Glial cells are functionally impaired in juvenile neuronal ceroid lipofuscinosis and detrimental to neurons

Homozygous Cln3 ^Δex1–6 mice (Cln3 ^−/− ) were used as a model of CLN3 disease [56] and cells isolated from early postnatal mice for tissue culture, as described below, and were also assessed histologically. For histological comparisons of the level of glial activation, homozygous Tpp-1-deficient mice (Tpp-1 ^−/− ) were used as a model of CLN2 disease (Late Infantile NCL) [84]. Wild type (WT) mice on the same strain (C57BL/6 J) background were used as controls. All animal housekeeping and procedures were carried out according to the UK Scientific Procedures (Animals) Act (1986). Cln3 ^−/− mice were analyzed histologically at 6.5 months (early symptomatic), 12 months (disease mid stage), and 22 months of age (severely affected), and Tpp-1 ^−/− mice histologically at 4 months of age (severely affected).

To investigate glial activation in the mouse brain, frozen sections from Cln3 ^−/− , Tpp-1 ^−/− and WT mice were prepared as previously described [8, 40, 68, 69]. To investigate glial activation in the human NCL brain, paraffin-embedded tissue blocks were prepared from the primary visual cortical region of human CLN2 and CLN3 autopsy tissue (n = 2 for each type of NCL), and cut into 8 μm sections, as previously described [18, 90]. Both mouse and human sections containing the primary visual cortex were immunostained with antibodies to glial fibrillary acidic protein (GFAP, 1:1000 for mouse tissue, 1:5000 for human tissue, rabbit polyclonal, Dako) to identify activated astrocytes and Cluster of Differentiation 68 (CD68, 1:150, Rat monoclonal, Serotec) to identify activated microglia [54, 59, 71]. Immunostaining was detected using VECTASTAIN Elite ABC Reagent (Vector Laboratories) and DAB substrate (Sigma) and human sections counterstained with hematoxylin [18, 90].

Microglial cells were activated by exposure to lipopolysaccharide (LPS, 1 μg/ml LPS, Sigma), while astrocytes were activated by exposure to LPS plus interferon-gamma (IFN-γ, 100 U/ml, Thermo Scientific) [12, 14]. The ability of mutant and WT glia to respond similarly to LPS and IFN-γ was assessed by studying the nuclear translocation of the downstream phosphorylated proteins, NF-κβ subunit P65 (P-P65) [22] or STAT1 (P-STAT1) [39] respectively, using phospho-specific primary antibodies (P-P65, 1:100; P-STAT1, 1:50, both from Cell Signaling).

Cultures were immunostained using standard protocols (see [9]). Where appropriate, nuclei were counterstained with DAPI (4′-6-Diamidino-2-phenylindole, 0.5-1 μg/ml, Sigma) and coverslips mounted using either Fluoromount G or Prolong gold (Southern Biotech). The composition of all cultures was assessed using cell-type specific markers. GFAP (rabbit polyclonal, 1:500, Dako) or glutamate synthetase (rabbit polyclonal, 1:500, Abcam) was used to identify astrocytes, O4 (monoclonal antibody, 1:100, Covance) to identify oligodendrocytes, CD68 (Rat monoclonal, 1:500, Serotec) to identify microglia and MAP2 (monoclonal antibody, 1:1000, Abcam) and/or NeuN (monoclonal antibody, 1:100, Chemicon) to identify neurons. For cytoskeletal analysis, phalloidin was used to visualize F-actin filaments and α- and β- tubulin antibodies to visualize microtubular organization (monoclonal and polyclonal antibodies respectively, both from Sigma and used at 1:1000). All secondary antibodies were obtained from Invitrogen, and used at a dilution of 1:1000 (Alexa 488, 546, 633 and biotinylated antibodies) or 1:5000 (Alexa 790, 680). Immunofluorescently stained cells were visualized using a Zeiss AxioImager Z1 fluorescence microscope (Carl Zeiss, Ltd) with a monochrome AxioCamMR3 camera using AxioVision 4.8. Imaging software (Carl Zeiss, Welwyn Garden City).

The overall cytotoxicity in co-cultures was evaluated by measuring lactate dehydrogenase (LDH) release using a Cytotox 96 assay kit (Promega) according to manufacturer’s instructions. Total LDH content (100% LDH) was determined by lysing cultures in 0.1% Triton X-100 for 30 min, and LDH release from cells was expressed as a percentage of total LDH (%LDH) in each sample. To reveal the identity of the cells undergoing cell death, a live/dead fixable cellular marker conjugated to a red fluorochrome (Invitrogen) was used, according to the manufacturer’s instructions, in association with relevant cell-type specific markers.

