Introduction

In numerous motor system diseases, it remains unknown as to whether upper motoneurons (originating in the brain) or lower motoneurons (originating in the periphery) first begin to degenerate¹. Historically, neuroanatomical changes in regions like the cortex have been considered symptomatic—not causal¹. This captured our interest when we serendipitously observed progressive and severe motor deterioration in genetically modified Rbp4-Cre mice with chronically-silenced cortical layer 5 (L5) projection neurons (Figure 1A-C)².

L5 neurons comprise the primary outputs from the cortex to subcortical and spinal regions involved in voluntary movement. Rbp4-Cre labels glutamatergic pyramidal neurons which, compared to pan-glutamatergic and pan-neuronal drivers, preferentially target L5 pyramidal tract (PT) neurons and bypasses more-pial intratelencephalic (IT) neurons (Figure 1B). L5-silencing has been widely employed in developmental studies; we previously observed normal initial cerebral and subcortical projections yet subsequent decline in synapse maintenance, demyelination, and altered glial distributions and morphologies². However, L5-silencing has not been systematically evaluated for putative long-term effects on neuroanatomical and motor integrity.

Here we report the emergence of progressive motor deficits upon chronically-silencing L5 neurons in mice. To further characterise this, we initiated a comprehensive multimodal investigation integrating behavioural assays, in vivo and ex vivo magnetic resonance imaging (MRI), and immunohistology (IHC) to capture whole-brain changes over time (Figure 1D-E). Our goal was to determine whether L5-silencing induces neurodegeneration, and to identify the spatiotemporal progression of anatomical and functional changes. We also aimed to establish methodological robustness through multimodal data alignment, and to use this approach to reveal convergent evidence of degeneration across behavioural, imaging, and histological domains.

Methods

We conducted experiments throughout postnatal weeks 4-21 using male and female L5-silenced (Rbp4-Cre⁺::Snap25^fl/fl::Ai14^fl/fl, N=12) and control mice (Rbp4-Cre⁺::Snap25^fl|+/+::Ai14^fl/fl, N=18) (Tg(Rbp4-cre)KL100Gsat/Mmucd Cre-driver, B6-Snap25^tm3mcw mutation, and B6;129S6-Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze}/J Ai14 RFP). L5-silencing was achieved via Rbp4-Cre-Snap25 truncation, wherein cells develop normally yet postnatally cannot communicate via regulated synaptic vesicle release (Figure 1C)². Transcriptomic analyses employed Seurat-formatted gene expression matrices using Allan Brain Atlas open-source data to highlight pyramidal tract specificity in Rbp4-Cre vs pan-glutamatergic lines (Scl17a7-IRES2-Cre/wt;Ai14(RCL-tdT)/wt) (Figure 1B).

In vivo manganese-enhanced MRI scans were acquired using a Bruker 7T scanner (BioSpec 70/20 USR) equipped with a 4-channel receive-only cryoprobe. Anaesthesia was induced with 4% isoflurane and reduced to 0.5-1.4% following subcutaneous medetomidine bolus (0.05mg/kg) administration. Continuous infusion thereafter (0.04mL/h/g) maintained ~120 breaths/min³. A custom heating system kept body temperature at 36-37°C. Scans lasted ~2rs45mins, followed by subcutaneous atipamezole bolus injection for recovery³. T1W acquisitions (multi-echo gradient echo, TR=60ms, TE=3.28ms, flip angle=30°, matrix size=220×160×180, 1×average, acquisition time=70mins, isotropic resolution=75μm) were processed using Mouse Imaging Centre Build Model (MBM) tools⁴.

Behavioural experiments quantified motor performance and home-cage activity. The CatWalk system and XT software (v.10.6.608, Noldus Information Technology) evaluated gait parameters over time (e.g., average speed) alongside a grip strength meter (San Diego Instruments). Qualitative metrics for hindlimb clasp, ledge walk, gait, and kyphosis were combined to generate composite assessments of motor phenotypes (CAMP) scores. Feeding and activity data was obtained using custom-built SnackerTracker devices and MATLAB scripts (R2024b).

