Connectomic and behavioural alterations in creatine transporter deficiency are partially normalized by gene therapy

Abstract

Creatine transporter deficiency (CTD) is an X-linked disorder due to the loss of SLC6A8 gene and presenting with low brain creatine, intellectual disability, autistic-like behaviour and seizures. No treatments are available yet for CTD, and little is known about the brain circuit alterations underlying its pathological endophenotypes.

Here, we tracked brain network and behavioural dysfunction in a murine model of CTD at two stages of disease progression. Functional MRI (fMRI) mapping revealed widespread disruption of brain connectivity in Slc6a8-knockout (KO) mice, with robust somatomotor hypoconnectivity in juvenile animals and weaker and more focal alterations of cortical and subcortical connectivity in adulthood. Notably, perinatal AAV-mediated expression of human SLC6A8 in Slc6a8-KO mice robustly prevented juvenile functional MRI (fMRI) hypoconnectivity, an effect accompanied by the regression of multiple translationally relevant phenotypes, including reduced stereotyped movements, improved declarative memory and increased body weight, all of which persisted into adulthood. However, early cognitive deficits, impairments in working memory and residual fMRI hypoconnectivity in adult mice were not ameliorated by gene therapy. Furthermore, significant cognitive impairments were observed in wild-type mice receiving gene therapy, highlighting a potential detrimental effect of ectopic expression of SLC6A8 in healthy brain circuits. Finally, multivariate modelling in adult mice revealed a basal forebrain network whose activity was associated with behavioural performance and modulated by brain creatine levels. This brain-behaviour relationship was disrupted in Slc6a8-KO mice.

Our results document robust network disruption in CTD and demonstrate that CTD pathology can be partially alleviated by perinatal genetic expression of SLC6A8, providing a foundation for the future development of experimental therapies for this genetic disorder.

Creatine transporter deficiency is a genetic disorder causing low brain creatine, leading to intellectual disability, autistic-like behaviours, and seizures. In a mouse model, Montani et al. identify widespread brain connectivity disruptions and show that gene therapy can improve symptoms, offering hope for future treatments.

Introduction

Creatine transporter deficiency (CTD, OMIM: #300352) is an X-linked disease originating from mutations of the solute carrier family 6-member 8 (SLC6A8) gene.¹  SLC6A8 encodes the protein responsible for cellular creatine (Cr) uptake. The clinical manifestation of CTD follows lack of Cr in the brain and is predominantly neurological. It includes global developmental delay, mild to severe intellectual disability (ID), disturbance of expressive and cognitive speech, psycho-motor impairment, autistic-like behaviour and seizures.^2,3 Despite our increased understanding of the aetiopathological cascade underlying CTD,^3,4 effective treatments for this monogenic disorder are lacking. Dietary supplementation with Cr, either alone or in combination with its synthesis precursors, has shown very limited success.^2,3 Indeed, loss-of-function of SLC6A8 not only prevents transport of Cr across the plasmatic membrane, but also affects its endogenous synthesis, which requires the coordinated action of multiple cell populations.^5,6 As a result, the current standard of care for CTD only involves pharmacological control of epilepsy and other CTD symptomatology.

Over the last few years, studies in murine models have revealed that neurological damage produced by CTD does not stem from overt neuronal death or degeneration, but rather reflects more subtle alterations in synaptic compartments. Specifically, our own work previously documented a loss of GABAergic synapses in the cerebral cortex of mice lacking Slc6a8 [Slc6a8-knockout (KO)]. This effect was associated with a dysfunction of parvalbumin-expressing GABAergic interneurons.^7,8 Supporting a key aetiopathological role of altered inhibitory transmission in CTD, we also described marked abnormalities in EEG neural oscillations, with decreased theta/alpha power and increased gamma activity in both Slc6a8-KO mice and CTD patients.⁹ Owing to the key contribution of oscillatory activity in coordinating large-scale patterns of brain synchronization and interareal communication,¹⁰ this observation suggests that the neurological alterations that characterize CTD may be partly due to impaired long-range functional connectivity.¹¹ This notion is consistent with neuroimaging observations in CTD patients, where abnormalities in anatomical connectivity (including white matter demyelination and corpus callosum thinning) have been reported.¹²

Here, we used functional MRI (fMRI) connectivity mapping^13,14 and behavioural testing in Slc6a8- KO mice to (i) longitudinally track the neural circuits primarily affected in CTD and (ii) relate connectivity alterations to the behavioural deficits that characterize this disorder. Importantly, we also probed whether perinatal adeno-associated virus (AAV)-mediated expression of a functional SLC6A8 transgene could restore proper creatine transporter (CRT) expression and Cr levels in the brain, thereby averting the manifestations of connectomic and behavioural phenotypes. Our results show that CTD is associated with distinct alterations in functional connectivity that evolve along its pathological trajectory. We also found that early intracerebral expression of human SLC6A8 (hSLC6A8) may partly ameliorate both fMRI hypoconnectivity and behavioural impairment in Slc6a8-KO mice. These results provide proof-of-concept evidence that CTD phenotypes may be partly reversed by genetic therapies aimed at reinstating homeostatic Cr levels.

Materials and methods

Full-length experimental procedures are available in the online Supplementary material.

Ethical statement

Animal research was conducted in accordance with Italian Law (DL 26/2014 of the Italian Ministry of Health implementing EU 63/2010) and the National Institutes of Health. Animal projects were approved by the Animal Care Committee of the University of Trento, Istituto Italiano di Tecnologia and Neuroscience Institute of Consiglio Nazionale delle Ricerche.

Animals

Slc6a8-KO mice were generated by the European Molecular Biology Laboratory (EMBL). Slc6a8^x/− females were crossed with wild-type male mice to generate Slc6a8^−/y males (Slc6a8-KO, hereafter KO). Slc6a8^+/y wild-type (WT) male littermates were used as controls. Mice were maintained in a controlled environment with humidity set at 60% ± 10% and temperature maintained at 21°C ± 1°C. Food and water were readily accessible ad libitum. Each standard cage provided nesting material and accommodated 2–5 mice.

Molecular cloning

The full-length human sequence of SLC6A8 (hSLC6A8) was cloned under the control of the small JeT promoter¹⁵ into a pAAV_WPRE.SV40. An HA tag at the C-terminal of SLC6A8 was used to detect the transgenic protein (pAAV_JeT-hSLC6A8-HA_WPRE.SV40). The JeT-hSCL6A8 sequence was synthesized by Twin Helix (Milano, Italy) as reported in Supplementary Table 1. To verify that the HA tag does not interfere with transporter function, we also generated a version of the pAAV_JeT-hSLC6A8_WPRE.SV40 plasmid without the HA tag.

