Benzoxazole-derivatives enhance progranulin expression and reverse the aberrant lysosomal proteome caused by GRN haploinsufficiency

Abstract

Heterozygous loss-of-function mutations in the GRN gene are a major cause of hereditary frontotemporal dementia. The mechanisms linking frontotemporal dementia pathogenesis to progranulin deficiency are not well understood, and there is currently no treatment. Our strategy to prevent the onset and progression of frontotemporal dementia in patients with GRN mutations is to utilize small molecule positive regulators of GRN expression to boost progranulin levels from the remaining functional GRN allele, thus restoring progranulin levels back to normal within the brain. This work describes a series of blood-brain-barrier-penetrant small molecules which significantly increase progranulin protein levels in human cellular models, correct progranulin protein deficiency in Grn^+/- mouse brains, and reverse lysosomal proteome aberrations, a phenotypic hallmark of frontotemporal dementia, more efficiently than the previously described small molecule suberoylanilide hydroxamic acid. These molecules will allow further elucidation of the cellular functions of progranulin and its role in frontotemporal dementia and will also serve as lead structures for further drug development.

Fig. 1. Small molecules induced PGRN protein production in a dose dependent manner in vitro and in vivo.

A Molecule structure and chemical name of active compounds. B Immunoblots of active compounds with increasing doses(Δ). Compound abbreviation listed to the side of each respective blot. C Quantification of treatment replicates. Biological replicates of n = 6 for C15, C19, C40 and C41, n = 4 for C66 and n = 3 for C105, C107, C116 and C127 are represented as mean value ± SD and *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; All significance calculated using Brown-Forsythe and Welch ANOVA with Dunnett’s T3 multiple comparisons test. D, E Seven days after pump implantation, Grn^+/− mice treated with C40, C41 and C127 showed rescue of PGRN protein levels compared to Grn^+/+ mice. (Grn^+/+ n = 4, Grn^+/− treated with ACSF n = 17, Grn^+/− treated with drug n = 4/drug) F, G I.P. delivery of 10 mg/kg C40 significantly increased PGRN protein levels in Grn^+/− mouse brains (n = 4) compared to vehicle treated Grn^+/− mice (n = 5) after 24 h but did not return PGRN levels to Grn^+/+ levels. Grn^+/+ mice (n = 5) treated with vehicle serve as a control for healthy PGRN levels. For D, G All biological replicates are summarized in mean values ± SD and * represents p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 with respect to Grn^+/+ ACSF; # represents p < 0.05, ^##p < 0.01, ^###p < 0.001, ^####p < 0.0001 with respect to Grn^+/− ACSF; ordinary one-way ANOVA with Tukey’s multiple comparisons test (E) or Dunnett’s multiple comparisons test (G) for significance was used.

Fig. 2. Structural optimization of the lead compound C40.

A Subdivision of lead compound C40 into core and linker region, R1 is the area altered in the first optimization round, R2 represents the area altered in the second optimization round after R1 was chosen. B Selected structures of functional groups used for R1 substitution. The dark red box indicates the moiety chosen in the first round of optimization based on the activity screen in. C Here, all analogs of the first optimization round of C40 were assessed for progranulin induction through the luciferase-based reporter assay after 48 h of treatment. Treatment concentration of 3 µM is shown normalized to the average value of DMSO controls (n = 4 biological replicates are summarized as mean ± SD). D In optimization round two, based on A03, R2 was substituted with shown functional groups. E Activity of new analogs were assessed via the luciferase-based reporter assay after 24 h of treatment and 3 µM treatment concentration normalized to the average value of DMSO controls is shown (n = 3 except for A24, A8, A29, A33, which are n = 2, biological replicates are summarized as mean ± SD). F Predicted LogS and LogP values of each analog of both optimization rounds as an indicator of solubility. Solubility scale of LogS and Lipophilicity scale of LogP are indicated, increasing blue color depicts increasing solubility. Predictions were made with SwissADME and the consensus LogP was used. Initial compounds are shown in black, new lead analogs are shown in red, first round of molecules are colored in pink, second round of molecules are colored in blue. G Principal component analysis of predicted molecular features of all analogs including measured luciferase activity. Closest analogs to A03 are circled. PCA was conducted with ClustVis and used parameters were predicted by SwissADME. H Hierarchical clustering analysis of predicted and measured molecular features of all analogs conducted via ClustVis. Rows are centered and variance scaling is applied. Correlation distance and average linkage is used for clustering both rows and columns.