Ex vivo scanning following standard sample preparation⁵. T2-weighted images were acquired using 3D FSE (RARE) sequences (TEeff/TR=12/350ms, echo spacing=12ms, 6 echoes, resolution=60x60x60µm) and analysed using MBM tools⁴. dMRI involved segmented EPI (12 segments, TE=30ms, TR=400ms, δ/Δ=6.7/13.5ms, 4×b=0s/mm² volumes with 1×reverse-phase-encoded, 30dir/shell for b=2,500s/mm² and 10,000s/mm², 100x100x100µm resolution). Processing pipelines corrected for Gibbs ringing, eddy currents, and image distortion before generating volumetric and diffusion outputs^{4, 5}. Multi-echo GRE images (IES=3ms; 20 echoes; 2×repeats, resolution=100×100×100µm) were processed using R2* and QSM scripts⁵. Images were co-registered into standard space⁴ to evaluate genotype effects over time.

For IHC, tissues were embedded in agarose, coronally sliced at 50μm using a vibratome, and stained using five protocols (Figure 1E). Primary antibodies: #1 RbαNeuN-NBP1-92716/MsαSMI32-801701; #2 RbαChAT-ab178850/MSαSMI32-801701; #3 RtαMBP-ab7349/MsαAT8-MN1020; #4 RbαIba1-019-19741/CkαGFAP-ab4674; and #5 RbαPV-PV27/BtVVA-B12352. Secondary antibodies: #1-2 DkαRb488-A21206/DkαMs647-A32787; #3 GtαRt488-A11006/GtαMs633-A150115; #4 GtαRb488-A11034/GtαCk647-ab150171; #5 DkαRb488-A21206/Cy5SAV650-A1011. For each, red and blue channels were populated with Ai14 and DAPI respectively. Slides were imaged using an Evident VS200 Slide Scanner (20X, 0.32μm resolution) and analysed using ImageJ (v2020b) and QuPath (v0.4.2).

Statistical analyses employed mixed-models with genotype and age as fixed effects using GraphPad Prism (v.9.3.1). Behavioural data involved 2-way ANOVA, while general additive models evaluated neuroimaging data. Random intercepts and slopes accounted for within-subject correlations and between-subject variability where applicable for repeated measures. FDR was controlled for using the Benjamini-Hochberg method. Analyses were conducted blind to animal genotype, sex, and age.

Results

In vivo T1-weighted (T1W) analyses revealed that L5-silencing induces volumetric reductions in both Cre-expressing and target regions (Figure 2A). Decreases were observed in L5 projection tracts (e.g., corticospinal, q_geno<0.001) (Figure 2B) and target regions (e.g., striatum, thalamus, and brainstem, q_geno<0.01) (Figure 2A). These decreases remain statistically significant after covarying for overall brain volume, which is decreased in L5-silenced mice relative to controls (as well as overall body mass, p_geno<0.001). However, covariance reveals relative increases in regions of Rbp4-Cre expression (e.g., throughout the cortex and dentate gyrus, q_geno<0.001).

CL5-silencing led to progressive deterioration in motor performance (Figure 2E-G). While no behavioural test showed significant genotype effects at 4wks, CatWalk analysis revealed notable decline in L5-silenced mice at 21wks: speed decreased (F_geno(1,210)=278.6, p_geno<0.0001; F_age*geno(18,210)=3.009, p_age*geno<0.0001) (Figure 2F), cadence faltered (F_geno(1,210)=152.5, p_geno<0.0001; F_age*geno(18,210)=0.410, p_age*geno=0.9848), and step-patterns shifted to favour alternate over cruciate sequences (controls spend more time in cruciate, F_geno(1,210)=159.6, p_geno<0.0001; L5-silenced mice spend more time in alternate, F_geno(1,210)=144.3, p_geno<0.0001). Grip strength declined (p_geno=0.006, p_age*geno=0.0011) alongside CAMP scores (p_geno<0.0001, p_age*geno<0.0001) (Figure 2G), which revealed the clear emergence of a hindlimb clasp (p_geno<0.0001, p_age*geno<0.0001) (Figure 2Ei-ii) and inability to successfully balance and walk along a ledge (p_geno<0.0001, p_age*geno<0.0001) (Figure 2Eiii-iv). Home-cage recordings showed no change in activity patterns (p_geno>0.05), which is consistent with previous studies indicating intact circadian rhythms. However, feeding behaviour periodicity shifted in CL5-silenced mice, conceivably due to deteriorating fine motor skills which are required to reach for food.

