Sex-specific regulation of the cardiac transcriptome by the protein phosphatase 2A regulatory subunit B55α

Experimental animals

Model 1: C57BL/6 N mice heterozygous (Het) for a Ppp2r2a^{tm2a(KOMP)Wtsi} ‘knockout-first’ allele (RRID: MMRRC:060623-UCD; see Fig. 1a) were generated by the Australian Phenomics Network (Monash University, Melbourne, Australia). Model 2: To generate cardiomyocyte-specific Ppp2r2a knockout (cKO) mice, mice heterozygous for the knockout-first allele were crossed with hemizygous transgenic mice expressing Flp recombinase (Flp^Tg/0 mice; RRID: IMSR_JAX:003800; see Fig. 1a). This removed the lacZ reporter and neomycin selection cassette and converted Ppp2r2a^{tm2a(KOMP)Wtsi} to a floxed Ppp2r2a allele. Next, offspring hemizygous for Flp and heterozygous for the floxed Ppp2r2a allele (Flp^Tg/0 Ppp2r2a^L/+) were crossed to remove the Flp transgene. Flp-negative floxed Ppp2r2a mice (Flp^0/0 Ppp2r2a^L/L) were then mated with hemizygous transgenic mice expressing Cre recombinase under the control of the αMHC promotor (αMHC-Cre^Tg/0 mice, FVB background)^20,27. Importantly, Cre expression in this strain does not lead to cardiotoxicity^20,28. Cre^Tg/0 Ppp2r2a^L/+ and Cre^0/0 Ppp2r2a^L/+ mice were mated to produce the following 6 cardiomyocyte-specific genotypes: Cre^Tg/0 Ppp2r2a^L/L (B55α knockout, cKO), Cre^0/0 Ppp2r2a^L/L (floxed control, FC), Cre^Tg/0 Ppp2r2a^L/+ (B55α heterozygote, cHET), Cre^0/0 Ppp2r2a^L/+ (half-floxed control), Cre^Tg/0 Ppp2r2a^+/+ (Cre control), Cre^0/0 Ppp2r2a^+/+ (wildtype control). FC and cKO mice for the RNA-seq analysis were generated from Cre^Tg/0 Ppp2r2a^L/L x Cre^0/0 Ppp2r2a^L/L breeding pairs. All aspects of animal care and experimentation were approved by The Alfred Research Alliance Animal Ethics Committee. Mice were housed in a temperature-controlled environment with a 12-hour light/dark cycle and fed standard chow ad libitum.

Timed matings for collection of embryos at 12.5 dpc

Het breeding pairs were mated. Pregnancy was identified via the presence of a vaginal plug, and midday was designated as 0.5 days post coitum (dpc). Pregnant females were anesthetised with pentobarbital (~30 mg intraperitoneal) and euthanised via cervical dislocation. An incision was made in the lower abdomen, and the uterine horn removed and immediately placed on sterile gauze in an ice slurry. Embryos were carefully dissected under microscope guidance and placed in cold 4% paraformaldehyde (PFA) with gentle shaking overnight.

Embryo histology

Embryos were dehydrated, paraffin-embedded in a sagittal orientation and 5 µm sections obtained at the level of the heart. Sections were stained with hematoxylin & eosin (H&E) and imaged using a Pannoramic Scan II (3D Histech, Budapest, Hungary), using a Carl Zeiss Plan-Apochromat 20x/NA 0.8 objective and a Point Grey Grasshopper 3 CCD monochrome camera. Digital slide images were viewed using SlideViewer software version 2.6 (3D Histech, Budapest, Hungary). Embedding and processing was conducted by the University of Melbourne Histology Platform (Melbourne, Australia) and slide scanning performed by Phenomics Australia (Melbourne, Australia). Additionally, four paraffin-embedded embryos (3 Hom, 1 Wt) were submitted to Phenomics Australia for histopathological evaluation^29–34 by two independent pathologists.

