High Resolution 1H NMR-based Metabolomics Indicates a Neurotransmitter Cycling Deficit in Cerebral Tissue from a Mouse Model of Batten Disease*

The neuronal ceroid lipofuscinoses (NCLs) constitute a range of progressive neurological disorders primarily affecting children. Although six of the causative genes have been characterized, the underlying disease pathogenesis for this family of disorders is unknown. Using a metabolomics approach based on high resolution 1H NMR spectroscopy of the cortex, cerebellum, and remaining regions of the brain in conjunction with statistical pattern recognition, we report metabolic deficits associated with juvenile NCL in a Cln3 knock-out mouse model. Tissue from Cln3 null mutant mice aged 1–6 months was characterized by an increased glutamate concentration and a decrease in γ-amino butyric acid (GABA) concentration in aqueous extracts from the three regions of the brain. These changes are consistent with the reported altered expression of genes involved in glutamate metabolism in older mice and imply a change in neurotransmitter cycling between glutamate/glutamine and the production of GABA. Further variations in myo-inositol, creatine, and N-acetyl-aspartate were also identified. These metabolic changes were distinct from the normal aging/developmental process. Together, these changes represent the first documented pre-symptomatic symptoms of the Cln3 mouse at 1 month of age and demonstrate the versatility of 1H NMR spectroscopy as a tool for phenotyping mouse models of disease.

The neuronal ceroid lipofuscinoses (NCLs) constitute a range of progressive neurological disorders primarily affecting children. Although six of the causative genes have been characterized, the underlying disease pathogenesis for this family of disorders is unknown. Using a metabolomics approach based on high resolution 1 H NMR spectroscopy of the cortex, cerebellum, and remaining regions of the brain in conjunction with statistical pattern recognition, we report metabolic deficits associated with juvenile NCL in a Cln3 knock-out mouse model. Tissue from Cln3 null mutant mice aged 1-6 months was characterized by an increased glutamate concentration and a decrease in ␥-amino butyric acid (GABA) concentration in aqueous extracts from the three regions of the brain. These changes are consistent with the reported altered expression of genes involved in glutamate metabolism in older mice and imply a change in neurotransmitter cycling between glutamate/glutamine and the production of GABA. Further variations in myo-inositol, creatine, and N-acetyl-aspartate were also identified. These metabolic changes were distinct from the normal aging/developmental process. Together, these changes represent the first documented pre-symptomatic symptoms of the Cln3 mouse at 1 month of age and demonstrate the versatility of 1 H NMR spectroscopy as a tool for phenotyping mouse models of disease.
Neuronal ceroid lipofuscinoses (NCLs) 3 are a series of autosomal recessive diseases that collectively constitute the most common cause of childhood neurodegeneration with an incidence of 1 in 12,500 (1,2). The disorders are typified by their progressive nature with symptoms including visual disturbances, psychomotor deterioration, mental impairment, worsening seizures, blindness, and, ultimately, premature death (3)(4)(5)(6)(7). Furthermore, all share the histopathological finding of accumulation of autofluorescent lipopigment in lysosomes, similar to the pigment lipofuscin found in normal aging brains (8, 9), and ceroid, found in pathological conditions (10). The mechanism underlying this aberrant accumulation and how it correlates with neurodegeneration is unclear. Indeed, one of the paradoxes of the NCLs is that the deposition of lipopigment does not apparently lead to disease in non-neuronal cell types.
NCLs have traditionally been divided into subtypes based upon age of onset and clinical course, although the identification of several of the underlying gene defects has enabled a more definitive genetic classification. The following Cln genes are responsible for each subtype: Cln1, infantile NCL (Santavuori-Haltia); Cln2, late infantile NCL (Jansky-Bielschowsky); Cln3, juvenile NCL (Batten disease); and Cln4, adult NCL (Kufs disease). There are also four variant late infantile forms: Cln5, Finnish variant; Cln6, Costa Rican variant; Cln7, Turkish Variant; and Cln8, Turkish variant and Northern epilepsy (11,12). Even though characterization of these genes has accelerated research into the disease pathology, the precise events downstream of these mutations and how they result in the similar clinical manifestations of NCLs, albeit on different time scales, remain elusive.
