than ataxia. Thus, SARA scores for Glut1DS patients reflect not only ataxia but also combined motor disabilities; hence, they indicate the extent of clumsiness in patients' movements. Scores range from 0 (no ataxia) to 40 (most severe ataxia). Patients with minimal dependence have scores ≤5.5, while totally dependent patients have scores ≥23 [
]. The SARA was assessed by the doctor in charge and was administered when patients were deemed to be stable, since the neurological status of patients with Glut1DS usually fluctuates and is affected by diverse factors.
2.6 Statistical analyses
The correlation coefficient between the DQ and SARA was calculated to confirm the reliability of the DQ. Subsequently, the associations between the DQ, SARA, and CSF parameters CSFG and CBGR at diagnosis were examined using simple linear regression analysis, which was equivalent to the derivation of the correlation coefficient. The Kruskal–Wallis test was performed to assess the differences in CSFG, CBGR, the DQ, and the SARA between genotypes, while the Mann–Whitney U test was employed to assess the differences between phenotypes.
In addition, the association between the DQ and CSF parameters stratified by age at CSF acquisition (before or after 1 year of age) was examined by creating separate graphical plots and simple regression lines broken down by CSF acquisition period.
Lastly, for patients with CSFL data, similar regression analyses were performed between the DQ, SARA, and CSF parameters, including CSFL. Multiple regression analyses were also performed to characterize the role of the CSF parameters CSFG and CSFL in determining developmental and neurological status. Statistical analyses were performed using IBM SPSS Statistics for Windows, version 28 (IBM Corp., Armonk, N.Y., USA).
3. Results
This study included 45 patients (24 females and 21 males) who visited Osaka University Hospital during the study period. (Table 2) Forty-one patients had a confirmed diagnosis and four had a probable diagnosis.
A, genotype A (hemizygosity); B, genotype B (nonsense, out-of-frame-shift, splice-site mutations); C, genotype C (in-frame-shift and missense mutations); d, day of age; DQ, developmental quotient; EOC, early-onset classical; KD, ketogenic diet; KSPD, Kyoto Scale of Psychological Development 2001; LOC, late-onset classical; m, month of age; NA, not available; NC, nonclassical; ND, not done; nd, not detected; Ref. References; SARA, Scale for the Assessment and Rating of Ataxia; y, year of age.
a Patient who initiated the ketogenic diet in infancy, data omitted from the simple linear regression.
Among the 41 patients with an SLC2A1 variant, three had genotype A, 16 had genotype B, and 22 had genotype C, including one patient with an in-frame insertion and 21 patients with missense variants.
This population included four familial cases. The families were named A (patients 5 and 19), B (8 and 10), C (27, 35, and 39), and D (36, 37, 38, 40, and 41). All patients had SLC2A1 variants. Family B included a monozygotic twin with a de novo variant. The founder was detected in families A and D through Sanger sequencing. The genetic trait of these families was autosomal dominant except for family B.
Patient 20 had a c.680-11G > A variant. Human Splicing Finder 3.1 suggested that this variant led to a broken acceptor site and most likely affected splicing. Subsequently, cDNA analysis confirmed that it caused an in-frame 9-base-pair insertion (TCCCCCCAG) upstream from exon 6 [
], and nine patients from five families had novel variants (A70T (patient 21), N34K (patient 23), G91C (patient25), P385S (patient 38), and P385L (family D, including patients 36, 37, 38, 40, and 41). These variants were not found in the Human Gene Mutation Database or the Genome Aggregation Database. In silico predictive algorithms showed that these five variants were possibly pathogenic (Table 3). Among these patients, we confirmed the variants were de novo in patients 21, 23, 25, and 38. Among the novel variants, a p.P385S variant was found in a familial case (family D). I-1 showed a low-level variant of c.1153C > T, as did his daughters (patients 40 and 41) and granddaughters (patients 36, 37, and 38: daughters of patient 41). This variant was determined as likely pathogenic (PM2, PP1, PP3, PP4, and PP5) according to the standards and guidelines of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG/AMP) [
Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology.
