Comparative proteomic profiling reveals mechanisms for early spinal cord vulnerability in CLN1 disease

CLN1 disease is a fatal inherited neurodegenerative lysosomal storage disease of early childhood, caused by mutations in the CLN1 gene, which encodes the enzyme Palmitoyl protein thioesterase-1 (PPT-1). We recently found significant spinal pathology in Ppt1-deficient (Ppt1−/−) mice and human CLN1 disease that contributes to clinical outcome and precedes the onset of brain pathology. Here, we quantified this spinal pathology at 3 and 7 months of age revealing significant and progressive glial activation and vulnerability of spinal interneurons. Tandem mass tagged proteomic analysis of the spinal cord of Ppt1−/−and control mice at these timepoints revealed a significant neuroimmune response and changes in mitochondrial function, cell-signalling pathways and developmental processes. Comparing proteomic changes in the spinal cord and cortex at 3 months revealed many similarly affected processes, except the inflammatory response. These proteomic and pathological data from this largely unexplored region of the CNS may help explain the limited success of previous brain-directed therapies. These data also fundamentally change our understanding of the progressive, site-specific nature of CLN1 disease pathogenesis, and highlight the importance of the neuroimmune response. This should greatly impact our approach to the timing and targeting of future therapeutic trials for this and similar disorders.

. Early and progressive spinal cord pathology in Ppt1 −/− mice. Stereological analysis of regional volume (A) in the spinal cord reveals significant reduction in total, grey and white matter in whole Ppt1 −/− cords at 3 and 7 months of age compared to wildtype (WT) controls. Thresholding imaging analysis (B) of sections stained for microglia (CD68) and astrocytes (GFAP) show a significant increase in both markers in the dorsal and ventral horns of the lumbo-sacral cord at early (3MO) and late (7MO) disease states in Ppt1 −/− cords compared to WT. Counts of neuron number in the lumbo-sacral cord (C) revealed a significant loss of Ppt1 −/− mouse spinal interneurons stained with calbindin and calretinin in laminae IV-IX as early as 3 months of age, compared to WT. However, spinal motor neurons in the ventral horns (CD-71) were only significantly lost at 7 months of age in Ppt1 −/− mouse spinal cords. Scale bars 200 µm. p values-**p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 0.0001; multiple two-tailed, unpaired, parametric t test with Bonferroni-Dunn correction. Values shown are mean ± SEM. (n = 5 mice/group).

Scientific RepoRtS
| (2020) 10:15157 | https://doi.org/10.1038/s41598-020-72075-7 www.nature.com/scientificreports/ Together, these data establish the spinal cord as a particularly vulnerable region of the CNS in CLN1 disease, with significant pathological changes evident earlier than the brain regions we have previously characterized [11][12][13] . The loss of interneurons before motor neurons is indicative of these being vulnerable neuronal populations, as seen in the forebrain 12 . However, the lack of rostro-caudal differences or differences between the dorsal and ventral horns, shows there to be a more widespread and less regionally selective pathology in the spinal cord.

Quantitative proteomic profiling of whole tissue lysates spinal cord and cortex at early and late stages (3 and 7 months).
To further characterize the changes occurring in the spinal cord in CLN1 disease, we processed whole wildtype and Ppt1 −/− cortical and spinal cord extracts using Tandem mass tagging (TMT) based quantitative proteomics 16,17 . We analyzed 6 groups in detail-3 month cortices and spinal cords as well as 7 month spinal cords from WT and Ppt1 −/− mice respectively ( Fig. 2A, https ://doi.org/10.7488/ ds/2750 (2020)). The 7,970 protein raw data outputs from LC-MS were filtered to include proteins with ≥ 2 peptides identified, considered more reliable observations 18 . These proteins were then cross-referenced with mouse (Mus musculus) protein sequences from UniProtKB/Swiss-Prot using the MASCOT search engine (Matrix Science, Version 2.2) through Proteome Discoverer (Version 1.4, ThermoFisher) for a total of 7,160 identified proteins (https ://doi.org/10.7488/ds/2750 (2020)) ( Fig. 2B). We then used the PANTHER gene ontology database 19 (Fig. 2C) and the Database for Annotation, Visualization and Integrated Discovery (DAVID) 20,21 to confirm that the list of proteins identified was representative of whole tissue lysates and not enriched in any particular sub-cellular compartment (Supplementary David Ontology File). LC-MS data for markers shown to be altered from spinal cord immunohistochemistry ( Fig. 1) were identified showing an increase in CD68 ratios. However, GFAP was not greatly increased in the 3-month-old spinal cord ratios, presumably due to its high expression in the white matter of the cord, even in WT tissue, but this increased at 7 months, and was also increased