The CSTB-deficient (Cstb^–/–) mouse strain used in this study is 129S2/SvHsd5-Cstb^tm1Rm, derived from the Jackson Laboratory strain 129-Cstb^tm1Rm/J (stock no. 003486)¹ (Pennacchio et al., 1998). Wild-type (wt) offspring from heterozygous matings were used as controls. Mice were genotyped for the Cstb^tm1Rm mutation using genomic DNA isolated from ear clippings and confirmed with DNA isolated from tail clippings following euthanasia. Mouse ages correspond to the following human developmental stages: P7 = early infancy, P14 = late infancy, P21 = childhood, P30 = adolescence, P120 = adulthood, based on age-dependent behavioral and hormonal signatures (Bell, 2018). The research protocols were approved by the Animal Ethics Committee of the State Provincial Office of Southern Finland (decisions ESAVI/10765/04.10.07/2015 and ESAVI/471/2019).

P14 and older mice were intracardially perfused with heparin (0.16 mg/ml in PBS) and 4% paraformaldehyde (PFA) under pentobarbital-induced terminal anesthesia (100–200 mg/kg; Mebunat Vet, Orion). P7 mice were sacrificed without perfusion. Brains were bisected along the midline and immersion-fixed in 4% PFA for 48 h before cryoprotection in 30% sucrose, 0.05% sodium azide in TBS. Frozen coronal and sagittal sections of 40 μm were cut through the cerebrum and cerebellum, respectively, and collected as series of free-floating sections in 30% ethylene glycol, 15% sucrose, 0.05% sodium azide in TBS. For immunohistochemistry, tissue sections were (1) incubated with 50 mM ammonium chloride in TBS for 30 min, (2) blocked with 10% normal donkey serum, 1% BSA, 0.1–0.3% TritonX-100 in TBS for 1 h, (3) incubated with primary and secondary antibodies in 10% normal donkey serum, 0.3% TritonX-100 in TBS at 6°C, overnight or RT for 2 h, respectively, (4) counter-stained with DAPI (Abcam) for 5 min, and (5) mounted on chrome alum-gelatin coated microscopy slides using Fluoromount aqueous medium (Merck).

Fluorescent tissue samples were imaged on a Zeiss LSM780 confocal system equipped with an Axio Observer.Z1 inverted microscope platform using ZEN 2.3 SP1 black acquisition software (Zeiss). The objectives utilized were a LD LCI Plan-Apochromat 25×/0.8 lmm korr DIC M27 (oil immersion), a Zeiss C Plan-Apochromat 40×/1.4 Oil DIC M27 and a Zeiss C Plan-Apochromat 63×/1.4 Oil DIC M27. DAPI, Alexa488 and Alexa594 were imaged with a 405 nm DPSS-laser, a 488 nm Argon laser and a 561 nm DPSS-laser and emission windows of 419–496 nm, 499–588 nm, and 588–733 nm, respectively. The pinhole was set to Airy 1 with unidirectional imaging and a pixel dwell time of 2 to 2.5 μs. No averaging was performed. Image analyses were carried out manually in ZEN 2.3 lite blue software (Zeiss) or Fiji/ImageJ 1.53 software except (1) H3cs1 and (2) Ki-67. Specifically, (1) the relative H3cs1 intensity per field was calculated by normalizing the fluorescence intensity of the H3cs1-positive cell nuclei [recognized automatically using the Fiji/ImageJ plugin for StarDist (Schmidt et al., 2018)] to the total nuclei count (counted manually using the DAPI field) and (2) Ki-67-expressing IBA1-positive cells were counted directly under the confocal microscope. For every biological replicate and condition, a minimum number of three random fields and 100 cells were analyzed. For standardization, images were obtained from the same areas in each cerebellar tissue sample.