]. Moreover, all except patient 41 showed Glut1DS symptoms; therefore, we judged that this variant was pathogenic. Accordingly, we judged I-1 to be the founder. Every other patient with novel missense variants had a change at an amino acid residue where other amino-acid changes had previously been established as pathogenic (A70T [
]), and patient 38 had a variant of P385L which was the same amino acid residue in family D. Therefore, these missense variants were determined as pathogenic (PS2, PM2, PM5, PP3, PP4 for patients 21, 23, and 25; and PS2, PM2, PP1, PP3, PP4, and PP5 for patient 38) according to ACMG/AMP guidelines [
Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology.
The MLPA study revealed a one-allele deletion in three patients who all had a deletion throughout exon 1–10. An SNP array was performed in these patients, which found interstitial deletions 1p34.2p34.1(43,109,278-45,174,793) (2.07 MB), 1p34.2(43,393,675-43,464,290) (71 KB), and 1p34.2(43,393,675-43,427,295) (34 KB) in patients 1, 2, and 3, respectively (Fig. 1). Within the largest deletion region in patient 1, no gene was found that causes a neurological deficit in a single-allele deletion.
Fig. 1Map of the deletion, size, extent, and genomic content in patients with heterozygous deletions involving the entire SLC2A1 gene. The RefSeq genes located in the region are shown.
The CSF data of 41 patients were available. The CSFG at diagnosis ranged 24–48 mg/dL (median, 34.0), while the CBGR ranged 0.24–0.53 (median, 0.38). CSFG, CBGR, and changes in them were plotted for patients whose CSF was collected multiple times (Fig. 2A and B ).
Fig. 2A CSFG values of all patients were arranged in chronological order, including changes in CSFG in patients who underwent cerebrospinal fluid collection multiple times. The values fluctuate; however, most are below the lower limit of the normal range. B CBGR values in the same patients. These values also fluctuate, and some are above the lower limit of the normal range. Moreover, the range of fluctuation is broader than that of CSFG. CBGR, cerebrospinal fluid glucose to blood glucose ratio; CSFG, cerebrospinal fluid glucose.
The median age at evaluation was 15.0 years (range, 1–55 years); the median age of onset was 6.0 months (range, 0 days to 27 years); and the median age at diagnosis was 6.0 years (range, 1 month to 48 years). Twenty-nine patients had the early-onset classical phenotype (64%), 15 had the late-onset phenotype (33%), and one had the non-classical phenotype (2%).
KD, including the modified Atkins diet, was administered to 38 patients immediately after diagnosis. The KD strength was adjusted to maintain beta-hydroxybutyrate levels at ≥2000 μmol/L. Most patients achieved this level with a ketogenic ratio between 1.8:1 and 3.0:1. The diet strength was later adjusted for some patients based on their condition. Three patients did not continue the diet owing to poor compliance, while six did not initiate the diet.
The DQ was plotted over time for patients whose DQ was evaluated multiple times (Fig. 3). The DQ values of the patients changed little as they aged.
Fig. 3DQ values of all patients were arranged in chronological order, including changes in the DQ in patients assessed multiple times with the Kyoto Scale of Psychological Development 2001. The values fluctuate but do not show a consistent trend. DQ, developmental quotient.
Six patients were diagnosed before 1 year of age. CSF acquisition was performed to rule out meningitis in febrile infants (patient 1) and apneic infants (patient 6), to screen for Glut1DS causing opsoclonus (patient 16), after seizure aggravation on starvation (patient 26), and as part of a routine neurological examination in cases of refractory epilepsy (patients 5 and 25). Their CSFG and CBGR at diagnosis were 24–40 mg/dL (median, 32) and 0.24–0.39 (median, 0.33), respectively. KD, which utilizes a specialized formula, was administered immediately after diagnosis. All patients maintained beta-hydroxybutyrate levels ≥4000 μmol/L. Four patients had high DQ scores (> 80), whereas the other patients had scores of 27 and 52. Therefore, not all patients in whom KD was initiated during early infancy attained good development.
