A mouse mutant deficient in both neuronal ceroid lipofuscinosis‐associated proteins CLN3 and TPP1

Late‐infantile neuronal ceroid lipofuscinosis (LINCL) and juvenile neuronal ceroid lipofuscinosis (JNCL) are inherited neurodegenerative diseases caused by mutations in the genes encoding lysosomal proteins tripeptidyl peptidase 1 (TPP1) and CLN3 protein, respectively. TPP1 is well‐understood and, aided by animal models that accurately recapitulate the human disease, enzyme replacement therapy has been approved and other promising therapies are emerging. In contrast, there are no effective treatments for JNCL, partly because the function of the CLN3 protein remains unknown but also because animal models have attenuated disease and lack robust survival phenotypes. Mouse models for LINCL and JNCL, with mutations in Tpp1 and Cln3, respectively, have been thoroughly characterized but the phenotype of a double Cln3/Tpp1 mutant remains unknown. We created this double mutant and find that its phenotype is essentially indistinguishable from the single Tpp1−/− mutant in terms of survival and brain pathology. Analysis of brain proteomic changes in the single Tpp1−/− and double Cln3−/−;Tpp1−/− mutants indicates largely overlapping sets of altered proteins and reinforces earlier studies that highlight GPNMB, LYZ2, and SERPINA3 as promising biomarker candidates in LINCL while several lysosomal proteins including SMPD1 and NPC1 appear to be altered in the Cln3−/− animals. An unexpected finding was that Tpp1 heterozygosity significantly decreased lifespan of the Cln3−/− mouse. The truncated survival of this mouse model makes it potentially useful in developing therapies for JNCL using survival as an endpoint. In addition, this model may also provide insights into CLN3 protein function and its potential functional interactions with TPP1.

in LINCL while several lysosomal proteins including SMPD1 and NPC1 appear to be altered in the Cln3 À/À animals. An unexpected finding was that Tpp1 heterozygosity significantly decreased lifespan of the Cln3 À/À mouse.
The truncated survival of this mouse model makes it potentially useful in developing therapies for JNCL using survival as an endpoint. In addition, this model may also provide insights into CLN3 protein function and its potential functional interactions with TPP1.

| INTRODUCTION
The neuronal ceroid lipofuscinoses (NCLs) are a group of more than a dozen genetically distinct but similar lysosomal storage diseases. 1 Characterized by an accumulation of autofluorescent storage material within the lysosome, NCLs are neurodegenerative and progressive diseases that are manifested by seizures, loss of vision, and eventual loss of mental capacity. Onset is typically in childhood and these diseases result in premature death. Two of the most frequently encountered NCL diseases are the classical late-infantile and juvenile forms (LINCL and JNCL, respectively). LINCL results from mutations in the gene encoding tripeptidyl peptidase 1 (TPP1, formerly designated CLN2), 2,3 a soluble lysosomal serine protease. [4][5][6] Disease in LINCL typically presents at around 4 years of age and lifespan is $8-15 years. 1 JNCL is caused by mutations in a gene encoding a lysosomal transmembrane protein, CLN3. 7 JNCL has a later onset and is more slowly progressing than LINCL, with first signs of disease (typically problems with vision) at around 8 years and patients frequently surviving into the second or third decade of life. 1 Despite differences in disease timeline and genetic etiology, LINCL and JNCL have a number of similarities including lysosomal storage of subunit c of mitochondrial ATP synthase (SCMAS). 8 There is considerable focus on the development of effective therapies for NCL diseases and LINCL is leading the way. Enzyme replacement therapy has been clinically approved for LINCL 9,10 and there is interest in gene therapy, with promising results obtained in animal models, [11][12][13][14][15] and clinical studies have been conducted. 16,17 Both enzyme replacement therapy (ERT) and gene-therapy rely upon the fact that TPP1 is a soluble lysosomal protein. 2 In ERT, exogenouslyadministered recombinant protein can be taken up by numerous cells by endocytosis and delivered to the lysosome, while in gene therapy, a only a proportion of cells are transduced but these can overproduce and secrete enzyme that is taken up by untransduced cells. In addition, there are LINCL animal models 18,19 with welldefined phenotypes that accurately recapitulate the human disease and these have been integral in testing treatment strategies.
There is currently no approved therapy for JNCL and this reflects several major obstacles. First, CLN3 is a transmembrane protein, 7 which excludes replacement treatment using exogenously administered recombinant protein. There is interest in gene therapy for JNCL 20 but because CLN3 is an integral membrane protein, nontransduced cells will not express the missing protein.
Thus, if the underlying metabolic defect is cell autonomous, cross-protection between transduced and untransduced cells may not be possible, and a very high proportion of cells will require transduction for effective therapy. Second, there is a fundamental lack of understanding of the cellular function of the CLN3 protein. It has been implicated in numerous cellular activities including lysosomal pH homeostasis, endocytosis, autophagy, apoptosis, lysosomal enzyme transport, and others. 21 While the precise function of CLN3 is yet to be definitively established, a recent study has established that glycerophosphodiesters accumulate within the lysosome in the absence of CLN3, 22 raising the possibility the CLN3 is a lysosomal transporter for these lipids. Regardless, mechanism-based approaches to treatment for JNCL are currently not an option. Third, animal models for JNCL do not present a robust phenotype, especially with respect to survival. There are several mouse models for JNCL [23][24][25][26] and their phenotypes are very similar but in all, disease is highly attenuated compared to mouse models of other NCL diseases.
In this study, we have created mice that are defective in both CLN3 protein and TPP1 with the rationale that the phenotype of a double mutant may provide insights into CLN3 protein function and possible functional interactions with TPP1. In terms of survival, pathology, and proteomic changes within the brain, the phenotype of the double mutant is essentially indistinguishable from mice that lack TPP1 alone. However, we find that the lifespan of the Cln3 À/À mouse model is shortened in a Tpp1 heterozygous background. While the physiological implications of this are unclear, this does potentially provide a mouse model with a clear JNCL survival phenotype that may be of value in therapeutic testing.