Fig. 3. Small molecules induce GRN mRNA and PGRN protein in human cells.

A Western blots of U-373 cells treated with analogs and A03 (n = 4) positive control. B Quantification of western blots from section A. Both treated concentrations (30 µM and 100 µM) of A21 (n = 6), A40 (n = 5) and A41 (n = 6) increased PGRN by more than two-fold in U-373 cells. Cells showed no response to A39 (n = 5). Biological replicates are represented as mean values ± SD and an ordinary one-way ANOVA with Dunnett’s multi comparisons test was used to calculate significance. * represents p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 compared to DMSO treated control (n = 7). C, D Western blots of HDF cells generated from a WT control patient (HDF X) and 2 different GRN haploinsufficient patients (HDF Y and 2). HDF cells were treated with A21 and A41 and probed for full length PGRN and processed GRN fragments 2/3. (*HDF Y A21 and A41 were run on same gel so DMSO is same. See source data for more information). E Quantification of HFD blots. Changes in GRN mRNA, PGRN protein and GRN fragments 2/3 protein quantified. A21 treatment (HDF X mRNA: all n = 3, PGRN and GRN2/3: DMSO n = 6, A21 treatment n = 5; HDF Y mRNA: DMSO n = 3, 50 µM n = 2, 100 µM n = 3, PGRN and GRN2/3: all n = 3; HDF (2) mRNA: DMSO n = 2, A21 treatment n = 3, PGRN and GRN2/3: all n = 5) and A41 treatment (HDF X mRNA: all n = 3, PGRN and GRN2/3: DMSO n = 3, A41 treatment n = 5; HDF Y all n = 3; HDF(2) mRNA: DMSO n = 4, A41 treatment n = 3, PGRN: DMSO and 100 µM n = 5, 50 µM n = 6, GRN2/3: all n = 5) are shown as mean values ± SD analyzed with two-way ANOVA with Dunnett’s multiple comparisons test each patient cell set individually. * represents p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 compared to DMSO treated control.

Fig. 4. Small molecule analogs lead to delayed induction of Grn mRNA and PGRN protein in vitro and in vivo.

A, B Western blot and quantification of Grn^+/− mice treated i.p. with analog. A21 (n = 6) and A41 (n = 7) significantly increased PGRN protein in the brain compared to Grn^+/− control. A21 and A41 also completely rescued Grn^+/− PGRN protein back to Grn^+/+ levels while A39 (n = 4) and A40 (n = 4) had no effect. All biological replicates are summarized in mean values ± SD and * represents p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 significance with respect to Grn^+/− control (n = 5); # represents p < 0.05, ^##p < 0.01, ^###p < 0.001, ^####p < 0.0001 significance with respect to Grn^+/+ control (n = 5); ordinary one-way ANOVA with Dunnett’s multiple comparisons test was used to determine significance. C N2A PRGN of A41 treatments at increasing time points (n = 2) with compound and after compound removal. Compound significantly increased PGRN protein and GRN mRNA levels at 18 and 24 h. Statistically significant increased levels of PGRN protein persisted for 12 h after compound removal. Biological replicates are shown as mean values ± SD and ordinary one-way ANOVA with Dunnett’s multiple comparisons was used to determine significance. * represents p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 in respect to DMSO treated control (n = 4). D, E Immunoblot and quantification of ^Grn+/− mice treated with 50 mg/kg A41 i.p. at increasing time points (n = 3 for each time point). A41 significantly upregulated PGRN protein at 20 and 24 h after treatment compared to controls (n = 6). (24 h data points included in quantification for Fig. 4B) Biological replicates are shown as mean values ± SD and ordinary one-way ANOVA with Dunnett’s multiple comparisons was used to determine significance. * represents p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 in respect to DMSO treated control (n = 7).

