Introduction

The neuronal ceroid lipofuscinoses (NCLs) are a group of progressive, pediatric neurodegenerative lysosomal storage disorders (LSDs). Among these, CLN3 disease, historically known as Batten disease, is the most common and arises from biallelic pathogenic variants in the CLN3 gene. CLN3 encodes a transmembrane protein that is expressed at low levels throughout the central nervous system (CNS). Recent studies have shown that loss of CLN3 expression results in accumulation of glycerophosphodiesters (GPDs) in lysosomes1,2,3. However, the precise function of CLN3 protein remains unclear. Deficient cells have secondary alterations in a number of crucial cellular processes such as endocytosis, intracellular trafficking and autophagy4. On histopathology, deficiency in CLN3 protein leads to accumulation of lipofuscin in lysosomes, a hallmark of NCL disorders.

Clinically, CLN3 disease typically begins with rapid-onset blindness5 and progresses with dementia6,7, seizures8, and movement disorders, often resulting in mortality in early adulthood9,10. Unfortunately, there are no approved disease-modifying therapies for CLN3 disease. While enzyme replacement11 and gene therapy approaches have progressed for other LSDs, effective treatments for neurological symptoms of CLN3 disease remain elusive9,10. Gene replacement using adeno-associated viral (AAV) vectors has been piloted in animal models12,13,14 and human patients (NCT03770572). Full results from this first-in-human study targeting early-stage patients have not been reported.

In general, trials targeting central nervous system (CNS) symptoms in LSDs have failed to translate into meaningful neurocognitive improvements if the therapy is given after symptom onset. For example, cerliponase alfa enzyme replacement therapy has demonstrated success in slowing CLN2 disease progression but does not reverse neurological symptoms present before treatment initiation17. This underscores the complexity of correcting established circuit-level functional deficits in the CNS.

Several mouse models of CLN3 disease have been created15,16,17,18. While these models replicate the accumulation of storage materials and neuroinflammation observed in patients, they typically exhibit normal survival rates. In addition, their behavioral traits are subtle and context-dependent19 and influenced by genetic backgrounds20. To address this limitation, robust and reproducible phenotypes has been identified through network-level electrophysiology studies in two CLN3 disease mouse models: a knockout model and a model with a recurrent 1.02kB human deletion mutation21,22. Specifically, abnormalities have been identified in the hippocampus of CLN3 disease mice as early as young adulthood17,21,23,24,25,26. This is a region important for learning and memory and is a rare location of adult neurogenesis, the birth of new neurons from progenitors. Unlike histopathological assessments, physiological measurements directly assess function and thus may provide a more effective indicator for the development of therapies.

Previously, we found that early gene replacement therapy22, given at p0, corrects network dynamics in the hippocampus of CLN3 disease mice. However, we hypothesized that hippocampal circuit dysfunction may arise very early in post-natal development, which could limit the therapeutic window when gene replacement may work. Here, we aimed to define the trajectory of abnormal hippocampal development in early postnatal life in CLN3 disease mice. Surprisingly, we found profound hypoexcitability of the hippocampus as early as p7. This work highlights that CLN3 disease is a neurodevelopmental disorder rather than a purely neurodegenerative condition.

Materials and methods

Mice

Cln3KO/KO (i.e., Cln3–/–), Cln3KO/WT(i.e., Cln3–/+), and Cln3WT/WT (i.e., Cln3+/+), mice were bred, housed, maintained, and genotyped in our laboratory as previously described17,21,22. The mice were a gift from the laboratory of Dr. Beverly Davidson at CHOP. All mice were maintained on a C57BL/J6 background. Male and female mice were used for all experiments.

