The generation of Klf10 KO mice has been previously described [22]. To be consistent with our previous studies performed only on female mice aged 3 months, we have also utilized the same type of 3 animals derived from heterozygous breedings. All mice were maintained in a temperature controlled room (22˚C ± 2˚C) with light/dark cycle of 12 hours. Animals had free access to water and were fed with standard laboratory chow ad libitum. The protocol was approved by the French ministry of higher education, research and innovation (Permit Number: DUO-4776), the local ethics committee Comité Régional d’Ethique en Matière d’Expérimentation Animale de Picardie (CREMEAP; Permit Number: APAFIS #8905-2021011109249708) and the local ethics committee Comité Régional d’Ethique en Matière d’Expérimentation Animale des Pays de La Loire (APAFIS #8186-2016121315485337) for behavior experiments.

WT (N = 3) and Klf10 KO (N = 3) mice were sacrificed by intracardiac perfusion of PFA (4% in PBS) followed by overnight post-fixation of the whole brain in a 4% PFA solution at 4˚C. 80 μm thick sagittal sections were prepared using a Leica VT1000S vibratome. Slices were permeabilized for 30 minutes in Permeabilization Buffer (PB: 1x PBS supplemented with 0.1% Triton X100 and 1% BSA), then incubated overnight with primary antibodies diluted in PB: Rabbit anti-MAP2 (Millipore, 1:1000) and Mouse anti-Neurofilament (BioLegend, clone [SMI-312], 1:1000), or Mouse anti-NeuN (abcam, clone [1B7], 1:500) and Rabbit anti-Parvalbumin (abcam, 1:500). The following day, sections were washed 3 times in 1x PBS and subsequently incubated in secondary antibody-containing PB (goat anti-mouse antibody, Alexa 488, 1:2000, life technology, and goat anti-rabbit antibody, Alexa 546, 1:2000, life technology). Nuclear DNA was stained using Hoechst 33258 (1:5000). Confocal images were acquired in 1024 × 1024 mode with a Nikon Ti-E microscope equipped with the C2 laser scanning confocal microscope. We used the following objective lenses (Nikon): 10× PlanApo; NA 0.45, 20× PlanApo VC; NA 0.75. Microscope control and image analysis was performed using the Nikon software NIS-Elements (Nikon).

MRI acquisition was performed on a 9.4T horizontal ultra-shielded superconducting magnet dedicated to small-animal imaging (94/20 USR Bruker Biospec, Wissembourg, France) and equipped with a 950 mT/m gradient set (Figure 1). A birdcage coil (35 mm inner diameter) was used for both proton transmission and reception.

Mice were anesthetized during the MRI experiment with 1.5% isoflurane and a mixture of O2/air (1:1) at an output of 0.5 L/min. Respiration was monitored during the entire experiment, and body temperature was maintained at 37˚C using a warm-water circulation system.

Axial and coronal images of the Klf10 KO (N = 10) and wild-type (WT) (N = 9) cerebellum were obtained to be quantified with texture analysis using a gradient echo (Flash) sequence with the following parameters: TE/TR = 6 ms/252ms, flip angle = 20˚, FOV size = 2 × 2 cm, matrix size = 256 × 256, bandwidth = 50 kHz, slice thickness = 570 µm, to display 78 × 78 µm² in plane resolution for a duration of 1 min (one accumulation). A region of interest (ROI) was chosen in the cerebellum and texture parameters (grey levels, run length, …) were analyzed with correspondence factorial analysis using our previously reported protocol [23].

Subsequently, a spin echo diffusion-weighted sequence was applied to 10 mice (Klf10 KO (N = 5) and WT (N = 5)) using the following parameters: TE/TR = 27/1500ms, flip angle = 90˚, FOV size = 2.5 × 2.5 cm, matrix size = 128 × 128, b values = 8, 200 and 400 s/mm² in x, y and z directions. Diffusion (D) coefficients Dx, Dy and Dz respectively in x, y and z direction were calculated with Bruker Paravision 5.1 software. They correspond to the slope of the decreasing line after the representation of ln(S/S₀) = f(b), where S₀ corresponds to the grey levels mean of the cerebellum ROI without diffusion and S the grey levels mean of the cerebellum ROI with diffusion.