Echocardiography

Echocardiography was performed on anaesthetised mice (1.8% isoflurane) using either an iE33 ultrasound machine with a 15-megahertz (mHz) linear array transducer (Phillips, Amsterdam, Netherlands) or a Vevo 2100 High-Frequency Ultrasound System with a MS550D transducer (Visual Sonics, Toronto, ON, Canada). Short axis M-mode images were used to measure interventricular septum (IVS) and left ventricular posterior wall (LVPW) thicknesses at diastole, as well as internal dimensions at diastole (LVID; d) and systole (LVID; s). Dimensions were measured from three beats per echocardiogram and averaged. Fractional shortening and LV mass were calculated using the following equations, respectively: [(LVID; d – LVID; s)/LVID; d] x 100%, [(IVS + LVPW + LVID; d)³ – LVID; d ³]. Investigators were blinded to genotype during image acquisition and analysis. All analyses were independently validated by the Baker Heart & Diabetes Institute Preclinical Microsurgery & Imaging Platform (Melbourne, Australia)³⁵.

Tissue collection

Adult mice were anesthetised with pentobarbital (~30 mg intraperitoneal) and euthanised via cervical dislocation. The heart was quickly removed, washed in cold PBS, excess moisture removed with sterile gauze, and weighed. The atria were dissected and weighed. Ventricles were cut along the short axis, separating the base and apex. The base was fixed in cold 4% PFA for processing and paraffin embedding. The apex was divided evenly along the long axis and snap-frozen in liquid nitrogen for protein and RNA extraction. Other tissues including lungs, liver, kidney, and spleen were weighed and snap-frozen in liquid nitrogen.

Gene expression by quantitative PCR

Total RNA was extracted from ventricles using TRI-Reagent (Sigma Aldrich, St. Louis, MO, US), according to the manufacturer’s instructions. Nanodrop spectrophotometry (Thermo Scientific, Waltham, MA) was used to quantify RNA. A High-Capacity RNA-to-cDNA kit (Thermo Scientific, Waltham, MA) was used to reverse transcribe 2 µg RNA to cDNA according to the manufacturer’s instructions. RT-qPCR was conducted using Taqman gene expression assays and Taqman Fast Universal Master Mix (Thermo Fisher Scientific, Waltham, MA) or primers and SYBR Green PCR Master Mix (Thermo Fisher Scientific, Waltham, MA). Targets were amplified using a Quant Studio ViiA7 Real-Time PCR System. The 2^–ΔΔCt method was used for the calculation of fold difference using hypoxanthine phosphoribosyltransferase 1 (Hprt1) as a reference gene. For SYBR Green reactions, primer efficiencies were calculated according to the following equation: E = 10^[-1/slope], using the slope of a standard curve generated using serial dilutions of cDNA. Melt curves were performed to assess the specificity of the amplification product. Samples were run in triplicate, and samples with replicates with a standard deviation exceeding 0.5 were excluded. Taqman gene expression assays from Thermo Fisher Scientific (Waltham, MA) were as follows: Nppa Mm01255747_g1, Nppb Mm01255770_g1, Myh6 Mm00440359, Myh7 Mm01319006, Col1a1 Mm00801666_g1, Col3a1 Mm00802300_m1, Ctgf Mm01192932_g1, Hprt1 Mm01545399_m1. SYBR Green primer sequences were as follows: Ppp2r2a fwd 5’-CCGTGGAGACATACCAGGTA-3’, rev 5’-AACACTGTCAGACCCATTCC-3’; Ppp2r2c fwd 5’-AGCGGGAACCAGAGAGTAAG-3’, rev 5’-GTAGTCAAACTCCGGCTCG-3’; Ppp2r2d fwd 5’- TTACGGCACTACGGGTTCCA-3’, rev 5’-TTCGTCGTGGACTTGCTTCT-3’; Hprt1 fwd 5’- TCCTCCTCAGACCGCTTTT-3’, rev 5’-CCTGGTTCATCATCGCTAATC’-3’.