Cell culture and transfection

HEK293T cells were maintained in Dulbecco’s Modified Eagle’s Medium with 10% fetal bovine serum and 1% penicillin/streptomycin. Cell cultures were incubated at 37°C in a humidified atmosphere with 5% CO₂. For analysis of Cr levels, upon reaching ∼80% confluency, cells were transfected using Lipofectamine 2000 at a 1:2 DNA/Lipofectamine ratio. Each well received 1.5 μg of plasmid DNA (JeT-hSLC6A8 or JeT-hSLC6A8-HA). The following day, cells were stimulated for 24 h with 125 μM Cr. After treatment, the cells were centrifuged and pellets were stored at −80°C.

Viral preparation

Serotype 9 adeno-associated viral (AAV) vectors containing pAAV_JeT-hSLC6A8-HA_WPRE.SV40 were produced by the University of Pennsylvania Vector Core (Philadelphia, PA). As control, we used an AAV9 containing pAAV_JeT-GFP_WPRE.SV40.

Experimental design

Investigations at IIT (Gozzi laboratory, Rovereto, Italy) were carried out on four groups of male mice as follows: Slc6a8-KO (n = 20) and WT (n = 10) injected with AAV-hSLC6A8, and Slc6a8-KO (n = 15) and WT (n = 23) injected with AAV-GFP as controls. The four groups of mice were administered with AAV vectors via intracerebroventricular (i.c.v.) injection at postnatal Day (P)1. The four experimental groups were subjected to fMRI scans, at P40 and P140. Forty-eight hours after each fMRI session, mice underwent two behavioural tests. The testing order consisted of self-grooming scoring, followed by the Y-maze test. The rest of the experimental mice were used for post-mortem Cr quantifications, using liquid chromatography-tandem mass spectrometry (LC-MS/MS).

We replicated the Y-maze experiments and conducted an Object Recognition Test (ORT) in a new group of mice treated with either AAV-hSLC6A8 or phosphate buffered saline (PBS) at P40 and P100. These tests were conducted at the CNR site (Baroncelli lab, Pisa). This part of the study involved four groups of male mice: Slc6a8-KO (n = 15) and WT (n = 17) receiving i.c.v. injections of AAV-hSLC6A8, and Slc6a8-KO (n = 18) and WT (n = 16) receiving PBS. At P100, a subset of animals (n = 4 per group) underwent post-mortem Cr quantifications, which at the CNR site were carried out using gas chromatography-mass spectrometry (GC-MS). GC-MS Cr measurements were also performed at P20 and P40 in separate groups of Slc6a8-KO (n = 4) and WT (n = 4) animals treated with AAV-hSLC6A8, along with corresponding control groups of Slc6a8-KO (n = 4) and WT (n = 4) animals.

Finally, an additional group of animals composed of Slc6a8-KO (n = 4) and WT (n = 4) mice treated with AAV-hSLC6A8, and Slc6a8-KO (n = 2), and WT (n = 2) mice receiving PBS as control, were sacrificed at P40 for analysis of transgene biodistribution and cellular targeting via immunohistochemistry.

AAV injection in newborn mice

Intracerebroventricular injection of AAV was performed as previously described.¹⁶ Briefly, at P1 pups were placed on ice and then moved onto a cooled stereotaxic frame. Coordinates were adjusted on the base of lambda point of repere (X, Y, Z) = (1, ±0.3, −2.0) mm. Pups received bilateral injections of 1 μl of AAV vector (3 × 10⁹ vg/mouse). The AAV solution was infused over a 60 s period and the pipette was removed after a delay of 30 s to prevent backflow.

Functional and anatomical MRI

Resting state fMRI (rsfMRI) data were acquired as previously described.^14,17-19 Briefly, mice were anaesthetized with isoflurane, intubated, and artificially ventilated. During resting-state scans, isoflurane was replaced with halothane (0.7%) to obtain light sedation. To assess potential differences in the sensitivity of Slc6a8-KO animals to the anaesthetic used, we measured the minimal alveolar concentration (MAC) in an independent group of adult mice, n = 10 WT, n = 5 Slc6a8-KO, as previously described.^20,21 We used a 7 T MRI scanner (Bruker), Bruker Paravision software version 6, a 72 mm birdcage transmit coil and a four-channel solenoid coil for signal reception.^20,22,23 For each session, in vivo structural images were acquired with a fast spin-echo sequence. Blood oxygen level-dependent (BOLD) rsfMRI time series were acquired using echo planar imaging (EPI) for 1920 volumes (total duration 32 min).

Analysis of fMRI time series

Resting-state fMRI BOLD time series were preprocessed as previously described.^14,20,22,24 Briefly, BOLD timeseries were despiked, motion corrected, and spatially registered to a common reference brain template. Mean ventricular signal and motion traces of head realignment parameters were regressed out from the time series. Finally, we applied a spatial smoothing and a band-pass filter to a frequency window of 0.01–0.1 Hz.

Network based statistics (NBS)^14,25 were carried out by extracting the fMRI signal in 85 anatomically parcellated regions (Supplementary Table 2). The list of employed regions can be found in.²⁶ We next computed an unpaired two-tailed Student's t-test for each element of the corresponding correlation matrix separately (t > 2.7–3.5). Family-wise error rate (FWER) correction at the network level was performed using 5000 random permutations (P < 0.05). The chord plots show the 85 parcels clustered in nine anatomical meta regions (Supplementary Table 2). rsfMRI connectivity was also probed using seed-based analyses.^22,27 A seed region was selected to cover the areas of interest, based on the NBS results (Supplementary Fig. 1). Voxel-wise intergroup differences in seed-based mapping were assessed using a two-tailed Student’s t-test |t > 2, P < 0.05) and family-wise error (FWE) cluster-corrected using a cluster threshold of P = 0.050.

Brain volume analysis

T2-weighted anatomical scans were manually skull-stripped using ITKSNAP. We used the nipype Python library²⁸ wrapper for ANTS²⁹ registration to compute a symmetric diffeomorphic transform³⁰ from the dsurqe³¹ reference template on individual images. The obtained transform was applied on cortex, hippocampus, striatum and thalamus using anatomical coordinates derived from the Paxinos/dsurqe ontology.³²

Immunohistochemistry

Conventional (single-label) immunohistochemistry assays were carried out as described in the Supplementary material, ‘Methods’ section. For double-labelling experiments, brain sections were co-incubated overnight with primary antibodies against the HA-tag targeting specific cell-type markers, including neuronal marker MAP2, astroglial marker GFAP, microglial marker Iba-1 and oligodendroglial marker BCAS1. Antigen-antibody interactions were visualized using Alexa Fluor-conjugated secondary antibodies (1:400, Invitrogen).