The generation of genetically accurate mouse models has made this species the principal platform for investigating the NCLs, offering numerous models encompassing a variety of NCL subtypes (5,13). In this present study we have applied a metabolomics approach to define biochemical abnormalities associated with a mouse model of juvenile NCL or Batten disease (14), which is caused by mutations in the Cln3 gene. Metabolic profiles derived from 1 H NMR spectroscopic analysis of biofluids and tissue extracts, in conjunction with multivariate statistics, have previously been demonstrated to be highly discriminatory for a number of neurological diseases, including a variant late infantile NCL (Cln8) (15). Using this approach, we have identified a number of metabolic deficits associated with this mouse model of juvenile NCL. In particular, changes in the concentration of glutamate, glutamine, and ␥-aminobutyric acid (GABA) were detected that are indicative of progressive perturbations between glutamate and glutamine cycling and the conversion of glutamate to GABA and that may explain some of the neurological deficits associated with this disease. * This work was supported in part by National Institutes of Health Metabolomics Roadmap Initiative Grant R21DK070288-01 and Commission of the European Communities Grant 503051. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1

MATERIALS AND METHODS
Tissues-Wild type and Cln3 knock-out mice on the same 129S6/Sv background were used in this study (14). All procedures were carried out in accordance with National Institutes of Health guidelines and the University of Rochester Animal Care and Use Committee Guidelines. All animals were housed under identical conditions. Tissue was taken rapidly (typically within 30 s) from animals killed by cervical dislocation. Specifically, tissue was taken from a stable mouse colony consisting of mice aged 1, 2, 3, and 6 months (n ϭ 5 for all ages). The cortex, cerebellum, and remaining cerebral tissue were snap-frozen using liquid nitrogen and stored at Ϫ80°C.
Solution 1 H NMR Spectroscopy-Samples were analyzed using a 400 MHz DRX Bruker Avance spectrometer with a triple axis inverse flow probe. Spectra were acquired over 128 scans using a conventional presaturation pulse sequence for solvent suppression based on the start of the nuclear Overhauser effect spectroscopy (NOESY) pulse sequence (relaxation delay ϭ 1.3 s; t1 ϭ 3 s; mixing time ϭ 150 ms; spectral width ϭ 12 ppm; time domain ϭ 32,000 data points; solvent presaturation was applied during the relaxation delay and mixing time) at 25°C. All spectra were processed using one-dimensional NMR manager software (Advanced Chemistry Development Inc., Toronto, Canada). Spectra were multiplied by an exponential weighting function and Fouriertransformed from the time to frequency domain. Spectra were phased, baseline-corrected, and referenced to the sodium (3-trimethylsilyl)-2,2,3,3-tetradeuteriopropionate singlet at ␦0 ppm.
Pattern Recognition-Spectra were integrated between 0.2 and 9.96 ppm over a series of 0.04 ppm integral regions, using an integration macro written within the one-dimensional NMR manager software package. To account for dilution or bulk mass differences between samples, individual integral regions were normalized to the total integral region following exclusion of the water resonance. Individual integrals were thereby standardized to the total integral of all low molecular weight metabolites (16,17). Data sets were imported into the SIMCA package (Umetrics, Umeå, Sweden) and pre-processed using Pareto scaling by weighting each integral region or variable by (1/S k ) 1 ⁄2, where S k represents the standard deviation of the variable. This increased the representation of lower concentration metabolites in the resultant data models while minimizing the contribution from noise.