3.4 Association between the DQ, SARA, and CSF parameters
The DQ was evaluated in 24 patients and SARA in 27. Pearson's correlation coefficient (r) was calculated between the SARA and DQ, and a strong correlation was found (r = −0.935, p < 0.001) (Fig. 4).
Fig. 4Simple linear correlation analysis between the DQ and SARA. DQ, developmental quotient; SARA, Scale for the Assessment and Rating of Ataxia.
Simple linear regression analyses between the DQ and CSFG, the DQ and CBGR, the SARA and CSFG, and the SARA and CBGR gave coefficients of determination (R2) of 0.601 (p < 0.001), 0.478 (p < 0.001), 0.619 (p < 0.001), and 0.597 (p < 0.001), respectively (Fig. 5A to D ). The relationship between these parameters was preserved when the data were stratified by the difference between the age at CSF collection and age at the assessment of development or cerebellar ataxia (Fig. 6A to D ).
Fig. 5Simple regression analysis of development, cerebellar ataxia, and cerebrospinal fluid parameters. A Simple regression analysis between the DQ and CSFG. B Simple regression analysis between the DQ and CBGR. C Simple regression analysis between the SARA and CSFG. D Simple regression analysis between the SARA and CBGR. CBGR, cerebrospinal fluid glucose to blood glucose ratio; CSFG, cerebrospinal fluid glucose; DQ, developmental quotient; SARA, Scale for the Assessment and Rating of Ataxia.
Fig. 6Stratified simple regression analysis of development, cerebellar ataxia, and cerebrospinal fluid parameters. A Simple regression analysis between the DQ and CSFG, showing the scatter plot stratified by the interval between CSF acquisition and DQ assessment. B Simple regression analysis between the DQ and CBGR. C Simple regression analysis between the SARA and CSFG, showing the scatter plot stratified by the interval between CSF acquisition and SARA assessment. D Simple regression analysis between the SARA and CBGR. CBGR, cerebrospinal fluid glucose to blood glucose ratio; CSFG, cerebrospinal fluid glucose; DQ, developmental quotient; SARA, Scale for the Assessment and Rating of Ataxia.
3.5 Correlations between biochemical parameters and genotype/phenotype
We observed a significant correlation of genotype and phenotype with CSF parameters and neurological status. The Kruskal–Wallis test showed an evident difference in CSFG, CBGR, the DQ, and the SARA among genotypes A, B, and C. In genotype C, all these parameters showed a mild phenotype. Their p-values were < 0.001, 0.002, 0.001, and < 0.001, respectively. Regarding phenotype, the Mann–Whitney U test showed a significant difference in CSFG, CBGR, the DQ, and the SARA between the early- and late-onset classical phenotypes. Moreover, the late-onset classical phenotype was always milder. The p-values were 0.002, 0.002, < 0.001, and < 0.001, respectively.
3.6 Influence of the CSF collection period on the association between the DQ and CSF parameters
For the six patients diagnosed before 1 year of age, we overlaid their DQ vs. CSFG and DQ vs. CBGR scatter plots on Fig. 5A and B, respectively (Fig. 7A and B ). Among these regression lines, analysis of covariance (ANCOVA) showed a significant difference between the patients diagnosed before 1 year of age and those diagnosed after 1 year of age for the DQ vs. CSFG (p = 0.002) and DQ vs. CBGR scatter plots (p < 0.001). In three patients whose CSF was reevaluated after 1 year of age, the CSFG and CBGR values increased, and the relation between the DQ and CSF parameters approached the regression line of patients diagnosed after 1 year of age (Fig. 7C and D).
Fig. 7Difference in development between patients diagnosed before and after 1 year of age. A Simple regression analysis between the DQ and CSFG in patients diagnosed during the first year of life versus later. The values of the patients diagnosed later than one year of age are stratified by the age of CSF acquisition. ANCOVA shows a significant difference between DQ in patients diagnosed during the first year of life versus later. (p = 0.002). B The corresponding result with regression between the DQ and CBGR; a more significant difference is observed (p < 0.001). C Change in CSFG at reevaluation after 1 year of age. All three patients exhibited an increase in CSFG from diagnosis. D Change in CBGR at reevaluation after 1 year of age. All three patients exhibited an increase in CBGR from diagnosis. ANCOVA, analysis of covariance; CBGR, cerebrospinal fluid glucose to blood glucose ratio; CSFG, cerebrospinal fluid glucose; DQ, developmental quotient; SARA, Scale for the Assessment and Rating of Ataxia.