Slice preparation for VSDI

400-µm thick hippocampal-entorhinal cortical slices, which preserve the hippocampal trisynaptic circuit21,22, were prepared as our standard protocols27 for VSDI studies. Slices were cut in a chilled high-sucrose cutting solution containing, in mM, 192 sucrose, 2.5 KCl, 1.25 NaH2PO4, 26 NaHCO3, 12.2 glucose, 3 sodium pyruvate, 5 sodium ascorbate, 2 thiourea, 10 MgSO4, and 0.5 CaCl2 chilled to become slurry. Slices were allowed to recover for 45 min at 37 °C and 45 min at room temperature prior to experiments. During recovery the slices were maintained in an artificial cerebrospinal fluid (ACSF) containing, in mM, 115 NaCl, 2.5 KCl, 1.4 NaH2PO4, 24 NaHCO3, 12.5 glucose, 3 sodium pyruvate, 5 sodium ascorbate, 2 thiourea, 1 MgSO4, and 2.5 CaCl2. For recording, slices were maintained in a standard ACSF solution containing, in mM, 128 NaCl, 2.5 KCl, 1.4 NaH2PO4, 26.2 NaHCO3, 12.2 glucose, 1 MgSO4, and 2.5 CaCl2.

VSDI acquisition

The VSD DI-3-ANEPPDHQ (Potentiometric Probes) was solubilized in 95% ethanol (0.020 mg/µL) and stored at − 20 °C. Slices were stained with the di-3-ANEPPDHQ (0.1 mg/mL, in aCSF) for 14–16 min. Slices were then transferred to a humidified interface chamber (BSC2, Scientific Systems Design), maintained at room temperature for recording.

VSDI experiments were completed as previously published21,22,27. In short, excitation light was provided by (LEX3-G Brainvision SciMedia) high-power LED with center wavelength of 530 nm. Fluorescence was recorded at 1000 frames per second with a fast video camera with 256 × 256 pixel resolution (MiCAM03-N256 SciMedia).

For each condition the perforant pathway (PP) and Schaffer collateral (SC) pathways were stimulated. Four stimuli were delivered as 0.1-ms pulses at 10 Hz through a concentric tungsten electrode (World Precision Instruments). All recordings were 1 s long, with a shutter delay of 500 miliseconds between recordings to allow fluorescence to recover from photobleaching (BV Workbench from Brainvision). Ten recordings of evoked activity were interleaved with 10 VSDI runs where no stimulus was delivered in order to normalize for background fluorescence.

VSDI analysis

Post-recording VSDI analysis was completed using the previously published VSDI toolbox27. The software facilitates inspection of VSDI data and allows for robust statistical analysis across both spatial and temporal dimensions. Unlike traditional VSDI protocols, which analyze only a small number of user-selected ROIs, this toolbox allows for an unbiased evaluation of the dynamics of the entire hippocampus.

In short, a computational algorithm is used to create an unbiased ROI covering the entire hippocampus for each slice. The fluorescence response in each ROI for each image is calculated. As previously published21,22,27, to compare between and average slices, linear interpolation was used to “stretch” rasters to be the same size. For all studies, a final stretched DG raster size was 44 ROIs, and a final stretched CA proper raster size was 70 ROIs.

Recordings from each condition per slice were averaged. A 2-dimensional raster plot showing the location, time, and fluorescence change for each recording protocol in each slice was generated. In addition to individual rasters for each slice, an average raster plot per condition was generated. The raster plots from different genotypes were compared using a permutation sampling method with n = 1000 permutations. A heatmap showing regions of the network with statistically significant (P < 0.05) differences was created. The average regional fluorescence response (DG, hilus, CA3, CA1) was calculated for each slice.

EEG acquisition

Recording electrodes were constructed by connecting 8 cortical leads, 1 reference lead placed over the cerebellum, 1 ground (0.004 inches, formvar-coated silver wire, California Fine Wire), and 2 hippocampal leads (0.005 inches bare, 0.008 inches coated, stainless steel wire, AM-Systems) to a micro connector (Omnetics). Electrodes were placed while mice were under inhaled isoflurane anesthesia after premedication and given 5 mg/kg meloxicam for pain management. Electrodes were implanted using the following stereotaxic coordinates (measurements relative to bregma): bilateral motor cortices: 0.5 mm rostral, 1 mm lateral, and 0.6 mm deep; bilateral barrel field cortices: 0.7 mm caudal, 3 mm lateral, and 0.6 mm deep; bilateral visual cortex: 3.5 mm caudal, 2 mm lateral, and 0.6 mm deep; bilateral auditory cortex: 2.7 mm caudal, 4 mm lateral, and 0.6 mm deep; and bilateral hippocampus (CA1 region): 2.2 mm caudal, 2 mm lateral, and 1.7 mm deep. The cerebellar reference lead was implanted posterior to lambda. The ground wire was wrapped tightly around a 1/8-inch self-tap screw (Precision Screws and Parts), which was inserted into the skull rostral to the motor cortex leads. A second screw was inserted caudal to lambda to help secure the recording headcap. The recording electrodes were secured with dental cement and supporting superglue, and the animal was allowed to recover for at least 24 hours before recording.