in the cortex at 3 months. There was little change in the expression of interneuron markers calbindin and calretinin, www.nature.com/scientificreports/ possibly due to the relatively low abundance of cells expressing these markers (Supplementary Table S1). We also validated these data using quantitative fluorescent western blotting (QFWB) of representative proteins from samples used for LC-MS (Supplementary Figure S2). Representative proteins that showed increased, decreased or relatively unaltered Ppt1 −/−/ WT abundance ratios, together with markers relevant to CLN1 disease pathology were chosen to appropriately validate the LC-MS data, including markers for synaptic proteins 22 , astrocytes, interneurons 11,12 and for oligodendrocytes (MBP). Overall our QFWB analysis showed similar trends to the predicted LC-MS data (Supplementary Table). SNAP25 which was significantly downregulated in the spinal cord as compared to the cortex, and synaptophysin showed little change across all groups. Little overall change in protein abundance was also observed in QFWB data for calbindin and cytochrome oxidase subunit IV (COX IV), consistent with LC-MS data (Supplementary Figure S2). These synaptic markers, calbindin and COX IV all showed an overall increased abundance in the cortex as compared to the spinal cord, as is expected given the higher density of cells expressing these markers in the cortex 15,23 . We also probed astrocyte markers-GFAP was significantly up regulated in the cortex at 3 months and spinal cord at 7 months, with only a trend to increased values at 3 months in the spinal cord, while glutamine synthetase did not show any significant change across groups, again consistent with the LC-MS data (Supplementary Figure S2). Lastly, Myelin basic protein (MBP) was more abundant in the spinal cord as compared to the cortex, which contains less white matter, but was decreased in 3 and 7 month Ppt1 −/− cords (Supplementary Figure S2). Functional proteomic analysis was carried out by filtering for proteins with increased stringency-either including those proteins with 1.2 fold (20%) expression change in Ppt1 −/− tissues, compared to wildtype 18,24 or by BioLayout Express 3D to cluster proteins that show similar changes in expression [16][17][18][25][26][27][28] (Fig. 2B). These data were then analyzed using the ingenuity pathway analysis (IPA) (Ingenuity systems) software to reveal the cellular pathways that may be altered in these different regions 16,18,[26][27][28] .  Table S2), which are anticipated given the intrinsic link between lysosomal function and autophagy [29][30][31] and the significant microglial activation observed by immunohistochemistry (Fig. 1). Furthermore, interferon signalling (p = 4.48E−05), antigen presentation (p = 5.86E−05) and T-lymphocyte mediated apoptosis (p = 4.08E−04) were also affected indicative of a pronounced humoral immune system involvement at this earlier time point (Fig. 3). IPA software was utilized to determine which cellular networks are most affected by interrogating curated list of known networks of protein interactions. This analysis further highlighted the significant inflammatory response, as the most enriched networks involved-Humoral Immune Response, Inflammatory Response and Nutritional Disease (Fig. 3, Supplementary Table S2, IPA-Score = 52, 28 focus molecules). The next most enriched networks were Cell Morphology, Cellular Assembly and Organization (IPA-Score = 42, 24 focus molecules) (Fig. 3, Supplementary Table S3) and Cell-To-Cell Signalling and Interaction, Nervous System Development, Cell Morphology (IPA-Score = 35, 21 focus molecules). The top predicted upstream regulators also included a predicted activation of interferon gamma (p = 8.37E−12, activation z-score (z) = 4.11) and Interferon alpha/beta receptor (IfNAR, p = 6.51E−14, z = 3.831), which are linked with a humoral immune response. There was also a predicted activation of STAT1 (p = 2.25E−06, z = 3.795), interferon regulatory factor 7 (IRF7, p = 8.85E−10, z = 2.837) and inhibition of TRIM 24 (p = 3.33E−13, z = − 3.819). These predicted regulatory alterations emphasize the findings of affected canonical pathways and highlight the extent of inflammatory changes in the Ppt1 −/− spinal cord at 3 months of age. Using the IPA analysis to further dissect which pathological disease processes are most affected, we found expected changes in lysosomal proteins (p = 8.98E−12, z = − 2.193), with other NCL proteins being upregulated as previously described 32,33 (Fig. 4). Other disease-associated processes included accumulation of lipid (p = 1.  Table S4).

Significant neuroimmune and inflammatory responses in
Overall, analyzing proteins whose expression changed by 1.2-fold or more at 3 months in Ppt1 −/− spinal cord revealed significant inflammatory changes, particularly a large humoral immune response. While the effect of infiltrating lymphocytes on the Ppt1 −/− mouse brain has been well documented 34,35 , this only occurred much later in disease progression.
