Knockdown of heterochromatin protein 1 binding protein 3 recapitulates phenotypic, cellular, and molecular features of aging

Abstract Identifying genetic factors that modify an individual's susceptibility to cognitive decline in aging is critical to understanding biological processes involved and mitigating risk associated with a number of age‐related disorders. Recently, heterochromatin protein 1 binding protein 3 (Hp1bp3) was identified as a mediator of cognitive aging. Here, we provide a mechanistic explanation for these findings and show that targeted knockdown of Hp1bp3 in the hippocampus by 50%–75% is sufficient to induce cognitive deficits and transcriptional changes reminiscent of those observed in aging and Alzheimer's disease brains. Specifically, neuroinflammatory‐related pathways become activated following Hp1bp3 knockdown in combination with a robust decrease in genes involved in synaptic activity and neuronal function. To test the hypothesis that Hp1bp3 mediates susceptibility to cognitive deficits via a role in neuronal excitability, we performed slice electrophysiology demonstrate transcriptional changes after Hp1bp3 knockdown manifest functionally as a reduction in hippocampal neuronal intrinsic excitability and synaptic plasticity. In addition, as Hp1bp3 is a known mediator of miRNA biogenesis, here we profile the miRNA transcriptome and identify mir‐223 as a putative regulator of a portion of observed mRNA changes, particularly those that are inflammatory‐related. In summary, work here identifies Hp1bp3 as a critical mediator of aging‐related changes at the phenotypic, cellular, and molecular level and will help inform the development of therapeutics designed to target either Hp1bp3 or its downstream effectors in order to promote cognitive longevity.


T-maze
Four to six weeks following viral injections, working memory was assessed on the T-maze. Mice were habituated to the testing room for three days prior to testing. Mice were placed in the start arm of the T and allowed to freely choose one of two goal arms. Upon entering an arm, mice were confined to that same arm for 30s. The mice were then placed back into the start arm and allowed to choose again. A correct alternation occurred when the mouse chose the new arm, not the arm to which it had previously been confined. Two trials per day were conducted over three days for a total of six independent trials. Results were combined to obtain a measure of average percent correct alternation.

Contextual fear conditioning
Similar to T-maze, mice were habituated to the testing room for three days prior to testing. Standard contextual fear conditioning was performed (Neuner et al., 2016;Neuner et al., 2015), with mice receiving four 0.9 mA shocks on training day. Twenty-four hours later, mice were returned to the training chamber and the percentage of time the mice spent freezing during a 10-minute test was recorded as an index of contextual fear memory.

Western blot
Whole hippocampal lysates were prepared as previously described (Neuner et al., 2016). Protein concentration was determined using a NanoDrop Spectrophotometer 2000 and 20 µg total protein was loaded onto a BioRad TGX Mini Protean gel. Proteins were separated by electrophoresis and transferred to a nitrocellulose membrane using the Trans-Blot Turbo system (BioRad). Membranes were blocked in 5% non-fat milk in PBST and primary antibodies for HP1BP3 (gifted by Drs. Benjamin Garfinkel and Joseph Orly) and GAPDH (Fitzgerald Industries #10R-G109A) were incubated at 4C overnight. Fluorescently labeled secondary antibodies were incubated for 1 hour at RT and bands were visualized using an Azure Biosystems gel imager. Observed double band staining is typical expression pattern for HP1BP3 (Garfinkel et al., 2015) and overlaps with positive control HP1BP3 overexpression lysate from human 293T cells (Abnova #H00050809-T02), which was used as a positive control. As such, both bands were included in our analysis. Band densities were measured using LI-COR Image Studio Lite (LI-COR Biosciences). Densities were first normalized to an internal B6 control, and then total HP1BP3 was adjusted for total protein loaded using GAPDH. Adjusted densities are plotted in Figure 1B and Figure S1.