Data were recorded using the Intan RHS2000 recording system (Intan Technologies). Electrophysiologic data was acquired at 2,500,000 Hz and then down sampled to 2,500 Hz using the Intan RHD2000 Recording System Software (http://intantech.com/downloads.html). Mice were recorded for at least 24 consecutive hours per animal. At least 6 animals per genotype group were recorded at p30-40.

EEG analysis

All analysis was completed in MATLAB (https://www.mathworks.com/products/matlab.html, Version2020a, ) with software designed by members of the laboratory21. Code is available upon request.

Data preprocessing

Data were notch filtered at 60 Hz to remove line frequency and filtered with a 6th order bandpass Butterworth filter. Detection of poor recording channels was completed as previously described21,22,28. In short, detection of a bad or noisy recording channel was completed by calculating the average of the root mean squared amplitude and skew of the voltage for each second of the recording. Channels were excluded from further analysis if they had a root mean squared amplitude less than 30 µV or more than 200 µV or a skew greater than 0.4. Data from each recording period were then divided into 5s epochs. Epochs with amplitude z-scores of greater than 3 were considered artifactual and were excluded from analysis, except for quantification of spikes.

Spike detection

Spikes were quantified as previously described21,22,28 Spikes were defined as voltage deflections that were greater than 5 standard deviations above the mean with a full width at half maximum amplitude of 5–200 ms. Spikes were eliminated if they occurred within 10 s of a deflection found to be an artifact.

Power analysis

To analyze mutation-induced differences in background EEG frequency composition, fast Fourier transform analysis was then completed on artifact-free epochs as previously described21,22,28. Quantification of the power in each of the major EEG frequency bands (delta 0.1–4.0 Hz, theta 4–8 Hz, alpha 8–13 Hz, beta 13–25 Hz, and gamma 25–50 Hz) was completed. Power in each band was normalized to total power. Power in each band across all epochs was averaged.

Isolation of mouse brain regions

A sex-balanced cohort of P11 mice were fasted for 3–4 h prior to tissue harvest. Mice were anesthetized with 5% isoflurane. The cortex, hippocampus, and cerebellum were dissected on ice. During extraction tissue was repeatedly washed with 1X PBS to prevent freeze burning and minimize blood contamination of the tissue. Tissue was then flash frozen in liquid nitrogen and stored at − 80 C for further processing.

Metabolite extraction from mouse brain

Regions of frozen brain were lyophilized overnight in a Labconoco FreeZOne Vaccume and weighed on Mettler Toledo XPR105 Analytical Balance (0.01 mg readability). Lyophilized tissue was then powdered using Precellys Lysing Kit CK28-R homogenizing tubes in the Bertin Precellys Evolution machine set to 7200 RPM, 3 cycles of 20s with 15s pauses with ‘Crolys: ON’ at 4°C. The amount of extraction solvent used was calculated based on the dry weight of the tissue for each individual sample. Extraction solvent was pre-prepared in a master mix for all samples with 40:40:20 Methanol: Acetonitrile: H20 (HPLC Grade) with 0.5% Formic Acid (ACS grade) with isotope labeled internal standards L-Leucine U-13C6 98% 15 N 98% (Cambridge Labs, CNLM-281-0.05) and L-Tryptophan U-13C11 98% (Cambridge Labs, CNLM-2475-0.1) each with final concentration of approximately 2ug/mL. For 5.00 mg of lyophilized tissue, 1 mL extraction solvent was added to the tissue and samples were homogenized on the Precellys Evolution machine at 7200 RPM for 6 cycles of 20s with 15s pauses with ‘Cryolys: ON’ at 4°C. Samples were incubated on ice for 10 min. For 5.00 mg lyophilized tissue, 84 uL 15% (w/v) NH4HCO3 was added to sample, vortexed, and incubated on ice for 20 min. Samples were centrifuged at ~ 17,000 g for 30 min at 4°C. The supernatant was transferred to a new 2mL tube and centrifuged for 30 min at 4 °C. The remaining pellets from both spin downs were used for downstream lipid extractions for future investigations.