To validate this proteomic finding of comparatively early spinal peripheral lymphocyte infiltration, we immunostained WT and Ppt1 −/− spinal cords sections at 3 and 7 months simultaneously for peripheral lymphocyte markers CD4 and CD8 35 . This staining showed positive lymphocytes in in the grey and white matter of Ppt1 −/− spinal cords, even at the early 3 month timepoint (Supplementary Figure S1), confirming the changes seen in the proteomic analysis. This is significantly earlier than similar changes are observed in the brains of these mice 35 . Furthermore, proteomic analysis revealed protein changes associated with abnormal myelination, which we confirmed by QFWB analysis of MBP (Supplementary Figure S2), as well as other pathophysiological mechanisms including nervous system development and movement disorders that have not been previously described for this disease. These serve to not only demonstrate the relatively early onset of pathology in this region of the CNS, but may also help understand the cause of such vulnerability.  Figure S3, Supplementary Table S10). In order to compare affected pathways between 3 and 7 month timepoints, we used IPA software to generate a comparative heatmap of altered canonical pathways revealing many more affected pathways with disease progression (Fig. 5A). While this might be expected, several processes including interferon signalling (3 month z = 2.236, 7 month z = 1.0), Production of Nitric Oxide and Reactive Oxygen Species in Macrophages (3 month z = 2.0, 7 month z = 0.78), Protein Kinase A signalling (3 month z = 0.81-, 7 month z = − 1.22) and TEC Kinase signalling (3 month z = 1.63, 7 month SC z = 0.94) were all unexpectedly more pronounced at 3 months of age. These data indicate that there is not simply a worsening of common phenotypes with progressing disease, but rather that there are certain pathological processes that are more florid at early symptomatic stages in the Ppt1 −/− cord, and whose involvement decreases with age. Studying those proteins whose expression ratio changed by 1.2-fold (20%) or more highlights the most significant pathological changes at each time point. However, to dissect which of these processes are distinct or common between 3 and 7 months in Ppt1 −/− cords, we also analyzed these datasets using BioLayout Express 3D [16][17][18][25][26][27][28] . We input the expression ratios (Ppt1 −/− /WT) of 7,160 proteins from LC-MS without a filtering cutoff for 3 and 7 month-old spinal cords (Fig. 2B). Individual protein clusters were grouped based on trends of expression as those that showed more changes at (a) 3 months (2,096 proteins); (b) 7 months (3,159 proteins); or (c) clusters that showed progressive or similar changes between timepoints (1,541 proteins) (Fig. 5B). IPA software was then used for a functional cellular pathway analysis of each group.  Table S11). There were also changes involving development of the CNS (p = 1.23E−6, z = − 2.035) and development of the synapse (p = 3.63E−08, z = 0.451).
Further dividing the BioLayout 3D clusters predominantly altered in the 3 month Ppt1 −/− spinal cord into those proteins that showed increased (993 proteins) or decreased expression (1,103 proteins) revealed that while mitochondrial dysfunction and oxidative phosphorylation changes were the main canonical pathways associated with increased expression in the 3 month Ppt1 −/− spinal cord in these protein clusters (p = 8.43E−24, 3.39E−18, respectively) (  Table S6). These findings are of particular importance as they highlight early and distinct disease-related changes in the CLN1 spinal cord that decrease with age. These changes are possibly indicative of either altered postnatal maturation processes 32 , and/or increased cell turnover due to cell loss in the 3 month old Ppt1 −/− spinal cord (Fig. 1).