3.7 Role of CSFG and CSFL in development and neurological status
A similar analysis was performed for the 16 patients whose CSFL was measured. CSFG, CSFL, and changes in these parameters for these patients are shown in Fig. 8A and B . Simple regression analyses between the DQ and CSFG, the DQ and CSFL, the SARA and CSFG, and the SARA and CSFL gave R2 values of 0.634 (p = 0.003), 0.680 (p = 0.002), 0.716 (p < 0.001), and 0.642 (p < 0.001), respectively (Fig. 9A to D ). The CSFG results were consistent in all 24 patients. In these patients, a simple linear correlation analysis between CSFG and CSFL gave an r of 0.479 (p = 0.061) (Fig. 10). A multiple regression model that included CSFG and CSFL gave an R2 of 0.881 for DQ and 0.899 for SARA. (Table 4).
Fig. 8A Changes in CSFG in patients whose CSFL was measured. B Changes in CSFL in the same patient population. Although most are at or near the lower limit of the normal range, and a few are clearly within the normal range, especially infantile patients.
Fig. 9Simple regression analysis of DQ, SARA, and CSF parameters in patients with known CSFL values. A Simple regression analysis between the DQ and CSFG. B Simple regression analysis between the DQ and CSFL. C Simple regression analysis between the SARA and CSFG. D Simple regression analysis between the SARA and CSFL. CSFG, cerebrospinal fluid glucose; CSFL, cerebrospinal fluid lactate; DQ, developmental quotient; SARA, Scale for the Assessment and Rating of Ataxia.
This high R2 value suggests that CSFG and CSFL both play important roles in psychomotor development and motor control. Moreover, the substantial increase in R2 observed with the inclusion of CSFL suggests that CSFL may also contribute to neurological status independently of CSFG.
4. Discussion
This study demonstrated a good proportional relationship between CSF biomarkers and disease severity as measured by continuous variables in patients with Glut1DS. CSFG and CBGR showed a moderate association with the DQ and SARA scores using a simple regression analysis. Therefore, our findings showed that developmental and neurological status can be predicted from CSF parameters.
In this population, CSFG was slightly superior to CBGR in the regression analysis. CBGR was found to fluctuate widely in patients when CSF was collected repeatedly. CSFG may directly represent the amount of glucose supplied to neurons, while CBGR may reflect GLUT1 activity. Blood glucose largely fluctuates according to meals, hormones, or stress, and CSFG also varies diurnally [
]; however, the fluctuation is faster and steeper in blood than in CSF. We presume that a relatively more stable CSFG would be preferable to CBGR as a determinant of developmental and neurological status.
Consistent with previous studies, we found that CSFG and CBGR vary widely in the first year of life (Fig. 2A and B) [
] and found a proportional trend between developmental status and CSF parameters in patients diagnosed before 1 year of age. Therefore, the difference in the regression equation between patients diagnosed before and after 1 year of age was examined. ANCOVA revealed a significant difference in the association between DQ and CSF parameters at the time of diagnosis. In three patients whose CSF parameters were reevaluated after 1 year of age, the relationship with DQ and CSF parameters at reevaluation was close to the regression line of other patients diagnosed after 1 year of age. As CSF parameters change significantly in the first year of life [
], we therefore believe that CSF data acquired during the first year is not suitable for direct application to the regression equation corresponding to patients diagnosed after 1 year of age. Since CSFG and CBGR are not affected by the introduction of KD [
], these values on reevaluation are considered to be equivalent to values collected at the same age without KD introduction. As innate disease severity is often difficult to assess from patient history or neurological findings in early infancy, regression analysis using CSF parameters would be a useful alternative solution to predict disease severity. To compare the severity of patients diagnosed before 1 year of age and those diagnosed after 1 year of age using these regression lines, severity plots and CSF parameters reevaluated after 1 year of age must be compared with regression lines calculated from patients diagnosed after 1 year of age. Besides this method, which requires another collection of CSF, another possible solution to predict disease severity is the establishment of a regression equation for patients younger than 1 year of age, but requires validation using additional data.