Protein analysis

Ventricles were homogenised in lysis buffer [20 mM Tris-HCl (pH 7.4), 127 mM NaCl, 10% glycerol, 1% IGEPAL, 20 mM NaF, 10 mM EGTA, 1 mM sodium pyrophosphate, 1 mM vanadate, 1 mM PMSF, 4 µg/mL pepstatin, 4 µg/mL aprotinin, 4 µg/mL leupeptin], incubated on ice for 15 min, and centrifuged at 4 °C for 15 min at 16,000 g to pellet cellular debris. Protein lysates (80 µg) and molecular weight marker (Precision Plus Protein All Blue Prestained Protein Standard, Bio-Rad, 1610373) were separated using 7.5% or 10% SDS-PAGE gels, transferred to PVDF membranes, and incubated with primary antibody overnight at 4°C or 1.5 hours at room temperature for HRP-conjugated secondary antibodies. Chemiluminescent signals were quantified using ImageJ pixel analysis (US National Institutes of Health) or Genetools analysis software (Syngene, Cambridge, UK). Data were expressed as fold difference relative to the control group. Antibodies were as follows: B55α (#sc-81606, Santa Cruz, 1:1000), FoxO1 (#2880, Cell Signaling Technology, 1:2000), pSer256 FoxO1 (#9461, Cell Signaling Technology, 1:1000), pThr24/Thr32 FoxO1/O3a (#9464, Cell Signaling Technology, 1:2000), HDAC5 (#20458, Cell Signaling Technology, 1:500), pSer259 HDAC5 (#3443, Cell Signaling Technology, 1:500), PP2A-C (#2038, Cell Signaling Technology, 1:1000), PP2A-A (#2039, Cell Signaling Technology, 1:250), α-tubulin (#2144, Cell Signaling Technology, 1:2500) and vinculin (#V9131, Sigma Aldrich, 1:5000). All uncropped blots are included in the Data Supplement.

RNA-sequencing

Left ventricle tissue (~20 mg) was mechanically homogenised in TRI-Reagent with a 5 mm steel bead at 30 Hz for 2 × 30 s (TissueLyser II, Qiagen) then centrifuged for 10 min at 10,000 g. RNA was extracted from the supernatant using spin columns with DNase-I treatment (Zymo DirectZol RNA Miniprep kit #R2050, Integrated Sciences, Australia). RNA concentration and purity was determined spectrophotometrically (Nanodrop, Thermo Fisher, Australia) and RNA integrity was determined by electrophoresis (Tapestation, Agilent). RNA integrity (RIN) values for all samples were ≥7.7. Stranded libraries were prepared from 500 ng total RNA input with poly-A+ selection (TruSeq Stranded mRNA, Illumina). Library sizes and concentrations were confirmed prior to sequencing (TapeStation, Agilent). At least 40 million 150 bp paired-end reads per library were generated on the Illumina NovaSeq 6000 platform (Australian Genome Research Facility, Melbourne, Australia). Raw read preprocessing was conducted on the Galaxy Australia platform³⁶. Reads underwent initial quality check with FastQC v0.74, then quality filtering and adapter trimming was conducted with Cutadapt v4.6. Filtered reads were aligned to the mouse reference genome (Mus musculus, Ensembl version GRCm39.104) using STAR v2.7.11a³⁷ in 2-pass mode. Gene expression was quantified at the gene level using HTseq-counts v2.0.5 and collated into a counts matrix. Analysis of differential expression was performed on genes with ≥1 counts per million (CPM) in all samples in DESeq2 in iDEP v2.0³⁸. False discovery rate (FDR) was calculated using Benjamini-Hochberg method and genes with an FDR < 0.05 were considered differentially expressed (DE). DE genes were analysed for pathway enrichment using Gene Ontology (GO) genesets in iDEP v2.0.

Fibrosis analysis

Paraffin-embedded left ventricles were cut (6 μm sections), stained with picrosirius red and imaged using a Pannoramic Scan II (3D Histech, Hungary) using a Carl Zeiss Plan-Apochromat 20x/NA 0.8 objective and a Point Grey Grasshopper 3 CCD monochrome camera. Images of scanned sections were obtained at 20x magnification (10 regions/heart) using SlideViewer software version 2.6 (3D Histech, Budapest, Hungary). The number of red pixels (collagen) was counted using ImageJ (NIH, Bethesda, USA) and divided by tissue area to calculate the percentage of fibrosis. Analysis was undertaken blinded to sex and genotype.