Similarly, IPA analysis of BioLayout 3D clusters with progressive or similar changes at 3 and 7 months showed RAC signalling, Axonal guidance, Rho family GTPase signalling, Stathmin1 regulation of breast cancer and ARP-WASP complex1 to be significantly altered (p = 7.09E−10, 1.19E−09, 2.91E−09, 3.84E−08, 1.03E−07, respectively) (Supplementary Figure S4B, Supplementary Table S11). Interestingly, these clusters were primarily comprised of   Figure S4C, Supplementary Table S12). Together, our comparative proteomic analyses reveal multiple pathological processes occurring in the Ppt1 −/− mouse spinal cord at 3 months, highlighting its significant involvement relatively early in disease. Interestingly, dissecting up-and down-regulated processes revealed the involvement of distinct early disease-related changes including mitochondrial dysfunction, cancer pathways, mTOR signalling and developmental processes. While these have previously been shown to be linked to PPT1 function in vitro 33,36 , and our data confirms this alteration in vivo for the first time at a significantly earlier timepoint than previously predicted. Early involvement of these cellular processes that play a role in spinal cord development and/or function may further help explain spinal cord vulnerability in CLN1 disease. BioLayout Express 3D clustering representation of proteomic expression data for 3-and 7-month-old Ppt1 −/− spinal cords, and a normalised WT group. Each sphere or node represents a single protein, and edges or lines represent the relatedness of proteins to one another, in this case, based on similarity in expression profile. Applying the Markov Clustering Algorithm in BioLayout Express 3D clusters these nodes based on similarity in expression profile. As a result, proteins exhibiting similar expression trends to one another are not only grouped closely in spatial proximity within the graph but also are coloured identically to represent membership within the same cluster. The expression profile behind each cluster can then be individually analysed to identify and isolate those that show more change at 3 months of age, 7 months of age or those that show a similar/progressive change. www.nature.com/scientificreports/ Comparing Ppt1 −/− cortical and spinal cord changes highlights conserved and divergent mechanisms. Although significant pathology is already underway in their spinal cord at 3 months, the cortex of Ppt1 −/− mice shows no significant glial activation, accumulation of storage material or neuron loss at this timepoint 11,12 . Therefore, having elucidated various altered cellular processes associated with spinal cord pathology, we compared the proteomic changes that occur in the spinal cord at 3 months with changes seen in the cortex at this age. Despite the obvious cytoarchitectural and biochemical differences between these tissues, such a comparison would reveal common and distinct mechanisms responsible for regional CLN1 disease progression. We first filtered proteins from the cortex at 3 months that showed a 1.2-fold (20%) change in Ppt1 −/− tissue. This yielded 705 proteins (343 up-and 362 downregulated) (compared to 249 proteins with a similar change in the 3 month cord) (Fig. 2B). While this may be due to greater coverage of cortical proteins with LC-MS and IPA analysis, there was a significant overlap of affected cellular processes between the cortex and spinal cord.  Table S14). These data highlight the great similarity in the most affected pathways in the cortex and spinal cord, with the cortex showing a higher expression fold change. The degree of change in cortical protein expression was particularly striking as there is little histologically detectable pathology observed at this stage in this brain region 11,12 . However, a marked exception to this broad similarity between regions was the degree of altered inflammatory pathways, especially the humoral immune response pathways, revealing a potential cause for spinal cord  Table S8). Many of these processes were also shown to be altered in the 3 month spinal cord, but to a lesser extent, highlighting the common pathological processes between these regions. Analyzing protein clusters that showed more changes in the spinal cord revealed affected canonical pathways-Dopamine-DARPP32 Feedback in cAMP Signaling (p = 7.91E−10), Breast Cancer Regulation by Stathmin1 (p = 2.06E−08), CREB Signaling in Neurons (p = 2.35E−08), CDK5 Signaling (p = 5.26E−08), GNRH Signaling (p = 9.36E−08), indicating that these cell signaling pathways may be uniquely dysregulated in the spinal cord (Fig. 7C, Supplementary Table S10).We then analyzed clusters for similar protein expression changes between the spinal cord and cortex, revealing common pathways-Synaptic Long Term Potentiation (p = 6.10E−07), G Beta Gamma Signaling (p = 1.05E−06), Opioid signaling Pathway (p = 6.19E−06), CREB Signaling in Neurons (p = 1.11E−05) and adrenergic Signaling (p = 2.46E−05), further highlighting the synaptic defects observed early in disease in these mice 34,35 , also revealing novel cell-signalling pathways (Supplementary Figure S6, Supplementary Table S15). www.nature.com/scientificreports/ Overall, these analyses show similarities in the cellular pathways impacted by disease in the cortex and spinal cord of Ppt1 −/− mice. Despite a lesser extent of histopathological phenotypes, the cortex showed a greater degree of change in terms of protein expression in these shared pathways as compared to the spinal cord. We cannot discount the differences in cytoarchitecture or cellular function between the cortex and spinal cord as influencing these data. Nevertheless, the extent of overlap of changes seen between the cortex and spinal cords of Ppt1 −/− mice demonstrate the significance of cellular processes that precede neurodegeneration. However, the most conspicuous difference we detected was the degree of inflammatory changes, particularly humoral immune response in the Ppt1 −/− spinal cord as compared to the cortex at this relatively early stage of disease, in addition to revealing the presence of unique cell signalling pathways within the spinal cord.

Discussion
CLN1 disease is a rapidly progressing form of NCL with no curative therapy available 1,5 . Despite the success of enzyme replacement therapy for CLN2 disease 37 , brain-directed therapies have not been as effective for CLN1 disease 1,5 . We recently showed significant spinal pathology in CLN1 disease patients and Ppt1 −/− mice, which contributes to disease outcome and unexpectedly precedes the onset of brain pathology 9 . Therefore, in this study, we firstly characterized the nature and progression of CLN1 disease in the spinal cords in more detail, showing significant early changes in regional volume, glial activation and interneuron number in these mice at 3 months of age. We then elucidated disease mechanisms that may explain the regional differences in the CNS of Ppt1 −/− mice.
