Abstract
The severity of Alzheimer's disease (AD) progression involves a complex interplay of genetics, age, and environmental factors orchestrated by histone acetyltransferase (HAT)-mediated neuroepigenetic mechanisms. While disruption of Tip60 HAT action in neural gene control is implicated in AD, alternative mechanisms underlying Tip60 function remain unexplored. Here, we report a novel RNA binding function for Tip60 in addition to its HAT function. We show that Tip60 preferentially interacts with pre-mRNAs emanating from its chromatin neural gene targets in the Drosophila brain and this RNA binding function is conserved in human hippocampus and disrupted in Drosophila brains that model AD pathology and in AD patient hippocampus of either sex. Since RNA splicing occurs co-transcriptionally and alternative splicing (AS) defects are implicated in AD, we investigated whether Tip60-RNA targeting modulates splicing decisions and whether this function is altered in AD. Replicate multivariate analysis of transcript splicing (rMATS) analysis of RNA-Seq datasets from wild-type and AD fly brains revealed a multitude of mammalian-like AS defects. Strikingly, over half of these altered RNAs are identified as bona-fide Tip60-RNA targets that are enriched for in the AD-gene curated database, with some of these AS alterations prevented against by increasing Tip60 in the fly brain. Further, human orthologs of several Tip60-modulated splicing genes in Drosophila are well characterized aberrantly spliced genes in human AD brains, implicating disruption of Tip60's splicing function in AD pathogenesis. Our results support a novel RNA interaction and splicing regulatory function for Tip60 that may underly AS impairments that hallmark AD etiology.
SIGNIFICANCE STATEMENT Alzheimer's disease (AD) has recently emerged as a hotbed for RNA alternative splicing (AS) defects that alter protein function in the brain yet causes remain unclear. Although recent findings suggest convergence of epigenetics with co-transcriptional AS, whether epigenetic dysregulation in AD pathology underlies AS defects remains unknown. Here, we identify a novel RNA interaction and splicing regulatory function for Tip60 histone acetyltransferase (HAT) that is disrupted in Drosophila brains modeling AD pathology and in human AD hippocampus. Importantly, mammalian orthologs of several Tip60-modulated splicing genes in Drosophila are well characterized aberrantly spliced genes in human AD brain. We propose that Tip60-mediated AS modulation is a conserved critical posttranscriptional step that may underlie AS defects now characterized as hallmarks of AD.
Introduction
Alzheimer's Disease (AD) is a chronic late-onset neurodegenerative disorder characterized by an accumulation of amyloid plaques and neurofibrillary tangles, memory impairment and cognitive decline (DeTure and Dickson, 2019; Knopman et al., 2021). The severity of AD progression is dependent in large part, by epigenetic histone acetylation mediated neural gene control mechanisms (Sanchez-Mut and Gräff, 2015; Killin et al., 2016; Nativio et al., 2018). Reduced histone acetylation resulting from decreased histone acetyltransferase (HAT) and/or increased histone deacetylase (HDAC) activity causes chromatin packaging alterations in neurons with concomitant transcriptional dysregulation that is a key initial step in AD etiology (Francis et al., 2009; Gräff et al., 2012; Peixoto and Abel, 2013; Lu et al., 2014). In this regard, we previously identified a neuroprotective role by the Tip60 HAT in AD (Zhu et al., 2007; Pirooznia et al., 2012; Johnson et al., 2013; Pirooznia and Elefant, 2013; Xu et al., 2014, 2016; Panikker et al., 2018; Karnay et al., 2019; H. Zhang et al., 2020; Beaver et al., 2021; Bhatnagar et al., 2023). Increasing Tip60 HAT levels in the brains of Drosophila that model AD-associated neurodegeneration protects against AD associated neuroepigenetic deficits that include reduced Tip60 and enhanced HDAC2 chromatin enrichment and concomitant transcriptional dysregulation and ameliorates multiple AD-associated phenotypes including Aβ plaque accumulation, neural apoptosis, synaptic plasticity, learning/memory, and longevity. Intriguingly, recent insights reveal that histone modifying enzymes, such as HDACs, not only determine which genes are expressed but also how the transcribed RNA is ultimately spliced (Luco et al., 2011; Rahhal and Seto, 2019; Agirre et al., 2021). Thus, while the role of Tip60 HAT activity in chromatin-mediated gene expression is well established, it remains to be studied whether Tip60 has the ability to modulate alternative splicing (AS) decisions that may in part contribute toward its neuroprotective abilities.
In addition to the catalytic HAT domain, Tip60 also contains an N-terminus chromodomain that acts as a code reader that recognizes distinct methylated-lysine histone tails (Sun et al., 2009; C.H. Kim et al., 2015). Elegant studies have shown that chromodomains within certain proteins have the ability to directly interact with RNA that likely aids in chromosomal recruitment and targeting (Akhtar et al., 2000; Bernstein and Allis, 2005; Morales et al., 2005; Bernstein et al., 2006; Shimojo et al., 2008; Ishida et al., 2012; Akoury et al., 2019). Interestingly, a closely related HAT belonging to the same MYST superfamily as Tip60, MOF, is dependent on its chromodomain-RNA binding for integration into chromosomal complexes and dosage compensation (Akhtar et al., 2000). Tip60 chromodomain has also been shown to be critical for its recruitment to chromatin-rich regions in human cell lines as chromodomain mutations cause Tip60 mislocalization (Sun et al., 2009; C.H. Kim et al., 2015). Additionally, chromatin-interacting heterochromatin protein 1 (HP1) has been recently shown to modulate alternative splicing decisions via its direct RNA binding function (Rachez et al., 2021). However, it remains to be elucidated whether Tip60 has RNA binding capabilities and if so, whether Tip60-RNA binding aid in chromatin recruitment and/or splicing modulation.
Here, we uncover a novel RNA binding function for Tip60 HAT that underlies RNA alternative splicing (AS) regulation in the brain. Genome-wide RNA immunoprecipitation (RIP) and sequencing of Tip60-bound RNA from Drosophila brain reveal Tip60 specifically targets RNAs enriched for critical neuronal processes implicated in AD. Strikingly, Tip60 targets pre-mRNA emanating from its chromatin gene targets and this function is conserved in human hippocampal tissues and disrupted in both Drosophila AD-brains that model AD pathology and in AD patient hippocampal samples. Over half of these Tip60 interacting RNAs from AD fly brains exhibit a multitude of mammalian-like AS defects enriched for in the AD-gene curated database. Notably, some splicing alterations are partially protected against by increasing Tip60 in the AD fly brain, suggesting that Tip60 modulates AS decisions for AD-associated RNA targets. Our results support an RNA splicing regulatory function for Tip60 that modulates AS decisions for its unspliced pre-mRNA targets and that disruption of this Tip60 function in the AD brain may underly AS impairments that hallmark AD etiology.
Materials and Methods
Fly stocks and crosses
All fly lines were raised under standard conditions at 25°C with 12/12 h light/dark cycle on yeasted Drosophila media (Applied Scientific Jazz Mix Drosophila Food, Thermo Fischer Scientific). The w1118, pan-neuronal driver elavC155-Gal4, transgenic UAS lines carrying human APP695 isoform (UAS-APP695), and Tip60-RNAi-mediated knock-down (UAS-Tip60 RNAi) were all obtained from Bloomington Drosophila Stock Center. Generation and characterization of the double-transgenic UAS- APP695; Tip60WT fly lines are described previously (Pirooznia et al., 2012). The elavC155-Gal4 driver line was crossed with either w1118 (wild-type control), UAS-APP695 (APP model), UAS-APP695; Tip60WT (APP; Tip60 model), or UAS-Tip60 RNAi (Tip60 RNAi model). For all experiments, transgene expression levels for APP and/or Tip60 in each UAS fly lines were revalidated using a quality control qPCR strategy with RNA extracted from the same pooled larval brains used for RNA immunoprecipitation and sequencing (RIP-Seq) or RIP-qPCR experiments. Although all transgenic fly lines have been well characterized with appropriate controls, we do not rule out the unpredictable physiological changes associated with addition of P[w+] transgene for generation of fly lines.