Data were analyzed using principal component analysis (PCA), projection to latent structures by partial least squares (PLS), and PLS-discriminant analysis (PLS-DA) within the SIMCA package (18). PCA is a technique that transforms the variables in a data set into a smaller number of new latent variables called principal components (PCs), which are uncorrelated with each other and account for decreasing proportions of the total variance of the original variables. Each new principal component is a linear combination of the original variables such that a compact description of the variation within a data set is generated. Observations are assigned scores according to the variation measured by the principal components with those having similar scores clustering together. For example, t1 is the score of a particular observation associated with prin-cipal component 1 (PC1). Where PCA proved inadequate to define a clustering, a supervised approach was used. PLS-DA is a supervised extension of PCA used to distinguish two or more classes, e.g. wild type and diseased, by searching for variables (X matrix) that are correlated to class membership (Y matrix). In this approach the axes are calculated to maximize the separation between groups and can be used to examine separation that would otherwise be across three or more principal components (18). Another supervised extension of PCA, PLS, is used to maximize the correlation between two data sets, e.g. spectral data (X matrix) and age (Y matrix), so that the response variable Y can be predicted from X (18). For each model built, the loading scores and the variable influence on projection (VIP) parameters were examined, in conjunction with the original spectra, to identify which metabolites contributed most to clusterings or a trend observed in the data. Loading scores describe the correlation between the original variables and the new component variables, whereas VIP parameters are essentially a measure of the degree to which a particular variable explains the Y variance (class membership or linear trend).
Model performance was evaluated using the R 2 and Q 2 parameters, both of which vary between 0 and 1 (18). R 2 provides an indication of how much of the variation within a data set can be explained by the various components of the model. The cumulative score, R cum 2 , records how much variation is represented by the total model. Typically, for a PCA plot the first three components can represent 60 -90% of the total variation found within a NMR-based metabolomic data set. Q 2 indicates how accurately the data, either classed or non-classed, can be predicted, and this term is more relevant to supervised pattern recognition processes. Q 2 scores over 0.08 indicate a model that is better than chance, whereas a score between 0.7-1.0 demonstrates a highly robust trend. This figure is calculated by iteratively leaving out samples from the model and predicting either class membership (e.g. presence or absence of disease) or the trend variable (e.g. age).

Increases in Glutamate and Decreases in GABA Characterize
Cerebral Tissue from Cln3 Mice Aged 1 Month-Solution state 1 H NMR spectral profiles from the cortex, the cerebellum, and the remaining brain tissue of wild type and Cln3 mice were examined for metabolic differences (Fig. 1). NMR profiles from the tissue of mice aged 1 month were readily separated by PLS-DA in all three brain regions (data not shown). Analysis of the loading plots revealed similar metabolite changes to be responsible for the clustering across all regions. In particular, concentrations of glutamate, myo-inositol, and creatine were increased in Cln3 tissue relative to wild type, whereas GABA concentrations were consistently decreased. Furthermore, metabolite perturbations including increases in glutamine, choline, N-acetyl-aspartate (NAA), and taurine, were also apparent but did not feature in all brain regions (TABLE ONE).
To investigate common metabolic changes across the whole brain regardless of metabolic differences between different regions, spectral profiles from all three brain regions were simultaneously analyzed by PLS-DA. This supervised pattern recognition process was necessary in order to minimize the variation associated with the different brain regions while maximizing the variation associated with the mouse model mutant/wild type status in the model. Analysis successfully categorized the profiles, with the corresponding loadings plot indicating that the same discriminatory metabolites identified in individual regions are responsible for clustering (Fig. 2).