| Animals
Mice were maintained and used following protocols approved by the Rutgers University and Robert Wood Johnson Medical School Institutional Animal Care and Use Committee ("Preclinical evaluation of therapy in an animal model for LINCL," protocol I09-0274-4). Tpp1 À/À mice 19 were either in an Nnt À/À27 C57BL/6J or Nnt +/+ C57BL/6 strain background. Cln3 À/À mice 28 were in a C57BL/6J genetic background. Animals arising from crosses between C57BL/6 Tpp1 À/À mice and C57BL/6J Cln3 À/À mice are in a mixed C57BL6 substrain background designated here as C57BL/6Â6J and they are a mixture of Nnt +/+ , Nnt +/À and Nnt À/À genotypes. Genotyping of Tpp1 À/À and Cln3 À/À mice was conducted as described. 19,28 Experimental cohorts contained $equal numbers of males and female animals and were analyzed at $120 days of age. For biochemical analyses, animals were deeply anesthetized with sodium pentobarbital/ phenytoin (a 1:4 dilution of Euthasol; Delmarva Laboratories, Midlothian, VA) and euthanized by transcardial perfusion with 0.9% saline. Brains were dissected and frozen on dry ice. For histopathology, mice were anesthetized, perfused with saline then perfusion-fixed with 4% paraformaldehyde in PBS. Brains were excised, fixed for 48 h in 4% paraformaldehyde in PBS, and then transferred to 30% sucrose/PBS at 4 C until they sunk.

| Quantitative mass spectrometry
Sample preparation and quantitative mass spectrometry on whole brain extracts were conducted as described previously 29 and parameters for mass spectrometry, peak list generation, and database searching are summarized in Table S1. In brief, detergent-solubilized extracts were prepared, proteins were digested in-solution sequentially with trypsin (specificity, carboxyl side of K and R) and endoprotease LysC (specificity, carboxyl side of K) and the resulting peptides labeled with TMT11-plex isobaric reagents (ThermoFisher Scientific). Labeled samples were pooled and analyzed by synchronous precursor selection MS3 on a Lumos Tribrid instrument (ThermoFisher Scientific).