Glycerophosphodiester analysis

Metabolites were measured via LC/MS analysis using a Vanquish Horizon UHPLC System coupled with an Orbitrap Q Exactive Plus Mass Spectrometer (Thermo Scientific). Hydrophilic interaction chromatography (HILIC) with a Waters XBridge BEH Amide XP Column (particle size, 2.5 μm; 150 mm (length) × 2.1 mm (i.d.)) was used for metabolite separation29. Column temperature was 25 °C. Autosampler temperature was 4 °C. Mobile phase A is 20 mM ammonium acetate and 22.5 mM ammonium hydroxide in 95:5 (v/v) water: acetonitrile (pH 9.45) and B is 100% acetonitrile. These were used for both positive and negative modes. The gradient is 90% B (0.0–2.0 min), 75% B (3.0–7.0 min), 70% B (8.0–9.0 min), 50% B (10.0–12.0 min), 25% B (13.0–14.0 min), 0.0% B (16.0–20.5 min), 90% B (21.0–25.0 min). Total runtime was 25 min. Flow rate was 0.15 mL/min. Sample injection volume was 5.0 uL. ESI source parameters for both negative and positive scans: sheath gas flow rate, 28 arb (arbitrary units); aux gas flow rate, 10 arb; sweep gas flow rate, 0 arb, spray voltage, 3.50 kV, capillary temperature, 320 °C, S-lens RF level, 65.0. Parameters for positive full scan: resolution, 70,000; AGC target, 3e6; maximum injection time, 200 ms and scan range, 118.5 to 950 m/z. Parameters for negative full scan: resolution, 70,000; AGC target, 3e6; maximum injection time, 200 ms and scan range, 70 to 950 m/z.

MS/MS spectra were obtained using a PRM scan with the following parameters: runtime, 10 to 14 min; polarity, negative; inclusion, on; resolution, 17,500; AGC target, 5e5; maximum injection time, 100 ms; isolation window, 4.0 m/z; (N)CE/ stepped (N)CE, nce: 35, 45, 55.

LC/MS raw files (.raw) were converted to mzXML format using ProteoWizard (https://proteowizard.sourceforge.io, version 3.0.20315)30. El-MAVEN (https://github.com/ElucidataInc/ElMaven, version 0.7.0 or 0.12.0)2 was used to generate a peak table containing m/z, retention time, and intensity for peaks. Isotope labeled internal standard peak intensities were corrected for natural 13 C and 15 N abundances using AccuCor23 package31,32.

LC/MS/MS raw (.raw) files were analyzed using Thermo XCalibur Qual Browser (https://www.thermofisher.com/us/en/home/industrial/mass-spectrometry/liquid-chromatography-mass-spectrometry-lc-ms/lc-ms-software/lc-ms-data-acquisition-software/xcalibur-data-acquisition-interpretation-software.html, version 4.1.31.9). Identity of glycerophosphodiester Species (GPDs) verified by comparing with spectra from Laqtom et al.1 Relative abundance of GPDs was calculated by normalizing respective peak intensity to the isotope labeled L-Leucine spike-in. Peak quality was assessed manually by examining shape, signal intensity (E5 or higher), RT compared to known metabolites, and the background signal.

Statistics

All statistical analysis of VSDI rasters was completed using the VSDI toolbox27 (code available at https://doi.org/10.1371/journal.pone.0108686). For summary statistics of regions or groups, GraphPad Prism (https://www.graphpad.com, version 10) was also used. In graphs, data are expressed as mean ± SEM. Statistical analysis of LC/MS data 2-Way ANOVA with Tukey’s multiple comparisons was performed between all genotypes within a brain region and between the same genotypes between different brain regions (p values; *< 0.05, ** <0.01, *** < 0.0005, **** <0.0001).