Mice were sacrificed by cervical dislocation. Following dissection, brains were bisected, rinsed with PBS and snap-frozen in liquid nitrogen. To prepare whole-tissue lysates, snap-frozen cerebella were lysed in 2% SDS, 12.5% glycerol, 60 mM Tris–HCl (pH 6.8) using a NG010 tissue grinder (Nippon Genetics), and further homogenized with a digital sonifier (Branson). Cytoplasmic and chromatin-bound protein extracts were prepared using a subcellular protein fractioning kit for tissues (Thermo Fisher Scientific, #87790). Specifically, (1) snap-frozen brains were homogenized with the loose pestle of a Dounce homogenizer (Jencons Scientific) for 13 strokes, filtered through a tissue strainer with a pore size of 250 μm and lysed in cytoplasmic extraction buffer as per manufacturer’s instructions. (2) Chromatin pellets were subsequently prepared from crude nuclei and lysed by micrococcal nuclease digestion for 30 min at room temperature. The presence of lysosomal proteins in the cytoplasmic extract was determined by immunoblotting for lysosomal marker LAMP1. Histone extracts were prepared using Histone Extraction Kit (Abcam, #ab113476). Protein concentrations were calculated using the BCA protein assay (Thermo Fisher Scientific). For western blotting, (1) protein samples were diluted in a sample buffer containing lithium dodecyl sulfate, glycerol, 50 mM DTT and SERVA Blue G250. (2) Protein samples were resolved using 4–12% Bis-Tris polyacrylamide gels in MES SDS running buffer (Bolt, Thermo Fisher Scientific). (3) Proteins were transferred to PVDF or nitrocellulose membranes with a pore size of 0.2 μm using TransBlot Turbo semi-dry system (Bio-Rad) in transfer buffer containing 39 mM glycine and 20% ethanol in 48 mM Tris (pH 9.2). (4) Target proteins were detected by enhanced chemiluminescence (ECL) using SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific) at antibody-dependent concentrations. (5) Molecular weights were calculated using two independent protein ladders. (6) For total protein normalization, membranes were stained with 0.1% amido black 10B (Thermo Fisher Scientific) and analyzed using Image Lab software (version 6.0, Bio-Rad).

Cathepsin L activity measurements were carried out using a substrate-based assay (Abcam, ab65306). Specifically, (1) assays were performed in the presence of 1.5 μM CA-074 (Merck) to prevent substrate cleavage by cathepsin B and (2) a standard curve with aminofluorocumarin (AFC, Abcam) was used to calculate AFC production from the fluorescent intensity readout of the assay. In situ β-galactosidase activity assays were carried out as previously described (Dimri et al., 1995). Specifically, (1) PFA-fixed free-floating cerebellar sections were incubated in staining solution, containing 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 2 mM magnesium chloride, 150 mM sodium chloride and 1 mg/mL X-gal in 40 mM citric acid—sodium phosphate buffer (pH 6) for 3 h and 30 min at 37°C, and (2) bright-field microscopy images (four random fields per biological replicate and experimental condition) were captured using a Zeiss Axio Imager M2 microscope equipped with 20× objective and Axiocam 503 camera (Zeiss) and analyzed using Fiji/ImageJ 1.53 software.

Rabbit α-H3cs1 [Cell Signaling Technologies, western blot (WB) 1:1,000, immunohistochemistry (IHC) 1:500, RRID:AB_2797961], mouse α-NeuN (Thermo Fisher Scientific, IHC 1:500, RRID:AB_2298772), rat α-GFAP (Thermo Fisher Scientific, IHC 1:300, RRID:AB_2532994), goat α-Olig2 (R and D Systems, IHC 1:1,000, RRID:AB_2157554), goat α-IBA1 (Novus, IHC 1:500, RRID:AB_521594), rabbit α-histone H3 (Abcam, WB 1:15,000, RRID:AB_302613), mouse α-GAPDH (Abcam, WB 1:10,000, RRID:AB_2107448), rat α-LAMP1 (DSHB, WB 1:1,000, RRID:AB_2134500), goat α-cathepsin B (R and D Systems, WB 1:2,000, RRID:AB_2086949), goat α-cathepsin F (Thermo Fisher Scientific, WB 1:2,000, RRID:AB_2576799), mouse α-cathepsin L (Abcam, WB 1:1,000, RRID:AB_305417), rabbit α-lamin B1 (Abcam, WB 1:2,000, RRID:AB_443298), rabbit α-p21^cip1 (Abcam, IHC 1:500, RRID:AB_2734729), rabbit α-γH2AX (Abcam, WB 1:3,000, IHC 1:10,000, RRID:AB_1640564), rabbit α-IBA1 (Wako, IHC 1:4,000, RRID:AB_839504), rat α-Ki-67 (Thermo Fisher Scientific, IHC 1:500, RRID:AB_10853185).

Real-time-qPCR assays were performed in accordance with MIQE guidelines (Bustin et al., 2009). Total RNA was isolated from cerebellum with NucleoSpin RNA Plus kit (Macherey-Nagel) and reverse-transcribed with iScript cDNA synthesis kit (BioRad). RT-qPCR were performed on a CFX96 Real-Time System (Bio-Rad) using TaqMan Fast Advanced Master Mix (Applied Biosystems). The following TaqMan probes were used for target amplification: Cstb (Mm00432769_m1), Lmnb1 (Mm00521949_m1), Ywhaz (Mm03950126_s1), and Rpl13 (Mm02526700_g1). Cstb and Lmnb1 expression data was normalized to that of both Ywhaz and Rpl13, chosen due to their stable expression in contexts of neural development (Daura et al., 2021) and neurodegeneration (Rydbirk et al., 2016), respectively. Relative gene expression was calculated using the 2^–ΔCt method.