Temporal Profile of Metabolic Changes Associated with Cln3-To evaluate any temporal changes associated with the Cln3 disease pro-gression, spectral profiles derived from mice aged 2, 3, and 6 months were also analyzed. Wild type and Cln3 spectra were successfully categorized for all age groups, although for 2-and 6-month-old mice separation was only evident in two of the three brain regions (TABLE ONE). The failure to find separation in all regions may result from the number of samples analyzed being too small to identify subtle changes or from the confounding effects of aging/development (see below) masking changes associated with the disease. From the accompanying loading plots, changes in glutamate and GABA agreed for all time points exam-ined with those observed at 1 month of age. Specifically, glutamate was increased in Cln3 tissue, whereas the concentration of GABA was decreased in all age groups, one exception being an increase in GABA in Cln3 cortex tissue from 6-month-old mice. However, at all time points and in all brain regions the ratio of glutamate/GABA was increased in Cln3 mice, even in regions where the pattern recognition analysis did not detect separation. Likewise, increases in myo-inositol and creatine, apparent in 1-month-old Cln3 mice, were also observed in mice aged 2, 3, and 6 months, with two exceptions at 3 months old. However, analysis FIGURE 1. High resolution 400 MHz 1 H NMR spectrum of an aqueous extract of cortex tissue taken from a Cln3 mouse aged 1 month. Each resonance corresponds to a chemical moiety within a particular metabolite, with the intensity being proportional to the concentration of that metabolite. Metabolic profiles were developed from the spectra and analyzed by pattern recognition techniques to identify differences between wild type and Cln3 mice. Peak 1, valine, leucine, and isoleucine; peak 2, lactate; peak 3, alanine; peak 4, GABA; peak 5, acetate; peak 6, NAA; peak 7, N-acetyl-glutamate (NAG); peak 8, glutamate and glutamine; peak 9, glutamate; peak 10, succinate; peak 11, citrate; peak 12, aspartate; peak 13, creatine; peak 14, choline; peak 15, phosphocholine; peak 16, taurine; and peak 17, myo-inositol.

Summary of metabolic deficits detected between wild type and Cln3 cerebral tissue
Changes are listed for each of the three brain regions analyzed at various ages. "No separation" indicates where a model could not be computed to distinguish tissue according to disease state.
of the loadings revealed that whereas several of the causative metabolic deficits were consistent with those identified in 1-month-old mice, other deficits followed an opposite trend (TABLE ONE). For example, glutamine concentrations in Cln3 mice were generally decreased in mice after 1 month, such that there was an overall decrease in the glutamine/glutamate ratio. Moreover, NAA was found to be increased in both 1-and 2-month-old Cln3 mice but decreased in Cln3 mice aged 3 and 6 months. No discernible trend in choline and taurine changes was identified.
Wild Type and Cln3 Mice "Metabolically Age" at the Same Rate between 1 and 6 Months Such That the Disease Process Is Not a Result of Premature Aging-Having analyzed the metabolic changes in the Cln3 mouse of particular age groups, a subsequent PLS analysis was conducted to consider the temporal metabolic variation associated with FIGURE 2. Pattern recognition model of metabolic changes identified across the whole brain between wild type and Cln3 mice aged 1 month. A, PLS-DA score plot distinguishing Cln3 (OE) and wild type (E) metabolic profiles derived from all brain regions (R 2 ϭ 36. 0%; Q 2 ϭ 41.7%). Cortex, cerebellum and remaining tissue profiles are represented by cx, cb, and ot, respectively. B, accompanying PLS-DA loading plot detailing the metabolites that contribute the most to this separation. Metabolites located toward the left are increased, and those toward the right are decreased in spectra of Cln3 mice relative to wild type mice. Major changes as indicated by the variable influence on projection (VIP) scores (metabolites that lie outside the 95% limit of these scores) are listed in the adjoining table along with the ppm values corresponding to the upper limit of the associated integrated bucket regions. DECEMBER 30, 2005 • VOLUME 280 • NUMBER 52 development/aging. Considering both wild type and Cln3 observations together produced a highly robust model that correlated 1 H NMR spectral profiles with age (Fig. 3A), indicating a metabolic aging process common to both wild type and Cln3 mice. Moreover, when further pattern recognition models using PLS were constructed whereby wild type and Cln3 NMR profiles were analyzed individually (data not shown), similar metabolites varied with age in all models. In particular, glutamate, glutamine, myo-inositol, creatine, and NAA concentrations increased with age, whereas lactate and acetate concentrations decreased with age. Because many neurological diseases are associated with an accelerated aging phenomenon, the relative rates of this apparent aging trend between wild type and Cln3 mice were compared. PLS models computed using either wild type or Cln3 spectral data were used to predict the ages of both strains of mice. If an accelerated aging phenomenon was present, the wild type model would be expected to predict Cln3 mice to be older than they were, whereas the Cln3 model would predict wild type mice to be younger than they were. In both models, there was no difference between the age prediction of wild type and Cln3 mice (Fig. 3, B and C). This result signifies that the metabolic processes that characterize the Cln3 mutant are distinct from those associated with normal aging.