Subsequently, we reconsidered the hypothesis that early KD introduction leads to good development. Some studies have shown that early diagnosis and KD initiation during early infancy are beneficial and contribute to normal development [
]. In the present study, however, some patients diagnosed during adulthood had average intelligence or were asymptomatic and had not received specific treatment for Glut1DS during childhood. Hence, we speculate that some patients with mild innate disease severity can achieve normal development without early dietary intervention. Moreover, some patients were diagnosed and started on a KD in their early infancy and maintained a strict KD thereafter. A few of these patients had normal development, while others did not. Proportional tendencies could be present between developmental status and CSF parameters in patients diagnosed in the first year. Accordingly, innate disease severity influences development, and we must consider innate disease severity when we speculate on the ground of normal development in patients with Glut1DS, even when they started KD in early infancy. For our three patients who started KD in early infancy, the difference between actual DQ and estimated DQ from a CSFG-based regression line at reevaluation indicates the degree of improvement achieved by early intervention (Fig. 7C). The results showed that all patients had attained better development; moreover, the regression line in the CBGR-based analysis also showed that two patients had attained better development (Fig. 7C and D). These differences in development might reflect the improvement in the early KD introduction and we should consider these differences when we assess the effect of intervention in patients with Glut1DS. This discussion relates to the effect of early KD introduction on development, which is only one aspect of neurological function. All our patients benefited from the initiation of KD in ways that could not be evaluated by the means adopted in this study, including decreased epileptic seizures, decreased fatigability, improved attention, and improved movement disorders.
Lastly, in patients whose CSFL was evaluated, a simple regression analysis was performed with respect to DQ, SARA, and CSF, and moderate associations were observed among all parameters. On the other hand, the correlation between CSFG and CSFL was moderate. Indeed, another study reported that CSFG and CSFL were not correlated [
]. Accordingly, the correlation between CSFG and CSFL is not necessarily strong in Glut1DS. This result prompted us to conduct a multiple regression analysis to investigate how the combination of CSFL and CSFG contributes to the DQ and SARA. For both DQ and SARA, the multiple regression model had much higher R2 than the simple regression models, and the regression coefficients for CSFG and CSFL were statistically significant. In multiple regression analysis, the regression coefficient for an explanatory variable is regarded as the change in the mean outcome (DQ and SARA) for the unit change of one explanatory variable when the other explanatory variables are fixed, and vice versa. In other words, even with the same CSFG value, differences in CSFL influences the outcome, and CSFL may have some roles that are independent of CSFG. However, this is simply a statistical observation and should be carefully interpreted from a biological point of view. This implies that lactate and glucose contribute to predicting neurological function through independent mechanisms. Since CSFL and CSFG were examined prior to KD introduction in our patients, the main determinants of developmental and neurological function in Glut1DS may be the innate factors that regulate CSFG and CSFL. There are two possibilities of interpretation for the association between CSFL and neurological severity. One is that lactate contributes to neurological function, and the other is that lactate elevation is caused by neurological activation.
Glucose enters the brain from blood vessels through GLUT1 and serves as a neuronal energy source. Two hypotheses describe how this energy may flow: 1) the astrocyte neuron lactate shuttle (ANLS) and 2) direct glucose diffusion through the astrocytes and interstitial spaces [
]. The ANLS hypothesis suggests that astrocytes synthesize glycogen from glucose, degrade this glycogen to lactate, and subsequently export lactate to the neurons. The lactate is utilized as neuronal fuel and as a signal molecule for several neural functions, including learning, memory, and decision making [
]. The direct glucose diffusion hypothesis proposes that glucose, which is directly imported to the neuron, is used as fuel and is required for certain aspects of memory such as contextual fear acquisition [
]. For the ANLS hypothesis, there is a counterargument in terms of energy flow in the brain: the direct glucose uptake hypothesis. However, in terms of signaling molecules, various experiments have proven the functionality of lactate [
A glycogen phosphorylase inhibitor selectively enhances local rates of glucose utilization in brain during sensory stimulation of conscious rats: implications for glycogen turnover.