For MRS acquisition, static B0 homogeneity was first adjusted with first and second order shims in a cubic (3.5 × 3.5 × 3.5 mm) voxel centered in the cerebellum with the Bruker Fastmap procedure [24]. The half-height linewidth achieved for tissue water was less than 18 Hz. A PRESS (Point Resolved Spectroscopy) sequence was used to record localized ¹H spectra in a (3 × 3 × 3 mm) voxel placed in fastmap cubic voxel with the following parameters (TR = 4 s, TE = 16 ms, 256 scan: 17 min, 2048 points, bandwidth = 4000 Hz) with water suppression using VAPOR (VAriable Pulse power and Optimized Relaxation delays) module and outer volume suppression [25]. Eddy current compensation and static magnetic field drift correction were applied during spectra acquisition.

MRS Spectra were analyzed with a time-domain method described in Ratiney et al. 2010 [26], using the CSIAPO software developed by Dr Le Fur (CRMBM, Marseille, France) and Dr Ratiney (CREATIS—INSA, Lyon, France) which is a spectral analysis tool for quantification algorithm QUEST (quantification based on quantum estimation).

QUEST is a linear combination model decomposing free induction decay (FID) signal in temporal domain. QUEST uses a database to fit a weighed combination of metabolite signals and then to obtain a quantification in arbitrary units (A.U.). This database is simulated through GAVA (GAmma Visual Analysis) which is used to model metabolite signals including their characteristics (chemical shift, amplitude and phase) and also experimental parameters (magnetic field and echo time).

The following cerebellum metabolites were then quantified: choline (Cho), creatine (Cre), glutamine (Gln), glutamate (Glu), myo-inositol (Myo), N acetyl aspartate (Naa) and taurine (Tau).

Prior to NMR analysis, the dried cerebellum samples (N_WT = 5 and N_{Klf10 KO} = 5) were reconstituted in 220 µL of deuterated buffer containing TSP (3-trimethylsilylpropionic acid) at 145 μmol·L⁻¹ and transferred to conventional 3 mm NMR tubes (CortecNet). As previously described, the extraction protocol of metabolites from cerebellum was adapted from the Folch type two-step procedure described by Wu et al. [27]. ¹H-NMR spectra were obtained with a Bruker DRX-600 AVANCE-III HD spectrometer (Bruker SADIS, Wissembourg, France), operating at 14 T, with a TCI cryoprobe. Standard water suppressed ¹H-NMR spectra were acquired at 298 K using a “noesypr1d” pulse sequence with relaxation delay of 20 s. Spectra were processed using Topspin version 3.2 software (Bruker Daltonik, Karlsruhe, Germany). As previously described [28], ¹H-NMR spectra were automatically reduced to ASCII files using AMIX Software package (Analysis of MIXture, version 3.9.14, Bruker Biospin, Karlsruhe, Germany). Spectral intensities were scaled to the total spectral intensity, and the resulting data table was analyzed by multivariate and univariate statistical analyses.

For multivariate analysis, Partial Least Square Discriminant Analysis (PLS-DA) was performed using SIMCA-P+ Software (version 13.0, Umetrics, Umeå, Sweden) on the NMR dataset containing WT (N = 5) and Klf10 KO (N = 5) cerebellum values. All data sets were scaled to unit of variance allowing all metabolites to become equally important. The PLS-DA is a classification and prediction method that allows the identification of spectral features (metabolites) that define the separation between experimental groups (phenotypes). Variable Importance in the Projection (VIP) > 1 were considered as the most contributive in the phenotype separation.

For univariate analysis, Student’s t-test was performed using MetaboAnalyst [29] for all variable importance on projection (VIP). A p-value < 0.05 was considered as significant.

Heatmaps were generated to allow for visualization of data. Each colored cell on the map corresponds to an intensity value in our dataset. Considering the VIP > 1 and a p-value < 0.05, a heatmap was displayed for the cerebellum.

The enzyme activities of citrate synthase (CS), mitochondrial NADH: coenzyme Q oxidoreductase (Complex I) and cytochrome oxidase (COX, Complex IV) were evaluated. Briefly, cerebellum tissues (N_WT = 6, and N_{Klf10 KO} = 6) were harvested and rapidly frozen in azote solution before storage at −80˚C. Subsequently, the extraction of enzymes were performed with an ice-cold buffer (50 mg·ml⁻¹; containing (in mM): HEPES 5 (pH 8.7), EGTA 1, dithiothreitol 1, and 0.1% Triton X-100) using a Bertin Precellys 24 homogenizer (Bertin, Montigny-le-Bretonneux, France) and the enzyme activities were measured using standard spectrophotometric protocols [30,31].

The in vivo tests were blindly performed in the same sequence for each mouse, with equivalent time of rests in between, and at the same time of the day. Mice were randomly assigned to 2 groups according to genotype Klf10 KO (N = 10) and WT (N = 7) as described above.