The experimental unit (n) in the analyses was one individual animal, except in the nuclear area measurements, where the experimental unit was defined as one individual cell. Statistical analyses were carried out with GraphPad Prism v8.4.2.679. Comparisons between two experimental conditions were made using Student’s t-test with Welch’s correction. Comparisons between three or more experimental conditions were made using one-way ANOVA with Sidak correction or two-way ANOVA with Tukey’s correction. The normality assumption of ANOVA was assessed with the Shapiro–Wilk test and by manually inspecting the distribution of ANOVA residuals. In datasets not conforming to the normal distribution, comparisons between three or more experimental conditions were made using Kruskal–Wallis test with Dunn’s correction.

Differences between groups were considered statistically significant when P < 0.05.

We previously showed that murine neural precursors cultured in vitro undergo limited proteolysis of the N-terminal tail of H3cs1 upon induction of differentiation and that CSTB modulates the cleavage event, with altered cleavage in CSTB-deficient cells (Daura et al., 2021). Moreover, we reported that H3cs1 during in vitro neurogenesis is mediated by cysteine protease cathepsins B and L. The increase in H3cs1 was also evident in embryonic mouse brain extracts. Here we investigated the significance of this mechanism in postnatal mouse brain.

We first asked whether histone tail cleavage occurs in a spatiotemporally regulated manner in the central nervous system. To characterize the distribution of H3cs1 in the wild-type mouse brain, we performed immunohistochemistry of serial brain sections at five timepoints spanning from early infancy (P7) to adulthood (P120). We quantified the intensity of H3cs1 in three anatomically distinct brain regions: the primary somatosensory cortex (S1), the striatum, and the cerebellum (Figure 1A). H3cs1 occurred at varying levels from P7 to P21, peaking at P14 in the primary somatosensory cortex and, to a lesser extent, in the striatum. In the cerebellum, we detected steadily high levels of H3cs1 from P7 to P21. On the contrary, H3cs1 was virtually absent at P30 and P120 in all brain regions. These observations indicate that H3cs1 normally occurs during postnatal mouse brain development and maturation.

In P14 brain, practically all NeuN positive neurons and GFAP positive astrocytes contained H3cs1 in their nuclei (Supplementary Figure 1). Instead, Olig2 positive oligodendrocyte and the IBA1 positive microglia/macrophage lineages showed both positive and negative cells. These data imply that H3cs1 is subjected to cell-type specific regulation in vivo. Examination of individual cell nuclei revealed that H3cs1 is largely excluded from DAPI-dense heterochromatin clusters (Figure 1B). This is in line with previous reports, which indicate that H3cs1 localizes in open chromatin regions implicating its regulatory role in gene expression (Santos-Rosa et al., 2009; Cheung et al., 2021).

We next sought to uncover the cysteine proteases, which cleave H3cs1 in mouse brain. We first analyzed subcellular cytoplasmic and chromatin fractions from P14 and P120 brains of wt mice by western blotting. In line with the immunohistochemistry analysis (Figures 1A,B), H3cs1 was clearly visible in the P14, but not in P120 chromatin fractions (Figure 1C). We did not detect the 15 kDa H3cs1 band in chromatin fractions using an antibody against the H3 C-terminus (Figure 1C), but a faint band corresponding to H3cs1 was observed in histone extracts from P14 brain (Supplementary Figure 2), indicating that H3cs1 only corresponds to the small proportion of the total H3cs1 in the wt mouse brain. We next probed the cytoplasmic and chromatin fractions with antibodies against three cysteine proteases with a previously reported nuclear localization, cathepsin L (Goulet et al., 2004), cathepsin B (Hämälistö et al., 2020), and cathepsin F (Maubach et al., 2008). Cathepsins B and F were only observed in lysosome-enriched cytoplasmic fractions, whereas cathepsin L was also present in chromatin fractions (Figure 1C). In cytoplasmic fractions, we detected three cathepsin L signals matching the pro-cathepsin L (39 kDa), and the mature single-chain (30 kDa) and double-chain (25 kDa) forms of cathepsin L (Ishidoh et al., 1998). In chromatin fractions, we detected the mature double-chain cathepsin L and a shorter, 37 kDa form of pro-cathepsin L in accordance with previous studies showing that the nuclear targeting of cathepsin L is accomplished through the alternative translation of the Ctsl mRNA into a shorter pro-cathepsin L devoid of a lysosome import signal (Goulet et al., 2004; Burton et al., 2017). The chromatin-associated shorter pro-cathepsin L was more abundant at P14 than at P120, whereas the cytoplasmic, longer pro-cathepsin L was more abundant at P120 than at P14 (Figures 1C,D). These findings imply that the subcellular targeting of de novo synthesized cathepsin L shifts from nuclear to cytoplasmic (lysosomal) in the adult brain, suggesting a role for nuclear cathepsin L in brain development and maturation. No evident differences between P14 and P120 were detected in the amount of either cytoplasmic or chromatin-associated mature double-chain forms of cathepsin L.