RNA sequencing
Total RNA was prepared from whole hippocampal homogenates using the Qiagen miRNeasy Mini kit (#217084). RNA quality was assessed using a BioAnalyzer and only samples with RNA Integrity Number (RIN) greater than 8.0 were included. Samples were submitted to The Jackson Laboratory Genome Technologies department, where mRNA libraries were prepared using the KAPA Biosystems mRNA Hyper Prep Kit while miRNA libraries were prepared using the Illumina TruSeq Small RNA Library Prep Kit. The Illumina HiSeq 2500 was used to sequence 75 paired-end reads, at a depth of 30 million reads for mRNA and 10 million reads for miRNA. Two pairs of mRNA/miRNA samples given identical barcodes during the initial sequencing, so all samples were re-sequenced and where available, fastq files from both runs were concatenated, providing expanded sequencing depth for 10/12 samples. Poor quality mRNA reads were filtered, trimmed, and aligned to either the C57BL/6J or DBA/2J genomes where appropriate using rsem v1.2.12, followed by bowtie2 v2.2.0 for expression estimate of trimmed reads (Li and Dewey, 2011). For miRNA, trimmed reads were aligned to the mouse miRNA-precursor miRBase reference (miRBase release 21) using bowtie v1.0.0 (Griffiths-Jones, 2006;Kozomara and Griffiths-Jones, 2014). Following alignment miRNA counts were extracted using featureCounts function from the subread package v1.5.2 (Liao et al., 2013).

Differential expression analysis
For both RNA types, expected read counts were used to perform differential expression analysis using DESeq2 according to established protocols (Love et al., 2014). miRNAs of interest were checked for single nucleotide polymorphisms (SNPs) using the Sanger Mouse Genomes database (Keane et al., 2011) to ensure observed differences were not due to strain-specific differences in miRNA sequence. For graphing of mRNA and miRNA differences across groups, normalized counts were extracted from DESeq2 using the counts function. miRNA counts were further log transformed using the rlog function.

Gene set enrichment analyses
For gene set enrichment analysis (GSEA), all genes nominally significantly differentially expressed relative to treatment (Hp1bp3 KD vs Ctrl unadjusted p-value < 0.05) were sorted into a ranked list by log 2 fold change, with the most positively differentially expressed genes at the top of the list. This pre-ranked gene list was uploaded into the GSEA desktop software (Subramanian et al., 2005). GSEA was used according to established procedures to identify significant overlap between genes positively or negatively changed by treatment and gene sets in the Molecular Signatures Database 3.0 (Liberzon et al., 2011). The same list of differentially expressed genes was uploaded into Ingenuity Pathway Analysis [IPA, Qiagen, (Kramer et al., 2014)] and the Core Analysis function was utilized to identify upstream miRNA regulators of observed mRNA changes. Specifically, we filtered results to include references from the Ingenuity Knowledge Base, included both direct and indirect connections, and considered only molecules and/or relationships which had been experimentally observed or had high predictive value. We then filtered the putative upstream regulators list to only include regulators of molecule type equal to "miRNA" or "mature miRNA". Finally, to perform simple overrepresentation enrichment analysis on small gene sets (e.g. miRNA target genes) as opposed to searching specifically for enrichment among up or down-regulated genes (as in GSEA), we utilized WebGestalt (Wang et al., 2013) to identify significantly enriched GO Biological Process terms.

Statistical analysis and data availability
All experiments and analysis were conducted with experimenters blind to treatment group. All statistics were performed using SPSS (IBM) or R and tests used included independent t-tests, two-way ANOVA, and two-way repeated measures ANOVA. Our LTP data used in the two-way repeated measures ANOVA violated Mauchly's test of sphericity (p < 0.05), so Greenhouse-Geisser corrected p-values were used for analysis. Unless otherwise stated, data values here are given as mean ± standard error. All raw and processed data has been deposited to Gene Expression Omnibus (GEO) and is available as SuperSeries GSE119321. Specifically, the mRNA sequencing data is available as GSE119318 and the miRNA sequencing data is available as GSE119319. All software packages used in R for data analysis are freely available.