Homology modeling and molecular visualization
3D protein structure of Drosophila Tip60 chromodomain was generated using SWISS-MODEL automated protein structure homology modeling server (Waterhouse et al., 2018). X-ray crystallized structure of Homo sapiens Tip60 chromodomain (PDB: 4QQG, chain A) with 79% coverage was used as a modeling template. After energy-minimization with YASARA Energy Minimization Server (Krieger et al., 2009) and stereochemical quality checks with ProCheck server (Laskowski et al., 1993), the resultant Drosophila Tip60 chromodomain model was exported as a PDB file. All visualization and molecular alignments were performed using PyMOL molecular viewing software (DeLano, 2002).
In silico RNA target predictions
Tip60's RNA interaction probabilities were calculated using the RNA-protein interactions prediction (RPISeq) server (Muppirala et al., 2011). RNA sequences of key genes involved in synaptic plasticity were obtained from the NCBI database. Input protein sequences containing either only the chromodomain region or the full protein Drosophila Tip60 protein (Q960X4) were submitted. Each interaction was scored between 0 and 1 using support vector machine (SVM) classifier.
Multiple sequence alignment and secondary structure prediction
Protein sequences of Esa1 in Saccharomyces cerevisiae (Q08649) and Tip60 from Drosophila melanogaster (Q960X4), H. sapiens (Q92993), Pongo abelii (Sumatran orangutan; Q5RBG4), Mus musculus (Q8CHK4), and Rattus norvegicus (Q99MK2) were obtained from UniProt Knowledgebase (UniProt Consortium, 2020). Protein sequences were aligned using Clustal Omega multiple sequence alignment tool (Sievers et al., 2011) with default parameters. Alignment results were visualized using Jalview bioinformatics software (Waterhouse et al., 2009). Secondary structure predictions were performed using protein sequence alignment of yeast Esa1 and Drosophila Tip60 on the “Easy Sequencing in PostScript” (ESPript) program (Robert and Gouet, 2014).
Polytene chromosome squashes, staining, and imaging
Polytene chromosomes (PC) were prepared from wild-type third instar larvae and were fixed and stained according to conventional squash technique using acid fixation as previously described (Johansen et al., 2009). For RNase treatment, salivary glands were incubated in PBS with 0.4% PBT for 5 min and RNase (500 μg/ml, Thermo Scientific EN0531) for 15 min before proceeding with fixation. Primary antibodies used were guinea-pig anti-dTip60 (1:5000; from Schirling et al., 2010), mouse anti-RNA polymerase-II (1:400, Sigma: 05-623), and rabbit anti-acetyl-histone H3 antibody (1:400, Sigma 06-599). Secondary antibodies used were Alexa Flour 488, 568, and 633 (Invitrogen) at 1:200. DNA on chromosomes was counterstained using DAPI dye. Confocal microscopy was performed using laser scanning Fluoview Olympus microscope (FV-1000, Olympus Lifesciences) at 60× magnification using z-stacks. Sequential scanning mode was used to detect fluorophores in two different phases to avoid cross talk. Images were processed using ImageJ software.
rMATS splicing analysis
For splicing analysis, clean reads from Input samples were aligned to the Drosophila melanogaster genome (Ensembl version BDGP6) using STAR (Dobin et al., 2013). Splice isoform switching events were detected using replicate Multivariate Analysis of Transcript Splicing (rMATS; S. Shen et al., 2014). Alternative splicing was quantified using the percent spliced in (PSI) metric that reports inclusion or splicing of an event such that PSI = Inclusion/(Inclusion + Exclusion). For genotypic comparisons, differences in relative isoform abundance were calculated as ΔPSI values: ΔPSI(APP vs wild-type) = PSI APP – PSI wild-type; and ΔPSI(APP;Tip60 vs APP) = PSI APP;Tip60 – PSI APP. Positive ΔPSI values indicate higher inclusion in APP over wild-type and APP;Tip60 over APP, respectively. Significant splicing events were identified using the cutoffs: false discovery rate (FDR) < 0.1 and |ΔPSI| ≥ 0.1. Conserved human orthologs were predicted using best match from DRSC integrative ortholog prediction tool (DIOPT; Hu et al., 2011).
RNA immunoprecipitation and sequencing (RIP-Seq)
Magna RIP RNA-binding protein immunoprecipitation kit (Millipore) was used for native, more direct RNA immunoprecipitation without protein cross-linking. A total of 200 Drosophila third instar larval brains were dissected in ice-cold PBS and teased apart with a Dounce homogenizer. The tissue was resuspended in RIP lysis buffer after centrifugation at 1500 rpm for 5 min. In each tissue lysate sample, 10% fraction was kept aside for total RNA purification (INPUT RNA) and the remaining 90% were used for RNA immunoprecipitation (IP RNA). Magnetic beads were prepared according to the protocol using rabbit-Tip60 antibody (Abcam ab23886, 7.5 μg), and normal rabbit IgG (7.5 μg) was used as a negative control. The pretreated beads and tissue lysate were mixed and incubated with rotation overnight at 4°C. After washing with RIP wash buffer for five times, protein was digested using proteinase K treatment. RNA was phenol– chloroform precipitated from IP and INPUT samples in parallel. RNA purity and integrity were assessed using Nanodrop spectrophotometer (Thermo Fisher Scientific) and RNA 6000 Nano assay on 2100 Bioanalyzer (Agilent Technologies). Whole transcriptome sequencing was performed on IP and INPUT RNA samples using DNBSEQ sequencing technology platform (BGI Genomics, China) with 100-bp paired-end reads. Low-quality raw reads were filtered out using in-house BGI genomics pipeline on SOAPnuke (BGI-flexlab; Chen et al., 2018). Clean RNA reads were aligned to the Drosophila melanogaster genome (Ensembl version BDGP6) using HISAT2 (D. Kim et al., 2019). Reads were mapped using Bowtie2 (Langmead and Salzberg, 2012) and gene expression was quantified using RNA-Seq by expectation-maximization (RSEM; B. Li and Dewey, 2011). Principal component analysis (PCA) and heatmap clustering (Euclidean distance) were performed to cluster the samples and identify the batch effects and sample heterogeneity. All plots were constructed using R/Bioconductor. Gene ontology biological processes and human disease relevance was assessed using FlyEnrichr, a gene list enrichment analysis tool for D. melanogaster (Chen et al., 2013). Read distribution was assessed using Resect RNA-seq Quality Control package (Wang et al., 2012) on individual BAM files and Drosophila dm6 RefSeq genome bed file (O'Leary et al., 2016), and the output was visualized using MultiQC modular tool (Ewels et al., 2016).
RNA immunoprecipitation and RT-qPCR (RIP-qPCR) on human hippocampal tissues
For all human studies, human hippocampal samples were obtained from the National Disease Research Interchange (NDRI), with informed consent by all donors. The control brains included three males with an age range of 70–85 years. The AD brains were from one male and two females with an age range of 73–87 years. For RNA Immunoprecipitation, frozen hippocampal tissues were disrupted in liquid nitrogen using Cryo-Cup Grinder (BioSpec Products). Lysate were processed with either rabbit-Tip60 antibody (Abcam ab23886, 7.5 μg) or normal rabbit IgG (7.5 μg). Protein was digested using proteinase K treatment and RNA was phenol–chloroform precipitated. For RT-qPCR analysis, cDNA was prepared using the SuperScript II reverse transcriptase kit (Invitrogen) according to the manufacturer's instructions with 1 µg of total RNA. RT-qPCRs were performed in a 10 µl reaction volume containing cDNA, 1 μm Power SYBR Green PCR Master Mix (Applied Biosystems), and 10 μm forward and reverse primers. Primers are listed in Table 1. RT-qPCR was performed using an ABI 7500 Real-Time PCR system (Applied Biosystems) following the manufacturer's instructions. Fold enrichment for all the respective genes was calculated relative to the nonspecific rabbit IgG antibody control.
Primer sequences used for human RIP-qPCR
Splice-specific qPCR
Total RNA was isolated from 40 staged third instar larval brains using the Quick-RNA Miniprep kit (Zymo Research). cDNA was prepared using the SuperScript II reverse transcriptase kit (Invitrogen) according to the manufacturer's instructions with 1 µg of total RNA. Isoform specific exon-exon junction primers were designed using NCBI Primer-BLAST. The primer pair specificity was analyzed using the reference sequence database of D. melanogaster (taxid: 7227). RT-qPCRs were performed in a 10-µl reaction volume containing cDNA, 1 μm Power SYBR Green PCR Master Mix (Applied Biosystems), and 10 μm forward and reverse primers. Primers are listed in Table 2. RT-qPCR was performed using an ABI 7500 Real-Time PCR system (Applied Biosystems) following the manufacturer's instructions. Fold change in mRNA expression was determined by the δ-δCt method relative to wild type using Rpl32 as housekeeping gene.
