Restoring neuronal progranulin reverses deficits in a mouse model of frontotemporal dementia

Abstract

Loss-of-function mutations in progranulin (GRN), a secreted glycoprotein expressed by neurons and microglia, are a common autosomal dominant cause of frontotemporal dementia, a neurodegenerative disease commonly characterized by disrupted social and emotional behaviour. GRN mutations are thought to cause frontotemporal dementia through progranulin haploinsufficiency, therefore, boosting progranulin expression from the intact allele is a rational treatment strategy. However, this approach has not been tested in an animal model of frontotemporal dementia and it is unclear if boosting progranulin could correct pre-existing deficits. Here, we show that adeno-associated virus-driven expression of progranulin in the medial prefrontal cortex reverses social dominance deficits in Grn+/- mice, an animal model of frontotemporal dementia due to GRN mutations. Adeno-associated virus-progranulin also corrected lysosomal abnormalities in Grn+/- mice. The adeno-associated virus-progranulin vector only transduced neurons, suggesting that restoring neuronal progranulin is sufficient to correct deficits in Grn+/- mice. To further test the role of neuronal progranulin in the development of frontotemporal dementia-related deficits, we generated two neuronal progranulin-deficient mouse lines using CaMKII-Cre and Nestin-Cre. Measuring progranulin levels in these lines indicated that most brain progranulin is derived from neurons. Both neuronal progranulin-deficient lines developed social dominance deficits similar to those in global Grn+/- mice, showing that neuronal progranulin deficiency is sufficient to disrupt social behaviour. These data support the concept of progranulin-boosting therapies for frontotemporal dementia and highlight an important role for neuron-derived progranulin in maintaining normal social function.

Keywords: FTD; dementia; experimental models; gene therapy; progranulin.

Figure 1

AAV-Grn reverses social dominance deficits in Grn^+/– mice. Wild-type and Grn^+/– mice were screened in the tube test for social dominance to confirm the presence of a losing phenotype prior to AAV injection. This losing phenotype was evident based on the total number of wins per genotype and the distribution of winning percentages from individual mice (A), with wild-type mice winning the majority of matches overall, and exhibiting a higher per-mouse winning percentage. The mice were then injected with an AAV expressing mouse progranulin with a C-terminal myc tag (AAV-Grn) into the mPFC. Six weeks after injection, the mice were again tested in the tube test. AAV-GFP-treated Grn^+/– mice maintained their losing phenotype against AAV-GFP-treated wild-type mice (B), but AAV-Grn-treated Grn^+/- mice no longer exhibited this losing phenotype (C). Consistent with these data, AAV-GFP-treated wild-type mice had a significantly higher winning percentage versus AAV-GFP-treated Grn^+/– mice than AAV-Grn-treated Grn^+/– mice (D). Additionally, AAV-Grn-treated Grn^+/– mice exhibited a dominant phenotype against AAV-GFP-treated Grn^+/– mice (E), while AAV-Grn-treated wild-type mice did not exhibit a dominant phenotype against AAV-GFP-treated wild-type mice (F). The win totals for the tube test were analysed by binomial test and the winning percentage was analysed by Mann-Whitney test. The winning percentage of AAV-GFP-treated wild-type mice versus the two Grn^+/– groups was analysed by Wilcoxon matched-pairs signed rank test. n = 22–31 mice per group. ^*P < 0.05, ^**P < 0.01. NS = not significant.

Figure 2

AAV-Grn boosts amygdala function in Grn^+/– mice. AAV-Grn increased the central amygdala (CeA) response to a novel, social environment in Grn^+/– mice (A, ANOVA genotype × virus interaction, P = 0.0419). Data were analysed by ANOVA with Dunnett’s post hoc test. n = 11–15 mice per group. ^*P < 0.05, ^**P < 0.01. Representative images from the ×4 images used for cell counting are shown in B, D and F, with the central amygdala enlarged in C, E and G. The scale bar in B = 100 µm, C = 50 µm.