Histopathological analysis of Ppt1 −/− mouse spinal cords at 3 and 7 months of age revealed progressive changes in regional volume that affect both the grey and white matter, a significant activation of astrocytes and microglia, and a selective loss of interneuron populations before motor neurons. This spinal pathology affected both dorsal and ventral horns and also affected different levels of the spinal cord to a similar extent.
The mouse spinal cord is fully developed by 3 months 38 . However, as the Ppt1 −/− spinal cord is already smaller in volume at this age, it raises the possibility that the spinal cord may not have developed normally. Whether this is due to defects in embryonic development or post-natal maturation is yet to be determined and may help explain this regional vulnerability.
The activation of microglia and astrocytes is a more accurate predictor of neuron loss in the NCLs than the accumulation of storage material 12,39 . Given the inherent defects of Ppt1 −/− mouse cortical astrocytes and microglia in vitro and their impact on neuron survival 40 , as well as the importance of neuroinflammatory changes in other neurodegenerative pathologies 41,42 , our data provide evidence for a distinct pathogenic role of glial activation in regional vulnerability in CLN1 disease. However, this awaits further experimental validation in vivo and it will be important to characterize the inflammatory networks involved in the early pathogenesis of CLN1 disease. To achieve this goal a comprehensive analysis of early cytokine and chemokine changes, in addition to defining the activation states of astrocytes and microglia and the role they play in regional neurodegeneration is currently ongoing.
The Ppt1 −/− mouse brain shows also selective neuron loss, with interneurons and large complex neurons such as lamina V cortical neurons and cerebellar Purkinje cells particularly vulnerable 12,13 . While Ppt1 −/− mouse spinal cords show significant overall neuron loss at 3 months 9 , our data now show spinal interneurons are lost significantly before motor neurons in the Ppt1 −/− spinal cord. To date this is the earliest significant cell loss seen in these mice, highlighting the severity of spinal pathology and revealing a different pattern of cellular vulnerability to other NCLs 43 and to diseases such as amyotrophic lateral sclerosis (ALS) and other motor neuron diseases (MNDs) [44][45][46][47] . Therefore, it will be critical to study the precise relationship between glial activation and neuron loss in the CLN1 disease spinal cord, and investigations into neuron-glial interactions in this especially vulnerable region of the CNS are currently ongoing. Together, our histopathological characterization has emphasized the extent and significant progression of spinal pathology in the cord of Ppt1 −/− mice, that precedes brain pathology and has identified spinal interneurons and glia as being particularly involved early in this disease.
We next undertook a comparative proteomic investigation of Ppt1 −/− and wildtype mouse spinal cords to investigate the mechanisms driving spinal cord vulnerability in CLN1 disease. Our comparison of proteins altered in 3 month old cords revealed significant alterations in autophagy, phagocytosis and neuroinflammatory processes. In particular, there was a profound humoral immune response in the Ppt1 −/− spinal cord which is progressive until 7 months. The status of the blood brain barrier (BBB) has previously been detailed in CLN1 disease 35,48 , with CD4 and CD8 + ve lymphocytes, seen infiltrating the Ppt1 −/− mouse optic nerve and brain as early as 3 months 35 . Taking into account our proteomic findings, we validated these findings by immunostaining 3 and 7 month spinal cord sections at the same time for both CD4 and CD8 + ve lymphocytes. This revealed increased infiltration of these cells at 3 months of age, significantly earlier than similar changes occur in the brains of these mice 35 . Unlike the BBB, the blood-spinal cord-barrier (BSCB) is relatively more porous and therefore potentially more vulnerable to lymphocyte infiltration [49][50][51] . However, the BSCB has yet to be studied in any form of NCL, and our data suggest performing a detailed investigation. Interestingly, the expression of PPT1 is most abundant in CD8 + ve dendritic cells and microglia in mice, with comparable expression in humans (www.biogp s.org (2020)) 52 . Therefore, as we have documented inherent defects in microglia and astrocytes 40 , peripheral blood lymphocytes may be also vulnerable to CLN1 disease and require further characterization.