Quantification of HA-tag biodistribution and analysis of co-localization

The right hemisphere of each section was imaged using a Zeiss laser-scanning Apotome microscope. Maximum intensity projections (MIPs) were generated from the four consecutive optical sections displaying the highest mean pixel intensity. A threshold was established based on negative control HA-tag staining from PBS-injected animals. Subsequently, the area fraction of HA-positive pixels in the whole hemisphere was quantified.

For more targeted analysis, regions of interest (ROIs) were drawn around specific areas and the area fraction of HA-positive pixels within these ROIs was measured.

Behavioural tests

To investigate spontaneous self-grooming, mice were individually placed in an open field arena for 20 minutes. Sessions were recorded and mice automatically tracked using the ANY-maze software. After a 10-min habituation period, the cumulative time spent by mice grooming themselves was scored for 10 min as an index of stereotypic behaviour.^7,33 Spatial working memory was tested using the Y-maze, as previously described.^7,34

We also assessed declarative memory using the ORT. The total movement of the animal (Total exploration time) and the total distance covered (Distance moved) were automatically computed. A discrimination index was computed as previously reported.³⁴

Creatine measurements

Liquid chromatography-tandem mass spectrometry

Brain samples were homogenized in PBS-Protease inhibitor (100:1) on ice. An aliquot of each brain homogenate was extracted with cold CH₃CN containing creatine-(methyl-d3) as internal standard and centrifuged at 21.100g for 20 min at 4°C. A calibration curve was prepared in PBS containing 20% CH₃CN. The calibrators were extracted as the brain homogenates. The supernatants of the extracted brain homogenates and calibrators were further diluted 100-fold with 2 mM NH₄OAc in H₂O (pH 8) and analysed by LC-MS/MS on a Waters ACQUITY ultra-high performance (UHP)LC-MS/MS system. Electrospray ionization was applied in positive mode. The compound-dependent parameters multiple reaction monitoring (MRM) transitions and collision energy were developed for the parent compound and the internal standard. The analyses were run on an ACE Excel 2 C18 (150 × 2.1 mmID) with an ACE Excel UHPLC pre-column filter at 40°C, using 2 mM NH₄OAc in H₂O (A) and CH₃CN (B) as mobile phases at 0.2 ml/min. All samples were quantified by MRM peak area response factor in order to determine the levels of Cr in the brain samples. Data were expressed as ng Cr per mg brain (ng/mg brain) and normalized by WT creatine content (%WT).

Gas chromatography-mass spectrometry

HEK293T pellets and brain samples were homogenized in 0.7 ml of ice-cold PBS buffer using an ultrasonic disruptor. Samples were centrifuged at 600g for 10 min at 4°C. A 50 µl aliquot of the supernatant was taken for protein content assessment, while the remaining volume was used for Cr analysis, as previously described.^7,35 Data were expressed as nmol Cr per mg protein (nmol/mg pr) and normalized relative to the creatine content measured in WT samples (% WT).

Partial least squares correlation

Partial-least square analyses were designed to maximize the covariance between imaging readouts (seed-based connectivity maps of the anterior cingulate), behavioural scores (i.e. self-grooming and spontaneous alternations) and brain Cr levels via the identification of latent components (LC) representing the optimal weighted linear combination of the original variables.³⁶ The null hypothesis that the observed brain-behaviour relationship could be due to chance was verified via permutation testing (1000 iterations) of the behavioural and biochemical data matrix. Permutations were performed within groups as per previous guidelines.³⁷ The stability of the contribution of each brain and behavioural element was assessed via bootstrapping, where bootstrap resampling was performed within each experimental group to avoid patterns being driven by group differences.³⁸

Results

Creatine deficiency disrupts fMRI connectivity in cortical and subcortical brain regions

To probe whether CTD affects interareal functional connectivity, we carried out rsfMRI measurements in Slc6a8-KO mice and control littermates (WT) at two different stages of disease progression. Specifically, we longitudinally mapped fMRI connectivity in the same mice at P40 and in adulthood (P140), corresponding to early and late pathological stage, respectively. Animals were scanned using an established sedation protocol that has been shown to preserve network organization in rodents, making it comparable to that mapped in awake conditions.^39,40 Moreover, the protocol has proven effective in identifying genotype-dependent effects in multiple genetic models of developmental disorders,⁴¹ revealing highly congruent functional connectivity changes in corresponding clinical populations.^19,20,42 To obtain a spatially unbiased mapping of genotype-specific differences in fMRI connectivity, we used NBS with a whole-brain network parcellation.^14,25 This analysis revealed the presence of diffuse hypoconnectivity in multiple brain regions of Slc6a8-KO mice both at P40 and P140 (Fig. 1). The spatial extent and magnitude of the observed changes, however, were largely different across the probed ages. In juvenile mice, fMRI hypoconnectivity was distributed across multiple cortical and subcortical regions, with a prominent involvement of somatosensory and motor cortices as well as hippocampal and thalamic areas (Fig. 1A and B). In contrast, fMRI hypoconnectivity in Slc6a8-KO mice at P140 was weaker and more focal, encompassing the motor cortex and subcortical regions such as the striatum, the hypothalamus and some thalamic areas (Fig. 1C and D). To further investigate the circuit-level substrates differentially affected by CTD, we next used a seed-based analysis to probe fMRI connectivity in some of the brain regions exhibiting large effect-size in NBS (Fig. 1). This investigation revealed robustly reduced inter-hemispheric connectivity in somatomotor areas of Slc6a8-KO mice at P40 (Fig. 2A–D), and to a lower extent also at P140 (Fig. 2E and H and Supplementary Fig. 1). At P40, Slc6a8-KO mice also showed reduced thalamo-hippocampal and frontocortical connectivity, but no alterations in striatal regions (Supplementary Fig. 2). In contrast, the striatum represented one of the primary hubs of fMRI hypoconnectivity in Slc6a8-KO mice at P140, with evidence of decreased coupling of this region with thalamic nuclei (Fig. 2F and H). Adult Slc6a8-KO mice also exhibited foci of reduced fMRI connectivity in thalamo-hippocampal regions (Supplementary Fig. 2). Importantly, these connectivity differences are unlikely to reflect a different sensitivity of Slc6a8-KO animals to the employed sedative regime, because Minimal Alveolar Concentration (MAC) testing in a separate group of animals revealed highly comparable sensitivity to halothane across the two genotypes (WT, MAC 1.70 ± 0.13; Slc6a8-KO, MAC 1.68 ± 0.08; Student’s t-test P = 0.87). Leveraging the longitudinal design of our investigation, we also examined the temporal evolution of connectivity in somato-motor and striatal circuits across the two timepoints tested (Supplementary Fig. 3). A pairwise quantification of connectivity in these networks did not reveal any difference in temporal trajectories across groups (Supplementary Fig. 3). Altogether, these findings reveal widespread disruption of brain connectivity in Slc6a8-KO mice, with robust alterations of somato-motor connectivity in juvenile mice evolving into more focal cortical and subcortical hypoconnectivity in adulthood.

