Introduction
Alzheimer’s Disease (AD), the most common form of senile dementia, is a neurodegenerative disease characterized clinically by irreversible memory loss and cognitive decline [
1]. The pathology of AD is comprised mainly of amyloid-β (Aβ) plaques [
2‐
5] and tau neurofibrillary tangles [
6]. However, the relationship of these pathological hallmarks to each other and to disease features such as cognitive decline and neurodegeneration remain incompletely understood [
7,
8].
Advances in understanding of chromatin architecture dynamics during normal aging and during neurodegeneration have revealed epigenomic dysregulation as a potential important link connecting aging, disease pathological features, and neurodegeneration [
9]. In one example, we and others have uncovered a disease-association for histone 3 acetylation at lysine residue 27 (H3K27ac), a histone modification catalyzed by the lysine acetyltransferases p300 (EP300) and CBP (CREBBP) [
10,
11]. H3K27ac is found primarily in enhancer regions, and promotes active transcription of genes [
12‐
14]. However, the balance of epigenomic H3K27ac gains and losses (genomic peaks measured using ChIPseq) is less clear, and indeed may be brain region-dependent and gene-dependent. Specifically, while we observed a substantial net increase in the number of H3K27ac peaks in the lateral temporal lobe of AD patients [
15], there was a net decrease in H3K27ac peaks in the entorhinal cortex (although a substantial number of H3K27ac peaks were still gained in AD patient samples compared to control) [
16]. Hence, the precise occurrence and role of enhancer-associated H3K27ac may be complex and gene-specific, and requires additional clarification.
Efforts to determine a functional relationship, whether disease-promoting or disease-ameliorating, between altered H3K27ac abundance and pathology are complicated by the complexity of brains and the difficulty in establishing causality in postmortem tissue samples, where pathology and epigenomic dysregulation are co-occurrent. We therefore sought to investigate these questions in an iPSC-neuron model. Methods for induction of cortical neurons from iPSCs (iPSC-neurons) through overexpression of the master neuronal transcriptional regulator NGN2 have been well-established to produce excitatory cortical neuronal cells with high speed, consistency, and efficiency [
17‐
19], and iPSC-neurons are increasingly utilized as a tractable model for neurodegenerative diseases [
19‐
22]. We leveraged these properties to generate direct epigenomic perturbations in a homogenous population of neurons, and to examine downstream effects on gene expression and AD-related pathology.
Here, we performed transcriptomic and functional characterization of iPSC-neurons derived from non-demented control (NDC) donors and from familial Alzheimer’s Disease patient donors harboring an amyloid precursor protein (APP) duplication (APPDup). Our results provide additional evidence for the important role played by chromatin-regulating histone acetylation in neurodegenerative disease, and reveal a homeostatic protective mechanism whereby neurons regulate gene transcription in response to AD pathology.
Discussion
There exist disparate lines of evidence regarding the nature of epigenomic dysregulation in neurodegenerative Alzheimer’s disease. We and others have reported differential H3K27ac in human brain tissue that is dependent on the brain region surveyed: H3K27ac increases in the lateral temporal lobe [
14], but decreases in the entorhinal cortex of AD patient brains [
15]. We also examined the histone acetylation mark H4K16ac and discovered a decrease in the brains of AD patients [
9], and studies in fly models of AD suggest a protective role of H4K16ac against AD pathology-related insults [
68,
69] Additionally, in a mouse model of AD, the histone deacetylase HDAC2 increases and H4K12ac decreases [
70], and an ameliorative effect was demonstrated in an AD mouse model of increasing acetyl-CoA synthetase 2 (ACSS2) [
71], which generates acetyl-coA and regulates histone acetylation in rodent hippocampus to promote memory [
72,
73]. These findings fuel speculation that histone deacetylase inhibitors could be potential therapeutics in Alzheimer’s disease and other neurodegenerative diseases, to overcome a disease-associated “epigenomic blockade” [
70,
74].