We also demonstrated alterations in a variety of cellular processes including mitochondrial dysfunction, oxidative phosphorylation and TCA Cycle function, EIF2, EIF4 and p70SK, mTOR pathway. These alterations have previously been described in CLN1 disease either in vitro in neuroblastoma cells 33 and cancer cell lines 36 or in a CLN1 mouse model that bears a different disease-causing mutation (Ppt1 △ex4 ) 32 . Our data confirm these shared pathological processes also occur in the more commonly used Ppt1 −/− mice, and that these pathological changes are ongoing at this relatively early time point in the spinal cord. Despite having specific disease-causing mutations and the unique pathomechanisms discussed here, the Ppt1 −/− mouse cortex and spinal cord also show alterations www.nature.com/scientificreports/ in biological processes that are significantly affected in a host of neurodegenerative conditions [53][54][55][56] . Such common pathogenic mechanisms may prove critical in understanding the downstream cellular consequences of lysosomal pathology and may also prove to be amenable to therapeutic intervention. Given the regional vulnerability of the spinal cord, cell populations within the cord may be more vulnerable to alterations in these processes than other cell types, however, this awaits experimental confirmation. Given marked differences in the onset of histologically detectable pathology in the spinal cord and cortex of Ppt1 −/− mice 9,12 , proteomic profiling revealed surprisingly similar changes in protein expression in these two CNS regions at 3 months of age. Despite the cortex being relatively spared in terms of glial activation and neuron loss at 3 months, there were many similar processes altered in this brain region as in the spinal cord and many of these were to a greater extent than in the cord. While this may be in part due to inherent differences in cellular and biochemical constitution, these data reiterate the similarity in disease processes across the CNS regions and shift the focus upon regional and cell-type vulnerability to these alterations. However, the most conspicuous difference between the Ppt1 −/− mouse cortex and spinal cord at 3 months was the remarkable extent of immune response present in the cord at this early stage of disease, and work into characterizing the exact nature and progression of these changes in currently ongoing. As already discussed, whether this is due to inherent cell-type defects or as a reaction to regional cues is yet to be determined, but these data help us understand the cause of such regional differences in vulnerability. Furthermore, our data raise the possibility of spinal cord targeted regional therapeutic interventions such as anti-inflammatories or neuroprotective agents 57,58 to improve disease outcome.
Taken together, this study provides a thorough characterization of early symptomatic spinal cord pathology in CLN1 disease, and using comparative proteomic profiling, elucidates significantly affected cellular mechanisms that contribute to this early regional vulnerability. In particular, we highlight spinal interneurons and glial cells as the earliest observed vulnerable cell populations. Early alterations in cell-signalling, developmental processes, mitochondrial dysfunction and a significant spinal cord immune response may play a major role in disease progression and regional vulnerability in Ppt1 −/− mice. It is now apparent that CLN1 disease does not just affect the brain. Characterizing and subsequently targeting the early inflammatory response observed in Ppt1 −/− spinal cords may therefore prove to be an effective strategy either alone or in combination with other therapies, as has become increasingly common across the NCLs 57,59,60 . Such characterization of the pathomechanisms of CLN1 disease and other forms of NCLs can directly inform therapeutic efforts in order to significantly improve disease outcomes.

Methods
Animals. The Ppt1 −/− mouse was created by targeted gene disruption that eliminates the last exon of the Ppt1/ Cln1 gene 61 . These mice were backcrossed with C57BL/6 mice for more than 10 generations and maintained as a homozygous breeding stock. Ppt1 −/− and age-matched C57BL/6 control mice were bred and housed in a barrier facility at the Institute of Psychiatry, Psychology and Neuroscience, King's College London (London, UK). Immunohistochemistry. Histological tissue processing. Mice at 3 and 7 months of age were anesthetized with sodium pentobarbitone (100 mg/kg) and perfused transcardially with heparinised PBS followed by 4% formaldehyde in PBS, pH 7.4. Whole mice were post-fixed in this 4% formaldehyde solution for a further 48 h before being transferred to a 50 mM Tris solution (TBS). Spinal cords were separated from the brainstem from the dorsal aspect, just below the foramen magnum. The entire spinal column was then dissected out from the surrounding tissue, and the spinal cord was dissected from the column by initial dorsal laminectomy followed by separation and removal of individual vertebrae. The spinal cords were cryoprotected in 30% sucrose in 50 mM TBS, pH 7.6. 40 µm coronal sections were then cut on a Microm HM430 freezing microtome (Microm International GmbH, Wallendorf, Germany).
Cresyl fast violet staining (Nissl staining). Cresyl fast violet staining was performed by mounting a one in 24 series of sections on to chrome-gelatin coated slides and leaving them to air-dry overnight at room temperature. The slides were then incubated in 0.1% cresyl fast violet solution with 0.05% acetic acid (VWR) overnight, before being differentiated by passing through a graded series of Industrial Methylated Spirit (IMS) solutions (70%, 80%, 90% and 2 × 100%), before being cleared in Xylene (VWR) and coverslipped in DPX (VWR).