Figure 3

AAV-Grn normalizes prefrontal cortical LAMP1 levels in Grn^+/– mice. We observed elevated LAMP1 immunoreactivity in the mPFC of Grn^–/– mice (A, ANOVA effect of genotype, P = 0.001), and noticed a trend for a more subtle elevation in Grn^+/– mice (P = 0.18). To better measure LAMP1 immunoreactivity in Grn^+/– mice, we performed confocal microscopy. This method revealed that Grn^+/– mice also have elevated levels of LAMP1 in mPFC (B, P = 0.0226). AAV-GFP-treated Grn^+/– mice exhibited a similar trend as uninjected mice (C), and AAV-Grn significantly reduced LAMP1 levels (C, ANOVA, P = 0.0497). Similarly, both Grn^+/– and Grn^–/– mice had elevated LAMP1 levels in the mPFC as measured by western blot (D, ANOVA effect of genotype, P < 0.0001). AAV-GFP-treated Grn^+/– mice replicated the phenotype of uninjected Grn^+/– mice, and AAV-Grn normalized these elevated LAMP1 levels (E, ANOVA, P = 0.0234). Data were analysed by ANOVA, followed by t-test due to the differences in effect size in Grn^+/– and Grn^–/– mice. Analysis of confocal LAMP1 images was performed by t-test. n = 8–20 per group for imaging experiments, and 12–26 per group for western blot experiments. Representative ×40 images of LAMP1 immunostaining are shown in A with 30 µm scale bars, and representative ×60 maximum intensity projections of LAMP1 immunostaining are shown in B and C with 10 µm scale bars.

Figure 4

AAV-Grn transduces neurons. Brains from AAV-treated mice were immunostained for progranulin to assess the spread of AAV-expressed progranulin (A). Strong progranulin immunoreactivity was detected throughout the mPFC of AAV-Grn-treated mice, as well as in other medial structures including striatum, septum, and ventromedial thalamus. Progranulin ELISA was performed to more quantitatively assess progranulin levels in mPFC and amygdala (B and C). Both wild-type and Grn^+/– mice exhibited dramatic increases in progranulin levels in the mPFC (B, ANOVA main effect of virus, P < 0.0001), with a similar level of expression in each genotype. AAV-Grn-treated mice also had a modest, but statistically significant increase in progranulin in the amygdala, restoring progranulin levels to around normal in Grn^+/– mice (C, ANOVA main effect of virus, P = 0.0001). In both regions, AAV-GFP-treated Grn^+/– mice exhibited progranulin insufficiency as expected (B inset, C). Brains from AAV-treated mice were double-immunostained for the myc tag on virally-expressed progranulin and markers for neurons (D, NeuN), astrocytes (E, GFAP), or microglia (F, Iba1), revealing that of these three cell types, AAV-Grn only transduced neurons. Immunostaining and ELISA data were analysed by ANOVA. Tukey’s post hoc test was used for immunostaining and amygdala ELISA data. Genotype differences within viral group from the prefrontal cortex ELISA data were analysed by t-test due to the large difference in progranulin levels between AAV groups. Scale bars in representative double-labelled images represent 5 µm. n = 10–20 mice per group for progranulin immunostaining and 6–16 per group for progranulin ELISA. ^**P < 0.01, ^***P < 0.001, ns = not significant.