| Database searching
Peak lists were generated using Proteome Discoverer 2.2 (ThermoFisher Scientific) and data were searched using a local implementation of the Global Proteome Machine (GPM) 30,31 using search parameters shown in Table S1. Protein assignments are shown in Table S2. Mass spectrometry files (mgf and raw files, GPM search files, and Excel files denoting protein assignments, peptide-spectrum matches, and corresponding reporter ion intensities) are archived in the MassIVE (http://massive.ucsd.edu) and ProteomeXchange (http://www.proteomexchange.org/) repositories in submission MSV000087613.

| Normalization and statistical analyses of mass spectrometry data
TMT-11 reporter ion intensities were normalized and analyzed as described previously. 32 In brief, reporter ion intensity data were extracted from peak list files using custom in-house scripts (https://github.com/cgermain/ IDEAA), then spectra were normalized to total reporter ion intensity per channel to correct for differences in labeling efficiency and/or amounts of protein labeled. Spectral assignments with reporter ion intensities are shown in Table S2. Peptides were first filtered for fully tryptic cleavage, no missed cleavage sites, and complete iTRAQ labeling of lysines and amino termini. Peptides were then filtered to remove those containing posttranslational modifications that may increase variability in the data (i.e., asparagine or glutamine deamidation, methionine dioxidation, tryptophan mono-and dioxidation, and isobaric labeling of tyrosine at positions other than the amino terminus). For selected comparisons, ratios of expression and q-values corrected for multiple comparisons using the Benjamini-Hochberg procedure 33 were generated using a nested procedure that accounts for variability at both peptide and protein level 32 (Table S3).

| Immunostaining
For immunohistochemistry, a one in six series of coronal brain sections from each mouse were stained using a modified immunofluorescence protocol 34 for the astrocyte marker glial fibrillary associated protein (GFAP, rabbit anti-GFAP, 1:1000, DAKO), and the microglial marker CD68 (rat anti-mouse CD68, 1:400, Bio-Rad). Briefly, 40 μm coronal sections were mounted on Superfrost Plus slides (Fisher Scientific) and air-dried for 30 min, slides were then blocked in a 15% serum solution in 2% TBS-T (1Â Tris Buffered Saline, pH 7.6 with 2% Triton-X100, Fisher Scientific) for 1 h. Slides were then incubated in primary antibody in 10% serum solution in 2% TBS-T for 2 h, washed three times in 1xTBS, and incubated with fluorescent Alexa-Fluor labeled IgG secondary antibodies (Alexa-Fluor goat anti-rabbit 488, goat anti-rat 546, Invitrogen) in 10% serum solution in 2% TBS-T for 2 h. Slides were washed three times in 1xTBS and incubated in a 1Â solution of TrueBlack lipofuscin autofluoresence quencher (Biotium, Fremont, CA) in 70% ethanol for 2 min before rinsing with 1xTBS. Slides were cover-slipped in Fluoromount-G mounting medium with DAPI (Southern Biotech, Birmingham, AL).

| Thresholding image analysis
To analyze the degree of glial activation in the gray matter (GFAP positive astrocytes + CD68 positive microglia) a semiautomated thresholding image analysis method was used with Image-Pro Premier software (Media Cybernetics). Briefly, stained sections were scanned using a Zeiss AxioScan Z1 slide scanner at the Washington University Center for Cellular Imaging (WUCCI), at 10Â magnification of each one in six series of sections per animal followed by demarcation of all regions of interest. Images were subsequently analyzed using Image-Pro Premier (Media Cybernetics) using an appropriate threshold that selected the foreground immunoreactivity above the background. This threshold was then applied as a constant to all subsequent images analyzed per batch of animals and reagent used to determine the specific area of immunoreactivity for each antigen. Measurements for histological processing were performed blind to genotype and statistical analyses were performed using GraphPad Prism version 8.0.0 for MacOS. Data were analyzed using two-way ANOVA with a post hoc Bonferroni correction, with a p-≤ 0.05 considered significant.