We previously showed that overexpression of CSTB abolished H3cs1 in differentiating mouse neuronal precursors derived from Cstb^–/– mice indicating that CSTB acts as a modulator of H3cs1 tail proteolysis in neural cells (Daura et al., 2021). Here, we investigated the impact of CSTB deficiency on H3cs1 tail proteolysis in the postnatal mouse brain. We concentrated the analysis to the Cstb^–/– mouse cerebellum for its early onset and striking neuronal degeneration and glial activation (Pennacchio et al., 1998; Tegelberg et al., 2012). We performed immunohistochemical staining of H3cs1 in serial cerebellar sections from Cstb^–/– mice at five postnatal timepoints (P7, P14, P21, P30, and P120) covering postnatal cerebellar development and maturation and compared them to the wt. The pattern of H3cs1 intensity in the examined timepoints varied significantly between the genotypes (Figures 2A,B). In wt cerebellum, the H3cs1 level decreased from P14 to P120, whereas in Cstb^–/– cerebellum, H3cs1 initially decreased from P14 to P21, but then increased from P21 onward (Figures 1A, 2A,B). A western blot analysis of cerebellar lysates showed that, in wt mice, the level of H3cs1 decreased by half, whereas in Cstb^–/– mice, it increased fourfold from P14 to P30 (Figure 2C). As an increase in Cstb mRNA expression is temporally correlated with a decrease in H3cs1 during in vitro differentiation (Daura et al., 2021), we asked the question whether the same correlation is also seen in postnatal brain and analyzed Cstb mRNA expression in P14, P30, and P120 wt mouse cerebellum by qPCR. We detected an increase in the levels of Cstb mRNA from P14 to P30 (Figure 2D), concomitant with the decrease in H3cs1 levels (Figures 2A–C) implying that CSTB functions as negative regulator of H3cs1 also in vivo.

Next, we sought to clarify whether the regulatory function of CSTB on H3cs1 could be mediated through its inhibitory effect on cathepsin L activity. We measured the enzymatic activity of cathepsin L in whole tissue lysates and in chromatin-associated protein fractions from wt and Cstb^–/– mice brains at P30 (Figure 2E). At the whole tissue level, cathepsin L activity did not differ between the genotypes, but it was higher in chromatin fraction of Cstb^–/– than of wt brain (Figure 2E). These findings suggest that CSTB serves as a chromatin-specific inhibitor of cathepsin L in vivo.

Histone H3 tail cleavage by cathepsin L intensifies during cellular senescence. Importantly, the ectopic overexpression of cleaved histone H3 is sufficient to induce senescence in fibroblasts and melanocytes (Duarte et al., 2014). We investigated whether sustained H3cs1 is associated with cellular senescence in the Cstb^–/– mouse brain. We focused the analysis on the cerebellum, the brain region that presents the highest overall H3cs1 both in wt and in Cstb^–/– mice (Figures 1A, 2A).

We first evaluated the acidic lysosomal beta-galactosidase (SA-β-Gal) activity, which has been considered an archetypical senescence-associated biomarker (Dimri et al., 1995), to elucidate whether mice presented signs of cellular senescence at the emergence of neurodegeneration (P30) or in the presence of significant neuron loss (P180). When compared to wt mice, an increase in the number of SA-β-Gal positive cells was detected in Cstb^–/– mice (Figure 3A). The increase was observed in the granule cell layer at P30 and at P180, and in the molecular layer at P180. Moreover, the SA-β-Gal positive cell count increased from early symptomatic P30 to late symptomatic P180 Cstb^–/– cerebellum.