Figure 5

CaMKII-Cre and Nestin-Cre selectively depletes neuronal progranulin. To confirm the expected neuron-specific progranulin depletion in CaMKII-Cre and Nestin-Cre N-KO mice, co-localization was assessed between progranulin and markers for neurons (NeuN, A and E) and microglia (Iba1, C and G). Co-localization was qualitatively scored between NeuN and progranulin in the frontal cortex, which showed a significant reduction in neuronal progranulin labelling with CaMKII-Cre (B, Chi-square, P < 0.0001). To avoid interference from neuronal progranulin, microglial progranulin labelling was scored in the corpus callosum (D, Fisher’s exact test, P > 0.9999), with no significant effect of CaMKII-Cre detected. Similarly, NeuN/progranulin co-localization was reduced in Nestin-Cre N-KO mice (F, Chi-square, P < 0.0001), while Iba1/progranulin co-localization was not changed (H, Fisher’s exact test, P = 0.6461). Representative ×60 double-labelled images are shown adjacent to the corresponding monochrome image of the green channel (progranulin), with 10 µm scale bars. For quantification of double-immunostaining, two images were counted per mouse, from four to six mice per genotype from each line. A total of 706–749 neurons and 91–97 microglia were counted per genotype in CaMKII-Cre N-KO mice. 356–609 neurons were counted for NeuN/progranulin co-labelling and 75–82 microglia were counted for Iba1/progranulin co-labelling in Nestin-Cre N-KO mice. Data were analysed by Chi-square test (for NeuN) and Fisher’s Exact test (for Iba1, used due to the presence of only two bins, as no microglia in either line had weak or absent progranulin immunolabelling). ****P < 0.0001; ns = not significant.

Figure 6

Neuronal progranulin depletion with CaMKII-Cre impairs social dominance, nesting, and the amygdala response to a novel, social environment, but does not produce gliosis or lipofuscinosis.CaMKII-Cre N-KO mice developed many of the same behavioural abnormalities that have been observed in global Grn^+/– mice, including a losing phenotype in the tube test versus Cre− littermates (A, P = 0.0325). CaMKII-Cre N-KO mice also exhibited a deficit in nesting behaviour (B, P = 0.0129). After exposure to a novel, social environment, CaMKII-Cre N-KO mice had fewer c-Fos-positive cells across the three major amygdala nuclei (C, ANOVA effect of genotype, P = 0.0142). Representative images of c-Fos immunostaining are shown with lines drawn to illustrate boundaries between amygdala nuclei (D). The boxed areas in the central amygdala are enlarged in the panels below the top images. The scale bars in the larger image in D represent 100 µm, and the scale bar in the smaller central amygdala image represents 50 µm. For assessment of pathology, brains from 20-month-old CaMKII-Cre N-KO mice and Cre− littermates were immunostained for markers of microgliosis (E, CD68) and astrogliosis (F, GFAP). The presence of autofluorescent lipofuscin granules was measured by imaging brain sections with an epifluorescence microscope (G). No differences were detected between CaMKII-Cre N-KO mice and controls (repeated measures ANOVA effect of genotype and genotype × region interaction were not statistically significant for all measures), showing that neuronal progranulin deficiency is not sufficient to produce pathology. A limited sample of brains from Grn^–/– mice were processed in parallel with the sections to serve as a positive control, but were not included in the statistical analysis given the small sample size. Representative ×20 images of CD68, GFAP, and lipofuscin are shown for each group. Scale bars = 50 µm. Nesting behaviour and tube test winning percentage were analysed by Mann-Whitney test, and c-Fos and pathology data were analysed by repeated measures ANOVA with Sidak’s post hoc test. n = 35–38 mice per group for the tube test, 26–29 per group for nesting, eight per group for c-Fos, and five to eight per group for pathology measures. BLA = basolateral amygdala; CeA = central amygdala; Ctx = cortex; HPC = hippocampus; MeA = medial amygdala; Thal = thalamus. ^*P < 0.05, ^**P < 0.01.