At first, we performed histological analyses of mouse brains (Figure 2). There was no change overall of the size and structure of the cerebellum, as the different lobules could be identified. We marked cerebellar neurons with the pan-neuronal marker NeuN, as well as the Purkinje-neurons marker Parvalbumin (Figure 2(A), Figure 2(B)). Overall we saw no change in the granular (gr), ganglionic (gg) and molecular (ml) layers organization in Klf10 KO mice compared to control littermates (Figure 2(A’), Figure 2(B’)). Layers appeared similar in size and a single layer of Purkinje neurons was clearly visible. In parallel, dendrites (MAP2) and axons (neurofilaments: SMI312) were stained. Confocal analyses revealed a normal segregation of the somatodendritic and axonal compartments (Figure 2(C), Figure 2(D), magnification in Figure 2(C’), Figure 2(D’)). Collectively, our results suggest that the global architecture of the cerebellum is normal in Klf10 KO animals. A future direction of the morphological analysis would be to examine the cellular level of the structure using higher resolution confocal microscopy or electron microscopy.

Figure 3 illustrates the results of the multiparametric analyses of the cerebellum texture profiles obtained for WT and Klf10 KO mice from correspondence factorial analyses (CFA) and dendrogram representations. Both results reveal 2 distinct groups of texture profiles as a function of mouse genotype. Hierarchical ascending classifications (HAC) reveal that class I (CI) has a majority of WT cerebellum (N = 8) compared with class II (CII), which has the most Klf10 KO cerebellum (N = 9). The most discriminating (p < 0.05) texture parameters were mean, run length distribution (0˚, 45˚, 90˚, 135˚) and grey level distribution (0˚, 45˚, 90˚, 135˚). These results demonstrate the differing texture profiles between WT and Klf10 KO cerebellum. The global values for the cerebellum texture analysis were found to be 89% (WT vs. Klf10 KO).

The diffusion coefficients measured within the WT and KO cerebellum for each direction are illustrated in Figure 4. Only the coefficient of diffusion (Dy), calculated in y direction, was significantly (p < 0.01) different as a function of genotype. The extreme, the mean (red cross), and the median (solid line in the box) values of Dx, Dy and Dz were plotted as a function of mouse genotype.

In vivo acquisition of the cerebellum metabolite is shown in Figure 5 with box plots representing the extreme, the mean (red cross), and the median (solid line in the box) values of the 7 metabolites as a function of mouse genotype. The amount of myo-inositol decreased significantly (p = 0.009) in Klf10 KO cerebellum compared to the WT mice. The remaining metabolites (choline, creatine, glutamine, glutamate, N acetyl aspartate, taurine) showed no significant difference between the WT and KO mouse cerebellum.

The PLS-DA model obtained from the NMR dataset containing the WT and Klf10 KO cerebellum (Figure 6) revealed a significant separation as a function of mouse genotype (predictive ability Q² = 0.7, goodness of fit R²Y (cum) = 0.76). The PLS-DA distribution confirms differences between WT and KO metabolic profiles.

The PLS-DA permitted the clustering of WT vs KO using 25 variables. Figure 7 depicts a heatmap generated with the top 25 features ranked by the PLS-DA models. Over these 25 variables, the PLS-DA model allows for the identification of 9 variables of importance in the projection (VIP > 1) demonstrating the biological relevant changes in the metabolome of the compared groups (Table 1). Deletion of Klf10 induced a significant up regulation of two metabolites (lactate, nicotinurate) with a trend for inosine + adenosine region and NAD+, and the significant down regulation of five metabolites (tyrosine, valine, isoleucine, alanine) with a trend for NAA + aspartate region.

NAD: Nicotinamide adenine dinucleotide. NAA: N-Acetylaspartic acid. NS: No significant.

Significant increase (P < 0.01) in the enzymatic activities for the citrate synthase (CS), the NADH: ubiquinone oxidoreductase (Complex I) and the cytochrome oxidase complex (COX, Complex IV) were detected in Klf10 KO cerebellum compared to WT cerebellum (Figure 8).

We have observed that clasping reflex experiments showed no significant difference between Klf10 KO (N = 10) and WT (N = 7) mice and the score equivalent to zero showed normal reflex retraction of the legs and paws. These results are in good agreement with the analysis of the spontaneous locomotion that showed no significant difference between the WT and Klf10 KO mice (Figure 9). With regard to these measures, there is no apparent effect on behavior in Klf10 deficient animals. An interesting result was observed using the grip test measurements. As shown in Figure 9, Klf10 KO mice showed lower grip strengths compared to WT mice.