]. Considering this interpretation, the lactate elevation could have only reflected neurological activation. We must keep in mind that the lactate in these MRS or histological studies consisted of intracellular or parenchymal concentration, which is different from CSFL. Since our study was observational, we could not determine whether ANLS or direct glucose uptake is the principal pathway of energy flow for neurons and whether lactate elevation in our patients is the cause or result of neurological activation [
]. Nevertheless, it would be beneficial to increase CSFL for patients with Glut1DS.
Moreover, our findings might offer novel insight into the therapeutic mechanism of the KD. KD is thought to be effective for Glut1DS because it supplies ketone bodies as energy to the brain and has antiepileptic properties [
], but ketone bodies do not metabolize to lactate; hence, they do not directly lead to lactate elevation. Instead, they inhibit pyruvate dehydrogenase, which leads to the accumulation of pyruvate and the final reduction to lactate [
]. The current study suggests that this indirect lactate elevation by the KD may contribute to KD efficacy in Glut1DS.
Furthermore, our results support an alternative or supplementary treatment strategy for Glut1DS of lactate supplementation to the brain, as previously suggested [
]. Lactate elevation has been traditionally linked to a negative implication of poor prognosis in ischemia and fatigue after exercise. However, lactate is recently emerging as a treatment for various conditions, supplying fuel for the brain in traumatic brain injury and the heart after cardiac surgery and providing volume expansion for hemorrhagic shock due to dengue fever in children [
]. The possible problems with lactate supplementation are thought to be related to the difficulty of enough lactate ingestion due to taste and the risk of acidosis, competition with ketones for transport across the blood-brain barrier [
]. In the previous studies, lactate was administered by oral intake as calcium lactate or by venous infusion as sodium lactate, and they caused only minor adverse effects, which did not lead to metabolic acidosis and did not necessitate the interruption of the studies [
]. Our experience with KD for patients with Glut1DS showed that the ketone body value can be high even under insulin detection. Since patients with Glut1DS did not have complete depletion of Glut1 activity, a small amount of glucose is delivered to the brain; thus, it is not necessary to satisfy all energy requirements of the brain by lactate. While KD is the only established treatment for Glut1DS, it does not ameliorate all symptoms [
]. Lactate can replenish pyruvate to the cycle, and its administration may be able to solve this issue. Therefore, if there is the prospect that lactate administration is safe, we would consider lactate to be given with caution to patients on a KD or as an alternative treatment option for patients who cannot adopt a KD.
This study had some limitations. First, considering its retrospective nature, the ages of clinical evaluation, CSF acquisition, and DQ and SARA evaluation varied among the patients. CSF data were acquired before KD introduction, though clinical evaluations by DQ and SARA were performed after KD introduction for practical reasons. DQ is usually clinically examined in toddlers or school-aged children. SARA can be examined in cooperative patients, and in our study population, the youngest child who could complete the SARA was 6 years old. Therefore, it would not be impossible to obtain DQ and SARA in a pre-treatment state in a certain number of patients. Although the effect of KD on intellectual functions is controversial, since we did not intend to mix pre- and post-treatment evaluations, we adopted all clinical evaluations after the KD introduction. In our experience, the KD improved not only epilepsy and cerebellar ataxia but also fatiguability and inattention. As all data were acquired on KD, these results were obtained without interruption by epileptic seizures or fatigue. This consistency is an advantage of adopting post-treatment evaluations. We were also concerned that the interval between CSF acquisition and clinical evaluation varies widely by case. Hence, we decided to stratify the simple regression analyses by the interval (Fig. 6), and we confirmed that even patients evaluated in different intervals showed the same tendency to distribution along the regression line. Second, the KSPD may be difficult to use outside Japan. However, the KSPD has a clear advantage: it can be used to evaluate development from infancy to adulthood. Moreover, its reliability has been confirmed by several studies that compared it with the third edition of the Bayley Scales of Infant and Toddler Development and the Wechsler Intelligence Scale for Children-III. [
]. In contrast, the SARA can be administered anywhere when patients cooperate. An excellent correlation was observed between the DQ and SARA; therefore, we consider that the KSPD is a reliable instrument for estimating developmental status. Third, only a limited number of patients underwent CSFL measurement. Regrettably, it is uncommon to evaluate CSFL on suspicion of Glut1DS in Japan. Therefore, further examination is necessary to confirm the hypothesis of CSFL influence on developmental and neurological status in Glut1DS. Fourth, it is uncertain whether CSFG or CSFL strictly reflect the respective concentrations of these metabolites in the microenvironment around the neurons.