Data presentation and statistical analysis

Data are presented as scatterplots with bars and lines indicating the mean +/− SEM (for qPCR data, as per Thermo Fisher Scientific’s Guide to Performing Relative Quantitation of Gene Expression Using Real-Time Quantitative PCR, 2008 edition) or median +/− IQR (all other data). Fold difference (qPCR and Western blot data) was calculated by dividing by the mean or median of the control group, as appropriate. Gene expression data were analysed using two-sided unpaired t-tests (2 groups) or two-way ANOVA followed by Tukey’s or Sidak’s post hoc tests, as appropriate. All other data were analysed using Mann-Whitney U tests (2 groups) or Kruskal-Wallis tests (6 groups). A P-value < 0.05 was deemed statistically significant. Animal numbers and data exclusions are reported in Supp Fig. 1.

Homozygous knockout of B55α causes embryonic lethality without overt cardiac defects

The expected Mendelian ratio of Wt : Het : Hom mice in the global knockout mouse strain (Model 1) was 1:2:1. However, no Hom offspring were born to Het breeding pairs, indicating lethality and resorption of Hom embryos in utero. Consistent with previous reports in different B55α knockout mouse strains^11,26, Hom embryos were smaller than Wt littermates at 12.5 dpc (Fig. 1b) and started dying between 12.5 and 14.5 dpc (26% Hom at 12.5 dpc, 6% Hom at 14.5 dpc). Histopathological evaluation at 12.5 dpc identified several abnormalities in Hom embryos, including arrested oral/nasal cavity development (Fig. 1c, left panel), arrested limb bud development (Fig. 1c, middle panel), and deficient skin layers (Fig. 1c, right panel). One of the three Hom embryos evaluated also had a smaller liver and displayed hepatocyte necrosis. Hearts of Hom embryos were assessed for the presence of congenital cardiac abnormalities, including atrial or ventricular septal defects, abnormal leaflet formation or myocardial non-compaction, however there were no obvious micromorphological differences compared with Wt (Fig. 1d).

Reduced expression of B55α did not lead to left ventricular hypertrophy or dysfunction in young mice but induced LV wall thinning in male mice with age

Ventricular B55α protein levels were ~45% lower in the male Het group vs Wt and ~70% lower in the female Het group vs Wt (Fig. 2a, b), allowing us to investigate the impact of reduced B55α expression on cardiac morphology and function in adult mice. There were no differences in the relative abundance of B55α between male and female control hearts (Supp Fig. 2). We assessed mice at 10–12 weeks of age (representing early adulthood) and at 12 months (representing middle age). By echocardiography, there were no differences in left ventricular wall thicknesses or internal dimensions between Wt and Het mice in the 10–12-week old cohort (Supp Table 1). Heart rate and fractional shortening, a measure of left ventricular contractility, were also not different between Wt and Het mice (Supp Table 1). At 12 months of age, male Het mice had thinner LV walls compared to Wt mice based on LVPW and IVS thicknesses (Fig. 2c). This phenotype was less apparent in females, with lower IVS thickness in Het mice but no difference in LVPW (Fig. 2d). No other differences in echocardiographic parameters were observed (Fig. 2c, d; Supp Table 2).

Assessment of organ weights at 10–12 weeks and at 12 months of age revealed no differences in heart weight or atria when normalised to body weight or tibia length, although atria tended to be smaller in male Het vs Wt in the younger cohort (Fig. 2e, f; Supp Table 3, Supp Table 4). Lung weight was within the normal range, consistent with the absence of cardiac pathology (which can lead to pulmonary congestion; Supp Tables 3, 4). Liver weight was lower in female Het vs Wt in the younger cohort (Supp Table 3) and tended to be higher in male Het vs Wt in the older cohort (see Supp Table 4). Variations in the liver were also observed in Hom embryos at 12.5 dpc (section “Homozygous knockout of B55α causes embryonic lethality without overt cardiac defects”), indicating a possible hepatic function for B55α. We did not measure blood pressure in these mice, which is a limitation of the study.