Fig. 5. LysoIP TMT-MS reveals loss of key lysosomal proteins in progranulin-deficient mouse and patient-derived fibroblasts.

A Diagram of 3xHA-tagged TMEM192 used for LysoIP. B Experimental design for LysoIP experiments. C Representative western blot of LysoIP experiment demonstrating specific isolation of lysosomes. Gene ontology of MEF (D) and HDF (E) LysoIP-derived proteins demonstrating enrichment of proteins under the KEGG term ‘lysosome’. Volcano plots of LysoIP isolated lysosomal proteomes from Grn^+/− MEFs (n = 6 Grn^+/+, n = 6 Grn^+/−) (F) and GRN^+/− HDFs (n = 3 GRN^+/+, n = 6 GRN^+/−) (G). H GRN^+/− vs GRN^+/+ fold changes of whole-cell lysate proteins correlate with fold changes in LysoIP-derived proteins (p = 1.16 × 10⁻⁵⁰; simple linear regression). I LysoIP-derived proteins are significantly altered in mouse (n = 6 Grn^+/+, n = 6 Grn^+/−) and human (n = 3 GRN^+/+, n = 6 GRN^+/−) progranulin haploinsufficient cell types (*q < 0.05 Two-stage step-up Benjamini, Krieger, and Yekutieli Test). All error bars represent stardard error of the mean.

Fig. 6. Progranulin-boosting drugs correct aberrant lysosomal proteome.

A, B Western blot validation demonstrates processed TPP1 levels in GRN^+/− HDFs is reduced, and 24-h treatment with 50 µM A41 upregulates TPP1 levels (n = 4 GRN^+/+, n = 5 GRN^+/−, n = 5 GRN^+/− + A41) (*p = 0.03; **p = 0.005 two-sided Student’s t-Test). C RT-qPCR demonstrates upregulation of GRN, TPP1, and DPP7 levels by 24- and 72 h treatment with 50 µM A41 (n = 5 GRN^+/+, n = 6 GRN^+/− + A41 24’, n = 5 GRN^+/− + A41 72’) (**p < 0.01; ***p < 0.001; One-way ANOVA, Holm-Šídák correction). D Representative fluorescent microscopy images of TFEB-GFP (green) and DAPI (blue) in 293 T cells treated with various concentrations of A41 or EBSS (positive control). Micrographs represent 3 independent experiments. E Quantification of TFEB-GFP nuclear/cytoplasmic ratios (*p = 0.014; ****p = 1.0 × 10⁻¹⁵; One-way ANOVA, Holm-Šídák correction). F Volcano plot of lysosomal proteomes from Grn^+/− MEFs treated for 24 h with 100 µM A41 compared to vehicle. G PCA depicting variations between DMSO treated Grn^+/+, DMSO or A41 treated Grn^+/−, and DMSO treated Grn^−/− MEF lysosomal proteomes. H Volcano plot of lysosomal proteomes from GRN^+/+ HDFs treated for 72 h with 50 µM A41 compared to vehicle. I PCA depicting variations between DMSO treated GRN^+/+ and DMSO or A41 treated GRN^+/− HDF lysosomal proteomes. J A41 treatment corrects the levels of several key lysosomal proteins in Grn^+/− MEF lysosomes (n = 6 Grn^+/−, n = 6 untreated Grn^+/−, n = 3 A41 treated Grn^+/−) (*q < 0.05; Two-stage step-up Benjamini, Krieger, and Yekutieli Test). K A41 treatment rescues the lysosomal expression of several proteins reduced in GRN^+/− HDFs (n = 5 GRN^+/−, n = 6 untreated GRN^+/−, n = 5 A41 treated GRN^+/−) (*q < 0.05; Two-stage step-up Benjamini, Krieger, and Yekutieli Test). All error bars represent standard error of the mean.