Development of a brain-penetrant G9a methylase inhibitor to target Alzheimer's disease-associated proteopathology

Abstract

Current Aβ-targeting therapeutics for Alzheimer's disease (AD) only slow cognitive decline due to poor understanding of AD pathogenesis. Here we describe a mechanism of AD pathogenesis in which the histone methyltransferase G9a noncanonically regulates translation of hippocampal proteins associated with AD pathology. Correspondingly, we developed a brain-penetrant inhibitor of G9a, MS1262, which restored both age-related learning & memory and noncognitive functions in multiple AD mouse models. Further, comparison of AD pathology-correlated mouse proteomes with those of AD patients found G9a regulates pathological pathways that promote Aβ and neurofibrillary tangles. This mouse-to-human overlap of G9a regulated AD-associated pathologic proteins supports at the molecular level the efficacy of targeting G9a translational mechanism for treating AD patients. Additionally, MS1262 treatment reversed the AD-characteristic expression or phosphorylation of multiple clinically validated biomarkers of AD that have the potential to be used for early-stage AD diagnosis and companion diagnosis of individualized drug effects.

Fig. 1. Investigating the role of G9a translational mechanisms in AD pathology.

A The overall design for deciphering G9a-mediated AD pathogenesis and the mechanism of G9a-target drug action. B ChaC-MS revealed a noncanonical function of G9a in the translational regulation of AD pathogenesis. The networks shared by 139 G9a (EHMT2) interactors that were identified in at least 3 of 4 ChaC experiments from 5xFAD mouse hippocampus and AD organoids (human). (See also Figures. S1-S3 & S6; Data S1A–C).

Fig. 2. MS1262 is a highly potent, selective, and brain-penetrant inhibitor of G9a and GLP.

A Discovery of MS1262. B Concentration-dependent inhibition of G9a (left) and GLP (right) by MS1262 in G9a and GLP enzymatic assays. UNC0642 was used as a control. Data shown are the mean ± SD from four independent experiments. C ITC titrations of MS1262 into G9a (top) and GLP (bottom). The calculated values represent the means ± SD from two independent experiments. D Activity of MS1262 against 21 other methyltransferases at 1 µM. Data are the means ± SD from two duplicate experiments. E Concentration-dependent reduction of the H3K9me2 level by MS1262 in K562 cells. K562 cells were treated with MS1262 at the indicated concentrations for 48 h. Western blot results are representative from at least two independent experiments. F Plasma and brain concentrations of MS1262 over 4 h following a single 5 mg/kg intraperitoneal injection of MS1262 in mice. Data shown are the mean ± SD from three tested mice per time point.

Fig. 3. Treatment of 5xFAD mice with MS1262 rescues behavioral deficits.

A Experimental timeline for drug administration and novel place recognition test. B Depiction of the paradigm used to test memory and affective related behavior. C Locomotion in an open field was unaffected under wild-type, 5xFAD, vehicle, and chronic MS1262 treatment. D Preference for the novel-located object during retrieval was significantly reduced in 5xFAD mice compared with wild-type controls and was completely rescued by chronic MS1262 treatment, as measured by discrimination ratio (see “Methods” for calculations). E Time spent in the center of an open field was unaffected under wild-type, 5xFAD, vehicle, and chronic MS1262. F MS1262 administration rescued anxiety-like behavior in 5xFAD mice back to wild-type levels demonstrated by increased time spent in the open arms of a zero maze. G Depressive-like behavioral deficits in 5xFAD mice were rescued to wildtype levels after chronic MS1262 treatment as measured by immobile time during the forced swimming paradigm. H Locomotion in an open field was unaffected under wildtype, APP^NLGF, vehicle, and chronic MS1262 treatment. I Preference for the novel-located object during retrieval was significantly reduced in APP^NLGF mice compared to wildtype controls and was rescued by chronic MS1262 treatment as measured by discrimination ratio. J Time spent in the center of an open field was unaffected under wildtype, APP^NLGF, while chronic MS1262 treatment led to an increase in time spent in the center. K MS1262 administration rescued anxiety-like behavior in APP^NLGF mice back to wildtype levels demonstrated by increased time spent in the open arms of a zero maze. L Depressive-like behavioral deficits in APP^NLGF mice were rescued to wildtype levels after chronic MS1262 treatment as measured by immobile time during the forced swimming paradigm. Data are visualized as mean +/- SEM with each individual displayed as a point. Wildtype (C57BL/6 J) (n = 7), 5xFAD (n = 10), 5xFAD vehicle (n = 14), 5xFAD MS1262 (n = 12) and Wildtype (C57BL/6J-A^w-J/J) (n = 7), APP^NLGF vehicle (n = 10), APP^NLGF MS1262 (n = 10), mice were utilized for behavioral studies. Significance was assessed by ANOVA and two-sided Tukey’s posthoc test between each condition. ns = not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (See also Figures. S4 & S5 & Data S3).