Astrocyte cultures were immunostained with GFAP and images of 10 random fields of cells, whose processes were not overlapping, were taken and cell soma size measured using ImageJ (National Institutes of Health, Bethesda, MD). The average cell soma size of Cln3 ^−/− astrocytes was normalized to the corresponding values from WT astrocytes. To assess the morphological response of microglia to activation, cells were classified into 3 subcategories [98]: type 1 cells – microglia with extended processes (non-activated); type 2 cells – microglia with retracted processes (partly activated); type 3 cells– rounded cells with a small soma (fully activated), and the percentage of each morphological type present (determined from counting 10 random fields per culture) was calculated for each culture condition.

The quantitative analysis of the levels of proteins secreted by Cln3 ^−/− and WT glial cells grown under basal conditions and at various time points (between 6 h and 96 h) after activation with LPS/IFNγ was carried out by Myriad RBM (Austin, TX, USA, RodentMAP version 2.0 cytokine analysis). Simultaneous analysis of 59 different proteins was carried out on three different biological samples for each treatment per genotype using an automated quantification system. The values obtained were normalized to the relative number of cells in the culture from which the medium was collected, as determined by counting DAPI stained nuclei.

The intracellular levels of reduced (GSH) and oxidized (GSSG) glutathione was determined in WT and Cln3 ^−/− astrocytes. Samples for these measurements were generated from stimulated (for 24 or 48 h) and non-stimulated cultures by trypsinization and resuspension of cells in 300μl of isolation medium (320 mM sucrose, 10 mM Tris, 1 mM EDTA, pH 7.4), and the sample split into two for testing. One half was used to quantify GSH levels by separating this antioxidant from other components within the sample using reverse-phase high performance liquid chromatography (HPLC) followed by detection using an electrochemical method [33]. The GSH levels obtained were normalized to the total amount of protein, as determined using a Lowry protein assay (Thermo Scientific). The other half of the sample was used to determine the presence of GSSG. This was carried out by treating samples with glutathione reductase (GR) in the presence of reduced nicotinamide adenosine dinucleotide phosphatase (NADPH) to convert GSSG to GSH [86], and the level determined by HPLC as before. This gives a measure of the total glutathione within the cell. The difference between total glutathione concentration and GSH concentration was then used to calculate the concentration of GSSG. Finally, glutathione levels in the culture medium were measured using the GSH-Glo glutathione assay kit (Promega), according to manufacturer’s instructions. In some experiments, the effect of actin depolymerisation on glutathione secretion was assessed by treating WT astrocytes with Cytochalasin D (1 μM), an inhibitor of actin polymerization.

Fluctuations in the levels of intracellular Ca²⁺ ([Ca²⁺]_i) were examined to determine the ability of WT and Cln3 ^−/− astrocytes to generate calcium waves when exposed to ATP (100 μM) as described previously [60]. To measure intracellular Ca²⁺ levels, cells were loaded with 5 μM of Fura-2-acetoxymethyl ester (Fura-2 AM, Invitrogen), which is a membrane permeable derivative of the ratiometric calcium indicator Fura-2, for 30 min at room temperature and excess reagent removed by washing. Fluorescence measurements were carried out at room temperature using an epifluorescence inverted microscope equipped with a 20X fluorite objective over a 30–45 min period. [Ca²⁺]_i was monitored in single cells, with the excitation light provided by a Xenon arc lamp, using a monochromator (Cairn Research) to excite fluorescence sequentially at 340, 380 nm (all at 10 nm bandwidth). Using a long pass filter from 510 nm, the emitted fluorescence light was reflected and then transferred to a frame transfer cooled CCD camera (Hamamatsu Orca ER).

The glutamate clearance capacity of WT and Cln3 ^−/− astrocytes was determined using a Glutamate Assay kit (Abcam), according to manufacturer’s instructions. Values were normalized to the amount of total protein in each sample, determined using a BCA protein assay kit (Thermo Scientific).

The ability of Cln3 ^−/− astrocytes to migrate was assessed by performing a scratch wound assay. A scratch was made in confluent astrocyte cultures grown on Essen Image Lock 24-well plates using an Essen Wound-maker, generating an 800–900 μm wide cell-free region. Cultures were then placed in the IncuCyte live cell imaging system (Essen) and the wound width measured every hour for 24 h [62]. The rate of migration was obtained by measuring the width of the existing wound over time.

Neurite complexity was analyzed in P0 cortical neuron cultures from WT and Cln3 ^−/− mice after 7 DIV using ImageJ software to analyze immunofluorescence images of MAP2-positive cortical neurons, measuring 40 cells per genotype per experiment. The number of primary, secondary, and tertiary neurites present on each neuron was counted, the area of its cell soma measured, together and the total length of all primary neurites and the length of the longest primary neurite (assumed to represent the axon). Similar measurements of neurite complexity and soma size were also obtained from neurons co-cultured with glial cells.