5. Conclusion
CSFG and CBGR showed good associations with developmental and neurological statuses in patients with Glut1DS, with CSFG showing a superior association. However, the regression analyses revealed that CSF parameters acquired during the first year of life, i.e., obtained at early diagnosis, should not be applied to the regression equation derived in older patients. Moreover, CSFL independently contributes to the developmental and neurological status in Glut1DS through other mechanisms than those of glucose. Thus, our results suggest a novel Glut1DS treatment strategy that involves increasing the level of lactate in the brain.
Funding
This work was supported by JSPS KAKENHI [grant numbers JP15K09620 and JP19K08322], the MHLW Research Program on rare and intractable diseases [grant numbers JPMH20FC1025 and JPMH20FC1039], and AMED [grant numbers JP18ek0109280, JP21ek0109486, JP21ek0109549, JP21cm0106503, and JP21ek0109493].
Declaration of Competing Interest
Shin Nabatame has received a research grant from Morinaga Houshikai and honoraria for lecture from the Internatinal Symposium on Genetic Role of Neurometabolic Diseases with Infantile Epilepsy (ISGNIE), the 22nd Annual Meeting of the Infantile Seizure Society.
Jun Natsume is affiliated with the endowed department from Aichi prefecture (Department of Developmental Disability Medicine).
Satoshi Hattori has received consulting fee for statistical advice from Chugai Pharmaceuticals, and honoraria for the BSJ award for outstanding scientific contribution from the Biometric Society of Japan. He also participates in the data safety monitoring board of Shionogi Pharmaceuticals and the scientific advisory committee of the Radiation Effects Research Foundation.
Acknowledgments
We are grateful to the patients and their families for their participation in this study. We thank the referring physicians; Hidehito Kondo and Takahiro Fujiwara for technical guidance; supporting dietitians Naoko Nagai, Yuko Yamamichi, and Chise Yamaguchi; and our successive residents in the clinical pediatric neurology group. This research used VaProS, a data-cloud developed by the Information Core of the Platform Project for Supporting Drug Discovery and Life Science Research (Platform for Drug Discovery, Informatics, and Structural Life Science) from the Japan Agency for Medical Research and Development (AMED). We thank Editage (www.editage.com) for English language editing.
References
De Vivo D.C.
Trifiletti R.R.
Jacobson R.I.
Ronen G.M.
Behmand R.A.
Harik S.I.
Defective glucose transport across the blood-brain barrier as a cause of persistent hypoglycorrhachia, seizures, and developmental delay.
Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology.
A glycogen phosphorylase inhibitor selectively enhances local rates of glucose utilization in brain during sensory stimulation of conscious rats: implications for glycogen turnover.
It indicates primers identical to the primer used for template preparation.
00057-6&id=ga1.jpg){kind=link}
00057-6&id=gr1.jpg){kind=link}
00057-6&id=gr2.jpg){kind=link}
00057-6&id=gr3.jpg){kind=link}
00057-6&id=gr4.jpg){kind=link}
00057-6&id=gr5.jpg){kind=link}
00057-6&id=gr6.jpg){kind=link}
00057-6&id=gr7.jpg){kind=link}
00057-6&id=gr8.jpg){kind=link}
00057-6&id=gr9.jpg){kind=link}
00057-6&id=gr10.jpg){kind=link}