Scientific RepoRtS
| (2020) 10:15157 | https://doi.org/10.1038/s41598-020-72075-7 www.nature.com/scientificreports/ Measurement of regional volume. Regional volumes were measured as previously described 10 , where whole spinal cord volumes and spinal cord grey matter volumes were measured in a one in 24 series of sections for each spinal cord. Using the Cavalieri estimator, estimates of volume in cubic µm were obtained by superimposing a 125 μm grid on Nissl-stained sections and using Stereo Investigator software (Microbrightfield Inc, Williston, Vermont) linked to a Zeiss Axioskop 2 MOT (Zeiss, Germany) with a DAGE-MTI CCD-100 camera (Dage-MTI, Michigan City, Iowa). White matter volumes were obtained as a subtractive value of the whole volume and grey matter volume of each spinal cord section, with reference to a spinal cord atlas 15 .
Thresholding image analysis. To analyze glial activation in the grey matter of sections stained for GFAP and CD68, as well as neuron density in Laminae I-III of sections stained for interneuron markers, we performed thresholding image analysis using Image Pro Premier software (Media Cybernetics, Chicago, IL, USA). Here, 30 non-overlapping images from 3 sections at 40× magnification for the glial markers and 50 non-overlapping images from 5 sections at 63× magnification for interneuron markers, were captured for each defined region, with all parameters of light intensity, camera setup and calibrations being kept constant. An appropriate threshold was applied to each set of images so as to select the positive foreground immunoreactivity over the background. Results were obtained as an average percentage area of positive staining per image 62 .
Interneuron and motor neuron counts. Due to the relatively low numbers of interneurons, stereological methods prove inefficient for counting them 63 . Counts for neurons in Laminae IV-X in a one in 24 series of sections immunostained for interneuron markers (Calbindin, Calretinin) and motor neuron markers (CD-71), was carried out manually using unbiased sampling. Each section was visualized at 10× magnification using a Axioplan microscope (Zeiss) and a live video camera (Luminera Infinity 3URM Colour camera), with Image Pro Premier software (Media Cybernetics, Chicago, IL, USA). Total estimates were obtained as the product of the counted cells, periodicity and the number of intervals.

Proteomic processing. Protein extraction for liquid chromatography-with tandem mass tagged-mass spectrometry (LC-MS/MS).
Mice at 3 and 7 months of age were euthanized in a CO 2 chamber, followed by isolation of the spinal column with scissors at the base of skull and at the pelvic bone. The skull was then dissected to remove the brain and the cortices were isolated. A 20 ml syringe with TBS and an 18G needle was then inserted into the rostral end of the vertebral column, ensuring a tight fit. The spinal cord was then extruded by hydraulic pressure into a glass petri dish containing TBS as previously described 64 . Collected cortices and spinal cords were then frozen at − 80 °C prior to proteomic analysis. Sample preparation for Tandem-Mass-Tag proteomic analysis was carried out as previously described 16,17 . Briefly, samples were homogenized in extraction buffer [100 mM Tris-HCl (pH7.6) 4% (w/v) SDS] containing 1% protease cocktail inhibitor (Thermo Fisher, UK). Post homogenization, samples were spun at 300×g for 2 min and left on ice for 20 min. Homogenates were transferred to Lo-Bind 1.5 ml tubes (Sigma Aldrich) and centrifuged at 20,000×g for 20 min at 4 °C with the soluble fraction of each sample then transferred to new Lo-Bind tubes. Protein determination using the bicinchoninic acid assay (BCA) (Pierce, UK) was carried out according to manufacturer's guidelines. Samples were then pooled into groups for a single tissue region at a single timepoint per genotype (n = 4 or 5 mice) as follows -3 month wild type (WT) cortex, 3 month Ppt1 −/− cortex, 3 month WT spinal cord, 3 month Ppt1 −/− spinal cord, 7 month WT cortex, 7 month Ppt1 −/− cortex, 7 month WT spinal cord and 7 month Ppt1 −/− spinal cord (https ://doi.org/10.7488/ds/2750 (2020)). Individual samples containing 20 μg of protein per sample within each group, i.e., equivalent amounts of protein from all samples-were pooled to generate a condition specific master sample of 100 μg protein.