| Survival statistics
Survival curves were compared using log-rank tests in GraphPad Prism 9.1 (GraphPad Software, San Diego, CA, www.graphpad.com).
2.8 | TPP1 enzyme assay TPP1 was measured using an endpoint assay with Ala-Ala-Phe-AMC substrate as described previously. 5 For TPP1 assay in the presence of glycerophosphodiesters, TPP1 activity was again measured by endpoint assay with 200 μM Ala-Ala-Phe-AMC and 1 or 10 mM L-α-Glycerophosphorylcholine (Sigma) or 100 μM Glycerophosphoinositol (Echelon), each dissolved in water.
Survival of mutant animals in Tpp1 +/À or Tpp1 +/+ backgrounds are shown in Figures 1B,C. Note that we collected limited survival data for wild-type animals and therefore include data from C57BL/6J animals from another study ("Yuan dataset"). 35 Survival of Tpp1 heterozygotes did not differ significantly (p = 0.1635) from the Yuan dataset of wild-type animals (median survival 901 days) ( Figure 1B). We did not collect survival data for Tpp1 +/+ ;Cln3 +/À animals but given that heterozygosity for Tpp1 did not affect survival, we predict no significant differences in survival of Cln3 heterozygotes when compared to wild-type. Survival of Tpp1 +/À ;Cln3 +/À double heterozygote mice (median 748 days) was similar to the single Cln3 À/À mice (p = 0.5719) and was also significantly shorter than wild-type animals (p = 0.0014). Tpp1 heterozygosity resulted in decreased survival of CLN3 À/À animals (median 584 days) which was significantly (p < 0.0001) shorter than that of the Tpp1 +/+ ;Cln3 À/À animals. Overall, these data indicate that heterozygosity for Tpp1 resulted in decreased survival of both Cln3 +/À and Cln3 À/À animals. Age at death (days) Survival of our Tpp1 +/+ ; Cln3 À/À mice (median 719 days) was not significantly different ( p = 0.2339) from historical data from a different but similar JNCL knockout mouse model (742 days) 36 ( Figure 1C). However, the survival of both Cln3 mutants was significantly reduced compared to wild-type mice (p < 0.0001).

| Pathology
In order to compare brain pathology in the double mutant Tpp1 À/-Cln3 À/À and single mutant Tpp1 À/À animals, animals were euthanized at their median lifespan of $120 days. Sections were analyzed by immunofluorescence for the simultaneous detection of markers of glial activation, (CD68 in microglia and GFAP in astrocytes) in two brain regions where pronounced glial activation is consistently observed in multiple NCL mouse models: the barrel field of the somatosensory cortex (S1BF) and the region encompassing the medial and lateral ventral posterior nuclei of the thalamus (VPM/VPL). As expected, immunostaining for either GFAP or CD68 was present at very low levels in animals with a Tpp1 +/À or Tpp1 +/+ genotype irrespective of Cln3 genotype. In these animals, only very faintly stained GFAP-positive astrocytes or CD68-positive microglia ( Figure 2) were present. Consistent with previous observations, 38 marked activation of both astrocytes and microglia was detected in the S1BF and VPM/VPL of Tpp1 À/-Cln3 +/+ animals, with pronounced upregulation of these markers and corresponding changes in cellular morphology. Similar levels of pronounced glial activation were detected in the S1BF and VPM/VPL of double mutant Tpp1 À/-Cln3 À/À animals, with similar degrees of cellular hypertrophy of both astrocytes and microglia apparent to those seen in Tpp1 À/-Cln3 +/+ animals. Thresholding image analysis confirmed these qualitative observations with significant elevation of both GFAP and CD68 immunoreactivity in both brain regions of Tpp1 À/-Cln3 +/+ and Tpp1 À/-Cln3 À/À animals compared to other genotypes ( Figure 3). Although the individual levels of GFAP and CD68 immunoreactivity differed to some extent between Tpp1 À/-Cln3 +/+ and Tpp1 À/-Cln3 À/À animals, there were no significant differences for either antigen in either T A B L E 1 Survival statistics for Tpp1 À/À animals. Note: Survival data were obtained from littermates from crosses between double Tpp1 and Cln3 mutant crosses unless indicated otherwise (* for single mutant crosses). Significance for the Log-rank tests conducted below was evaluated after correction of p values for multiple comparisons using the Bonferroni method.
brain region between animals of these genotypes. This broad similarity between Tpp1 À/-Cln3 +/+ and Tpp1 À/-Cln3 À/À animals in terms of the extent of glial activation is consistent with the similarities observed in the survival phenotypes of these mutants. While we did not measure autofluorescence, lysosomal storage, and neuronal loss, we would predict these to be similar or identical in the Tpp1 À/-Cln3 +/+ and Tpp1 À/-Cln3 À/À animals given the similarities in glial activation pathology and survival phenotypes. Overall, Cln3 genotype had no observable effect on the pathology of Tpp1 À/À animals at the age tested ($100 days). This provides further evidence that Cln3 defects do not exacerbate the LINCL phenotype in these animals.