Vertebrate animals approval

All animal studies were approved by the Institutional Animal Care and Use Committee at the Children’s Hospital of Philadelphia. All experiments were performed in accordance with relevant guidelines and regulations and reported in accordance with the ARRIVE guidelines.

Results

Glycerophosphoinositol accumulates very early in the brain of CLN3 disease mice

To evaluate for very early accumulation of disease relevant GPDs, we performed targeted HILIC LC/MS on brain tissue homogenates from P11 mice. All five GPD species were detected in negative and positive scanning mode. GPI, GPG, GPE, GPS were reported from negative scanning mode as done in Laqtom et al., except for GPC which was more abundant in positive scanning mode. All five GPDs had moderately good to good quality peaks across the samples. We found that glycerophosphoinositols (GPIs) were significantly increased in tissue homogenates from all p11 homozygous Cln3–/– brain regions as compared to wildtype Cln3+/+ or heterozygous Cln3+/– mice (Fig. 1a). There were no significant differences between disease and control mice in any the brain regions for any of the other GPDs including glycerophosphoglycerol (GPG, Fig. 1b), glycerophophoethanolamine (GPE, Fig. 1c), glycerophosphocholine (GPC, Fig. 1d) or glycerophosphoserine (GPS) (Fig. 1e). Of note, both GPG and GPE species demonstrated a trend towards higher abundance in the cerebellum as compared to cortex. These species also had a higher total abundance than GPS species.

Fig. 1
figure 1

CLN3 disease mice accumulate GPIs, but not other GPDs, by p11 throughout the brain. HILIC method with full scan in negative mode was performed on individual brain regions from Cln3−/− (blue), Cln3−/+ (gray), and Cln3−/+ (black) mice. (A) Gycerophosphoinositol (GPI) (B) glycerophosphoglycerol (GPG), (C) Glycerophosphocholine (GPC), (D) glycerophosphoethanolamine (GPE), and Glycerophosphoserine (GPS). Data are mean ± s.e.m. (n = 6 mice per genotype; 3 female, 3 male). 2-Way ANOVA with multiple comparisons was performed between all genotypes within a brain region and between the same genotypes between different brain regions (p values; *< 0.05, ** <0.01, *** < 0.0005, **** <0.0001).

Full size image

CLN3 disease mice demonstrate abnormal early postnatal hippocampal circuit development

Our previous studies have shown that CLN3 disease mouse models exhibit pronounced and progressive defects in neuronal network dynamics. Specifically, we found that in CLN3 disease mice, the dentate gyrus of the hippocampus demonstrates hypoexcitability, as detected through voltage sensitive dye imaging, by two months of age21. This was surprising, as this is before significant storage accumulation or cell death occurs in this model. This led us to hypothesize that early hippocampal development is altered in CLN3 disease.

To evaluate how CLN3 disease affects early postnatal hippocampal development, we completed voltage sensitive dye imaging in hippocampal slices obtained from Cln3−/− and Cln3+/+ mice at p7, p14, and p21. Fluorescence alterations across the entire hippocampus were quantified using the VSDI statistical toolbox previously reported by our group27. First, we evaluated the response to direct perforant path (PP) stimulation. The PP provides the major input into the hippocampus from the entorhinal cortex (Fig. 2a, stimulation site noted by asterix). By p7, there is striking hypoexcitability of the Cln3- /- dentate gyrus (DG) in response to PP stimulation as compared to Cln3+/+ hippocampal slices (Fig. 2b). This DG hypoexcitability was associated with decreased excitation of the rest of the hippocampus including the hilus, CA3 and CA1 regions. At p14 (Fig. 2c) and p21 (Fig. 2d) the wildtype DG matured to effectively serve as a gate to limit the spread excitation from the DG to the CA proper. In Cln3−/− mice, excitability of the DG improves by p14 (as compared to p7 Cln3−/− mice); however, the CA proper remains hypoexcitable as compared to Cln3+/+ mice. Relative hypoexcitablity worsened by p21, mirroring our previous findings in older Cln3−/− hippocampus21.