Figure 1.

Reduced long-range functional MRI connectivity in Slc6a8-knockout mice. Functional MRI (fMRI) connectivity alterations in Slc6a8-knockout (KO) mice as assessed with network based statistics (NBS) at postnatal Day (P)40 (A and B) and P140 (C and D; t > |2.7| for both comparisons). Histograms represent the number of links exhibiting reduced connectivity in Slc6a8-KO mice at P40 and P140, respectively. Circular plots in B and D illustrate corresponding patterns of altered interareal connectivity. In these plots, individual brain areas have been grouped into nine anatomical meta regions. The thickness of links in circular plots is proportional to the relative number of affected links. The complete list of brain areas (and abbreviations) used for NBS is reported in Supplementary Table 2. WT = wild-type.

Figure 2.

Functional networks exhibiting altered functional MRI connectivity in Slc6a8-knockout mice. (A–D) Seed-correlation mapping highlighted disrupted somato-motor connectivity in Slc6a8-knockout (KO) mice at postnatal Day (P)40. (E–H) Functional MRI (fMRI) connectivity in motor regions was also disrupted at P140. At this age, we also found a prominent reduction of fMRI connectivity between striatal regions and the thalamus. Red/yellow indicates areas in the brain maps exhibiting significant (t > 2.1) fMRI connectivity with seed regions (indicated with red lettering). Blue indicates between-group connectivity differences (t-test, t > 2.1). Corresponding quantifications of connectivity changes in the two groups are reported in C and D for P40 (S1 t = 4.02, P = 0.0003; M1 t = 4.3, P = 0.0001) and G and H for P140 (M1, t = 2.06, P = 0.046; Striatal-Thal, t = 3.55, P = 0.001). All statistics are family-wise error cluster-corrected. *P < 0.05, **P < 0.01, ***P < 0.001. Error bars indicate standard error of the mean and dots represent individual values. L = left; R = right; M1 = motor cortex; S1 = somatosensory cortex; Thal = thalamus.

AAV-mediated expression of human SLC6A8 prevents fMRI hypoconnectivity in juvenile Slc6a8-KO mice

To probe the efficacy of genetic expression of exogenous CRT as putative therapy for CTD, we conducted i.c.v. injections of a novel AAV9-vector encoding a functional copy of the non-codon optimized human SLC6A8 gene under the control of JeT promoter (AAV9/JeT-hSLC6A8, herein referred to as AAV-hSLC6A8). The AAV-hSLC6A8 vector was designed to leverage the wide biodistribution of AAV following i.c.v. perinatal injection⁴³ and to allow broad and ubiquitous transgene expression in the brain via the use of JeT promoter.⁴⁴ Our experimental goal was to induce early hSLC6A8 expression to minimize or prevent the emergence of pathological CTD-related phenotypes. Before proceeding with in vivo efficacy testing, we conducted a series of control studies to probe the function, transduction efficiency and tropism of the employed vector.

We first verified that the HA-tagged version of CRT we employed would not affect its function. To this end, we compared Cr uptake in 293T cells transfected with either the original plasmid encoding the HA-tagged transporter or a newly developed plasmid encoding the native (untagged) transporter. Notably, we observed highly comparable Cr uptake between the two conditions (204.00 ± 13.32 and 208.00 ± 16.64 nmol/mg protein for the tagged and untagged version, respectively, one-way ANOVA, post hoc Tukey’s test, P = 0.97), suggesting that the incorporation of an HA-tag in the hSLC6A8 gene does not impair CRT function.

We next mapped in vivo transduction patterns of exogenous CRT using an anti-HA-tag antibody (Fig. 3A and Supplementary Fig. 4A). We found that transgenic CRT was expressed in a punctate pattern along cell membranes as expected. The transduction pattern of exogenous CRT exhibited widespread extension in both WT and Slc6a8-KO mice (Fig. 3A and Supplementary Fig. 4A). Expression was particularly prominent in the deeper layers of the cortex and extended to subcortical regions, including the hippocampus and the striatum. Quantitative analyses indicated an overall transduction efficiency of 7.7% across experimental groups (7.6% ± 3.7% and 7.8% ± 1.3% in Slc6a8-KO and WT animals, respectively; Fig. 3B and Supplementary Fig. 4B). Transduction efficiency was not affected by genotype (two-way ANOVA, genotype effect, P = 0.37) but exhibited regional variability, with lower efficiency in deep parenchymal regions (e.g. striatum, two-way ANOVA, region effect, P = 0.01). To better define the cellular targets of hSLC6A8 gene transfer and determine whether transduction differentially affects neurons and glial cells, we next performed double-labelling of HA-tag with cell-type-specific markers, including MAP2 for neurons, GFAP for astrocytes, IBA1 for microglia, and BCAS1 for oligodendrocytes. Our analysis demonstrated a strong tropism of the AAV-hSLC6A8 vector for neurons (Fig. 3C and Supplementary Fig. 4C), as well as oligodendrocytes and microglia, whereas no astrocytes were found to be double-labelled with the HA tag (Supplementary Figs 5–7).

Figure 3.

Exogenous creatine transporter protein is widely distributed throughout the brain and colocalizes with neurons in different brain regions in Slc6a8-knockout mice. (A) Representative coronal sections (×20 magnification) stained to quantify HA-tagged distribution of transgenic creatine transporter (CRT) protein in the striatum, hippocampus and cerebral cortex of Slc6a8-knockout (KO) mice treated with AAV-hSLC6A8. Scale bars = 1 mm (top), 100 μm (bottom). (B) Quantification of the area positive for HA-tag staining (%) in the three considered brain regions and in the right brain hemisphere (total) of Slc6a8-KO mice. Error bars indicate standard error of the mean. (C) Representative images illustrating colocalization of HA-tag staining (green signal) along with the neuronal marker MAP2 (red signal) across different brain regions. Far left: ×20 magnification (scale bars = 50 μm). Second left to far right: ×63 magnification (scale bars = 20 μm, 50 μm and 5 μm from left to right, respectively). AAV = adeno-associated virus.