| Proteomic analysis
We conducted a proteomic analysis on brain samples from the Tpp1 and Cln3 mutant mice with the goal of identifying changes that might correlate with disease survival and may shed light on the respective cellular functions of these proteins. In addition, given that earlier proteomic studies failed to identify any informative brain expression changes in Cln3 À/À mice, 29 we hypothesized that changes related to Cln3 mutation could be exacerbated and thus possibly easier to detect in a Tpp1 mutant background. Proteins in brain extracts from animals at $105 days were identified and quantified using isobaric labeling mass spectrometry. In Figure 4, we compared various genotypes of interest using volcano plots, a scatter plot that allows comparison of effect size with the probability of significant difference. Q values (i.e., p values adjusted for multiple comparisons found using the Benjamini-Hochberg false discovery rate method) were calculated by comparing reporter ion intensities for all individual spectra assigned to a given protein from all biological replicates for each genotype.

| TPP1 activity in mutant mice
Previous studies have reported elevated TPP1 activities in the brains of Cln3 knockout mice. 26 Consistent with these earlier studies, we find that TPP1 activity in the Cln3 mutant is $2-fold higher than that measured in wild-type animals, while no activity was detectable in Tpp1 À/À animals irrespective of Cln3 genotype ( Figure 6). While previous analyses 19 indicated that TPP1 activity in Tpp +/À animals is, as expected, $50% of wildtype, TPP1 activity in Tpp +/-Cln3 À/À animals is $100% of wild-type. This again indicates that loss of CLN3 results in an $2-fold increase in TPP1 activity.