Fig. 2
figure 2

CLN3 disease hippocampus is significantly hypoexcitable in response to perforant path (PP) stimulation by p7. (A) The PP was stimulated, and the response was measured using voltage sensitive dye imaging in hippocampal slices from (B) p7, (C) p14, and (D) p21 Cln3−/− and Cln3+/+ mice. Raster plots of average fluorescence change (DF/F, warm colors excitation, cool colors inhibition) over time (x axis) and location within hippocampus (y axis) are shown. Stimulation of Cln3+/+ slices results in robust excitation of the dentate gyrus (DG) that spreads to the CA proper (hilus, CA3, CA1) at p7. In Cln3+/+ slices from older (p14 and p21) animals demonstrates increased gating function of the DG, as the spread of excitation from the DG to the CA proper is reduced. In Cln3−/− animals, the DG is significantly hypoexcitable at p7. DG excitability improves by p14, but by p21 the CA proper region becomes hypoexcitable, consistent with prior studies of older Cln3−/− mice. For each age group, Cln3−/− vs. Cln3+/+ rasters were compared using a permutation sampling method with 1000 iterations. Regions of significant (p < = 0.05) hypoexcitability are shown in purple. N = at least 10 slices from 5 animals per condition.

Full size image

Next, we evaluated how CLN3 disease disrupts maturation of excitability of the distal hippocampus in response to Schaffer collateral (SC) stimulation (Fig. 3a, stimulation site noted by asterix). Interestingly, we found that at early timepoints including p7 (Fig. 3b) and p14 (Fig. 3c), there was a small increase in backpropagation of the signal into the hilus after SC stimulation. However, as early circuit pathology progresses by p21 (Fig. 3d), the CA proper becomes hypoexcitable to SC stimulation, similar to what is seen after PP stimulation in p21 Cln3−/− slices (and similar to what we previously reported in older Cln3−/− hippocampus)21.

Fig. 3
figure 3

CLN3 disease hippocampus is initially hyperexcitable then hypoexcitable to Schaffer collateral (SC) stimulation. (A) The VSDI response after SC stimulation of (B) p7, (C) p14, and (D) p21 Cln3−/− and Cln3+/+ hippocampus is shown. Raster plots of average fluorescence change (DF/F, warm colors excitation, cool colors inhibition) over time (x axis) and location within hippocampus (y axis) are shown. Stimulation results in both forward and backward propagation of excitation. At p7 and p14 there is a trend toward hyperexcitability in Cln3−/− hippocampus, but by p21, both forward and backward propagation of excitation is decreased. For each age group, Cln3−/− vs. Cln3+/+ rasters were compared using a permutation sampling method with 1000 iterations. Regions of significant (p < 0.05) hypoexcitability are shown in purple. N = at least 10 slices from 5 animals per condition. Location of stimulus shown by an asterix.

Full size image

Significant in vivo EEG background slowing arises by P30 in Cln3 −/− hippocampus

To better understand in vivo circuit dynamics in the young Cln3−/− hippocampus, we completed longitudinal in vivo EEG recordings in p30 mice with hippocampal electrodes implanted in the CA1 region of the hippocampus (Fig. 4a). Previously, we had shown that rates of interictal spikes are increased in aged Cln3−/− mice. At p30 (Fig. 4b), there is a trend towards increased spiking rate, that does not reach statistical significance. We had also found shifts in the background frequency composition of hippocampal EEG in aged Cln3−/− mice. Here, we found that this hippocampal circuit dysfunction starts very early in the disease process. While the absolute power (Fig. 4c) of the CA1 EEG signal does not change significantly at p30 in the Cln3−/− hippocampus, there is a shift in normalized power, with a statistically significant increase in (Fig. 4d) normalized slow delta power (0.1–4 Hz), and a decrease in (Fig. 4e) normalized alpha power (8–12 Hz). These changes lead to an (Fig. 4f) increased delta to alpha ratio, an established EEG measure of pathologic slowing seen in other disease processes including ischemic injury and autism.