Ex vivo diffusion analyses revealed decreased fractional anisotropy (FA) in projection tracts (e.g., internal capsule, q_geno<0.01) (Figure 2Ci), specific mean diffusivity (MD) decreases (p_geno<0.0001, q_geno>0.05) (Figure 2civ), and mean kurtosis (MK) increases in prefrontal cortical areas (p_geno<0.0001, q_geno>0.05) (Figure 2Cv), with greater axial than radial contribution to both. While not all assumptions for modelling various tissue compartments hold in grey matter for NODDI whole-brain analysis, brain maps show putatively increased intracellular volume fraction (ICVF) in L5 (p_geno<0.05) (Figure 2Cii) and increased orientation dispersion indices (OD) in L5, particularly in prefrontal and motor areas (p_geno<0.05) (Figure 2Ciii,vi). Ex vivo MGE image analysis identified trends towards decreased susceptibility (χ) in Rbp4-expressing regions (t>2) (Figure 2Di) and increased R2* in white matter tracts (t>2) (Figure 2Dii), which could collectively indicate the changes in calcium deposition or, more conceivably, demyelination.

Preliminary qualitative histological findings compliment behavioural and MRI insights. Rbp4-Cre Ai14 signals clearly show cell and axon degeneration in L5-silenced mice at 21wks (Figure 2H-I,Ji,Kiii), as well as in downstream lower motor neurons at 21wks which appear healthy at 4wks yet punctate at 21wks (Figure 2Jii,Kiv). Interestingly, this deterioration is not observed in all ChAT⁺ nuclei (e.g., dorsal tegmental and pedunculopontine nuclei)—only in brainstem sensorimotor (e.g., motor trigeminal) and spinal nuclei. No changes in SMI32⁺ neurofilament or MBP⁺ myelin IHC were qualitatively obvious, however AT8 immunoreactivity was observed in areas of Rbp4-Cre projection, most notably the cerebral peduncles and corticospinal tract, indicating pathological tau hyperphosphorylation within cortical projection neurites. Iba1 and GFAP staining showed increased microglial and astrocyte infiltration within the cortex, projection tracts, and spinal cord grey matter.

Conclusion

Our results strongly suggest that chronically silencing L5 neurons induces a cascade of progressive degenerative changes affecting both cortical and L5-projection regions, corroborating a top-down cortical contribution to motor system integrity. Our multimodal strategy enhances methodological confidence by observing parallel trends across independent experiments, enabling the characterisation of degeneration patterns that may otherwise remain undetected.

In vivo T1W MRI scans collectively indicate selective volumetric decreases in regions targeted by Rbp4-Cre neuron projections (Figure 2A-B), yet relative volumetric increases in regions of Rbp4-Cre neuron expression. Downstream deterioration may be due to histologically confirmed axon degeneration or activity-dependent effects on targeted brain regions, while cortical and hippocampal increases were unexpected given established Rbp4-Cre soma and neurite deterioration (Figure 2H-I). However, localised volumetric increases may be explained by compensatory neuroinflammatory mechanisms such as microglial and astrocyte infiltration observed via IHC.

Higher-resolution ex vivo MRI findings compliment in vivo results. Increased FA and decreased MD (Figure 2i,iv) typically indicate increased white matter organisation and restricted water movement along tracts. However, given that this is observed in the cortex, changes may rather indicate tissue damage via pathologically enhanced connectivity or compensatory reorganisation in L5-silenced animals. Neuroinflammation could contribute to this, especially given increased MK in Rbp4-Cre-expressiong regions which reflects the pattern of MD decreases (i.e., in the prefrontal cortex) (Figure 2Cv). Increased MK indicates that water diffusion is more restricted or complex (non-Gaussian) than expected, often corresponding with increased cellularity, barriers, or deviant microstructure. Increased OD in similar regions (Figure 2Ciii,vi) suggests that L5-silenced mice demonstrate an increased number of neurites which are more randomly oriented, particularly in the prefrontal cortex. Yet, increased pan-cortical voxelwise ICVF in what appears to be L5 (Figure 2Cii) suggests microstructural changes via an increased proportion of water molecules restricted within cell membranes, whether due to cytotoxic edema (which is unlikely given histological observations in Figures H-K) or hypercellularity (which is histologically supported).