Primer sequences used for splice-specific qPCR in Tip60 RNAi-mediated knock-down
Experimental design and statistical analysis
All statistical analysis were performed using GraphPad Prism version 9.4.0 software package. Statistical analysis of RNA-Seq data differences between two groups were considered statistically significant with q < 0.05 [false discovery rate (FDR) < 0.05, controlled by Benjamini–Hochberg]. For identification of Tip60-RNA targets significantly enriched in IP over Input, a threshold cutoff of adjusted p-value < 0.05 was used. Volcano plots comparing Tip60's RNA targets significantly enriched in wild type, APP, and APP;Tip60 were generated using a threshold cutoff of adjusted p-value < 0.05 and log2 Fold Change of ≤−0.583 and ≥0.583). Alternative splicing events significantly altered between genotypes were identified using FDR < 0.1. Volcano plots depicting difference in relative isoform abundance between genotypes were generated using a threshold cutoff of FDR <0.1 and |ΔPSI| ≥ 0.1, where PSI = percent spliced in. For splice-specific RT-qPCR, statistical significance between the two groups was calculated using unpaired Student's t test with p < 0.05. For Tip60 IP fold enrichment in healthy versus AD human tissues, two-way ANOVA with Sidak's multiple comparison test was used with p < 0.05.
Results
Structural homology and evolutionary conservation of Drosophila Tip60 RNA binding residues across mammalian species supports their critical functional significance
Chromodomains are protein–RNA interaction modules (Akhtar et al., 2000), yet it remains to be determined whether the Tip60 chromodomain structure is primed for an RNA-binding function. Structural studies on Esa1 HAT, the common ortholog of Tip60 and MOF HATs in yeast, have mapped its RNA-binding activity to a specific helical turn structural motif (η2) in its chromodomain (Shimojo et al., 2008). Using protein secondary structure predictions, we found structural conservation between Tip60 and Esa1 chromodomains, especially at the RNA-binding helical turn motif (η2; Fig. 1A). Similarly, protein structure superimposition shows Tip60 chromodomain folds into an almost identical 3D structure as the Esa1 chromodomain with minimal structural divergence of 1.02 root mean square deviation (Fig. 2A). Importantly, Tip60 chromodomain contains the predicted tudor-knot conformation and the RNA-binding helical turn motif essential for RNA-binding in Esa1 (Shimojo et al., 2008), supporting functional similarity for putative Tip60-RNA binding. Since Tip60 plays a crucial role in synaptic plasticity (Sarthi and Elefant, 2011; Beaver et al., 2020), we next assessed whether key mRNA involved in synaptic plasticity are predicted to interact with the Tip60 chromodomain and full protein. Using the in silico RNA-protein interaction prediction server (Muppirala et al., 2011), we identified several mRNA candidates strongly predicted to interact with the Tip60 chromodomain (Fig. 1B). These results support an RNA-binding function for Tip60's chromodomain that is predicted to target mRNA enriched for synaptic plasticity.
Tip60 secondary structure conservation with Esa1 HAT and putative RNA targets. A, Tip60 chromodomain is predicted to fold in a similar secondary structure as Esa1 chromodomain with conserved RNA-binding helical turn (η2) and four proven RNA-binding residues from Esa1 (denoted with *). Drosophila melanogaster Tip60 (Q960X4) and Saccharomyces cerevisiae Esa1 (Q08649) sequence similarities and secondary structure information were analyzed using ESPript. B, Several key mRNA involved in synaptic plasticity are putative Tip60 targets. RNA-protein interactions prediction (RPISeq) server was used to score Drosophila Tip60 (Q960X4) protein interactions with mRNA candidates. A probability of >0.5 suggests a strong possibility of the mRNA candidates being a target of the Tip60 chromodomain or the full protein, respectively.
Tip60's structural homology with known RNA-binding Esa1 HAT uncovers distinct conserved RNA-binding and histone-binding sites. A, Structural homology between Drosophila Tip60 chromodomain (cyan, SWISS-MODEL) and known RNA-binding yeast Esa1 chromodomain (orange, PDB: 2Ro0). B, Amino acid residues in the Drosophila Tip60 chromodomain predicted for RNA binding (red), histone binding (green), or both functions (magenta). C, Evolutionary conservation of chromodomain RNA-binding and histone-binding residues across mammalian species. Multiple sequence alignment with Clustal Omega was used to align Drosophila Tip60 chromodomain (UniProt: Q960X4) with Esa1 yeast (Q08649) and Tip60 from Homo sapiens (Q92993), Pongo abelii (Sumatran orangutan; Q5RBG4), Mus musculus (Q8CHK4), and Rattus norvegicus (Q99MK2).
Prior work using Esa1 mutational screens identified four precise RNA-binding residues in the chromodomain that completely abolished Esa1's RNA-binding ability when mutated (Shimojo et al., 2008). Notably, these four RNA-binding residues were found to be conserved in Tip60's chromodomain and are exposed at the surface near the RNA-binding turn (Fig. 2B). The polar nature of all four amino acids, Tyr 57, Tyr 60, Asn64, and Arg66, suggests these residues interact with RNA via hydrogen bonding, which is a typical characteristic of protein-RNA interactions (Teplova et al., 2011; Corley et al., 2020). Additionally, the positively charged Arg66 is able to complement the negatively charged RNA for ionic bonding, another common observation with protein-RNA binding (Chen and Varani, 2005). In contrast, the proven and predicted Tip60's histone-binding residues- Trp39, Phe56, Val58, His59, Tyr60, Val61, Phe63, Leu67, Val71, Asp75, Leu76 (Sun et al., 2009; Letunic and Bork, 2018; Y. Zhang et al., 2018) are mostly nonpolar amino acid residues positioned inside the chromodomain core. Specifically, Tyr60, and Phe63 amino acids together form an aromatic cage in the Tip60 chromodomain that recognizes methylated lysine residues for histone acetylation (Y. Zhang et al., 2018). The close proximity of RNA-binding residues with this aromatic cage and the indispensable role of Tyr60 in both, RNA-binding (Shimojo et al., 2008) and histone-binding functions (Sun et al., 2009) suggests that these two functions do not occur simultaneously. Lastly, all RNA-binding residues and most histone binding residues were found to be evolutionary conserved across mammalian species, indicating their functional importance (Fig. 2C). Together, these results strongly support a novel, conserved RNA-binding function for Tip60 chromodomain that is likely mutually exclusive from its well-studied histone-binding function.
Tip60 interacts with protein encoding RNAs enriched for neuronal processes implicated in cognition and neurodegenerative disorders in vivo
Since our structural and molecular findings support an RNA-binding function for the Tip60 chromodomain, we asked whether Tip60 directly interacts with RNA molecules in vivo. For genome- centric identification of Tip60-RNA interactions, we used a noncrosslinking, native RNA immunoprecipitation (RIP) technique to extract Tip60-bound RNA from wild-type Drosophila larval brains. We confirmed the presence of RNA molecules in the Tip60-immunoprecipitate that revealed RNAs at specific nucleotide size (∼100–2000 base pairs) in different biological replicates (Fig. 3A). Notably, RNA was not detected with nonspecific rabbit IgG and less RNA was immunoprecipitated with RNAi-mediated Tip60 knock-down, indicating Tip60's RNA binding is specific in vivo (Fig. 3B). Further, nucleic acid bands were completely lost after RNase treatment, confirming presence of RNA in the immunoprecipitate samples. Our results uncover a specific and reproducible RNA-binding function for Tip60 in Drosophila brain, in vivo.
Tip60-RNA immunoprecipitation (RIP) assay controls. A, Bioanalyzer gel shows immunoprecipitated RNA at specific nucleotide sizes with increasing Tip60 antibody concentrations of 5 μg (lanes 1, 2), 7.5 μg (lanes 3, 4), and 10 μg (lanes 5, 6). RNA is not immunoprecipitated with Rabbit IgG control (lanes 7, 8). RNA is separated based on nucleotide sizes from larger molecules at top to smaller molecules at bottom. B, Tip60-RNAi-mediated knock-down reduces amount of RNA immunoprecipitated (red) when compared with wild-type (blue). Sample peak is lost after RNase treatment of wild-type sample (magenta), confirming presence of RNA in the immunoprecipitate samples. RNA migration time (seconds, x-axis) of constant marker dye and samples are plotted against fluorescence intensity (y-axis). RNA concentration is determined based on the time corrected area underneath each sample peak and the upper marker in each sample. [nt]: nucleotide sizes; [FU]: fluorescence intensity.