All quantitative data was collected using Microsoft Excel spreadsheets, and analyzed using Graphpad PRISM. Where appropriate the data were normalized to values from untreated WT cultures. Most frequently, to allow comparisons of groups, one-way ANOVA with Bonferroni correction was used to test for statistical significance. However, when two groups were compared with each other a Student’s T-test was used. In general, three technical replicates were used, and independent experiments were repeated at least three times (unless otherwise stated). Data was presented as mean ± SEM and changes were considered significant with a p-value of ≤0.05. P-values ≤0.05 marked with *, P-values ≤0.01 marked with **, P-value ≤0.001 marked with ***.

In moderately affected Cln3 ^−/− mice the reactive response of glia, judged by hallmark morphological changes, appears attenuated compared to earlier onset forms of NCL [68, 69]. To investigate this possibility further we extended our analysis to more aged and severely affected Cln3 ^−/− mice, and also investigated the extent of glial activation in the same cortical region in human CLN3 disease autopsy tissue.

We compared the extent of gliosis in the primary visual cortex (V1) of Cln3 ^−/− mice, (from 6.5–21 months of age) with that of wildtype controls and Tpp-1 ^−/− mice (at 4 months of age, representing disease end stage), a model for CLN2 disease [84], an earlier onset and more rapidly progressing type of NCL [4]. In V1 of these severely affected Tpp-1 ^−/− mice, the morphological features characteristic of reactive astrocytosis were evident within these astrocytes, with intense GFAP immunoreactivity, thickened processes and pronounced hypertrophy (Fig. 1a). This astrocytosis in Tpp-1 ^−/− mice displayed laminar specificity, being most pronounced in laminae II and III, V and VI.

In comparison, in V1 of Cln3 ^−/− mice, GFAP immunoreactivity revealed a markedly different extent of astrocytosis, with substantial differences in astrocyte morphology. Apart from a small population of darkly immunostained astrocytes present in the most ventral portion of lamina VI, the majority of astrocytes in V1 of presymptomatic 6.5 month old Cln3 ^−/− mice displayed a protoplasmic morphology with numerous thin processes, compared to the characteristic appearance of fully activated astrocytes in Tpp-1 ^−/− mice (Fig. 1a). Initially largely confined to laminae I and IV, these protoplasmic astrocytes became more widespread within V1 of Cln3 ^−/− mice with disease progression, being present in lamina II and most of laminae V and VI at 12 months of age, and additionally in lamina IV by 21 months of age. Although the intensity of GFAP immunoreactivity progressively increased with time, only a small proportion of these astrocytes (mostly within lamina VI) displayed thickened processes or an enlarged cell soma, with most retaining a protoplasmic appearance with thin processes, even in severely affected 21 month old Cln3 ^−/− mice. These data suggest that at least a subset of Cln3 ^−/− astrocytes do retain the ability to transform morphologically, but only in a protracted manner towards disease end-stage.

A similarly attenuated activation of microglia was also evident in the cortex of Cln3 ^−/− mice (Fig. 1b). Although CD68 immunoreactive microglia were clearly more darkly immunostained in V1 of Cln3 ^−/− mice than in WT controls at all ages examined, these microglia retained a relatively small cell soma and numerous long thin processes. Only in severely affected Cln3 ^−/− mice by 21 months, were more overtly swollen and activated microglia were seen, and these were largely restricted to the more ventral portion of lamina VI and V. Even then, many of these swollen CD68 positive cells retained long thickened processes. This is in marked contrast to the V1 of 4-month-old Tpp-1 ^−/− mice that were at disease end-stage, where numerous intensely immunostained, completely rounded, and fully activated amoeboid microglia were evident.

Taken together, these data suggest that loss of Cln3 impairs the morphological transformation of both astrocytes and microglia, which is limited and only occurs late in the disease, in more severely affected Cln3 ^−/− mice. Having described a relatively limited glial response in the hippocampus of human CLN3 disease cases [90], we next investigated whether a similar phenotype was also evident in the primary visual cortex of CLN3 disease patient brain autopsy material (Fig. 1c). Numerous intensely immunostained GFAP positive hypertrophied astrocytes with many thickened processes were observed in V1 of CLN2 cases. In contrast, only a few weakly immunostained GFAP expressing astrocytes with thin processes were present in this region of CLN3 disease cases (Fig. 1c, top row). Similarly, dramatically fewer CD68 positive microglia were observed in V1 of CLN3 disease cases compared with CLN2 disease cases (Fig. 1c, bottom row).