TMT proteomic analysis of the pooled samples was performed by the FingerPrints Proteomics facilities at the University of Dundee. Pooling according to sample type allows a reduction in potential noise in the system generated through inter individual differences, subtle post-mortem handling differences and/or sample isolation. The inclusion of an equivalent proportion of each individual sample into a readily comparable pool allows the generation of a molecular fingerprint representative of each condition and enables subsequent analysis of individual patient variability in the resulting validatory work (as a deviation from the population signal, as previously described 17,18,65,66 ). Indeed, we have previously demonstrated that pooling can enable the reliable identification of up to 3× more proteins in complex samples 26 .
As per 16,17 , following labelling with TMT, pooled samples were desalted, and dried in a speed-vac at 30 °C, re-dissolved in 200 μl ammonium formate (10 mM, pH 10) and peptides were fractionated using an Ultimate 3,000 High Performance Liquid Chromatography column (Thermo-Scientific) containing an XBridge C18 column (XBridge peptide BEH, 130 Å, 3.5 μm, 2.1 × 150 mm) (Waters, Ireland) with an XBridge guard column (XBridge, C18, 3.5 μm, 2.1 × 10mm) (Waters, Ireland). Buffers A (10 mM ammonium formate in milliQ water) and B (10 mM ammonium formate with 90% acetonitrile) were adjusted to pH 10 with ammonia. Fractions were collected using a WPS-3000FC auto-sampler (Thermo-Scientific) at 1 min intervals. Column and guard column www.nature.com/scientificreports/ were equilibrated for 20 min at a constant flow rate of 0.2 ml/min. 175 μl per sample was loaded onto the column at a rate of 0.2 ml/min, and the separation gradient was started 1 min after sample was loaded onto the column. Peptides were eluted from the column with a gradient of 2-5% Buffer B in 6 min, and then from 5 to 60% Buffer B in 50 min. The column was washed for 16 min in Buffer B and re-equilibrated at 2%. The fraction collection started 1 min after injection and stopped after 80 min (total 80 fractions, 200 μl each). The total number of fractions concatenated was set to 15 and the content of the fractions was dried and suspended in 50 μl of 1% formic acid prior to analysis with LC-MS/MS.  (2020)).
Biolayout express 3D. Unfiltered lists comprised of protein IDs with corresponding expression ratios at each time point relative to the equivalent timepoint wildtype value (available here: https ://doi.org/10.7488/ds/2750 (2020)) were imported separately into BioLayout Express 3D and clustered based on relative expression profile in two separate comparisons-(a) 3 month Ppt1 −/− spinal cord vs. 7 month Ppt1 −/− spinal cord; and (b) or 3 month Ppt1 −/− spinal cord vs. 3 month Ppt1 −/− cortex. Algorithms in BioLayout Express 3D generate a visual network to represent each data set, in this study utilizing spatial proximity to represent the similarity in expression profile between nodes 25 . It was therefore possible to isolate clusters of proteins grouped by similarity in terms of expression profiles over time. The resultant visual networks were utilized to distinguish expression clusters that followed trends in expression that favored either group or both groups (3 vs. 7 month spinal cord analysis-altered only at 3 months, altered only at 7 months and similar/progressive alterations; 3 month cortex vs. 3 month spinal cord analysis-differential detection only in cortex, differential detection only in spinal cord and similar alterations in both cortex and spinal cord at 3 months). These clusters were analyzed and input into Ingenuity Pathway Analysis (Ingenuity Systems). www.nature.com/scientificreports/ Ingenuity pathway analysis (IPA). The IPA software (Ingenuity systems) contains a library of biological pathways published in the literature that are ranked by the significance of the association between the dataset and the canonical pathway. This significance is defined by two parameters: (a) the ratio of the number of proteins from the input dataset that are pertaining to a particular pathway divided by the total number of genes ascribed by the Ingenuity Knowledge Database to that canonical pathway and (b) a P value calculated using Fischer's test that determines whether the probability of association between component proteins in the input dataset and the canonical pathway are due to chance. Prediction activation scores (z score) are a statistical measure of the match between an expected relationship direction and the observed protein expression within the input dataset. A positive z score indicates activation while a negative z score indicate inhibitiony 16,18,68 . A 1.2 fold-change (20%) threshold filter was applied in IPA to the datasets for 3 month cortex, 3 month spinal cord and 7 month spinal cord (Ppt1 −/− vs. wildtype respectively), analyzed and observed interactions were selected for this analysis.
Statistical analysis. All measurements for histological processing were performed blind to genotype. Statistical analysis for histological measurements of regional volumes, neuron counts, cell areas and thresholding image analysis as well as for QFWB analysis between two groups at a given time-point was calculated by multiple two-tailed, unpaired, parametric t test with Bonferroni-Dunn correction where p ≤ 0.05 was considered significant using GraphPad Prism version 8.0.0 for MacOS (GraphPad Software, San Diego, CA, www.graph pad.com (2020)).