Fig. 4. Intermittent MS1262 treatment increases frequency but not amplitude of sEPSCs in DG granule cells of 5xFAD mice without altering intrinsic excitability.

A Quantification of membrane capacity between WT mice and 5xFAD mice treated with vehicle and MS1262 (n = 15/21/22 cells for WT/Vehicle/MS1262) ANOVA followed by Tukey’s test visualized as mean ± SEM. B Quantification of input resistance between WT mice and 5xFAD mice treated with vehicle and MS1262 (n = 15/21/22 cells for WT/Vehicle/MS1262) ANOVA followed by Tukey’s test visualized as mean ± SEM. C Quantification of resting membrane potential between WT mice and 5xFAD mice treated with vehicle and MS1262 (n = 15/21/22 cells for WT/Vehicle/MS1262) ANOVA followed by Tukey’s test visualized as mean ± SEM. D Mean ( ± SEM) number of action potential elicited in response to increasing step current from dentate granule cells of WT mice and 5xFAD mice injected with vehicle (black) or MS1262 (blue). (n = 15/21/22 cells for WT/Vehicle/ MS1262. ANOVA followed by Tukey’s test. E Representative traces of sEPSCs recorded from dentate granule cells derived from WT mice and 5xFAD mice injected chronically with vehicle (left) or MS1262 (right). F, G Cumulative probability (f) and average amplitude distribution (g) of sEPSCs (WT, n = 15 cells from 3 mice, 5xFAD-Vehicle, n = 16 cells from 3 mice; 5xFAD-MS1262, n = 15 cells from 3 mice) ANOVA followed by Tukey’s test visualized as mean ± SEM. H, I Cumulative probability distribution (h) and average frequency (i) of sEPSCs (WT, n = 15 cells from 3 mice, 5xFAD-Vehicle, n = 16 cells from 3 mice; 5xFAD-MS1262, n = 15 cells from 3 mice. P = 0.0004, Kruskal-Wallis test, Dunn’s test for multiple comparisons p < 0.0001 and p = .0453) visualized as mean ± SEM. J, K fEPSP slope (J) as measured as a percent of baseline and LTP visualized in a bar graph (k). (WT, n = 8 slices from 3 mice, 5xFAD-vehicle, n = 4 slices from 3 mice; 5xFAD-MS1262, n = 9 slices from 4 mice. ANOVA followed by Tukey’s test *p < 0.05, ****p < 0.0001 Data are visualized as mean ± SEM. (See also Data S3).

Fig. 5. MS1262 treatment reverses AD-driven changes to the hippocampal m⁶A methylome.

A Violin plot showing an increase in m⁶A level in the hippocampus of 5x-FAD mice compared to age-matched wildtype controls (AD vs. WT), which is reversed upon MS1262 treatment (AD-tr vs. AD). Hippocampus tissue from four wild type (WT) mice, four 5x-FAD (AD) mice, and four 5x-FAD mice treated with MS1262 (AD-tr) was used for m⁶A sequencing. A two-tailed paired t-test (****p < 0.0001) was used, with the number of transcripts (n = 3172) for which the m⁶A modification ratio could be calculated for both comparisons indicated at the top. B A three-way comparison showing that MS1262 treatment reverses AD-mediated changes in the m⁶A methylome of hippocampus. Hippocampus tissue from four wild type (WT) mice, four 5x-FAD (AD) mice, and four 5x-FAD mice treated with MS1262 (AD-tr) was used for m⁶A sequencing. Weighted fold-change, from all m⁶A peaks mapping to a gene, is shown for AD vs WT (x-axis) and AD-tr vs AD (y-axis) mice. Select transcripts with higher (red, n = 461) or lower (blue, n = 57) m⁶A-modification in the hippocampus of AD mice, compared to age matched wildtype controls, that was reversed following MS1262 treatment are highlighted. Linear regression showing overall trend (blue line with 95% CI highlighted by grey shading on either side), Pearson correlation ($R$) and associated P-value are shown. (C) Enrichment analysis for AD-dysregulated, MS1262 reversed, m⁶A modified transcripts highlighted in (B). Top 20 terms with significant over-representation are shown, with grey cells indicating a lack of significant enrichment. P-values were calculated using cumulative hypergeometric distribution followed by Benjamini-Hochberg correction to account for multiple testing. (See also Figure. S7 & Data S2).

Fig. 6. MS1262 inhibition of G9a reversed AD proteopathology (AD-correlated proteome and phosphoproteome).