To investigate if the mild LV wall thinning observed in Het mice at 12 months of age was accompanied by differences in the expression of genes involved in pathological cardiac remodelling, we performed qPCR on ventricular lysates (Supp Fig. 3a, b). Levels of cardiac stress markers atrial and B-type natriuretic peptides (ANP and BNP, encoded by Nppa and Nppb, respectively) did not differ in Het mice compared to Wt in either sex. There were no differences in the expression of myosin heavy chain isoforms αMHC or βMHC (encoded by Myh6 and Myh7, respectively). Expression of type I and III collagens (Col1a1 and Col3a1) and connective tissue growth factor (Ctgf) were comparable between genotypes, suggesting an absence of cardiac fibrosis. We also assessed the expression of blood and lymphatic vessel markers (Cd31 and Lyve1, respectively) as impaired angiogenesis or lymphangiogenesis in the heart is deleterious^39,40, and a previous study reported vascular and lymphatic vessel defects in skin of embryos lacking B55α¹¹. We observed a significant reduction in mRNA, but not protein, levels of the lymphatic vessel marker Lyve1 in male Het ventricles vs Wt (Supp Fig. 3c, d), while Cd31 expression was not different between genotypes (Supp Fig. 3c). Lyve1 mRNA and protein levels were not different in female Het and Wt hearts (Supp Fig. 3e, f). Cd31 mRNA expression tended to be lower in female Het mice vs Wt, however the effect was small (~10% decrease, P = 0.05, Supp Fig. 3e).

Loss of B55α in cardiac myocytes does not impact postnatal cardiac growth, but alters the transcriptional profile in male mice

To circumvent the effects of B55α deletion on embryonic development and investigate the role of cardiomyocyte B55α in postnatal heart growth, we next characterised the phenotype of mice with cardiomyocyte-specific knockout (cKO) of B55α using an αMHC-Cre line (Model 2, Fig. 1a). We confirmed loss of B55α in ventricular tissue of 10–12-week old cKO mice by Western blotting. As expected, B55α protein levels were significantly lower (~90% decrease in expression) in both male and female cKO hearts compared to floxed controls (FC, Fig. 3a–d). We also examined the expression of the other components of the heterotrimeric PP2A holoenzyme. We observed a significant decrease (~35%, P = 0.03) in PP2A-A protein levels in female cKO hearts, but no differences in the expression of PP2A-C (Fig. 3a–d). Deletion of Ppp2r2a in the heart was not associated with differences in the expression of other PP2A B55 subunit isoforms, i.e. Ppp2r2c (encoding B55γ) or Ppp2r2d (encoding B55δ; Fig. 3e, f). We also observed no differences in HDAC5 phosphorylation at Ser259 between genotypes in either sex (Fig. 3g, h), suggesting that HDAC5 is not phospho-regulated by PP2A-B55α in the heart under normal physiological conditions.

At 10–12 weeks of age, there were no detectable differences in LV wall thicknesses or chamber dimensions between any of the genotypes (Fig. 4a, b; Supp Table 5). Fractional shortening tended to be lower in male cKO vs FC, but the effect was small (~10% decrease, P = 0.10). Heart, atria and lung weight of both male and female cKO mice were comparable with controls (Fig. 4c, d; Supp Table 6), indicating the absence of pathological cardiac hypertrophy and dysfunction.

Interestingly, we observed differences in the expression of genes involved in extracellular matrix and myofilament composition in ventricular tissue of male mice (Fig. 5a). Expression of type I collagen (Col1a1) and connective tissue growth factor (Ctgf) were modestly increased in male cKO hearts compared with FC (~30% increase in Col1a1, P = 0.04; ~70% increase in Ctgf, P = 0.02, Fig. 5a). Type III collagen expression also tended to be higher (~20% increase in Col3a1, P = 0.08, Fig. 5a). Myh7 (encoding βMHC) was also significantly higher in male cKO hearts vs FC (~70% increase, P = 0.02, Fig. 5a). Upregulation of the foetal βMHC isoform is frequently observed in cardiac disease settings^16,41,42. Coupled with our observation of lower fractional shortening in male cKO mice vs FC, this gene expression profile could indicate the early onset of pathology in male cKO mice, although increased expression of the βMHC isoform is typically accompanied by a reciprocal decrease in the expression of the αMHC isoform.