To test whether perinatal AAV-mediated expression of hSLC6A8 represents a viable gene therapy for CTD, we carried out longitudinal fMRI and a behavioural investigation. The AAV-hSLC6A8 vector was injected perinatally in newborn (P1) Slc6a8-KO mice and WT littermates (Fig. 4A). An analogous vector expressing GFP protein (AAV-GFP) was used as the control condition in both genotypes. Notably, NBS analysis of fMRI measurements revealed that AAV-hSLC6A8 administration prevented the manifestation of fMRI hypoconnectivity in Slc6a8-KO mice at P40 (Fig. 4B and C). This effect entailed a marked increase in fMRI connectivity in virtually all the neocortical and subcortical regions that we found to be hypo-connected in Slc6a8-KO mice injected with AAV-GFP. Seed-based mapping corroborated this result, showing that gene therapy could effectively ameliorate somatomotor (Fig. 4D–G) and cingulate hypoconnectivity in Slc6a8-KO mice at P40 (Supplementary Fig. 8). In contrast, no appreciable connectivity normalization was observed in adulthood in the same mice, using NBS analysis (P > 0.9, t < 0.1, all links). Accordingly, seed-based mapping did not reveal any functionally relevant effect in treated Slc6a8-KO mice at P140 (Fig. 4H–K and Supplementary Fig. 8). Taken together, these results show that neonatal expression of hSLC6A8 robustly improved juvenile fMRI hypoconnectivity in Slc6a8-KO mice, but not the corresponding alterations in fMRI connectivity observed at a later pathological stage (i.e. P140).

Figure 4.

Perinatal AAV-hSLC6A8 injection prevents onset of functional MRI hypoconnectivity in juvenile Slc6a8-knockout mice. (A) Experimental design. (B) Network based statistics (NBS) highlighted increased functional MRI (fMRI) connectivity at postnatal Day (P)40 in AAV-treated Slc6a8-knockout (KO) mice compared to controls (red, t > |3|). (C) Connectivity links exhibiting a significant amelioration of connectivity in AAV-treated Slc6a8-KO mice and regional quantification in circular plots. (D–G) Seed-correlation mapping revealed increased interhemispheric connectivity in the somato-motor cortex of AAV-treated juvenile Slc6a8-KO mice. Bottom: Red/yellow colouring denotes the amelioration of fMRI connectivity produced by AAV-hSLC6A8 treatment (t-test, t > 2). Plots report quantification of connectivity strength between seed and regions exhibiting amelioration of connectivity (E, M1: two-way ANOVA, Genotype × Treatment interaction F = 9.43, P = 0.0031; t-test, ^#P < 0.05, ^###P < 0.001; G, S1: two-way ANOVA, interaction F = 4.06, P = 0.048; Mann–Whitney test, ^#P < 0.05; t-test, ^###P < 0.001). The same analysis at P140 revealed no effect of gene therapy in either motor cortex (I; two-way ANOVA, genotype factor F = 11.05, P = 0.0015, interaction F = 1.30, P = 0.26) or the striatum-thalamus network (K; two-way ANOVA, genotype factor F = 11.77, P = 0.0011, interaction F = 1.16, P = 0.28). All statistics are family-wise error cluster-corrected. For two-way ANOVA, *P < 0.05, **P < 0.01, ***P < 0.001. Error bars indicate standard error of the mean, and dots represent individual values. The complete list of brain areas (and abbreviations) used for NBS is reported in Supplementary Table 2. AAV = adeno-associated virus; ICV = intracranial volume; L = left; M1 = motor cortex; R = right; S1 = somatosensory cortex; Thal = thalamus; WT = wild-type.

We finally tested whether our gene therapy would affect brain anatomical alterations in Slc6a8-KO mice. To this aim, we analysed longitudinal anatomical T2-weighted images acquired from the same mice used for fMRI at P40 and P140. Whole-brain volumetric analysis revealed a small but highly significant reduction in brain volume in Slc6a8-KO mice at both ages (P40: Δ = −4.9%; P140: Δ = −5.9%; Supplementary Fig. 9). Normalized regional volume quantifications showed no significant differences in all of the examined regions (thalamus, cortex, striatum, hippocampus) at P40 and P140. Importantly, AAV-mediated hSLC6A8 expression did not affect these volumetric reductions at either age (Supplementary Fig. 9). These findings reveal a generalized reduction in brain volume in Slc6a8-KO animals, which remained unaffected by gene therapy.

Human SLC6A8 prevents onset of autism-like behaviour and improves declarative memory in SLC6A8-KO mice

We previously reported that Slc6a8-KO mice exhibit a number of traits and phenotypes of translational relevance for CTD, including reduced body weight, cognitive impairment and autism-like behaviour.^7,12 To assess whether perinatal gene therapy could prevent the onset of CTD-relevant pathological manifestations, we probed working memory (using the Y-maze task) and stereotyped movements (via self-grooming scoring), and recorded body weight at P40 and P140 in mice injected with either AAV-hSLC6A8 or AAV-GFP (Fig. 5).

Figure 5.

Perinatal AAV-hSLC6A8 injection increases body weight and prevents the onset of autistic-like stereotypies in adult Slc6a8-knockout mice. (A–C) Time spent self-grooming (A, two-way ANOVA, Genotype × Treatment interaction F = 1.33, P = 0.25), spontaneous alternation in the Y-maze task (B, two-way ANOVA, interaction F = 8.55, P = 0.005) and body weight (C, two-way ANOVA, genotype factor F = 77.81, P < 0.0001, interaction F = 2.32, P = 0.13) in the four experimental groups at postnatal Day (P)40. (D) Quantification of brain creatine levels at P140 (two-way ANOVA, genotype factor F = 1068, P < 0.0001, interaction F = 39.63, P < 0.0001). (E) AAV-hSLC6A8 injection prevented the onset of aberrant grooming behaviour in adult (P140) Slc6a8-knockout (KO) mice (two-way ANOVA, interaction F = 2.09, P = 0.15), while no effect was found in the Y-maze (F, two-way ANOVA, genotype factor F = 4.23, P = 0.044, interaction F = 0.34, P = 0.56). (G) Body weight of treated Slc6a8-KO mice was significantly increased compared to controls (two-way ANOVA, genotype factor F = 420, P < 0.0001, interaction F = 1.08, P = 0.30). (H) Brain creatine levels negatively correlate with self-grooming scoring in Slc6a8-KO mice (Pearson r = −0.42, R-squared = 0.17, P = 0.022). This correlation was carried out using the same data as in E. Note that the treated KO group includes fewer animals because some mice were randomly selected for immunohistochemical assessments to verify viral expression. Error bars indicate standard error of the mean, and each dot represents a mouse. Tukey’s multiple comparison test, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. AAV = adeno-associated virus; WT = wild-type.