A Venn diagrams summarizing total number of proteins and phospho-sites identified in 5x-FAD and APP-NLGF mouse models in this study. Briefly, global/phospho-proteomic experiments were performed using proteins extracted from hippocampus of age matched wild-type controls (WT; n = 3), 5x-FAD/APP^NLGF mice at mid/late-stage Alzheimer’s (AD; n = 4), and AD mice treated with MS1262 (AD-tr; n = 4). Samples were TMT-labeled and run in quadruplicate. B Bar charts summarizing number of proteins (global proteomics) and phospho-sites (phosphor-proteomics) showing statistically significantly (log2(FC) ± 0.2; p < 0.05; two-tailed unpaired Student’s t test) changes in comparison of AD vs WT and AD-MS1262 treated vs AD mice. C Heatmap summarizing pathway activity z-scores, calculated by Ingenuity Pathway Analysis (IPA) using differentially regulated proteins and phospho-sites shown in (B) for indicated comparisons (AD vs WT & AD-MS1262 treated vs AD) in 5x-FAD and APP-NLFG mice. MS1262 treatment reverses/ameliorates pathway activity changes seen in both mouse models of AD, when compared to age matched wild-type (WT) controls. D Heatmap summarizing MS1262-affected diseases and functions based on dysregulated proteins/phosphosites shown in (B). Z-scores were calculated using Ingenuity Pathway Analysis (IPA). E The network of indicated signaling pathways (see also Figures. S8-S9 & Data S1D–E).

Fig. 7. MS1262 treatment reversed expression or phosphorylation of protein markers of AD risk or early AD stage.

A Heatmap of select AD-risk markers showing patient-correlated protein expression in the hippocampus of 5x-FAD & APP^NLGF mice whose expression was reversed following MS1262 treatment. Global proteomics data from three age-matched wild-type controls (WT), four 5x-FAD/APP^NLGF mice at mid-/late-stage Alzheimer’s (AD), and four AD mice treated with MS1262 (AD-tr), with each sample run in quadruplicate. Rows are clustered into three groups based on pattern of inhibitor effect: (1) “down_up” = markers down in AD mice, compared to WT controls, whose expression is back up following MS1262 treatment, (2) “up_down” = markers whose expression increases in AD mice, compared to WT controls, but is back down upon treatment, and (3) “not_affected” = inhibitor did not affect expression of these markers. The left annotation column (mouse_model) shows the proteomics dataset (orange = 5x-FAD, green = APP^NLGF) a marker belongs to. Right annotation columns [log2(AD/WT) & log2(AD-tr/AD)] show fold change for indicated comparisons. Rightmost heatmap shows AD-marker expression in 39 patients from the Banner Sun cohort (LPC = 12, HPC = 6, MCI = 6, and AD = 15) that were pooled and measured in duplicate using TMT-LC-MS/MS. LPC = controls with low pathology of plaques and tangles; HPC = controls with high Aβ pathology but no detectable cognitive defects; MCI = mild cognitive impairment with Aβ pathology and a slight but measurable defect in cognition; AD = late-stage AD with high pathology scores of plaques and tangles; FC = fold change. B MS1262 reversed cerebrospinal fluid (CSF) proteome of early-stage AD. Heatmap showing AD markers (identified from CSF of symptomatic/non-symptomatic AD patients) dysregulated in 5xFAD and APP^NLFG mice, compared to DMSO treated wild-type controls, whose expression is reversed following MS1262 treatment. The annotation column on right indicates ‘up’ (red) or ‘down’ (blue) regulation of said marker in symptomatic AD patients compared to age matched non-symptomatic patients. Protein names highlighted in red have AD-dysregulated and G9a reversed expression pattern that shows mouse-to-human conservation. (see also Data S1F).

Fig. 8. MS1262 treatment of AD mice reversed AD patient proteome or phosphoproteome.

A Heatmap depicting AD/G9a co-regulated proteins/phosphoproteins showing dysregulation in 5x-FAD/APP-NLGF mice, compared to age-matched controls, whose expression/phosphorylation pattern is reversed following MS1262 treatment. Right annotation columns depict log2(fold-change) for indicated comparisons (AD/WT & AD-tr/AD). B, C Heatmaps summarizing results of GO/pathway enrichment (B) and disease & function (C) analyses for proteins and phospho-proteins belonging to the two clusters identified in (A). P-values were calculated using cumulative hypergeometric distribution followed by Benjamini-Hochberg correction to account for multiple testing. All genes in the genome were used as enrichment background. D Volcano plot illustrating the relationship between expression levels of various proteins and the Aβ peptides (residues 6–28) in the brain of AD patients (n = 488). The plot shows the biweight mid-correlation (bicor) coefficient and the corresponding BH adjusted p-value for each protein. Cluster 1 (blue) and cluster 2 (red) proteins from (A) are highlighted with point shape denoting proteins (circle) and phospho-proteins (triangle). Names of select AD markers are shown. E Network illustrating the relationship between AD/G9a-coregulated proteins/phosphoproteins identified in (A) and various AD-patient correlated modules (‘M’) that are dysregulated in the brains of AD patients. Modules showing dysregulation at protein level only, without concomitant change at the transcriptional level in AD patients, are highlighted using cyan outlines (see also Figure. S10 and Data S1D, F–H).