To identify Tip60's RNA targets in vivo, we performed whole transcriptome sequencing (RIP-Seq) on both the Tip60-immunoprecipiated RNA (IP RNA) and total RNA present in tissue before immunoprecipitation (Input RNA) for enrichment comparison (Fig. 4A). Heatmap of RNA enrichment across samples shows two important observations (Fig. 4B). First, similar RNAs are immunoprecipitated in all three IP RNA samples, indicative of highly specific Tip60-RNA interaction. Second, several RNAs enriched in the Input RNA were not immunoprecipitated in the IP RNA, suggesting Tip60 does not equally favor binding to all RNA molecules. Next, we performed a principal component analysis (PCA) to observe variation between the samples (Fig. 5A). As expected, the IP RNA and Input RNA samples clustered in two separate groups, suggesting limited sample-to-sample variation. In contrast, there was major variation between the IP and Input RNA clusters, validating that Tip60's RNA binding function is highly selective and reproducible in different biological samples. To further confirm Tip60's RNA-binding selectivity, we used the MA scatter plot that shows log fold change of IP over Input RNA (Fig. 5B). Using a threshold cutoff of adjusted p-value < 0.05, we identified RNAs significantly different between IP and Input (red scatters). We then selected for RNAs significantly enriched in IP over input, which we refer to as “Tip60 RNA targets.” Our RIP-Sequencing identified a total of 2884 Tip60-RNA targets, of which 35 are noncoding RNAs and 2849 are protein encoding RNAs (Fig. 4C; Extended Data Table 4-1). To study biological pathways and disease relevance of the top 2000 Tip60 RNA targets, we used FlyEnrichr, a gene list enrichment analysis tool for D. melanogaster (Chen et al., 2013). Our analysis revealed that Tip60 RNA targets were enriched for critical dynamic neuronal processes that included axon guidance and axonogenesis, transcription and development (Fig. 4D). Notably, the majority of these processes are linked to several human diseases with tauopathy and Alzheimer's Disease displaying highest prevalence (Fig. 4E; Extended Data Table 4-2). Together, our results identify a highly specific RNA-binding function for Tip60 in the brain in vivo that favors interaction with protein encoding RNAs that mediate critical neuronal processes linked to neurodegenerative diseases.
RNA Immunoprecipitation and Sequencing (RIP-Seq) reveals a highly specific, selective, and reproducible RNA-binding function for Tip60 in the Drosophila brain. A, RIP-Seq schematic: Tip60-bound RNA molecules are immunoprecipitated and extracted (IP RNA) along with the total RNA (INPUT RNA) from Drosophila larval brains for RNA Sequencing. B, Hierarchically clustered heatmap depicting RNA homogeneity within replicates and variability between IP and INPUT groups from three wild-type (WT) biological replicates. C, Classification of Tip60 RNA targets as protein coding or noncoding RNA. D, Gene ontology biological processes and (E) human diseases enriched for the top 2000 Tip60 RNA targets. Refer to Extended Data Table 4-1 for Tip60 RNA targets significantly enriched in IP and Extended Data Table 4-2 for Gene Ontology analysis and human disease relevance.
Tip60-RNA target identification in the wild-type (WT) samples. A, Principal component analysis (PCA) plot showing samples variation between RNA enriched in the IP samples (triangles) and respective Input samples (circles) from three wild-type (WT) Drosophila biological replicates. B, MA scatter plot shows log fold change of IP RNA over Input RNA (y-axis) and average expression between the groups (x-axis). Red scatters are present below the threshold cutoff of adjusted p-value < 0.05. WT: wild type; IP: immunoprecipitate RNA; INPUT: Input RNA.
Extended Data Table 4-1
Tip60-RNA targets significantly enriched in the immunoprecipitate fraction of Drosophila wild-type, APP, and APP;tip60 larval brains. Using a threshold cutoff of Benjamini–Hochberg adjusted p-value (padj) < 0.05, RNA significantly enriched in Tip60-immunoprecipiated RNA (IP RNA) over the total RNA (Input RNA) were identified for (A) wild type, (B) APP, and (C) APP;Tip60. The Tip60 RNA targets were classified as coding mRNA (RefSeq category: NM) or noncoding RNA (RefSeq category: NR). Download Table 4-1, XLS file.
Extended Data Table 4-2
Gene Ontology and human disease relevance for Tip60 RNA targets in Drosophila wildtype larval brains. A, Gene ontology biological processes and (B) human disease relevance was assessed using FlyEnrichr gene list enrichment analysis tool for Drosophila melanogaster. Top 2000 Tip60 RNA targets significantly enriched in the immunoprecipitate fraction of wild-type Drosophila larval brains were used as input query. Download Table 4-2, XLS file.
Increased Tip60 partially rescues Tip60-RNA targeting alterations in the APP AD associated neurodegenerative brain
We previously showed that early-preclinical mild cognitive impairments (MCI) and late-stage AD pathologies in humans are tightly conserved both epigenetically and pathologically in the extensively characterized AD associated human amyloid precursor protein (APP) Drosophila model (APP AD; Zhu et al., 2007; Pirooznia et al., 2012; Johnson et al., 2013; Pirooznia and Elefant, 2013; Xu et al., 2014, 2016; Panikker et al., 2018; Karnay et al., 2019; Beaver et al., 2020, 2021; H. Zhang et al., 2020; Bhatnagar et al., 2023). This high degree of disease conservation allows for general principles learned from the AD APP fly to be applied to mammalian systems. Our prior studies revealed reduced Tip60 HAT levels with concomitant altered patterns of chromatin histone acetylation and neuronal gene expression in the brains of our APP AD Drosophila model that contribute to cognitive deficits and are prevented by increased Tip60 levels. Thus, we asked whether Tip60's RNA-binding function is also perturbed in APP AD flies and that this defect can be ameliorated by genetically increasing Tip60 levels. To address this question, we assessed Tip60-RNA interactions under pan-neuronally expressed human APP695 isoform alone (APP AD model) or in combination with Tip60 wild-type protein (APP;Tip60 model) using RIP-Seq on Drosophila larval brains. Transcriptomic sequencing of Tip60-IP RNA and Input RNA revealed that similar to wild-type, Tip60-RNA binding is specific and selective for only certain RNA molecules from the entire Input RNA pool in APP and APP;Tip60 genotypes (Fig. 6A,B). Further, using a threshold cutoff of adjusted p-value < 0.05, we identified Tip60 RNA targets enriched in IP that clustered separately between genotypes, suggesting variations in Tip60-RNA binding in different genotypes (Fig. 6C; Extended Data Table 4-1). Although the majority of RNA targets are shared by all three genotypes, we found certain RNAs that are uniquely targeted only in wild-type, APP, or APP;Tip60, supporting Tip60 RNA target divergence in different genotypes (Fig. 6D). Together, these results demonstrate that although specificity and selectivity of Tip60's RNA binding functions remains unaltered, Tip60 targets partially different sets of RNA in wild-type, APP, and APP;Tip60 Drosophila models.
Tip60's RNA targets comparison between Drosophila wild-type, APP, and APP;Tip60 conditions. Heatmap depicting RNA enriched in IP and INPUT biological replicates from (A) APP and (B) APP;Tip60 Drosophila larval brains. C, Principal component analysis (PCA) plot showing variation between Tip60's RNA targets that are specifically enriched in the immunoprecipitate RNA (IP RNA) between the wild-type (square), APP (circles), and APP;Tip60 (triangles). D, Venn diagram shows distribution of Tip60's RNA targets that are unique or shared between wild type (blue), APP (purple), and APP;Tip60 (red). WT: wild type; IP: immunoprecipitate RNA; Input: Input RNA.