Figure 7

Neuronal progranulin depletion with CaMKII-Cre and Nestin-Cre significantly reduces total brain progranulin.CaMKII-Cre N-KO mice exhibited significant reductions in progranulin RNA in the frontal cortex and amygdala (A, ANOVA effect of genotype, P = 0.0096). Progranulin protein levels were also reduced in the frontal cortex (FC) and amygdala (AMG) of CaMKII-Cre N-KO mice (B, ANOVA effect of genotype, P = 0.0189). Grn RNA and protein levels were not reduced in the thalamus of CaMKII-Cre N-KO mice, which is consistent with low CaMKII-Cre expression in this brain region. Progranulin immunostaining (C and D) was used to provide a semiquantitative screen of progranulin levels across the brain, and revealed progranulin depletion in the cortex (cingulate and insula), hippocampus (CA1 and CA3), and basolateral amygdala of CaMKII-Cre N-KO mice (repeated measures ANOVA effect of genotype P = 0.0039, genotype × region interaction P < 0.0001). Nestin-Cre N-KO mice exhibited even stronger reductions in total progranulin levels throughout the brain. Total progranulin RNA (E, repeated measures ANOVA effect of genotype, P < 0.0001) and protein (F, repeated measures ANOVA effect of genotype, P < 0.0001) were reduced in frontal cortex (FC), amygdala (AMG), and thalamus (Thal). Analysis of progranulin immunoreactivity (G) showed strong reduction of progranulin throughout the forebrain in Nestin-Cre N-KO mice (H, repeated measures ANOVA effect of genotype, P < 0.0001, genotype × brain region interaction, P < 0.0001). ELISA and quantitative PCR data were analysed by ANOVA, with one-tailed t-test used to compare progranulin levels between Cre+ N-KO mice and Cre− littermates within each brain region. Immunostaining data were analysed by repeated measures ANOVA, followed by Sidak’s post hoc test. n = 8–17 per genotype for ELISAs and 9–10 per group for immunostaining. ^*P < 0.05, ^**P < 0.01, ^****P < 0.0001. ACc = nucleus accumbens core; ACs = nucleus accumbens shell; AMG = amygdala; ARC = arcuate nucleus; BLA = basolateral amygdala; CeA = central amygdala; Cg = cingulate cortex; DG = dentate gyrus; FC = frontal cortex; IC = insular cortex; MD = mediodorsal thalamic nucleus; MeA = medial amygdala; Thal = thalamus; VMH = ventromedial hypothalamus; VPM = ventral posteromedial thalamic nucleus.

Figure 8

Pan-neuronal progranulin deficiency with Nestin-Cre impairs social dominance, nesting, and the amygdala response to a novel, social environment, and produces cortical astrogliosis.Nestin-Cre N-KO mice developed a losing phenotype in the tube test for social dominance (A, P = 0.0007), and a trend for reduced nesting behaviour (B, P = 0.057). Nestin-Cre N-KO mice also exhibited impaired activation of the amygdala after exposure to a novel, social environment (C, repeated measures ANOVA effect of genotype, P = 0.0092), with specific impairments in the central (P = 0.0167) and medial (P = 0.0225) amygdala. Representative images of c-Fos immunostaining are shown with lines drawn to illustrate boundaries between amygdala nuclei (D). The boxed areas in the central amygdala are enlarged in the panels below the top images. The scale bars in the larger image in D = 100 µm, and the scale bar in the smaller central amygdala image = 50 µm. For assessment of pathology, brains from 18–19 month-old Nestin-Cre N-KO mice and Cre^– littermates were immunostained for markers of microgliosis (E, CD68) and astrogliosis (F, GFAP). The presence of autofluorescent lipofuscin granules was measured by imaging brain sections with an epifluorescence microscope (G). No differences were detected between in CD68 immunoreactivity (E) or lipofuscin accumulation (G) showing that neuronal progranulin deficiency is not sufficient to reproduce most of the pathology of Grn^–/– mice. However, Nestin-Cre N-KO mice exhibited cortical astrogliosis (F, repeated measures ANOVA genotype × brain region interaction P = 0.0012). A limited number of brains from Grn^–/– mice were processed in parallel with the sections to serve as a positive control, but were not included in the statistical analysis given the small sample size. Representative×20 images of CD68, GFAP, and lipofuscin are shown for each group. Scale bars = 50 µm. Tube test and nesting data were analysed by Mann-Whitney test, and c-Fos and pathology data were analysed by repeated measures ANOVA with Sidak’s post hoc test. n = 23–24 mice per genotype for the tube test, 24 per genotype for nesting, 16 per genotype for c-Fos, and 9–14 per genotype for pathology. BLA = basolateral amygdala; CeA = central amygdala; MeA = medial amygdala; Ctx = cortex; HPC = hippocampus; Thal = thalamus. ^*P < 0.05, ^**P < 0.01.