DISCUSSION
In this study, solution 1 H NMR has been used to examine metabolic profiles from brain tissue of a Cln3 knock-out mouse model. Profound metabolic changes have been detected across the three brain regions investigated, providing the first evidence of metabolic changes at 1 month of age in the Cln3 mouse model, prior to neuronal cell loss detected from 7 months of age (14).
The NMR based metabolomic approach used here to phenotype brain tissue from the Cln3 mouse model was both relatively rapid, requiring ϳ7 min in full automation, and robust in terms of the reproducibility of the results. The analysis is also cheap on a per sample basis, making it ideal for screening of large populations. One criticism is that it is difficult to draw conclusions about dynamic processes such as flux rates through metabolic pathways from static observations. This is not limited to metabolomic data, but is also a problem of transcriptional and proteomic analyses where a snapshot is obtained of a cell, tissue, or organism. However, direct measurements of rates such as the glutamate/glutamine cycling rate in the intact brain by 13 C NMR spectroscopy are time consuming, costly, and pose considerable technical challenges. Furthermore, metabolic control analysis suggests that the measurement of metabolite concentration changes will often be more sensitive to metabolic perturbations than flux rate measurements.
An important factor of this analysis is that the multivariate statistical approach considers all of the metabolic changes simultaneously. Furthermore, Pareto scaling of the data allows both high and low concentration metabolites to make a significant contribution to the pattern recognition models. Despite the absolute variations in the concentrations of glutamate, glutamine, and GABA being very different, the multivariate approach was able to consider relative changes of metabolites with respect to one another. It is these perturbations that drive the separation detected in the models rather than absolute changes in concentration. A simple univariate analysis of metabolite concentrations is often less sensitive to metabolic changes.
In the present study, concentrations of glutamate were consistently increased across the three regions examined, whereas GABA concentrations were decreased in the brains of Cln3 mice relative to wild type. In addition, glutamine concentrations were also decreased from 2 months of age in Cln3 mice such that the glutamine to glutamate ratio was reduced. The increase in glutamate concentration relative to both GABA and glutamine is consistent with and extends the temporal scope of a previous investigation, which revealed a similar increase in glutamate concentration at 3 months of age (19). Such an increase was proposed to result from autoantibodies to glutamic acid decarboxylase (GAD65), which have been detected in serum from Cln3 knock-out mice and inhibit the enzyme's conversion of glutamate to GABA. These . PLS analysis of temporal metabolic variation associated with development/aging of wild type and Cln3 mice. A, correlation between mean PLS component (t1) score (E), representing spectral data from both wild type and Cln3 mice, and the age of mice. t1 scores were derived from a PLS model whereby spectral data (X matrix) were regressed against age (Y matrix) (R 2 ϭ 68.6% and Q 2 ϭ 66.3 for first component). For each age group, the mean t1 score Ϯ S.E. was plotted. B and C, similar PLS models were computed using either only the wild type (R 2 ϭ 60.6% and Q 2 ϭ 58.6% for first component) (B) or Cln3 (R 2 ϭ 79.3% and Q 2 ϭ 74.7% for first component) (C) spectral data (data not shown). The ages of both wild type (E) and Cln3 (OE) mice were then predicted using either the wild type data PLS model (B) or the Cln3 data PLS model (C). For each model, actual age was plotted against the mean predicted age Ϯ 2 ϫ S.E. autoantibodies can traverse the blood-brain barrier and have been detected in the cerebrospinal fluid of Cln3 mice. 4 Although the decrease in GABA concentration appears reasonable following GAD65 inhibition, in GAD65 knock-out mice, which are completely devoid of GAD65 activity, no decrease in GABA concentration is observed, strongly suggesting some kind of compensatory mechanism, most likely by the second GAD isoform, GAD67 (20). Another explanation for the decrease in GABA is the