In this study, we utilized iPSC-neurons derived from familial AD patients with an APP duplication as a model to study the functional effect of reducing the key acetyltransferases EP300 and CBP. Knockdown of the acetyltransferases reduces acetylation at H3K27, and lowers acetylation of other histone and non-histone proteins [
10]. As we do not observe differences in EP300 or CBP expression between APP
Dup and NDC neurons (Additional file
2: Fig. S4A,B) and did not assess for gross changes in H3K27ac between APP
Dup and NDC neurons, we cannot determine if higher levels of H3K27ac at baseline is a fundamental characteristic of APP duplication. However, in our iPSC-neurons, EP300 or CBP knockdown reduced total H3K27ac abundance (Fig.
3D, E), which correlated with widespread downregulation of gene transcription in both APP
Dup and NDC neurons (Fig.
4A, B, Additional file
2: Fig. S5A, B), including important genes in Alzheimer’s disease pathways (Fig.
4F, G, Additional file
2: Fig. S5F, G, Additional file
2: Fig S6, Additional file
2: Fig. S7). Knockdown of the two enzymes resulted in broadly similar transcriptional changes (Fig.
4C), and learning and memory and neuron-related processes were affected in both EP300 and CBP knockdown (Fig.
4H, I). Previous studies of EP300/CBP in the context of AD focused primarily on the well-established roles played by these acetyltransferases in learning and memory [
75]. Our study uncovers a novel link between EP300/CBP and amyloid pathology in AD.
We found that a subset of genes in the APP-Aβ pathway responsible for amyloid clearance and prevention are strongly upregulated in APP
Dup neurons compared to NDC neurons (Fig.
6B), suggesting activation of a homeostatic genetic response to compensate for increased abundance of APP and production of downstream Aβ species. Compared to knockdown in NDC neurons, knockdown of either EP300 or CBP in APP
Dup neurons resulted in stronger downregulation of genes with amyloid-reducing function (Fig.
6C, D), including the APP
Dup-upregulated, compensatory genes mentioned above. Importantly, likely as a consequence of downregulation of these amyloid-reducing gene programs, knockdown of EP300 or CBP causes an increase in abundance of disease-associated Aβ42 secreted by APP
Dup neurons (Fig.
6E). Moreover, the knockdown of CBP appears to more strongly affect the amyloid reduction pathways compared to knockdown of EP300, which is especially interesting in light of our previous observation of disease-specific activation of CBP and not EP300 in postmortem AD brains [
15].
Our findings reveal several pathways potentially impacted by EP300/CBP KD leading to increased toxic Aβ42. LPL is strongly expressed by neurons [
76] and is involved in extracellular association and sequestration of amyloid-β. Interestingly, we discovered that LPL was among the most significantly downregulated genes in neurons with EP300/CBP KD (Fig.
4A, 4B). Consistent with this finding, a small molecule inhibiting EP300/CBP catalytic activity strongly decreases LPL expression in mouse adipocytes [
77]. LPL is an amyloid-β-binding protein promoting cellular uptake and clearance of amyloid-β in astrocytes [
63]. Taken together, downregulation of LPL by KD of EP300/CBP may lead to increased amyloid-β. Although glial cells are considered to be primarily responsible for uptake and clearance of amyloid-β [
22,
78], our results open the possibility that LPL may have glial cell-independent effects on amyloid reduction, warranting further analysis. Decreased expression of BDNF may have contributed to the increase in Aβ in EP300/CBP KD cells. BDNF reduces Aβ production [
49] by promoting α-secretase activity to the exclusion of β-secretase activity [
47,
48] We find that EP300/CBP knockdown reduces BDNF expression, suggesting it could be a candidate for amyloid-targeting therapeutic approaches.
While our transcriptional and phenotypic data suggest a physiologic amyloid-reducing role of EP300/CBP, we did identify EGFR as a downstream central interactor of both EP300- and CBP-controlled gene expression (Fig.