Slc6a8-KO mice at P40 did not show aberrant grooming behaviour (Fig. 5A), but exhibited significantly decreased spontaneous alternation in the Y-maze (Fig. 5B), and reduced body weight compared to WT mice (Fig. 5C). The same mice showed significantly increased grooming (Fig. 5E), decreased spontaneous alternations (Fig. 5F) and robustly reduced body weight at P140 (Fig. 5G), thus recapitulating our previous findings.⁷

Perinatal gene therapy produced a significant amelioration of some of these pathological traits. Specifically, at P40 we found a moderate (≈15%) increase of body weight of Slc6a8-KO mice, but no effect on behavioural performance in the spontaneous alternation test. Remarkably, at P140 AAV-hSLC6A8 treatment completely prevented the appearance of increased grooming in Slc6a8-KO mice. This effect was accompanied by a partial (≈20%) gain of body weight. However, as seen in juvenile mice, gene therapy failed to improve behavioural performance in the spontaneous alternation task at this disease stage. The inclusion of a group of WT mice receiving an AAV-hSLC6A8 injection allowed us also to investigate the behavioural effect of CRT overexpression. We noted a slight deterioration of behavioural performance in the Y-maze task in WT mice receiving the AAV-hSLC6A8 vector (Fig. 5B and F), as well as increased baseline exploratory activity (i.e. number of arm entries) in both Slc6a8-KO and WT mice injected with AAV-hSLC6A8 (Supplementary Fig. 10).

To further strengthen our behavioural investigations, we replicated the Y-maze experiments, and conducted the ORT in a new group of mice treated with either AAV-hSLC6A8 or PBS at two distinct developmental time points (P40 and P100; Fig. 6). These tests were conducted at a separate site (CNR, Pisa) using an independent mouse colony, allowing us to assess the generalizability of our primary behavioural findings. Consistent with our previous results (obtained at IIT, Rovereto), perinatal AAV-hSLC6A8 treatment did not improve Y-maze performance in Slc6a8-KO mice at either time point (Fig. 6A and D). Additionally, we observed significantly reduced Y-maze performance in AAV-hSLC6A8-treated WT mice at P40, with a similar, though non-significant, trend towards poorer performance at P100 (Fig. 6A and D).

Figure 6.

Perinatal AAV-hSLC6A8 injection partially restores creatine levels and cognitive function in adult Slc6a8-knockout mice. (A) At postnatal Day (P)40, AAV-hSLC6A8 treatment impaired Y-maze performance in wild-type (WT) mice, while no changes were detected in Slc6a8-knockout (KO) mice (two-way ANOVA, genotype factor F = 15.83, P = 0.0002, treatment F = 7.37, P = 0.0086, interaction F = 8.55, P = 0.0048). At the same age, no effect of gene therapy was observed in either WT or hSLC6A8-KO mice in the Object Recognition Test (B, two-way ANOVA, genotype factor F = 8.58, P = 0.0048). (C) Quantification of brain creatine levels at P40 (two-way ANOVA, genotype factor F = 225, P < 0.0001, treatment F = 23.5, P = 0.0004, interaction F = 17.0.9, P = 0.0014). (D) At P100, AAV-hSLC6A8 injection did not ameliorate behavioural performance in Slc6a8-KO mice in the Y-maze (two-way ANOVA, genotype factor F = 5.04, P = 0.0284), but significantly improved cognitive performance in the Object Recognition Test (E, two-way ANOVA, genotype factor F = 7.62, P = 0.0076, treatment F = 9.54, P = 0.003). (F) Quantification of brain creatine levels at P100 (genotype factor F = 56.03, P < 0.0001, treatment F = 7.81, P = 0.0174). Error bars indicate standard error of the mean, and each dot represents a mouse. Tukey’s multiple comparison test, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. AAV = adeno-associated virus; Cr = creatine.

Importantly, AAV-hSLC6A8 administration significantly mitigated cognitive deficits of Slc6a8-KO mice in the ORT at P100 (discrimination index; Fig. 6E). No significant therapeutic effect was instead observed at P40 (Fig. 6B). Total exploratory behaviour was comparably increased in both WT and Slc6a8-KO mice treated with AAV-hSLC6A8 at P100, thus ruling out genotype-specific confounding effects of general locomotor activity in the ORT (Supplementary Fig. 11). Taken together, these behavioural results demonstrate that perinatal gene therapy in Slc6a8-KO mice can alleviate specific pathological traits of high translational relevance for CTD, including cognitive dysfunction and autism-like motor stereotypies.

Gene therapy with AAV-SLC6A8 increases brain creatine levels

Brain Cr depletion underlies the pathological cascade leading to CTD.^2,3 Consequently, the therapeutic effects of AAV-hSLC6A8 treatment we observed on fMRI connectivity, motor stereotypies, declarative memory and body weight should reflect higher brain Cr levels. To verify this, we measured post-mortem Cr levels in distinct groups of animals at P20 (Supplementary Fig. 12), P40 (Fig. 6C), and in a randomly selected subgroup of mice at the end of behavioural investigations at P100 and P140 (Figs 5D and 6F). Our findings revealed a significant increase in brain Cr levels in AAV-hSLC6A8-treated KO mice. At the earliest time point (P20), Cr levels in Slc6a8-KO mice reached approximately 70% of WT values (Supplementary Fig. 12). This increase stabilized at around 50%–55% of WT levels by P40, and remained reliable throughout the study (Figs 5D and 6C and F). These results indicate that AAV-mediated expression hSLC6A8 encodes significant amounts of functionally-active CRT protein. It should be noted that the employed gene therapy was not able to fully reinstate physiological levels of Cr in Slc6a8-KO mice, as the average Cr content measured in mutant mice receiving AAV-hSLC6A8 was approximately half the amount measured in WT mice. In spite of this, we found an inverse relationship between Cr brain levels and self-grooming scoring in Slc6a8-KO mice treated with AAV-GFP or AAV-hSLC6A8 (Fig. 5H; Pearson’s r = −0.42; P = 0.022). A similar inverse relationship and slope was also noticeable in Slc6a8-KO mice receiving AAV-hSLC6A8 injections, although the effect in this group did not reach statistical significance (Fig. 5H; Pearson’s r = −0.35; P = 0.18). These results suggest that perinatal AAV-hSLC6A8 administration can effectively increase central levels of Cr, resulting in domain-specific amelioration of CTD pathology.

Mesolimbic-prefrontal connectivity predicts behavioural changes and is modulated by creatine levels

To obtain an unbiased mapping of the putative circuits underlying behavioural disruption in our CTD mouse model, we used a multivariate model to associate cross-subject variance in fMRI connectivity with corresponding behavioural profiles. This was done using a partial least-squares correlation (PLS) analysis^36-38,45 on fMRI connectivity data extracted from the cingulate cortex. The use of multivariate approaches like PLS reduces the bias related to the use of univariate brain/behaviour correlations in relatively small samples, an approach that has been recently shown to be highly prone to false positives.⁴⁶ Another advantage of PLS is the possibility of modelling Cr levels as a continuous variable within the same framework. Thus, using PLS we sought to identify areas whose connectivity covaries with (and thus putatively explains) the behavioural profile across multiple tests.