To better understand how Tip60's RNA-binding function is altered in different genotypes, we compared the distribution and intersection of Tip60's RNA targets between APP versus wild-type and APP versus APP;Tip60 (Extended Data Table 7-1). Using volcano plot analysis, we first identified Tip60's RNA targets that are significantly enriched in IP in APP (red scatters), wild-type (blue scatters), or both (black scatters; Fig. 7A, left) and APP;Tip60 (red scatters), APP (blue scatters), or both (black scatters; Fig. 7A, right). We found similar number of Tip60-RNA targets that are either significantly enriched or depleted in binding in APP versus wild-type (192 and 171, respectively) and APP;Tip60 versus APP (554 and 607, respectively). Next, we used an UpSet plot to visualize intersections between these four differential Tip60-RNA target comparisons that are represented as individual rows (Fig. 7B). Out of the 171 RNA that interacted with Tip60 in wild-type but not in APP brains (row 1), increased Tip60 in the APP;Tip60 genotype restored Tip60-RNA interactions with 56 (32.8%) of these RNAs (row 1 and row 3 overlap, purple bar). Similarly, out of the 192 RNA that Tip60 mistargeted in APP but not in wild-type (row 2), increased Tip60 in the APP;Tip60 genotype resulted in reduced inappropriate interactions in 43 (22.4%) of these RNAs (row 2 and row 4 overlap, purple bar). Together, increased Tip60 in the APP;Tip60 genotype restored 27.6% (99/363) Tip60-RNA target interactions that were altered in the APP brain that we refer to as 'Tip60 rescued RNA targets'. To identify the biological processes these Tip60 rescued RNA targets are involved in, we performed functional annotation clustering on this set of RNAs using FlyEnrichr gene ontology analysis (Fig. 7C). Top enriched cellular processes included chromatin assembly and remodeling, axon and dendrite guidance, protein modification, intracellular transport, and proteolysis, and intriguingly, RNA transport and splicing. Together, our results point to a functional role for Tip60-RNA binding in the brain that is disrupted under APP neurodegenerative conditions and is partially protected against by increased Tip60.
Extended Data Table 7-1
Tip60-RNA target alterations between Drosophila wild-type, APP, and APP;Tip60 larval brains. Tip60-RNA targets selectively enriched in immunoprecipitate that were either significantly enriched or depleted in binding between genotypes were identified as: (A) enriched targeting in APP over wild type (up APP vs WT); (B) less targeting in APP over wild type (down APP vs WT); (C) enriched targeting in APP;Tip60 over APP (up APPTip60 vs APP); (D) less targeting in APP;Tip60 over APP (down APPTip60 vs APP). Download Table 7-1, XLS file.
Tip60 RNA targets are altered in Drosophila APP neurodegenerative model and partially rescued by Tip60 overexpression. A, Volcano plots depicting Tip60's RNA targets specifically enriched in immunoprecipitate RNA (IP RNA) between APP versus wild-type (left) and APP;Tip60 versus APP (right) Drosophila larval brains. Tip60-RNA targets with significantly enriched binding (red), reduced binding (blue) or nonsignificant binding alterations (black) between genotypes are depicted (cutoff: adjusted p-value < 0.05; log2 fold change of ≤−0.583 and ≥0.583). B, UpSet plot representing the distribution and intersection of Tip60's RNA target alterations between APP versus wild type and APP;Tip60 versus APP. Rows represent the total number of Tip60-RNA targets in each comparison that are either unique (black dots) or overlapping (connecting line) with other comparisons. Purple columns represent Tip60 rescued RNA targets. C, Biological pathways enriched for Tip60-rescued RNA targets that are either excluded in wild-type, targeted in APP, and excluded again in APP;Tip60 (left) or targeted in wild type, excluded in APP, and targeted again in APP;Tip60 (right). Some genes appear in more than one GO category. WT: wild type; NS: not significant. See Extended Data Table 7-1 for Tip60's RNA targets comparison between Drosophila wild-type, APP, and APP;Tip60 conditions.
Tip60 interacts with pre-mRNAs that emanate from Tip60's chromatin gene targets
We previously reported that Tip60 displays a nuclear cytoplasmic distribution pattern in both the Drosophila and mammalian brain. Thus, we asked whether Tip60 primarily interacts with unspliced pre-mRNA in the nucleus or mature spliced mRNA in the cytoplasm. We performed RSEQC read distribution analysis (Wang et al., 2012) on our Tip60-IP RNA and Input RNA samples from RIP-Seq to calculate the distribution pattern of mapped reads over different genome features, such as like coding DNA sequence (CDS) exon, 5′ untranslated region (UTR) exon, 3′ UTR exon, intron, and intergenic regions (Fig. 8A). As expected from the Input samples containing both pre-mRNA and mature mRNA, the majority of the reads mapped to CDS exonic regions and UTR regions, while the remaining mapped to introns and intergenic regions. Strikingly, we observed similar read distribution in Tip60-IP RNA samples, including mapping to intronic regions, suggesting the presence of pre-mRNA in the RNA population that Tip60 specifically interacts with. Further, read mapping at intronic regions revealed higher enrichment of RNAs from Tip60-RIP samples when compared with their respective Inputs for all genotypes, indicating that Tip60 preferentially targets unspliced pre-mRNAs that reside in the nucleus. Finally, since RNA splicing is predominantly co-transcriptional and occurs in close proximity with the chromatin loci it originates with, we asked whether there is any overlap between Tip60 target RNAs and chromatin gene loci. To address this question, we compared our Tip60 RNA targets from RIP-Seq with Tip60 chromatin gene targets we previously published using ChIP-Seq (Beaver et al., 2021). Importantly, RIP-Seq and CHIP-Seq were performed using identical staged larval brains for wild-type, APP, and APP;Tip60 genotypes. Remarkably, we observed a significant overlap (78–79%) between Tip60's RNA and gene targets for wild-type (p < 2.263e-06), APP (p < 7.115e-17), and APP;Tip60 (p < 6.523e-28, hypergeometric test; Fig. 8B; Extended Data Table 8-1). These results suggest that Tip60 regulates identical genes at both the chromatin and RNA level, potentially by interacting with nascent RNA as it is transcribed.
Tip60 targets identical gene loci at chromatin and RNA levels. A, Reads from Tip60-immunoprecipitated RNA (IP) and total RNA (Input) samples in wild type, APP, and APP;Tip60 were mapped to corresponding genomic features in the following priority order: CDS exons (blue) > UTR exons (black/green) > Introns (orange) > Intergenic regions. CDS: coding DNA sequence; TSS: transcription start site; TES: transcription end site. B, Overlap between Tip60's RNA targets and its gene targets identified via chromatin immunoprecipitation and sequencing (ChIP-Seq) in wild-type, APP, and APP;Tip60 Drosophila larval brains. C, Drosophila salivary polytene chromosomes stained for DAPI (blue), histone H3 pan-acetyl (yellow), RNA polymerase-II (green), and Tip60 (red) antibodies wither in the presence or absence of RNase. Refer to Extended Data Table 8-1 for overlap between Tip60's RNA targets and gene targets at the chromatin level.
Extended Data Table 8-1
Overlap between Tip60's RNA targets and gene targets at the chromatin level. Tip60's RNA targets identified via RNA-immunoprecipitation and sequencing (RIP-Seq) were compared to its gene targets identified via chromatin immunoprecipitation and sequencing (ChIP-Seq) in (A) wild-type, (B) APP, and (C) APP;Tip60 Drosophila larval brains. Download Table 8-1, XLS file.
The significant overlap between Tip60's gene targets at the chromatin and RNA level prompted us to ask whether Tip60's interaction with RNA is required for its chromatin interaction. To address this question, we assessed whether RNase treatment misclocalizes Tip60 in polytene chromosomes (PCs) within fly salivary glands. PCs are a well-established model to study functional chromosomes because of large size and prominent banding pattern (Johansen et al., 2009; Vatolina et al., 2011). As expected, we found that Drosophila Tip60 localizes to the less-compact interbands in PCs, representing regions of highly active gene transcription (Fig. 9; Schirling et al., 2010) and accordingly, co-localizes with RNA Pol II. Treatment of PCs with RNase reduced Tip60 staining, supporting a putative role for Tip60-RNA interaction occurring in close proximity to chromatin (Figs. 8C, 10). Together, our results demonstrate that Tip60 primarily targets identical gene loci at both the chromatin and RNA level and that Tip60's RNA binding function is at least in part, required for Tip60's chromatin interaction.
Tip60 localizes to the actively transcribed gene regions on polytene chromosomes. Polytene chromosomes from wild-type Drosophila salivary glands are squashed and co-stained with DAPI (blue), RNA polymerase-II (green), and Tip60 (red). Tip60 co-localizes with RNA polymerase II in the interbands region (merged yellow) representing sites of active gene transcription between densely packed heterochromatin. RNAPII: RNA-polymerase II.