These morphological observations suggest that the basic biology of glia, the neuronal support cells, may be impaired in CLN3 disease. To begin investigating how loss of CLN3 expression could influence glial cell function, we compared the properties of Cln3 ^−/− and WT glial cells using primary cultures of either astrocytes or microglia, having first defined the composition of our astrocyte (Additional file 1: Figure S1 and Additional file 2: Figure S2) and microglial monocultures (Additional file 3: Figure S3). One week after plating microglial cultures showed over 99% of DAPI stained cells expressed CD68 (99.97 ± 0.02% and 99.97 ± 0.02% CD68 + ve in WT and Cln3 ^−/−, respectively; with only 0.03 ± 0.02% and 0.03 ± 0.03% being GFAP + ve). One week after plating over 98% of DAPI positive cells in our astrocyte cultures were positive for GFAP (WT: 98.80 ± 0.28% GFAP + ve, 1.90 ± 0.17% CD68 + ve, and 0.10 ± 0.10% O4 + ve; Cln3 ^−/−: 98.86 ± 0.10% GFAP + ve, 1.05 ± 0.13% CD68 + ve, and 0.10 ± 0.03% O4 + ve). With subsequent time in culture an increased fraction of DAPI stained, but GFAP negative cells were apparent in our astrocyte cultures, especially those from WT mice (e.g. see Fig. 3A.). However, at these later time points virtually all these DAPI labelled cells immunostained positively for a second astrocyte marker glutamine synthetase (see Additional file 2: Figure S2 for an example taken 48 h later), suggesting a dynamic down-regulation of GFAP over time under some culture conditions. Nevertheless, a minor contamination of our astrocyte cultures with ependymal or endothelial cells cannot be excluded.

To determine whether the attenuated morphological response of Cln3 ^−/− glia observed in vivo was mimicked in vitro, we stimulated monocultures of either microglia and astrocytes and assessed their ability to undergo morphological changes. To do this we exposed cultures to standard inflammatory stimuli that up-regulate pathways associated with immune and injury-related functions [38], either the bacterial endotoxin lipopolysaccharide (LPS) alone to activate microglia, or to LPS combined with interferon-γ (IFN), which synergizes with the effects of LPS, to activate astrocytes [26].

We first confirmed that Cln3 ^−/− glia were able to activate relevant downstream signaling pathways (phosphorylated NF-κβ subunit p65 and phosphorylated STAT-1, as downstream effectors of LPS and IFNγ stimulation respectively, [22, 94]) (Additional file 4: Figure S4). Next, to characterize the morphological transformation of microglia, WT and Cln3 ^−/− microglial cultures were immunostained with CD68 at various time-points after activation (2, 6, 12, 24, 48, 72 and 96 h), dividing these cells according to their morphology into Type 1 (non-activated), Type 2 cells (partly activated) and Type 3 cells (fully activated) (see methods). Even under basal conditions, CD68 immunoreactivity appeared more intense in Cln3 ^−/− vs. WT microglial cultures, with more rounded cells present (Fig. 2A). Upon stimulation, microglia of both genotypes morphologically transformed, but many fewer Cln3 ^−/− microglia changed shape after 24 h (Fig. 2A). When these changes were quantified (see Fig. 2B), there were more type 2 cells in Cln3 ^−/− vs. WT microglial cultures under basal conditions (Fig. 2C, compare panels a and b), suggesting a higher level of basal activation, but a slower morphological transformation of CLN3 disease microglia. A transformation of Type 1 cells into Type 2 cells occurred in microglial cultures of both genotypes upon stimulation, however, Cln3 ^−/− microglia responded more slowly than WT microglia, with a slower decline in the number of Type 1 cells (Fig. 2C, a, b) and a slower increase in the number of Type 2 cells (Fig. 2C, compare panels c and d). Until 48 h very little change was observed in the percentage of Type 3 cells under any condition (Fig. 2C), but by 72 h there was a dramatic increase in the proportion of this fully activated cell type within both WT and Cln3 ^−/− microglial cultures under all conditions (Fig. 2c–f). This change was accompanied by a reduction in the percentage of both Type 1 and Type 2, suggesting a morphological transformation into Type 3 cells with increased time in culture.

The morphological response of astrocytes to stimulation (LPS/INFγ treatment for 24 h or 48 h) was assessed in GFAP immunostained cultures. Even under basal conditions, untreated Cln3 ^−/− astrocytes had a strikingly different morphology to WT astrocytes, appearing larger and flatter, with disrupted intermediate filaments (Fig. 3). Upon stimulation, WT astrocytes already began to morphologically transform after 24 h; changing from broad, non-process bearing, flat cells into cells with a shrunken soma and multiple branched processes (as described in [53]) (Fig. 3A, c arrowheads). These changes become more apparent with time (Fig. 3A, e). In contrast, no significant morphological transformation of Cln3 ^−/− astrocytes could be detected until 48 h stimulation, when soma size began to decrease and some cells developed processes (Fig. 3A, f). To quantify these changes the soma size of WT and Cln3 ^−/− astrocytes were compared (Fig. 3B). After activation for 24 h or 48 h, the cell soma of WT astrocytes became smaller, and this was statistically significant after 24 h (30.5% ± 3.3 decrease). After 24 h of stimulation the soma size of Cln3 ^−/− astrocytes remained unchanged, but after 48 h of stimulation was not statistically different to that of stimulated WT astrocytes (Fig. 3C).

These data demonstrate that Cln3 ^−/− astrocytes and microglia are attenuated in their ability to change their morphology upon stimulation, suggesting that these cells retain at least some of their in vivo disease characteristics when cultured.