4I, J), and showed that both EP300 and CBP knockdown reduced EGFR expression in APP
Dup and NDC neurons (Fig.
4K). EGFR has been implicated in mediating neurotoxicity downstream of Aβ, primarily via utilization of EGFR inhibitors, which reduce amyloid-associated inflammation and ameliorate memory resulting from introduced amyloid [
44,
79‐
81]. Upregulation of the EGFR ligand, epidermal growth factor (EGF), reduces amyloid-related deficits without affecting levels of Aβ itself [
82]. Therefore, it is likely that EGFR-mediated neurotoxicity is dependent on but does not contribute to amyloid pathology, and occurs after Aβ accumulation has already been established, falling downstream of the scope of our study. Nevertheless, these results indicate that although EP300/CBP regulate a protective amyloid-reducing pathway, they concurrently regulate EGFR, an Aβ toxicity-exacerbating interactor, underlining that there is a complex relationship between histone acetylation, gene activation, and the ultimate phenotype of Aβ pathology and neurotoxicity.
The limitations of our study reflect the limitations inherent in our model system. While an established single gene mutation, like APP duplication in the familial AD lines used here, likely leads to a larger and more consistent effect size on measurements compared to the use of sporadic AD lines, the low number of cell lines used in our study poses a valid concern for generalizability. Additionally, our model lacks astrocytes or other glial cell types, which may explain discrepancies between our study and others conducted in heterogenous brain tissue. For example, while we observed increased AD pathology upon KD of EP300/CBP in the form of Aβ42 secretion (Fig.
6E), it is possible that the exclusively neuronal makeup of our model prevented us from observing potential beneficial effects of EP300/CBP KD. such as the downregulation of noxious agent EGFR, which appears to enact its Aβ42-worsening effect through neuroinflammation and its interaction with glial cells [
80,
81]. Of particular relevance, our previous study reporting a net increase in H3K27ac in AD brains was performed in whole brain tissue, which does indeed contain glial cell types. Thus, it is possible that different cell types experience different degrees of histone acetylation change; for example, a report examining histone acetylation changes in individual cell types identified oligodendrocytes as the primary cell type to experience H3K27ac peak increases associated with amyloid load [
83].
Additionally, we limited the scope of our experiments to the effect of EP300 and CBP KD on H3K27ac and amyloid pathology. We did not investigate the relationship of these mechanisms to tau, though tau acetylation has been shown to contribute to tau pathology and dementia [
84,
85]. Whether similar compensatory mechanisms against tau pathology progression exist in tau mutant iPSC-neuron models, and whether EP300/CBP and H3K27ac drives the expression of those homeostatic programs in those models, remain interesting potential future directions. Additionally, while we demonstrated that neither CBP nor EP300 KD impacted neuronal viability (Fig.
6F), we did not investigate in detail the potential loss of normal neuron function, an especially important consideration, given the well-established role of EP300 and CBP in learning and memory [
40,
41,
86].
In summary, our findings suggest a complex mechanism of disease-associated EP300/CBP dysregulation and provide further evidence of the crucial role EP300/CBP plays in AD-related pathological processes. In contrast to postmortem brain tissue, in which functional experiments are not feasible, utilization of an iPSC-neuron in vitro model allowed us to address, via perturbation experiments, the role HATs play in AD. In particular, using iPSCs derived from familial AD patients with an APP duplication allowed us to isolate the role EP300/CBP plays in Aβ pathology. We show that, in neurons, EP300/CBP KD lowers H3K27ac, inhibits the expression of genetic programs compensating for increased Aβ load, and leads to increased amyloid-β secretion. Future strategies targeting the reduction or increase of histone acetyltransferases as a potential therapeutic mechanism should thus consider the broad role played by H3K27ac as a general activator of transcription and cell-type dependent effects.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.