| DISCUSSION
The overall goal of this study was to determine the phenotype of a double Tpp1 and Cln3 mutant mouse model, F I G U R E 3 Quantitative analysis of glial activation in 120-day old mice. Thresholding image analysis confirms the different levels of glial fibrillary associated protein (GFAP, astrocytes) and CD68 (microglia) in the primary somatosensory cortex (S1BF) and ventral posterior thalamic nuclei (VPM/VPL) of animals of different genotypes. These data confirm the significantly elevated levels of both antigens in Tpp1 À/-Cln3 +/+ and Tpp1 À/-Cln3 À/À mice compared to animals of other genotypes, which displayed very low levels of immunoreactivity for these markers. *p < 0.05 and ***p < 0.001 when compared to combined controls Tpp1 +/+ ;Cln3 +/À and Tpp1 +/À ;Cln3 +/+ using one-way ANOVA with Dunnett's test for multiple comparisons. Bars indicate mean and standard deviation.
Glial activation is a consistent pathological feature in NCL diseases, including both LINCL and JNCL. 39,40 Typically, localized glial activation precedes the onset of neuron loss and serves to predict where neurodegeneration subsequently occurs. Indeed, there have been suggestions that glial dysfunction may contribute to neuron loss in multiple NCLs. 40 Quantitative analysis of Tpp1 À/-Cln3 +/+ and Tpp1 À/-Cln3 À/À mice revealed very similar extents of astrocytosis and microglial activation between mice of these genotypes. In future studies, it will be important to determine whether this similarity extends to the degree of neuron loss in these mice. Such investigations lie beyond the scope of the current study but would further F I G U R E 5 Correlation between significant proteomic changes in animals of different genotypes. (A) Tpp1 À/ ; À Cln3 +/+ compared to Tpp1 À/À ;Cln3 À/À ; goodness of fit, R = 0.7901. Note that NNT is censored (see Results). (B) Tpp +/+ ;Cln3 À/À compared to Tpp1 +/À ; Cln3 À/À ; goodness of fit, R = 0.5752. Compared to wild-type, proteins shown are significantly altered in at least one model (FDR 1%), have a magnitude of change of ≥1.5 fold in at least one model, and have a consistent direction of change in both models. Names of select proteins of interest are shown in red (lysosomal) or black (other). Filled symbols indicate proteins that are significantly altered in both genotypes being compared, open symbols indicate proteins that are significantly altered in one of the two genotypes being compared.
F I G U R E 6 TPP1 activity in mutant mice. Activities were measured in brain extracts from 120-day old mice and genotypes were compared to wild-type using one-way ANOVA with Dunnett's test for multiple comparisons when compared to combined controls Tpp1 +/+ ;Cln3 +/+ . *p < 0.05; ***p < 0.001; ****p < 0.0001. Bars indicate mean and standard deviation. raise the translational significance of the newly developed mouse model.
Lifespan of the JNCL mouse model (median: 719 days) was somewhat shortened compared to the wild-type (median: 901 days) and this is consistent with previous analysis of a different Cln3 knockout mouse model (Cln3 À/À , median 742 days, wild-type median 868 days, 36 ). Unexpectedly, we found that heterozygosity for Tpp1 was associated with shortened survival of Cln3 À/À mice, further decreasing lifespan to a median of 584 days. In addition, survival of a double heterozygote Tpp1 +/À ;Cln3 +/À was also significantly reduced compared to wild-type.
One possible explanation for this observation is that heterozygosity for Tpp1 exacerbates the Cln3 phenotype due to substrain effects given that the Cln3 knockout mouse is in an Nnt À/À C57BL/6J background whereas the Tpp1 knockout is Nnt +/+ C57BL/6. As a result, the Tpp1 +/-Cln3 À/À mutant was predicted to be in a mixed Nnt +/+ , Nnt +/À and Nnt À/À C57BL/6Â6J background, and this was confirmed by genotyping Nnt in 15 Tpp1 +/-Cln3 À/À mutants: we found the majority (9/15) to be Nnt À/À , 5/15 to be Nnt +/À and 1/15 to be Nnt +/+ (data not shown). Thus, most of the Tpp1 +/-Cln3 À/À animals were in an Nnt À/À strain background like the Cln3 À/À mutant. It is also worth noting that an isogenic Nnt À/À C57BL/6J background shortens the lifespan of Tpp1 À/À animals compared to the same mutation in an Nnt +/+ C57BL6 background ( Figure 1). Thus, if substrain effects were to play a role in survival, we would predict that the introduction of a wild-type Nnt allele into the Tpp1 +/-Cln3 À/À animals would actually increase rather than shorten survival of the Cln3 À/À mouse. Taken together, these observations indicate that substrain effects do not play a role in the shortened survival of the Tpp1 +/-Cln3 À/À mouse.