Since morphological changes require cytoskeletal rearrangements, and GFAP immunostaining suggested that intermediate filament organization was perturbed in Cln3 ^−/− astrocytes (Fig. 3A), we also immunostained astrocytes for α- and β-tubulin to visualize microtubules and with phalloidin to visualize F-actin filaments and a similar cytoskeletal analysis was performed with microglia.

Both intermediate filaments and F-actin filaments appeared less defined and highly disorganized in Cln3 ^−/− vs. WT astrocytes (Fig. 4, a, b for GFAP, and c, d for F-actin, examples marked with arrowheads). Most Cln3 ^−/− astrocytes lacked F-actin filaments that spanned the cell body, a common morphological feature of cultured WT astrocytes (Fig. 4c, d). However, the α- and β-microtubular organization of WT and Cln3 ^−/− astrocytes appeared similar (Fig. 4, e, f and g, h, examples marked with arrows). These data reveal that enlarged Cln3 ^−/− astrocytes have an abnormally organized actin and intermediate filament cytoskeleton, whilst their microtubule organization appears normal. No overt changes were observed in the cytoskeletal organization of Cln3 ^−/− microglia (data not shown).

The secretion of soluble factors is a key feature of both microglia and astrocytes under both physiological and pathological conditions [51], therefore we compared the secretion profiles of WT and Cln3 ^−/− glia under basal conditions and after stimulation.

Cln3 ^−/− and WT microglia did not display any differences in the levels of factors secreted under basal conditions (data not shown), but five proteins were secreted at significantly lower levels by Cln3 ^−/− microglia upon LPS stimulation (Fig. 5). These proteins included three chemokines (MIP-1-γ, MIP-2 and RANTES), a glycoprotein (vWF) and a matrix metalloproteinase (MMP-9). These data indicate that Cln3 ^−/− microglia retain their capacity to secrete the majority of soluble factors into their environment. Equally, we observed no difference in the phagocytic properties of Cln3 ^−/− microglia grown under basal conditions or after stimulation (data not shown).

Under basal conditions there was no significant difference between WT and Cln3 ^−/− astrocytes in the secretion pattern of the majority of proteins analyzed (Additional file 5: Table S1). However, upon exposure to LPS/IFN-γ a broad range of secreted proteins were detected at significantly reduced levels in Cln3 ^−/− astrocyte supernatants (Additional file 5: Table S2). Indeed, out of the 59 screened factors none (at 6 h), 19 (at 24 h) and 42 (at 72 h) factors were secreted at significantly lower levels by Cln3 ^−/− astrocytes compared to WT astrocytes upon activation (Additional file 5: Table S1 and S2). These included the reduced secretion of several mitogens (M-CSF, IL-3, FGF2, GM-CSF, FGF-9, TPO and IL-5), chemokines (Eotaxin, MIP1α, MCP-3, MCP-1, KC/GROα, MIP-3α, MIP-2, IP-10, RANTES, MDC, MCP-5, MIP-1γ, GCP-2 and MIP-1β), anti- and pro-inflammatory cytokines (IL-17α, IL-6, IL-12p70, IL-1α, TNFα, IL-1β, IL-10, IL-2). Furthermore, these cells also showed an impaired ability to secrete a range of proteins shown to have neuroprotective properties: IL-6 [95], IL-3 [101], GM-CSF [55], IL-10 [7], LIF [57], MCP-1 and RANTES [15].

There were however exceptions to this pattern of altered protein secretion. Two chemokines, macrophage inflammatory protein-1γ (MIP-1γ) and granulocyte chemotactic protein 2 (GCP-2) were both secreted significantly less by Cln3 ^−/− astrocytes under basal conditions, while one protein, tissue factor (TF), was secreted more by Cln3 ^−/− astrocytes under basal conditions. In contrast fibrinogen and the mitogen CRP were secreted significantly more by Cln3 ^−/− astrocytes upon stimulation. Examples of astrocyte secretion profiles for specific proteins are shown in Fig. 6.

The altered secretion profile of Cln3 ^−/− astrocytes led us to consider that these cells may also fail in their antioxidant support of neurons, particularly in their ability to secrete glutathione (GSH), one of the brain’s most important neuroprotective factors [31, 78].