well documented loss of GABAergic interneurons, which are particularly vulnerable in Batten disease. However, the first signs of neuronal cell loss do not occur until 7 months of age (14). Thus, the decrease in GABA concentration observed here results either from an inhibition of GAD65 without compensation by GAD67 and/or from the early stages of metabolic impairment within GABAergic neuronal cell loss. Furthermore, the decrease in the ratio of glutamine to glutamate in Cln3 mice after 2 months indicates that there are no compensatory mechanisms for the increase in the excitatory glutamate produced as a result of the decrease in GABA production. Taken together, these findings are suggestive of a profound perturbation in neurotransmitter recycling/production in terms of the changes in the concentrations of glutamate, glutamine, and GABA and the balance between excitatory and inhibitory neurotransmitters.
The metabolism of glutamate, glutamine and GABA are linked via two pathways between neurons and glial cells (21) (Fig. 4). In general, the neurotransmitters glutamate and GABA are cleared from the synaptic cleft largely by uptake into surrounding glial cells. Within glial cells, glutamate is converted to glutamine by the glia-specific enzyme, glutamine synthetase (22). Glutamine is then released from glial cells to be transported into nerve cell terminals, where it is converted back to glutamate by glutaminase to replenish neurotransmitter pools. Our results indicate an imbalance in the excitatory and inhibitory neurotransmitters glutamate and GABA, respectively, in addition to a perturbation in the glutamate to glutamine ratio, which is strongly suggestive of a deficiency in neurotransmitter cycling. Consistent with this, the expression of glutamate dehydrogenase, glutamin-ase, and glutamine synthetase, key enzymes in the glutamate-glutamine cycle, have been reported to be markedly altered in the Cln3 mouse (19,23,24). Moreover, impaired cycling has been observed/suggested in several neurological disorders including epilepsy and Huntington disease (25)(26)(27). Thus, dysfunction of the glutamate-glutamine cycle in the brain may have a significant impact on the underlying NCL pathophysiology in the Cln3 mouse.
myo-Inositol and creatine, both of which were increased in Cln3 tissue relative to controls, also provided a unique insight into the disease process. myo-Inositol is a glial cell marker that has previously been implicated in the osmoregulation of astrocytes (28,29). The activation of microglia and astrocytes (gliosis) is a well known pathological response readily observed in neurodegenerative diseases (30 -32). During gliosis, astrocytes undergo hypertrophy of astrocytic processes. In contrast, microglia undergo a graded response that proceeds via a distinct series of steps that vary in their morphological and molecular profile. These steps include the transformation from resting ramified cells to phagocytotic brain macrophages and are tightly coupled to the underlying disease pathology (33). Although severe astrocytic hypertrophy and microglial transformation have not been detected in older Cln3 knock-out mice, a low level glial activation has previously been reported at 5 months of age (34). Therefore, the increase in myo-inositol detected in the present study from 1 month of age most likely reflects an early stage glial activation or at least up-regulation of metabolic processes within glial cells. Creatine, on the other hand, is thought to have a multifaceted role in the brain. Apart from being involved in energy homeostasis, it has recently been implicated in brain osmoregulation (35,36). An increase may therefore reflect increased metabolic activity coupled to a higher level of excitatory neurotransmission in Cln3 tissue, of which similar increases have been detected in frontal lobe epilepsy (37) or, indeed, gliosis accompanied by a high metabolic activity of the glial cells (38,39). Alternatively, an increase in creatine may be suggestive of a redistribution of cerebral osmolytes (39). Indeed, during neuronal cell death the decrease in ATP production impedes the normal transmembrane homeostasis of Na ϩ and K ϩ ions such that there are  Step 1, glutamate (glu) is synthesized from glutamine (gln) by the enzyme glutaminase and further decarboxylated to produce GABA by glutamic acid decarboxylase.