The trajectory of volumetric changes over time largely reflects and even precedes the extensive motor decline observed behaviourally (e.g., qualitatively comparing Figure 2B to Figure 2F-G). Together alongside the microstructural changes detected ex vivo, which reflect patterns of histologically observed Rbp4-Cre neuron degeneration and glial infiltration, this suggests that neuroimaging may be used to identify structural or microstructural changes before behavioural effects materialise and without histological confirmation. Thus, while our findings corroborate that L5 neuron degeneration can precipitate progressive functional deficits, thereby supporting top-down theories of motor degeneration¹, our experimental strategy holds greater implications for the scope of advanced MRI in preclinical studies and, if expanded to human brain imaging and mapping more generally, clinical settings.

This multimodal framework—combining behaviour, MRI, and histology—provides a powerful means of studying neurodegeneration via methodological cross-validation. Clinically, our results suggest that early cortical involvement may underlie or exacerbate neurodegenerative disorders traditionally considered subcortical in origin, such as amyotrophic lateral sclerosis. Future research should explore the mechanisms by which cortical silencing initiates degeneration and test whether restoring cortical activity can mitigate motor decline.

Acknowledgement

MM was supported by the Rhodes Scholarship, Clarendon Scholarship, and Goodger and Schorstein Scholarship (University of Oxford). JO was supported by the Clarendon Scholarship and UKRI-BBSRC grant BB/T008784/1. GG was supported by the NIH NINDS pre-doctoral IRTA. CT was supported by Wellcome Trust grant 2027888/Z/16/Z and Linacre College (University of Oxford). AV and ZM were supported by a St John’s College Research Centre grant (University of Oxford). The work in ZM’s laboratory is additionally funded by MRC grant MR/W029073/1, BBSRC grant BB/X008711/1, and Einstein Siftung Berlin as part of ZM’s Einstein Fellowship at Charité Universitätsmedizin Berlin. VV was supported by BBSRC grant BB/X008711/1. The Centre for Integrative Neuroimaging was supported by core funding from the Wellcome Trust (203139/Z/16/Z and 203139/A/16/Z).

Reference

1. van den Bos MAJ, Geevasinga N, Higashihara M, Menon P, Vucic S. Pathophysiology and Diagnosis of ALS: Insights from Advances in Neurophysiological Techniques. Int J Mol Sci. 2019;20(11):2818. Published 2019 Jun 10. doi:10.3390/ijms20112818

2. Hoerder-Suabedissen A, Korrell KV, Hayashi S, et al. Cell-Specific Loss of SNAP25 from Cortical Projection Neurons Allows Normal Development but Causes Subsequent Neurodegeneration. Cereb Cortex. 2019;29(5):2148-2159. doi:10.1093/cercor/bhy127

3. Adamczak JM, Farr TD, Seehafer JU, Kalthoff D, Hoehn M. High field BOLD response to forepaw stimulation in the mouse. Neuroimage. 2010;51(2):704-712. doi:10.1016/j.neuroimage.2010.02.083

4. Chakravarty MM, Steadman P, van Eede MC, et al. Performing label-fusion-based segmentation using multiple automatically generated templates. Hum Brain Mapp. 2013;34(10):2635-2654. doi:10.1002/hbm.22092

5. Tisca C, Tachrount M, Lange F, Wang C, Qiu LR, Barallobre-Barreiro J, Fava M, Mayr M, Lerch JP, Martins-Bach AB, et al. Vcan mutation leads to sex-specific changes in white matter microstructure in mice. ISMRM & SMRT. 2013;34(10):2635-2654.