To analyze the ability of astrocytes to make and secrete GSH, cultures were washed, lysed and the intracellular levels of the reduced (functional) form of GSH measured. Under basal conditions, no significant differences in GSH levels were detected between WT and Cln3 ^−/− cultures (Fig. 7a). However, following stimulation the intracellular levels of reduced GSH decreased significantly in WT cultures (56.3 ± 12.0% reduction after 24 h exposure, 59.9 ± 12.6% reduction after 48 h, Fig. 7a), but were markedly increased in Cln3 ^−/− cultures (63 ± 33.6% increase after 24 h, 28.1 ± 19.3% increase after 48 h, Fig. 7a). Thus, intracellular GSH levels in LPS/IFNγ treated WT astrocytes were significantly lower than in LPS/IFNγ treated Cln3 ^−/− astrocytes after 24 (59.6 ± 12.2% decrease in WT vs. Cln3 ^−/− ) and 48 h (51.9 ± 14.3% decrease in WT vs. Cln3 ^−/− ). These data suggest a failure of Cln3 ^−/− astrocytes to secrete GSH, consistent with the observation that significantly higher levels of extracellular GSH could be detected in the medium from WT vs. Cln3 ^−/− astrocytes after stimulation (Fig. 7b).

Since the method described above only measures the reduced form of glutathione, it is conceivable that these results could be explained by an opposing change in the level of oxidized glutathione (GSSG). This possibility was excluded by indirectly measuring intracellular GSSG levels by converting GSSG to GSH using glutathione reductase (GR). Subsequent HPLC analysis revealed no measurable difference between the intracellular levels of GSH and total glutathione (after GR treatment) in any of these samples, suggesting that all the intracellular glutathione in both WT and Cln3 ^−/− astrocytes is present in the reduced form (data not shown).

The actin cytoskeleton is important for exocytosis in astrocytes [70], and it appears abnormally organized in Cln3 ^−/− astrocytes (Fig. 3A). Therefore, the possible connection between the disrupted actin cytoskeleton and impaired glutathione secretion was examined by treating WT astrocytes with cytochalasin D (1uM for 30 min before the 8 h measurement period) to inhibit the polymerization of actin. This resulted in a significant reduction (50.1 ± 12.7%) in the levels of extracellular glutathione in LPS/IFNγ treated WT astrocytes (Additional file 6: Figure S5). Thus, a normal actin cytoskeleton is essential for glutathione secretion by astrocytes.

Since the defects in Cln3 ^−/− astrocytes appeared more profound than those in microglial cells, and the cytoskeletal disruption observed in these cells could impact many of their functions, we investigated whether Cln3 ^−/− astrocytes could perform other key tasks effectively.

Qualitatively, the distribution of MAP2 immunoreactivity appeared different in neurons of different genotypes, appearing to be more intense within the apparently smaller cell soma of Cln3 ^−/− neurons compared to the more even distribution within the soma and processes of WT neurons (Fig. 11A). Cell area measurements revealed Cln3 ^−/− neuron soma to be significantly smaller than WT cells (Fig. 11B), and their neurite complexity was altered (Fig. 11C). Although Cln3 ^−/− neurons had slightly more branching points than WT neurons, these differences were not statistically significant. However, the length of the longest primary neurite was significantly shorter in Cln3 ^−/− neurons (Fig. 11C, e), which also displayed a significantly shorter network of primary neurites (Fig. 11C, f).

The finding that Cln3 ^−/− astrocytes, and to a lesser extent microglia, have a compromised biology suggested that Cln3 ^−/− glia could potentially have a detrimental effect on neuronal health. To test this hypothesis, primary cortical WT or Cln3 ^−/− astrocytes and microglia were seeded on top of WT or Cln3 ^−/− primary cortical neuronal cultures. Glia in these co-cultures were not exposed to LPS or LPS/IFNγ. As a readout of neuronal health we analyzed the same neuronal phenotypes defined above, and assessed if there were any impact upon neuronal survival, and neurite complexity in these co-cultures.

Co-cultures were monitored for 7 days, and over this time, those composed of WT mixed glia and WT neurons remained healthy, with very few dying cells present (Fig. 12A, panel a). However, from day 2 onwards MAP2 immunoreactivity in processes of WT neurons grown with Cln3 ^−/− glia became increasingly punctate, suggesting compromised neuronal health, and by day 7 many dying neurons with red nuclei were observed (Fig. 12A, panel b). The morphology of the surviving WT neurons was also dramatically altered, showing a reduction in soma size and neurite complexity (Additional file 6: Figure S5). The co-culture combination of Cln3 ^−/− neurons with Cln3 ^−/− glia was the most detrimental for neuronal health, since by day 7 most of the Cln3 ^−/− neurons in these co-cultures were either dead (Fig. 12A, panel d), or appeared severely compromised, as judged by their morphology (Additional file 7: Figure S6). Interestingly, WT glia had a positive influence on both the survival and morphology of Cln3 ^−/− neurons (Fig. 12A, panel c).

In these co-cultures, we observed that in the presence of Cln3 ^−/− neurons, Cln3 ^−/− astrocytes had smaller cell bodies and longer, more numerous processes (reminiscent of activated astrocytes in culture), when compared to Cln3 ^−/− astrocytes grown with WT neurons (Additional file 8: Figure S7). No such morphological change was evident when Cln3 ^−/− neurons were co-cultured with WT glia, suggesting that Cln3 ^−/− astrocytes are more sensitive to the environment than their WT counterparts. Under all culture conditions the morphology of microglia were heterogeneous with some cells bearing processes and others being fully rounded (data not shown).