Tip60's chromatin recruitment is sensitive to RNase treatment. Polytene chromosomes from wild-type Drosophila salivary glands were squashed and co-stained with DAPI (blue), histone H3 pan-acetyl (yellow), RNA polymerase-II (green), and Tip60 (red) antibodies. A, In absence of RNase treatment, polytene chromosomes are saturated with histone H3 pan-acetyl, RNA polymerase-II, and Tip60 staining. B, After RNase treatment, RNA polymerase-II, and Tip60 staining are partially lost while histone H3 pan-acetylation staining remains unaffected on polytene chromosomes. RNAPII: RNA-polymerase II; H3Ac: histone H3 pan-acetylation.
Tip60 mediates alternative splicing selection of neural pre-mRNA targets associated with Alzheimer's disease
Our published ChIP studies showing Tip60 chromatin enrichment at intergenic and intronic regions within genes (Beaver et al., 2021) in conjunction with our RNA read map analysis revealing enrichment of RNAs within intronic regions of these same genes in Tip60- RIP samples support an interaction between Tip60 and unspliced pre-mRNAs. In neurons, AS of pre-mRNA is a central mechanism used to increase the genetic plasticity and proteomic diversity required for dynamic neuronal processes, making these tissues particularly suspectable to splicing defects (Q. Li et al., 2007; Su et al., 2018). Given that AS is primarily co-transcriptional and that AS defects hallmark AD, we hypothesized that Tip60-RNA interaction mediates AS of pre-mRNA targets emanating from Tip60's chromatin gene loci and that this process is disrupted in the APP AD larval brain. To assess Tip60's involvement in potential AS defects in the APP brain, we applied replicate multivariate analysis of transcript splicing (rMATS) analysis on RNA-Sequencing data from Drosophila wild-type, APP, and APP;Tip60 Input samples (Extended Data Table 11-1, Table 11-2). The relative abundance of each isoform was quantified as percentage spliced in (PSI); ΔPSI quantified the difference in relative isoform abundance between genotypes. Our analysis uncovered a multitude of differential mammalian-like AS alteration events including skipped exons (SE), alternative 5′ splice site (A5SS), alternative 3′ splice site (A3SS), mutually exclusive exons (MXE), and retained introns (RI) between genotypes (Fig. 11A). We identified a total of 698 and 517 significant splicing defects between APP versus wild-type and APP;Tip60 versus APP comparisons, respectively, that affected every AS event category, suggesting genotype-dependent splicing modifications at a global level (Fig. 11B). Although MXE accounts for most significant AS alterations, ∼80% of these alterations are present in Dscam1 gene that encodes for 38,106 distinct proteins via AS of 95 variable exons (Graveley, 2005). Strikingly, a comparison of RNAs showing altered splicing in APP or APP;Tip60 with Tip60 wild-type RNA targets identified from RNA-IP Sequencing data reveal that >50% of RNA undergoing significant AS alterations in APP (186/358) and APP;Tip60 (162/284) brains are bona-fide Tip60-RNA targets in the wild-type brain (Fig. 11C). Moreover, ∼30% of human orthologues for these Tip60-targeted AS genes in APP (54/177) and APP;Tip60 (44/152) were found to be enriched for AD in the DisGeNET curated database of gene-disease associations (Piñero et al., 2016; Tables 3, 4). Therefore, our results strongly suggest that Tip60 possesses the ability to influence AS decisions of unspliced pre-mRNA targets implicated in AD pathogenesis.
Extended Data Table 11-1
Alternative splicing results from rMATS analysis on Drosophila APP versus wildtype larval brains. Differential splicing isoforms were identified and compared between total RNA samples from APP and wildtype genotypes using rMATS. Junctions counts (JC) are quantified for each type of alternative splicing event: (A) skipped exon (SE), (B) alternative 5' splice site (A5SS), (C) alternative 3' splice site (A3SS), (D) mutually exclusive exons (MXE), and (E) retained intron (RI). The percent spliced in (PSI) values are reported as inclusion level for APP (Sample 1) and wild type (Sample 2). Significant splicing events were identified using the following cutoffs: false discovery rate (FDR) < 0.1 and inclusion level difference (ΔPSI) ≥ 0.1 or ≤ 0.1. Download Table 11-1, XLS file.
Extended Data Table 11-2
Alternative splicing results from rMATS analysis on Drosophila APP;Tip60 versus APP larval brains. Differential splicing isoforms were identified and compared between total RNA samples from APP;Tip60 and APP genotypes. Junctions counts (JC) are quantified for each type of alternative splicing event: (A) skipped exon (SE), (B) alternative 5' splice site (A5SS), (C) alternative 3' splice site (A3SS), (D) mutually exclusive exons (MXE), and (E) retained intron (RI). The percent spliced in (PSI) values are reported as inclusion level for APP;Tip60 (Sample 1) and APP (Sample 2). Significant splicing events were identified using the following cutoffs: false discovery rate (FDR) < 0.1 and inclusion level difference (ΔPSI) ≥ 0.1 or ≤ 0.1. Download Table 11-2, XLS file.
Mammalian conservation of Tip60-RNA targets with altered splicing in Drosophila APP and APP;Tip60 models
Tip60-RNA Targets with altered splicing enriched in DisGeNET Alzheimer's disease database
Tip60-RNA targets differentially spliced in Drosophila APP and APP;Tip60 models are implicated in Alzheimer's disease (AD). A, Summary of total differential alternative splicing events detected between APP versus wild type and APP;Tip60 versus APP using rMATS. Alternative splicing events are classified as Skipped Exons (SE), Alternative 5′ Splice Site (A5SS), Alternative 3′ Splice Site (A3SS), Mutually Exclusive Exons (MXE), and Retained Intron (RI). B, Volcano plot depicting splicing events significantly altered between APP versus wild-type (left) and APP;Tip60 versus APP (right) Drosophila larval brains. The relative abundance of each isoform was quantified as percentage spliced in (PSI) for every genotype; ΔPSI (x-axis) quantified the difference in relative isoform abundance between different genotypes. All events are significant [false discovery rate (FDR) < 0.1 and |ΔPSI| ≥ 0.1]. C, Significantly altered splicing events were mapped to fly genome and filtered for direct Tip60 RNA targets identified via RIP-Seq in wild type. Conserved human orthologs were predicted using DIOPT and compared with DisGeNET database for AD relevance. See Extended Data Tables 11-1 and Tables 11-2 for complete splicing results from rMATS analysis between APP versus wild-type and APP;Tip60 versus APP.
We next investigated whether restoring Tip60 levels in APP;Tip60 is sufficient to protect against AD-associated AS defects observed in Drosophila APP neurodegeneration. To test this, we screened the 54 Tip60-targeted AD-associated AS defects in APP for reversal in APP;Tip60 such that (ΔPSI)APPTip60 ≅ (ΔPSI)APP at the same or a nearby genomic location. Remarkably, we identified 15 triaged Tip60-rescued AS events mapping to 12 genes that are altered in APP and rescued by restoration of Tip60 levels in APP;Tip60 (Table 5). Only 1 out of the 34 MXE events in Dscam1 gene were included in the main list to avoid repetition (Table 6). These triaged Tip60-rescued AS events are distributed over all five types of AS and have no preference toward exon inclusion or exclusion, suggesting Tip60 is acting as a global splicing regulator (Fig. 12A). Individual schematic representations show splicing defects for each AS type that is reversed at the exact exonic/intronic genomic locations (Fig. 12B). For example, exon 6 in the heph gene is preferentially included in wild-type (PSI = 0.81), skipped in APP (PSI = 0.39), and included back again in APP;Tip60 (PSI= 0.69). Similarly, long isoforms of Dlg1 exon 1 (A5SS) and Rab3-GEF exon 4 (A3SS) are favored in wild-type (PSI= 0.96; 0.44), relatively excluded in APP (PSI = 0.71; 0.00) and restored in APP;Tip60 (PSI = 1.00; 0.34). Likewise, Dscam1 gene contains mutually exclusive exons at position 6 where a specific exon in wild-type (PSI = 0.88) is skipped for another exon in APP (PSI= 0.05) but included again in APP;Tip60 (PSI = 0.67). Lastly, the intron between exons 3 and 4 in Adar gene is less included in wild-type (PSI = 0.77) as compared with APP (PSI= 0.96) that is restored in APP;Tip60 (PSI = 0.60). Finally, to test whether the AS defects we observed in the APP neurodegenerative fly brain are modulated by Tip60-RNA targeting, we used splice-specific RT-qPCR on larval brains from wild-type and RNAi-mediated Tip60 neural knock-down samples (Fig. 12C–E). Remarkably, the expression of predominant RNA isoform for heph (t(4) = 6.797, p = 0.0012, unpaired Student's t test), Dscam1 (t(4) = 7.707, p = 0.0008, unpaired Student's t test), and Adar (t(4) = 2.530, p = 0.0323, unpaired Student's t test) in the wild-type larval brain was found to be significantly reduced on Tip60 RNAi-mediated knock-down. In conclusion, neural Tip60 knock-down is sufficient for inducing the exact AS defects identified in heph, Dscam1, and Adar under APP neurodegeneration., therefore validating the role of Tip60 in modulating AS decisions of its pre-mRNA targets.