Fig. 4
figure 4

p30 Cln3 disease hippocampus shows background slowing with an increased delta / alpha ratio on EEG. (A) Long-term in vivo EEG recordings from the CA1 region of the hippocampus from p30 mice shows no significant difference in (B) interictal spikes or (C) total raw power between Cln3+/+ (black) and Cln3−/− (blue) hippocampus. However, there is significant background slowing in Cln3−/− hippocampus as indicated by a (D) increase in slow delta band (0.1–4 Hz) power, a (E) decrease in alpha band (8–12 Hz) power, and a resulting (F) increased delta: alpha ratio. Data were analyzed via 2-way ANOVA, with p-value for genotype as a source of variation shown, *p < = 0.05, **p < = 0.01. N = 7 Cln3+/+ and 6 Cln3−/− mice. Spikes were quantified as spikes / 30 min period averaged from 24 h of recording. Average power in each frequency band was calculated for 12 h of daytime and 12 h of night recordings from electrodes implanted into the right and left hippocampi.

Full size image

Discussion

Here, we demonstrate for the first time that early postnatal neuronal network development differs in CLN3 disease mouse models, challenging the previously held belief that CLN3 disease is purely a neurodegenerative disorder. Furthermore, we found that GPIs are significantly elevated in all brain regions studied very shortly after birth, as early as postnatal day 11.

Glycerophosphoinositol as early biomarkers in Cln3 disease

Laqtom et al. previously reported the accumulation of glycerophosphodiesters (GPDs) in lysosomes of 1-year-old Cln3del78/del78 mice, a model of the common 1.02 kb deletion of exons 7 and 8 found in most CLN3 patients1. These findings suggest that CLN3 may play a role in the extrusion of GPDs from the lysosome to the cytoplasm. This phenotype has been replicated in other models, and glycerophosphoinositols (GPIs) have been detected in patient blood2,33. High levels of GPDs are known to inhibit glycerophospholipid catabolism in the lysosome, although the specific impact of individual GPD pathway disruptions on neuropathology remains unclear34. Laqtom et al. reported that GPI showed the highest fold change in GPD levels, followed by GPG, GPC, GPE, and GPS, with the lowest abundance observed in GPS. Our data during very early development at p11 support this trend for GPI and GPG, suggesting that differences in their levels may correlate with disease onset. The early accumulation of GPIs makes them a promising candidate as an early biomarker for future mechanistic and therapeutic studies. These findings point to a potential early role for the GPI pathway in hippocampal dysfunction that may contribute to disease pathology. It remains uncertain whether GPIs, their precursors, or alternative pathway analytes are involved in this process.

The major lipid precursor to GPIs, phosphoinositol (PI), is also involved in the biosynthesis of glycosylphosphoinositol -anchored proteins (GPI-APs), which are crucial for synaptic development35. Additionally, GPIs and their phosphorylated forms, such as glycerophosphoinositol-4-phosphate (GroPI4P) and glycerophosphoinositol 4,5-bisphosphate (GroPI45P), are implicated in T-cell signaling and chemotaxis36. The precursors of these phosphorylated species, membrane phosphoinositides (PtdIns), play significant roles in membrane dynamics, organelle specification, vesicle trafficking, and endocytosis37,38. PtdIns are also involved in calcium sequestration and release through inositol trisphosphate (IP3) signaling38. Therefore, the accumulation of GPIs in CLN3 disease may disrupt intracellular signaling. Further investigations are needed to evaluate if GPI alterations can directly induce differences in neuronal excitability and development.

We also observed that GPG abundance, by weight, is higher in the cerebellum than in the cortex in both Cln3−/− and Cln3-/+ mice, but not in wild-type (WT) controls. These differences may need to be examined at later developmental stages to determine whether WT mice exhibit similar trends, or if this finding is specific to this mouse strain or haploinsufficiency. Previous studies have shown that different brain regions have distinct lipid signatures, suggesting that byproducts of lipid metabolism would also differ in abundance as development progresses39.

Disruption of early hippocampal network development in CLN3 disease models

Previous studies have highlighted progressive defects in neuronal network function in CLN3 disease mouse models. Notably, we have identified hypoexcitability in the dentate gyrus (DG) of the hippocampus in CLN3 disease mice as early as two months of age, as detected by voltage-sensitive dye imaging. This observation was unexpected, occurring before significant storage accumulation or neuronal cell death, suggesting that early hippocampal development is impaired in CLN3 disease21,22.