These morphological findings correlated well with measurements of released LDH from the different co-cultures (Fig. 12B). The lowest LDH levels were observed when WT glia and neurons were co-cultured, but these levels increased dramatically when Cln3 ^−/− glia were co-cultured with WT or Cln3 ^−/− neurons (Fig. 12B). However, when Cln3 ^−/− neurons were co-cultured with WT, rather than Cln3 ^−/− glia, a lower level of LDH release was observed, possibly due to the supportive influence of the WT cells (Fig. 12B). As might be expected, there was significantly more LDH released in Cln3 ^−/− glia/Cln3 ^−/− neuron co-cultures than in WT glia/WT neuron co-cultures (Fig. 12B).

These results suggest that Cln3 ^−/− glia are detrimental to the health of both WT and Cln3 ^−/− neurons, with Cln3 ^−/− neurons being the most vulnerable. In contrast, WT glia appeared to have a positive influence on Cln3 ^−/− neurons, not just on survival, but also upon neurite complexity.

Despite concerted efforts, the normal function of CLN3 remains poorly understood and it is unclear how its deficiency relates to cellular dysfunction, including that of astrocytes or microglia. Despite microglia accumulating large amounts of storage material, which is also present in astrocytes, the current view is that it is not the accumulation of storage material per se that directly causes cellular dysfunction and death. Instead it appears that other, as yet unknown, consequences of Cln3-deficiency are responsible. Our data suggest that these negative consequences of Cln3-deficiency are also evident in glia, rather than being confined to neurons, and it will be important to gain in vivo correlates of the data we have found in tissue culture.

Nevertheless, all the biological defects we found associated with cultured Cln3 ^−/− astrocytes and microglia can plausibly be linked to known features of CLN3 disease pathogenesis, including the potential involvement of glutamate mediated excitotoxicity and oxidative stress. Indeed, although in vitro systems do not necessarily accurately reflect the in vivo situation, a series of similarities between our tissue culture observations and other reports exist. For example, the attenuated ability of Cln3 ^−/− glia to respond morphologically to stimulation is also evident in the Cln3 ^−/− mouse brain in vivo (Fig. 1, and [68, 69]), and a comparatively lower level of glial activation is evident in human CLN3 disease ([90], this study). This is in marked contrast to the robust glial activation and morphological transformation observed in all other forms of murine [24, 40, 47, 59, 64, 77, 93], ovine [61], or human NCL [90].

The impaired ability of Cln3 ^−/− astrocytes to take up extracellular glutamate is consistent with the reduced expression of the glutamate receptor EAA2 in human CLN3 brain tissue [36], the reduced GLAST and glutamine synthetase levels evident in Cln3 ^Δex7/8 mice [16], and presynaptic elevation of glutamate in Cln3 ^−/− mice [20]. Furthermore, neurons in these mice appear particularly vulnerable to AMPA-and NMDA-receptor stimulation [32, 43], perhaps because of excitotoxicity due to these elevated levels of glutamate, and different classes of glutamate antagonists provide some therapeutic benefit in Cln3 ^−/− mice [41‐44].

Another pivotal astrocyte function is the synthesis and secretion of the anti-oxidant glutathione (GSH) that plays a crucial role in protecting neurons against oxidative stress [31, 33]. Our data reveal that Cln3 ^−/− astrocytes can still make, but fail to secrete, glutathione. Indeed, Drosophila lacking CLN3 function are hypersensitive to oxidative stress [89], and oxidative damage has also been reported in both mouse [6] and human CLN3 disease [3].

Our calcium signaling studies revealed that Cln3 ^−/− astrocytes fail to generate a calcium wave after exposure to ATP, providing further evidence that intercellular signaling between CLN3 astrocytes may be compromised [16], and it will be important to study calcium signaling in acute slice preparations. This could impact upon the control of neurotransmission in the JNCL brain and perhaps contribute to the seizure activity observed in this disease [58, 67, 88].

All these functional problems associated with Cln3 ^−/− astrocytes and some of the phenotypes seen in vivo may at least partially be explained by their disrupted cytoskeleton, since expression of glutamate receptors at the cell surface [48], calcium signaling among astrocytes [27] and secretion by astrocytes [45] have all been shown to require a functional actin cytoskeleton. Indeed, the altered shape of Cln3 ^−/− astrocytes, along with the difficulties they exhibit in changing their morphology in vivo and in our culture, may plausibly results from their disrupted cytoskeleton and it will be important to study this in more detail and determine their importance in vivo. How these defects in the cytoskeleton are related to Cln3-deficiency is unclear, but a functional interaction of CLN3 with non-muscle myosin-IIB has been reported [34], and a migration defect in Cln3 ^−/− mouse embryonic fibroblasts that is consistent with our novel data for the impaired migration of Cln3 ^−/− astrocytes.