Another possibility is that the loss of the CLN3 protein could have adverse downstream effects on TPP1 function or access to its substrates, potentially causing or contributing to JNCL disease. If this is the case, then heterozygosity for Tpp1 would be predicted to exacerbate disease. The fact that both LINCL and JNCL accumulate SCMAS, a likely primary substrate for TPP1, 41 and the fact that TPP1 activity is elevated in JNCL ( Figure 6), 26,42 a possible compensatory response, is consistent with this model. It is also possible that the compensatory increases in TPP1 activity detected in the absence of CLN3 might actually provide a neuroprotective function thus heterozygosity for Tpp1 would diminish such a protective role, resulting in a shorter lifespan for the Cln3 À/À mouse. Finally, a recent study 22 indicated that loss of CLN3 results in a lysosomal accumulation of glycerophosphodiesters, and it is possible that increased levels of these lipids directly inhibit TPP1 function. To test this possibility, we conducted assays of increasing amounts of TPP1 in the presence of 100 μM glycerophosphoinositol or 1 mM and 10 mM glycerophosphocholine, two glycerophosphodiesters shown to accumulate in the absence of CLN3. The presence of these lipids had no detectable effect on TPP1 activity in vitro (Figure 7) thus we conclude that they do not directly act as inhibitors at the concentrations tested. However, we cannot exclude the possibility that locally higher concentrations of these accumulated lipids within in the lysosome, or complexes between these lipids and other lysosomal material, could have an inhibitory effect on TPP1 under physiological conditions. Regardless of the underlying basis for its shortened lifespan, decreased survival of the Tpp1 +/-Cln3 À/À mutant may be useful for testing of therapeutic strategies for JNCL. In developing any therapeutic strategy, survival provides a clear and objective endpoint: for example, in evaluating several different strategies for LINCL, survival studies in mouse models have highlighted promising approaches (e.g., gene therapy, intrathecal cerebrospinal fluid [CSF]-mediated ERT, bloodstream mediated ERT) [43][44][45][46][47] while survival studies in the dog model 12 helped pave the way for the approval of enzyme replacement therapy. LINCL animal models have a markedly shortened lifespan thus proof of principle for treatments in terms of survival is readily achievable. In contrast, the survival of Cln3 mutant mice approaches that of wild-type, complicating survival as an endpoint. As a result, behavioral models have been extensively employed to analyze disease progression and the effects of potential treatment in Cln3 mutant mice. While numerous studies have characterized locomotor deficits in Cln3 mouse mutants, 36,48-52 behavioral phenotypes are subtle and are dependent on mouse strain and gender. 49 Evaluation of potential therapeutics for JNCL in Cln3 À/À ;Tpp1 +/À mouse mutants would provide a relatively robust survival phenotype to measure efficacy. In addition, it is likely that behavioral phenotypes would be exacerbated at a younger age. Effective therapy that addresses the loss of CLN3 (e.g., widespread gene therapy throughout the brain) would be predicted to increase survival of this mouse mutant to resemble that of the Tpp1 heterozygote, which is indistinguishable from wild-type. A similar approach has been conducted using mutant Cln3 mice that express human amyloid precursor protein (APP) with familial Alzheimer disease mutations. 53 However, the mutant APP confers a severe survival phenotype even without CLN3 defects, thus positive treatment for CLN3 defects would essentially ameliorate a severe phenotype, not restore wild-type survival.
One goal of this project was to determine whether the double Tpp1 and Cln3 knockout could highlight proteomic changes in the brain that could potentially provide a platform for clinically useful biomarkers in either or both LINCL and JNCL. Neurofilament light chain (NEFL in mice) has been proposed as a treatmentresponsive biomarker based on studies of plasma from CLN2 patients and a dog model 54 and previously, we found levels of this and other neurofilaments proteins (NEFH and NEFM) to be elevated in mouse CSF. 29 However, we found brain levels of NEFL, NEFH, and NEFM to be unchanged or reduced in the CLN2 and CLN3 mouse models. This is consistent with the finding that brain NEFL is unchanged in mouse models of other neurodegenerative disease exhibiting elevated NEFL plasma and CSF levels. 55 However, this study does reinforce several proteomic changes in LINCL mouse brain that were identified previously including elevated GPNMB, LYZ2, and SERPINA3. 29 Further studies on Tpp1 À/À mouse models will indicate if any of these proteins are also significantly altered in more accessible biological samples including blood and cerebrospinal fluid. We also found that several lysosomal proteins (CTSF, SMPD1, and NPC1) were significantly altered in the Cln3 À/À animals and while changes in expression were modest, they may be more evident in older mice (animals were analyzed at $120 days in this study). These changes may be compensatory responses to the loss of CLN3 and could potentially provide useful information regarding its biological function. In addition, soluble lysosomal proteins SMPD1 and CTSF may also warrant further investigation as potential biomarkers in JNCL.

CONFLICT OF INTEREST STATEMENT
Peter Lobel and David E. Sleat have received royalty payments as Inventors on Patent 8029781 "Methods of treating a deficiency of functional tripeptidyl peptidase I (CLN2) protein," which is licensed to BioMarin Pharmaceutical Inc. Other authors have declared that no conflicts of interest exist.

ETHICS STATEMENT
Ethics approval was not required for this research study.

ANIMAL RIGHTS
Animals were maintained and used following protocols approved by the Rutgers University and Robert Wood Johnson Medical School Institutional Animal Care and Use Committee protocol I09-0274-4.