For each animal we modeled behavioural scores out of spontaneous alternation and grooming tests, and Cr levels measured at P140 (Fig. 7). By including Cr levels as a continuous variable, we were able to assess whether the identified circuits are sensitive to Cr modulation. We probed connectivity of the cingulate cortex because this region projects to multiple brain areas that are relevant to the tests examined, including ventral hippocampus for spontaneous alternation and striatal and mesolimbic areas for self-grooming.^20,47

Figure 7.

Multivariate modelling reveals disrupted brain-behaviour relationship in Slc6a8-knockout mice. Partial least squares (PLS) analysis revealed a significant latent component (LC1) that explained 40% of covariance between self-grooming, cingulate-striatal-insular network and creatine levels across mice. (A) Red/yellow colouring represents bootstrap ratio (BSR) z-scores obtained from 1000 bootstrapping iterations. The highlighted basal-forebrain and mesolimbic regions are those whose connectivity with the anterior cingulate exhibit robust covariance with behavioural performance and creatine levels as per the employed partial least squares (PLS) model. (B) Creatine and behaviour loadings for LC1. Columns indicate the contribution of each observed variable to the LC and the directionality of covariance. Error bars indicate 5th to 95th bootstrapping percentiles. Asterisk indicates a significant contribution to the overall PLS relationship for LC1 (relationship is non-zero, i.e. 95% confidence intervals do not encompass 0 loading). (C) PLS correlation between individual brain and Behavioural-Creatine scores for LC1, where the groups are represented by the symbols defined in the legend. The Behavioural-Creatine PLS variable indicates a latent dimension that captures shared variance between self-grooming, spontaneous alternation measures and creatine levels. Each symbol indicates an individual mouse. KO = knockout; WT = wild-type.

PLS analysis revealed one highly significant latent component (LC1; P = 0.002) accounting for 40% of the covariance between brain connectivity, behavioural performance and Cr levels. Regional mapping of the identified component revealed a prefrontal-basal-forebrain (i.e. insula, nucleus accumbens) circuit whose connectivity positively covaried with stereotyped behaviour, but not spontaneous alternation in WT mice (Fig. 7A and B). This result was consistent with the known organization of the distributed network of regions that control stereotyped activity in rodents⁴⁷ and suggested that the connectivity measurements we employed are behaviourally relevant. Interestingly, this brain-behaviour relationship was disrupted in Slc6a8-KO mice, where we found reversed covariance between behavioural scores and connectivity in the same network.

Importantly, Cr levels were also found to covary (and thus potentially modulate) the activity of this circuit, albeit with different directional effects in WT and Slc6a8-KO mice (Fig. 5B and C). In control mice, Cr levels were inversely related to the activity in these regions, suggesting that non-homeostatic levels of Cr can negatively affect this circuit to impact behaviour. In contrast, Cr levels were found to be positively related to behavioural performance in Slc6a8-KO mice, a finding consistent with the ameliorating effect of AAV-mediated hSLC6A8 expression.

These results uncover a behaviourally relevant prefrontal-basal-forebrain network whose activity is differentially modulated by Cr levels. This brain-behaviour relationship was disrupted in Slc6a8-KO mice, hence implicating aberrant prefrontal-mesolimbic connectivity in the pathology of CTD.

Discussion

Recent advancements in understanding the clinical manifestations and aetiopathological course of CTD^2,3,12 have not been paralleled by a deeper comprehension of the systems-level dysfunction and mechanisms underlying this genetic syndrome. To fill this knowledge gap, we have employed longitudinal functional imaging, behavioural phenotyping and genetic therapy in an established murine model of CTD to investigate how Cr depletion affects brain connectivity and leads to the circuit dysfunctions that characterize CTD. By using longitudinal mapping, we were able to track the longitudinal evolution of fMRI dysconnectivity in the same animal group, with minimal invasiveness and high precision. Our fMRI studies revealed robust functional hypoconnectivity in both juvenile and adult Slc6a8-KO mice, suggesting that Cr depletion leads to impaired large-scale interareal communication. We found fMRI hypoconnectivity mostly affected cortical regions during early pathological stages, subsequently evolving into more prominent subcortical hypoconnectivity (especially in striatal, mesolimbic and thalamic areas) in adulthood. Although fMRI hypoconnectivity may reflect multifactorial mechanisms,⁴⁸ our results are consistent with prior evidence of disrupted function and maintenance of synaptic circuits in Slc6a8-KO mice.⁸ Synaptic coupling is a key determinant of long-range synchronization underlying fMRI connectivity, and recent investigations have revealed fMRI hyperconnectivity in autism-relevant mouse models exhibiting aberrant synaptic signalling,²⁰ as well as fMRI hypoconnectivity in mouse models showing reduced synaptic density.^13,14 In this respect, our results corroborate the notion that brain connectopathy is a hallmark endophenotype of multiple neurodevelopmental disorders associated with synaptic pathology.^20,41

The observation of fMRI hypoconnectivity in Slc6a8-KO mice is also in agreement with our previous study showing that these mice exhibit a severe epileptic phenotype and significant changes in neural oscillations, with lower power of theta/alpha EEG frequencies and increased power of beta/gamma bands.⁹ Chemogenetic and pharmacological manipulations have indeed shown that fMRI hypoconnectivity can reflect reduced low-frequency EEG power and concomitant broad-band increased in higher-frequency activity as a result of increased excitability and asynchronous firing.^22,27 Thus, the effect of Cr deficiency on EEG power aligns with the observed reduction in fMRI connectivity (i.e. a measure of infraslow synchronization), potentially indicating a shift from slow, synchronized neural coherency to high frequency (yet asynchronous) activity. Similar spectral changes are present in the EEG of CTD children with respect to age-matched controls,⁹ underscoring the translational value of the present study.