The protein secretion profiles of both Cln3 ^−/− astrocytes and microglia was altered following activation, with astrocytes being more severely affected, showing significantly reduced levels of secretion of a range of proteins (Additional file 5: Tables S1 and S2). Our data are consistent with the reported evidence that LPS stimulation also results in a lower level of cytokine secretion by microglia derived from Cln3 ^Δex7/8 mice bearing the 1 kb deletion that is present in most CLN3 disease cases [99]. Intriguingly, these authors suggest that the responses of Cln3-deficient microglia are stimulus-dependent, with ceramide or neuronal cell lysates resulting in an increased inflammasome activation and expression of a wide array of proinflammatory cytokines and chemokines [99]. However, it should be borne in mind that these authors used Cln3 ^Δex7/8 ‘knock-in’ mice rather than the Cln3 ^−/− mice used in our study, and this may influence the different phenotypes observed.

The biological significance of our data showing altered secretion profiles of Cln3-deficient glia remains unclear, but the downstream effects are likely to be complex given the multiple and possibly synergistic effects of secreted proteins on different cell types under both physiological and pathological situations [1, 19, 73, 74, 83]. It is also important to emphasize that what we have detected in vitro may not reflect the in vivo situation. Nevertheless, our data raise the possibility that cell-cell communication via secreted factors may potentially be perturbed in the CLN3 disease brain. In addition, many of the neuroprotective proteins routinely secreted by WT astrocytes after activation [15, 30, 57, 76, 95, 100], are also significantly reduced in cultures of activated Cln3 ^−/− astrocytes, with two of these proteins (MCP-1 and RANTES) also being secreted at significantly lower levels by Cln3 ^−/− microglia. The reduced expression of IL-6, RANTES and MCP-1, which can protect neurons against NMDA receptor-mediated excitotoxicity, may be especially relevant given the increased sensitivity of Cln3-deficient neurons to AMPA receptor-mediated excitoxicity [43, 65]. Thus, defects in glial-glial and glial-neuronal interactions have the potential to have a significant impact on neuronal health in CLN3 disease, a suggestion that prompted us to grow mixed glial co-cultures with neurons.

Until in vivo data regarding the relative levels of chemokines and cytokines become available, our in vitro data demonstrating altered secretion levels should be interpreted with caution, especially as these data come from pharmacologically stimulated cultures. However, the reduction in chemokine secretion by stimulated Cln3 ^−/− glia in culture may also have a detrimental effect on the recruitment of microglia to sites of inflammation [72, 81], and partly explain the limited infiltration of monocytes and lymphocytes in CLN3 disease [50]. This reduced chemokine expression may also be associated with the attenuated microglial activation observed in vivo ([68, 69, 90], this study). Conversely, Cln3 ^−/− astrocytes showed a reduced ability to secrete anti-inflammatory cytokines, such as IL-4, IL-10 and IL-2, which could also prove harmful. Both genetic and pharmaceutical approaches to attenuate the adaptive immune response have been shown to result in a significant improvement in the pathology of Cln3 ^−/− mice [80].

Defects in glial biology have been associated with neuronal dysfunction and loss in many neurodegenerative diseases, see [28, 29, 66, 75, 85]. Both positive and negative roles for astrocytes have been proposed, and recently, more active roles for astrocytes and microglia in directly influencing neuron survival have been postulated [49]. Using a co-culture approach, we have shown here that Cln3 ^−/− astrocytes and microglia can indeed influence neuronal health, affecting the size and neurite complexity of both WT and Cln3 ^−/− neurons, but also causing the death of the latter, which appear to be inherently compromised by Cln3 deficiency. From our data it is not clear whether it is the Cln3 ^−/− astrocytes or microglia, or a combination of both cell types that negatively influence neuronal heath. It has been suggested that astrocytes can be primed by microglia to become toxic to neurons [49], and it will be important to determine if similar mechanisms operate in CLN3 disease, especially in an in vivo context. However, it is apparent that despite any overt intrinsic survival defect in these short-term cultures, Cln3 ^−/− neurons appeared to be compromised in terms of their morphology, and it will be important to investigate their functional status.

In other lysosomal storage disorders, introducing astrocyte-specific gene mutations is sufficient to harm neurons [21, 28], and correcting these defects is beneficial [102]. In our studies co-culturing Cln3 ^−/− neurons with healthy glia improved many of their morphological defects, and resulted in an increase in their survival, suggesting that healthy glia appear to have a neuroprotective effect. This is in marked contrast to the apparently negative influence of Cln3 ^−/− glia upon both healthy and mutant neurons. It remains to be seen how accurately our data from cultures reflect the in vivo situation, and for this reason we have generated cell-type specific mutant mice in which we can inactivate Cln3 in defined cell types in our future studies. Nevertheless, our data from this in vitro study suggests that therapies that target glia in addition to neurons may be an important step forward in treating this devastating disease.