To investigate whether CSTB-deficient mouse brains show signs of cellular senescence before the first signs of neuron loss and onset of myoclonus at P30, we focused our further analyses on the time window from the P14 to P30. H3cs1 is highest in both genotypes at P14, the timespoints when the first signs of gliosis appear in Cstb^–/– brain (Tegelberg et al., 2012), while at P30, H3cs1 is downregulated to very low levels of wt adult brain but remains high in Cstb^–/– brain (Figure 2A). To evaluate whether Cstb^–/– cerebellum displays enlarged cell nuclei, another well-established hallmark of cellular senescence (Mitsui and Schneider, 1976), we measured in tissue sections of Cstb^–/– and wt mice the nuclear area of cerebellar granule cells, the cell type that show marked progressive loss in Cstb^–/– brain (Pennacchio et al., 1998). We found that cerebellar granule cell nuclei were on average 26% larger in Cstb^–/– than in age-matched wt controls (Figure 3B). This phenotype was observed at P14, P21, and P30, indicative of changes in nuclear architecture manifesting at least 2 weeks prior to the symptomatic onset of the disease. However, lamin B1, a major component of the nuclear lamina, whose downregulation is usually linked to nuclear swelling during cellular senescence (Freund et al., 2012), showed no differences between the genotypes either in mRNA or protein levels, nor in the cleavage pattern (Figure 3C and Supplementary Figure 3).

Next, we performed a series of immunohistochemical analyses to look for expression changes in cyclin-dependent kinase (CDK) inhibitors, a group of proteins with an essential role in inducing the stable proliferative arrest characteristic of cellular senescence (reviewed in Martínez-Zamudio et al., 2017). CDK inhibitors p27 and p16^INK4a showed no differences between the genotypes (data not shown). Increased expression of p21^CIP1 was observed in glial fibrillary acidic protein (GFAP) positive Bergmann glia at P14 (Figure 3D). In wt mice, p21^CIP1 positive cells occurred in low frequency, suggesting that a subpopulation of these radial glia exits in cell cycle during normal cerebellar development. In Cstb^–/– mice, the number of p21^CIP1 positive cells was increased sixfold in comparison to wt controls (Figure 3D). However, by P30, no p21^CIP positive cells we observed in either genotype.

To find out whether Cstb^–/– mice present altered or persistent DNA-damage responses, another characteristic feature of senescent cells (Sedelnikova et al., 2004; Collin et al., 2018), we analyzed the level and staining pattern of histone variant H2AX phosphorylated at Ser139 (γH2AX), a DNA double strand break response marker (Kuo and Yang, 2008). Western blot analysis of whole-tissue lysates showed that the levels of γH2AX decreased significantly in both genotypes after the second postnatal week, in accordance with previous findings (Barral et al., 2014), with no apparent differences between genotypes (Supplementary Figure 4A). The number of cells presenting a pan-nuclear γH2AX immunoreactivity in wt cerebellum decreased after P14 (Figure 3E and Supplementary Figure 4B). On the contrary, in Cstb^–/– mice, the pan-nuclear γH2AX-positive cell count was constantly elevated in all three timepoints analyzed (Figure 3E and Supplementary Figure 4B). Pan-nuclear γH2AX responses mediate DNA damage events coupled to cell death by apoptosis (Ding et al., 2016). Accordingly, we observed that, in many of these cells, the γH2AX signal organized into an apoptotic ring conformation, an early marker of apoptosis (Solier and Pommier, 2009; Supplementary Figure 4C). Further analysis revealed that the cells displaying pan-nuclear γH2AX immunoreactivity were either Olig2 positive oligodendrocytes or GFAP positive astrocytes, but not NeuN positive neurons (Figure 3E).

In Cstb^–/– mouse brain, microglia become activated before the onset of clinical symptoms (Tegelberg et al., 2012). Therefore, this cell type is likely to play an important, yet so far unknown role in the disease pathogenesis. A growing body of evidence indicates that many chronic neurodegenerative diseases involve reactivation of microglial proliferation (Olmos-Alonso et al., 2016), a pathologic feature that eventually elicits microglial senescence by proliferative exhaustion (Hu et al., 2021). To assess the potential effect of CSTB-deficiency on microglia senescence through proliferative exhaustion (Hu et al., 2021), we quantified the proportion of IBA1 positive microglia expressing the proliferation antigen Ki-67 in cerebellar tissue sections from wt and Cstb^–/– mice (Figure 3F). In wt cerebellum, the percentage of dividing microglia decreased from approximately 16% at P14 to 1% at P21 and at P30, in line with previous findings (Nikodemova et al., 2015). In Cstb^–/– cerebellum, the proliferation of microglia was approximately twofold higher than in age-matched wt controls at P14 and presented no significant decline at P21 or P30.