From a cellular standpoint, these alterations may be linked to prior studies pointing to significant heterogeneity of Slc6a8 expression across cell populations^5,49,50 and the observation that synaptic alterations in CTD mostly affect GABAergic interneurons.^7,8 These alterations may in turn lead to a generalized reduction of cortical inhibitory tone and a disruption of neural circuitry efficiency, two key pathological determinants associated with multiple neurodevelopmental disorders.⁵¹ Given the crucial role of parvalbuminergic (PV) fast-spiking interneurons in the regulation of long-range functional synchronization,⁵² the presence of morpho-physiological dysfunction of PV synapses in Slc6a8-KO mice⁸ might represent a plausible cellular correlate of the observed fMRI desynchronization. Accordingly, previous studies have shown that alterations in the excitatory/inhibitory ratio may perturb brain oscillatory activity and fMRI connectivity, with potential specific contributions from different cell populations.^14,27

The putative translational relevance of these imaging findings is further corroborated by the observation of decreased inter-hemispheric connectivity in Slc6a8-KO mice at both P40 and P140. Given the robust anatomical foundations of functional connectivity²⁶ and the causal role of the corpus callosum in driving the synchronization of the two brain hemispheres,^53-55 these findings may be functionally linked to previous anatomical MRI studies revealing corpus callosum thinning in CTD patients.¹²

Our study also serves as a preclinical investigation of the therapeutic potential of early gene therapy in preventing the pathological and behavioural manifestations of CTD. The chronic progressive morbidity of CTD represents a significant unmet clinical need for which gene therapy could potentially offer a lifelong treatment option. In this respect, CTD is an ideal target for gene additive therapy for three reasons: (i) it is a monogenic condition; (ii) the replenishment of brain Cr is effective in ameliorating the clinical manifestations of two other disorders caused by alterations in Cr metabolism^56,57; and (iii) the expression of functional Slc6a8 and the severity of CTD phenotypes exhibit a dose-dependent relationship, with heterozygous females for Slc6a8 mutations showing moderate Cr reduction and partial cognitive deficits.^58,59 Interestingly, intracisternal delivery of an AAV2/9 vector carrying the rat Slc6a8 gene under a CMV promoter (targeting the cerebellum, medulla oblongata and spinal cord) prevented myocytic and locomotor impairments in a rat model of CTD but failed to improve other behavioural domains (e.g. cognition or stereotyped movements).⁶⁰ This result adds to the present findings to suggest that key aetiopathological phenotypes can be partly rescued by viral-mediated gene therapy in CTD.

In the present study, a single intraventricular infusion of AAV-hSLC6A8 in newborn Slc6a8-KO mice was sufficient to achieve widespread expression of the transgene across multiple brain regions. The observed transduction is broadly consistent with previously reported efficiency of AAV9-mediated vectors following intracerebroventricular administration,^61,62 resulting in a sustained and stable increase in cerebral Cr levels. Importantly, although Cr concentration in Slc6a8-KO brains remained approximately 50% lower than physiological levels measured in WT mice, postnatal reinstatement of CRT function led to partial or full rescue of key CTD-relevant phenotypes. These included functional brain hypoconnectivity in juvenile mice, reduced body weight, impaired declarative memory and the manifestation of autistic-like stereotyped behaviours in adulthood. The independent assessment of cognitive functions in two separate laboratories strengthens the robustness and reproducibility of these behavioural findings.

Interestingly, our PLS analysis linked these endophenotypes to the activity of a fronto-mesolimbic network whose function was disrupted in Slc6a8-KO mice, thus revealing a neural circuitry potentially central to CTD pathology and whose function was modulated by Cr levels in adulthood. These findings are consistent with previous studies showing that reinstatement of cerebral Cr metabolism via oral Cr supplementation can mitigate clinical symptoms in other neurodevelopmental syndromes associated with mutations of Cr synthetic enzymes.^63-65 However, a limitation of our PLS approach is that the identified latent variable represents a statistical combination of distinct behavioural measures rather than a single coherent behavioural domain. Future studies incorporating additional behavioural assessments are required to further refine the brain-behaviour associations reported here.

Rather unexpectedly, AAV-mediated delivery of hSLC6A8 did not fully restore cognitive performance in Slc6a8-KO mice, with significant improvements only observed in the ORT at P100. Several factors may account for this finding. Since optimal AAV expression requires about 2 weeks, our perinatal injection strategy might have failed to provide physiologically-relevant Cr levels during an early time window of critical importance for the maturation of higher-order cognitive functions.⁶⁶ Additionally, the viral titre employed in our study might not have reached the transduction threshold sufficient to fully prevent the onset of CTD pathological endophenotypes. Indeed, creatine levels in AAV-hSLC6A8-treated Slc6a8-KO mice remained well below the physiological concentrations observed in WT controls, suggesting that complete restoration of physiological creatine concentrations may be necessary to achieve normal cognitive functioning in CTD. In keeping with this, we recently reported a strictly dose-dependent therapeutic effect of cyclocreatine (a lipophilic creatine analogue capable of substituting its metabolic function) on cognitive defects and stereotyped behaviour in Slc6a8-KO mice, possibly reflecting distinct metabolic constraints of the different underlying brain processes.⁶³ Moreover, we cannot exclude the possibility that different cellular populations may require distinct physiological Cr levels for optimal function (e.g. PV interneurons⁸). If this is the case, ubiquitous transgene expression driven by JeT promoter might result in non-optimal (supraphysiological) expression of CRT in some cell populations and insufficient expression in others. This might also explain the partial impairment of mnemonic performance observed in AAV-hSLC6A8-treated WT mice at P40, an effect consistent with a potential detrimental effect of non-homeostatic CRT expression levels.

Finally, the discrepancy between the lack of efficacy in the Y-maze and the partial improvement observed in the ORT may reflect differing metabolic Cr demands across brain regions. Given that declarative memory primarily relies on prefrontal and perirhinal cortical circuits,⁶⁷ whereas the Y-maze probes hippocampal-dependent spatial working memory,⁶⁸ hippocampal circuits might require complete Cr replenishment to achieve full functionality. In contrast, prefrontal and cortical circuits may retain partial efficacy at lower Cr concentrations. These requirements could interact with, or be entirely driven by, differences in the maturation timeline of these circuits.^69,70 Further preclinical studies are warranted to optimize CRT expression to achieve physiological replenishment of Cr levels in the CNS of Slc6a8-KO mice.

We finally note that a limitation of this study is the lack of electrophysiological analyses of the epileptic phenotype. Due to the fragility of KO animals and the relative tardive onset of the epileptic phenotype, in our study we prioritized earlier timepoints and longitudinal behavioural studies with the aim to enable direct brain-behaviour correlations.

In conclusion, our results document translationally relevant, systems-level disruption of brain activity in a murine model of CTD and provide key proof-of-concept evidence that early gene therapy holds potential as a disease-modifying strategy for CTD. We expect these findings to help in the development of experimental therapies for this severe genetic disorder.

Data availability

All imaging data will be made publicly available to researchers upon acceptance for publication of our study.

Funding

This work was supported by the Telethon Foundation (GGP19177 to L.B. and A.G.). A.G. was also supported by Simons Foundation Grant (SFARI 400101), European Research Council (ERC—DISCONN, No. 802371, BRAINAMICS, No. 101125054). L.B. was also supported by project ECS_00000017 MUR Directoral Decree n.1055, 23 June 2022, CUP B83C22003930001, ‘Tuscany Health Ecosystem—THE’, Spoke 8, Italian Ministry of Reseach and University and Italian Ministry of Health, RC 2023, Intesa Sanpaolo Charity Fund.

Competing interests

The authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.