To investigate the early impact of CLN3 disease on hippocampal development, we performed voltage-sensitive dye imaging in hippocampal slices from Cln3−/− and Cln3+/+ mice at postnatal days 7, 14, and 21. We first assessed the response to direct perforant path (PP) stimulation, which is the primary input to the hippocampus from the entorhinal cortex. At p7, Cln3−/− mice exhibited pronounced hypoexcitability in the DG in response to PP stimulation compared to Cln3+/+ controls. This hypoexcitability in the DG led to reduced excitation in downstream regions, including the hilus, CA3, and CA1. By p14 and p21, the wildtype DG matured to function effectively as a gate, limiting the spread of excitation to the CA regions. In contrast, while DG excitability in Cln3−/− mice improved by p14 compared to p7, the CA regions remained hypoexcitable relative to wildtype controls. By p21, this hypoexcitability worsened, consistent with previous findings in older Cln3−/− mice24.

We next examined how CLN3 disease affects the maturation of excitability in the distal hippocampus, specifically in response to Schaffer collateral (SC) stimulation. At early timepoints (p7 and p14), we observed a small but notable increase in the backpropagation of the signal into the hilus following SC stimulation. However, as circuit pathology progressed by p21, the CA regions became hypoexcitable to SC stimulation, a pattern similar to that observed with PP stimulation in p21 Cln3−/− slices and consistent with our earlier observations in older CLN3 disease models21. One explanation for this change may be that at early time points when the DG is very hypoexcitable to the major input pathway stimulation, the hilus and CA3 regions compensate by becoming hyperexcitable relative to wildtype slices. These findings collectively suggest that CLN3 disease leads to early and persistent disruptions in hippocampal network maturation. Despite partial recovery of DG excitability at later stages, the overall hypoexcitability of the CA regions points to a fundamental defect in hippocampal circuit development that precedes more severe degenerative changes.

In vivo hippocampal circuit dynamics in early CLN3 disease

To validate that the hippocampus is also dysfunctional in vivo during early life, we conducted longitudinal hippocampal EEG recordings in 30-day-old mice. We previously found that aged Cln3−/− mice, had increased rates of hippocampal spikes22. Although we observed a trend toward an increased spiking rate at p30, this difference did not reach statistical significance. In addition, we had previously documented shifts in the background frequency composition of hippocampal EEG recordings from older Cln3−/− mice. Here, we demonstrate that these circuit alterations emerge early in the disease process. While the absolute power of the CA1 EEG signal did not differ significantly in p30 Cln3−/− mice, there was a notable shift in normalized power. Specifically, we observed a significant increase in normalized slow delta power (0.1–4 Hz) and a decrease in normalized alpha power (8–12 Hz). These alterations resulted in a higher delta-to-alpha ratio, a well-established EEG marker of pathological slowing, commonly seen in other neurodegenerative and neurodevelopmental disorders, including ischemic injury and autism40,41. These findings highlight early and subtle disruptions in hippocampal network function in the Cln3−/− model, preceding more pronounced circuit dysfunction observed at later stages of disease progression.

Future studies in prenatal CLN3 mice could help to elucidate if the GPI accumulation and network abnormalities begin before birth. In addition, it will be important to evaluate if early gene replacement therapy can normalize GPI levels in the developing brain.

In summary, our findings support the idea that CLN3 disease is marked by early metabolic and excitability disturbances that precede significant neurodegeneration and storage accumulation. Notably, alterations in GPD levels, particularly GPIs, emerge early and may serve as promising biomarkers for early diagnosis. Changes in hippocampal excitability provide insight into the pathophysiological mechanisms underlying cognitive and memory deficits in CLN3 disease. These findings emphasize the need for early intervention to preserve hippocampal network function and mitigate long-term neurodegenerative consequences. Future research should explore the temporal progression of metabolic and electrophysiological changes and investigate therapeutic strategies to restore hippocampal function before irreversible damage occurs.