Genomic coordinates of the triaged Tip60-rescued splicing events from rMATS analysis
Tip60-Rescued mutually exclusive splicing events in Dscam1
Tip60 modulates alternative splicing decisions of Alzheimer's disease (AD)-associated genes. A, The 15 triaged Tip60-rescued splicing defects in AD-associated genes. x-axis represents differences in relative isoform abundance (ΔPSI, where PSI = percent spliced in) between APP versus wild type (red) and APP;Tip60 versus APP (black). B, Schematic representation of Tip60-rescued splicing events from every type of alternative splicing. Percentage represents the relative isoform abundance for the indicated genotype obtained via rMATS analysis. C–E, Splice-specific RT-qPCR on Drosophila larval brains from wild-type and Tip60 RNAi-mediated knock-down (n = 3) to detect alternatively spliced isoforms in heph, Dscam1, and Adar genes. Histogram represents relative fold change in mRNA expression calculated using ddCT method using Rpl32 as the housekeeping gene. Statistical significance was calculated using unpaired Student's t test. *p < 0.05, **p < 0.01, ***p < 0.001. Error bars indicate SEM.
Tip60's RNA-binding function is conserved in the human brain and altered under AD pathology
The human orthologues of the 12 triaged Tip60-rescued AS genes perform diverse yet critical neuronal functions that go awry in AD pathogenesis (Fig. 13A). For example, ADAM10 (fly kuz) is a α-secretase that cleaves APP to promote the nonamyloidogenic pathway and reduce Aβ plaque load (Niemitz, 2013; Yuan et al., 2017). Accordingly, ADAM10 is the third most significant AD-associated gene in DisGeNET database (Piñero et al., 2016) and is currently being tested as a potential AD treatment (Manzine et al., 2019). Likewise, DLG1 (fly dlg1) is a scaffolding protein known to interact with APP intracellular C-terminal domain (AICD; Silva et al., 2020) and regulate APP metabolism by recruiting ADAM10 to the synapse (Marcello et al., 2013). Additionally, RBFOX1 (fly Rbfox1) and PTBP1 (fly heph) are two key splicing regulators of neuronal-specific AS in the mammalian brain (D. Li et al., 2021) that directly regulate AS of APP exon 7 (Smith et al., 2011; Alam et al., 2014) and therefore, control APP695 production and Aβ plaque load (Belyaev et al., 2010). Similarly, ADARB1 (fly Adar) is a major adenosine to inosine RNA editing enzyme in mammals that also regulates splicing (Solomon et al., 2013) and its function is reduced in human AD hippocampus (Khermesh et al., 2016; Annese et al., 2018). Moreover, HSPG2 (fly trol) extracellular matrix protein is observed to be co-deposited with Aβ plaques in the brains of AD patients (Van Gool et al., 1993; G.L. Zhang et al., 2014) where it accelerates Aβ oligomerization and aggregation (C.C. Liu et al., 2016). Besides, MADD (fly Rab3-GEF) and HDAC4 show altered expression in AD pathology (Del Villar and Miller, 2004; X. Shen et al., 2016) and have been proposed to serve as novel AD pharmacological targets (Mielcarek et al., 2015; Hassan et al., 2021).
Tip60-RNA targeting is conserved in human hippocampus and impaired in Alzheimer's disease (AD) patient hippocampus. A, Conserved human orthologs for the 12 triaged AD-associated Tip60-rescued genes were predicted using DIOPT. Previously reported splicing defects in postmortem human AD brain tissues were identified in the literature: +Marques-Coelho et al. (2021), *Adusumalli et al. (2019), ̂Tollervey et al. (2011), “Raj et al. (2018). B, Tip60-bound RNA immunoprecipitation and RT-qPCR (RIP-qPCR) was performed on hippocampus obtained from healthy controls or AD patients (n = 3 brains). Histogram represents Tip60 IP fold enrichment for each gene relative to Rabbit IgG (negative control). Statistical significance was calculated using two-way ANOVA with Sidak's multiple comparison test. ****p < 0.001. Error bars indicate SEM.
Importantly, and consistent with our data, several studies have previously reported AS defects in postmortem human AD-brains in 7 out of these 12 triaged Tip60-rescued AS genes (Tollervey et al., 2011; Raj et al., 2018; Adusumalli et al., 2019; Marques-Coelho et al., 2021). To test whether Tip60's RNA binding function is conserved in the human brain and whether such putative Tip60-RNA binding is altered under human AD pathology, we performed RIP-qPCR on RNA isolated from postmortem human hippocampal tissues obtained from healthy controls and AD patients (Fig. 13B). Remarkably, we found that Tip60 interacted with RNAs corresponding to the seven human Tip60-rescued AS Drosophila orthologs that exhibit known AS defects in the human AD brain. Importantly, as compared with healthy controls, Tip60 enrichment for RNA transcripts encoded by each of these 7 loci was significantly reduced in AD patients (F(7,32) = 3.775, p = 0.0043, two-way ANOVA with Sidak's multiple comparison test). Notably, Tip60 enrichment for ADAM10 transcripts in the healthy brain is most significantly reduced in AD brain (t(32) = 5.756, p = 1.10045E-06), supporting a role for Tip60 in mediating RNA processing of genes critical for keeping AD neurodegeneration in check. Together, our results reveal a novel RNA splicing regulatory function for Tip60 that mediates AS decisions for its unspliced pre-mRNA targets enriched for AS impairments that hallmark AD etiology.
Discussion
The selective interaction of Tip60 with protein coding neural mRNAs is disrupted in AD brain
Tip60, the second most highly expressed HAT in the mammalian brain, drives neural function and neuroprotection in AD but studies to date have conventionally focused on its role in chromatin dynamics and neural gene control, leaving additional mechanistic functions unexplored. Here, we report a highly specific, selective, and reproducible RNA-binding function for Tip60's chromodomain in the Drosophila brain in vivo. Our findings are not unprecedented as chromodomains within multiple chromatin regulatory proteins have been shown to directly interact with RNA (Akhtar et al., 2000; Morales et al., 2005; Bernstein et al., 2006; Shimojo et al., 2008; Ishida et al., 2012; Akoury et al., 2019). Chromodomains within MOF HAT and chromobox-7 achieve dosage compensation by targeting roX noncoding RNA at the male X chromosome and Xist noncoding RNA at the female X chromosome in Drosophila and mammalian cells, respectively (Akhtar et al., 2000; Bernstein et al., 2006). Our findings confirm and extend these studies by being the first to sequence and characterize a complex array of neural RNAs that are specifically bound to Tip60 in the fly brain. We find that Tip60 primarily targets protein encoding RNAs that mediate dynamic neuronal processes and are enriched for human diseases such as tauopathy and AD, indicating disruption of Tip60-RNA binding is involved in these cognitive disorders. In line with these findings, we observe that Tip60-RNA targeting is disrupted in the AD fly brain and in AD human hippocampal samples supporting a functional role for Tip60-RNA binding in AD pathology. Remarkably, increasing Tip60 levels in the AD fly brain partially protects against Tip60-RNA targeting alterations that are enriched for dynamic neuronal processes including chromatin assembly and remodeling, axonal guidance, protein modification processes, and RNA splicing and transport. We speculate that such Tip60-RNA binding disruptions lead to transcriptomic alterations that ultimately contribute significantly to AD pathologies but can be protected against by increased Tip60 levels.