Step 2, glutamate and GABA are released into the synapse in GABAergic/glutamatergic neurons in response to nerve depolarization and cleared by uptake into surrounding glial cells.
Step 3, GABA is metabolized by the GABA shunt and the tricarboxylic acid (TCA) cycle.
Step 4, glutamate is either metabolized in the TCA cycle or converted to glutamine by glutamine synthetase.
Step 5, glutamine is synthesized and returns to neuronal cells where it is converted back to glutamate.
Step 6, perturbation of glutamate-glutamine cycling creates an imbalance of excitatory and inhibitory neurotransmitters leading to excitotoxicity.
Step 7, prolonged excitotoxicity, in turn, leads to reactive gliosis and, step 8, neuronal cell loss accompanied by changes in osmotic state.
large changes in the osmotic state. Moreover, cationic imbalances have been documented in a yeast model of Batten disease whereby a defect in the transport of the basic amino acids lysine and arginine into the vacuole has been identified (40). Such a transport defect at the lysosomal membrane in neuronal cells may also warrant compensatory changes in osmolyte distribution (41).
Perturbations in NAA concentration also discriminated between wild type and Cln3 cerebral tissue, showing an increase in 1-month-old mice and decreasing in 2-, 3-, and 6-month-old mice. NAA is synthesized in the mitochondria and has been implicated in lipid synthesis (42) and acetyl-CoA buffering (43). The increase at 1 month may therefore reflect an up-regulation of lipid synthesis resulting from metabolic changes associated with lipid storage and degradation, known to accompany the NCLs. Alternatively, we have shown previously the degradation of NAA to be slower in a mouse model of Cln8 (15). In addition, NAA is commonly thought of as a biomarker for neuronal density (44) and, as such, decreases have previously been linked to neuronal cell loss in numerous experimental and clinical conditions (45)(46)(47). Moreover, recent research has also suggested a role in cellular osmoregulation (35). In mice aged 2, 3, and 6 months, when glial cell activation is more advanced it is likely that the early stages of neuronal cell loss and neuronal mitochondrial dysfunction will be detected.
Because many of the metabolic changes detected between wild type and Cln3 mice also featured in terms of the metabolic changes detected as part of a common aging trend between the two mice, it was tempting to speculate an accelerated aging phenomenon; for example, oxidative stress may constitute part of the disease pathogenesis. Indeed, this has been shown to be the case in certain neurological diseases including Alzheimer and Huntington disease, as well as in animal models of accelerated aging (48). The effect has also recently been suggested in Cln5-deficient mice (49). However, the decrease in GABA concentrations did not feature in the aging trend, and there was no significant difference found between wild type and Cln3 age predictions during PLS analyses. Thus, we believe the disease process in young Cln3 knock-out mice is distinct from normal aging. To verify conclusively that accelerated aging plays no part in the disease warrants a similar analysis conducted using older mice.
Using a combined approach of 1 H NMR-based metabolomics and multivariate statistics, we have identified a profound perturbation in glutamate, glutamine, and GABA concentrations in brain tissue in a mouse model of Batten disease from 1 month of age. In addition, metabolite changes were identified that were consistent with the early stages of glial cell activation. Whether these changes really reflect glial activation, however, or some other metabolic shift will still need to be determined to properly define the order of events during the disease pathogenesis. Nonetheless, these results represent the earliest pathological hallmark in the Cln3 mouse model, suggesting that the pathological trigger of the disease is occurring very early in development in these mice. That such a 1 H NMR based metabolic profiling approach has detected these deficits suggests it will be a useful technique in comparing and contrasting all the mouse models of the NCL disorders in order to better define the pathology of this class of diseases.