Tip60's bi-level gene regulation at the chromatin and RNA level mediates rapid fine-tuning of neural gene expression
Tip60 is a key mediator of activity-dependent gene expression underlying dynamic neuronal processes and is shuttled from cytoplasm into nucleus on neuronal stimulation for histone acetylation (Xu et al., 2016; Karnay et al., 2019). This raises the possibility that Tip60 could be binding with RNA emanating from its activity-dependent genes at various stages of RNA processing to dictate ultimate protein isoforms and function in the brain. However, whether Tip60 interacts with nascent RNAs in the nucleus or mature RNA in the cytoplasm and whether Tip60's interacting RNAs are transcribed directly from Tip60 chromatin targets remains unclear. Here, we show enrichment of intronic regions in the Tip60-IP bound RNA samples, indicative of Tip60 primarily targeting unspliced pre-mRNAs that reside in the nucleus. Further, we found reduced Tip60 staining on polytene chromosomes after RNase treatment, strongly suggesting that Tip60 interacts with newly transcribed pre-mRNA in close proximity to chromatin. Consistent with this finding, we identified a significant overlap between Tip60's RNA targets and its chromatin gene targets, suggesting Tip60 is regulating expression and function of identical neural targets via targeting at both the chromatin and RNA levels, respectively. Although our findings are unprecedented for a histone acetyltransferase, HP1 chromosomal protein has been shown to dissociate with heterochromatin to bind with newly synthesized RNA owing to its greater affinity for RNA over histones (Keller et al., 2012). In support of this concept, we find Tip60 is unlikely to bind with both histone and RNA concurrently because of steric hindrance at interacting sites. Therefore, we are the first to propose a model by which Tip60 rapidly fine-tunes its neural targets for dynamic gene regulation by orchestrating a bi-level switching mechanism such that Tip60 recruitment to chromatin allows for histone acetylation-mediated gene activation as well as targeting of newly synthesized RNA that may further stabilize binding in a positive feedback loop (Fig. 14).
Model for Tip60's novel bi-level gene regulation at the chromatin and RNA level. Our results support a model by which Tip60 regulates both, the expression and splicing of a similar set of neural targets via its functions at the chromatin and the RNA, respectively. A, Tip60 promotes neural gene expression via histone acetylation at the chromatin that increases chromatin accessibility for the transcriptional machinery. B, Tip60 targets the newly transcribed pre-mRNA to modulate its alternative splicing decision by either altering splice site accessibility, assembling a complex that affects splicing, or tethering it to the chromatin for splicing regulation via histone acetylation. The model figure was created using BioRender.
Tip60-mediated alternative splicing selection may underly splicing defects characterized as hallmarks of Alzheimer's disease
RNA splicing abnormalities have recently emerged as a widespread hallmark in AD and AS defects in major disease candidate genes, including APP, Tau, PSEN, and ApoE, have since been linked to AD pathology (Love et al., 2015; Jakubauskienė and Kanopka, 2021; D. Li et al., 2021). Although causes remain unclear, dysregulation of epigenetic mechanisms under AD pathology (X. Liu et al., 2018; Nativio et al., 2018) and their recent convergence with co-transcriptional AS regulation (Luco et al., 2011; Rahhal and Seto, 2019; Xu et al., 2021) strongly suggest a causative role for epigenetic regulators/modifications in AS defects underlying AD pathology. In support of this concept, here we show that the Tip60 HAT doubles as an RNA splicing modulator and mediates AS selection of its pre-mRNA targets associated with AD. We discovered a multitude of differential mammalian-like AS alteration events in the APP AD fly brain, with over half of these altered RNAs identified as bona-fide Tip60-RNA targets enriched for AD that are partially protected against by increasing Tip60 levels. Moreover, consistent with a previous finding that shows Tip60 knock-down in epithelial cells alters AS of a key integrin subunit (Bhatia et al., 2020), we find Tip60 neural knock-down is sufficient for inducing AS defects identified under APP neurodegeneration, therefore underscoring criticality of Tip60-mediated AS regulation in AD pathogenesis. Further, since several Tip60-rescued fly AS genes show splicing defects in postmortem AD human brains (Tollervey et al., 2011; Raj et al., 2018; Adusumalli et al., 2019; Marques-Coelho et al., 2021) and we find Tip60-RNA binding is altered in AD hippocampus, we propose that Tip60-mediated AS modulation is a conserved critical posttranscriptional step that is disrupted early in AD etiology. In particular, we find Tip60-RNA binding of ADAM10, a constitutive α-secretase, is significantly lost under AD pathology. Interestingly, reduced ADAM10 activity in the postmortem human AD brain has been linked to AS induced isoform change without a change in the overall gene expression (Marques-Coelho et al., 2021), suggesting Tip60 modulated ADAM10 splicing could be central to AD pathogenesis. Thus, we are the first to uncover distinct histone and RNA binding capabilities for Tip60 that mediate its function in neural gene control and RNA splicing, respectively, and may underly the chromatin packaging and splicing defects that are now characterized as hallmarks of AD.
One target, two functions: Tip60 HAT as a novel therapeutic target for Alzheimer's disease
Pharmacological treatments aiming to restore histone acetylation via HDAC inhibition are currently a research hotspot for developing AD cognition enhancing drugs (Gräff and Tsai, 2013; Simões-Pires et al., 2013; Mielcarek et al., 2015). Although promising in reinstating cognition, HDAC inhibitors are known to exhibit side effects because of nonspecific global hyperacetylation (Didonna and Opal, 2015; Yang et al., 2017). Alternatively, enhancing activity of specific HATs in promoting cognition associated histone acetylation serves as an exciting new therapeutic strategy that remains to be fully explored (Caccamo et al., 2010; Selvi et al., 2010; Valor et al., 2013). In support of this concept, here we identify a novel splicing modulation function for Tip60 that likely complements its histone function for neuroprotection, therefore highlighting Tip60 as unique dual-functioning therapeutic target for ameliorating both, histone and splicing aberrations in AD. Strikingly, mutations only in RNA-binding and not histone-binding residues in the Esa1 HAT chromodomain are lethal (Shimojo et al., 2008), strongly supporting that the RNA function of HATs are nonredundant and critical for viability. Although precise mechanisms underlying Tip60-mediated AS regulation remain to be elucidated, we propose three probable mechanisms. First, similar to RNA-binding proteins (Herzel et al., 2017; Rachez et al., 2021), Tip60 could be binding at a splice site or an accessory site influencing transient RNA folding, and therefore, may modulate the timing of splice site exposure to the splicing machinery. Second, since Tip60 typically interacts with additional proteins in a complex for gene regulation (Ikura et al., 2000; Frank et al., 2003), Tip60-RNA binding could trigger assembly of a secondary complex that ultimately modulates AS decisions. Third, since histone acetylation modifications have been implicated in AS regulation (Hnilicová et al., 2011; Rahhal and Seto, 2019), Tip60's HAT function might modulate AS decisions while it tethers the target pre-mRNA to the chromatin. Further, we observed Tip60-rescued AS events in two major splicing regulators, RBFOX1 and PTBP1 that could in turn impact splicing of other neural genes. Nevertheless, we do not rule out potential additional mechanisms. Our findings strongly underscore supplementing current histone acetylation targeted therapeutics with splice-switching strategies, such as the use of antisense oligonucleotides (AO) that for desired pre-mRNA processing (Quemener et al., 2020; D. Li et al., 2021). Currently, six splice-switching AO have been approved by the US FDA for mRNA manipulation in rare diseases (D. Li et al., 2021; Raguraman et al., 2021). Although successful in reducing Aβ production and ameliorating cognition in AD pre-clinical models (Huynh et al., 2017; Chang et al., 2018), further studies are needed to corroborate the effectiveness and safety of splice-switching AO in AD. Dissecting apart Tip60's histone versus RNA function and further elucidation of mechanisms underlying Tip60-mediated AS modulation should provide earlier, safer, and more selective ways for AD therapeutics in the clinical setting.
Footnotes
The research was supported by the National Institutes of Neurological Disorders and Stroke of the National Institutes of Health under Award Number R01NS095799 (to F.E.). We thank Dr. Ann Ehrenhofer-Murray for generously contributing the Drosophila-Tip60 antibody. We also thank Dr. Harini Sreenivisappa for overseeing microscopic imaging at Drexel University's Cell Imaging Center.
The authors declare no competing financial interests.
- Correspondence should be addressed to Felice Elefant at fe22{at}drexel.edu