We and others have previously shown that the αMHC-Cre transgenic mouse strain used in this study does not have increased collagen expression or fibrosis at 6-8 months of age^20,28. However, as another αMHC-Cre strain (with much higher levels of Cre expression) develop cardiac fibrosis²⁸, we compared Col1a1, Col3a1 and Ctgf expression in hearts of male FC, cKO and Cre control mice. Our analyses revealed no differences in expression between the FC and Cre control groups (Supp Fig. 4), confirming that the modest increases in expression observed in cKO hearts were due to deletion of B55α and not an off-target effect of Cre expression.

In female cKO mice, increases in fibrotic gene expression were not present (Fig. 5b). We observed reduced expression of Myh6, a known MEF2 transcriptional target encoding the dominant adult αMHC isoform (~40% decrease, P = 0.004, Fig. 5b). Expression of the natriuretic peptide genes Nppa and Nppb were not different in male or female cKO hearts compared to control.

To further investigate differences in cardiac gene expression arising from loss of cardiomyocyte B55α, we performed RNA-seq analysis on a separate cohort of FC and cKO mice. Interestingly, and consistent with the findings of our qPCR analysis (Fig. 5a, b), we observed a greater number of genes that were differentially expressed due to B55α deletion in the male heart compared with the female heart (Fig. 6a, b; Supp Table 7). Remarkably, only 5 genes (in addition to Ppp2r2a) were affected by B55α deletion in the female heart (Fig. 6b), while 543 were differentially expressed in male cKO vs FC (FDR < 0.05, Fig. 6a). Eya1 was the only gene that was significantly upregulated in both male and female hearts with B55α deletion (Fig. 6a, b). Eya family members are transcriptional co-activators that have been shown to physically interact with B55α, directing PP2A activity towards transcription factors such as c-Myc⁴³. The physiological significance of this interaction has not been investigated in the heart.

Pathway enrichment analysis of genes that were upregulated in male cKO vs FC hearts revealed an over-representation of Gene Ontology terms associated with cell/tissue/organism development and morphogenesis (Biological Process), extracellular matrix (Cellular Component), and ECM structural constituents and binding (Molecular Function, Fig. 6c; Supp Table 8). Amongst genes that were downregulated in male cKO vs FC hearts, there was an over-representation of terms associated with oxidative phosphorylation and energy production (Biological Process), mitochondria (Cellular Component), and activity of the electron transport chain (Molecular Function, Fig. 6d). These data suggest a role for B55α in the maintenance of ECM composition and mitochondrial energy metabolism in the male heart.

Analysis of LV cross-sections stained with picrosirius red revealed that upregulation of ECM-associated genes in male cKO mice translated to a small increase in collagen protein deposition (Fig. 6e). Surprisingly, we detected a similar increase in LV collagen content in female cKO mice vs FC, despite no differences in the expression of Col1a1 or Col3a1 (Figs. 5b and 6b). Examination of other collagen species quantified by RNA-seq revealed a significant effect of genotype and sex for Col1a2, Col4a3, Col4a4 and Col14a1 (P < 0.05 for all, Fig. 6f–i). Col4a3 and Col4a4 expression tended to be higher in female cKO hearts vs FC, although these differences were not as pronounced as in male hearts and were not statistically significant (Fig. 6g,h). Col14a1 expression was significantly higher in female hearts compared with male hearts (~2.7-fold increase in female FC vs male FC, P < 0.001, Fig. 6i). Collectively, these data identify cardiomyocyte B55α as a regulator of the cardiac ECM.

In addition, we investigated the phosphorylation status of FoxO1, a reported target of B55α (at Ser256 and Thr24)¹⁸ and key transcription factor regulating mitochondrial homeostasis and cardiac energy metabolism^44,45. FoxO1-dependent gene transcription is regulated by phosphorylation, with Ser256 dephosphorylation promoting the nuclear accumulation of FoxO1 and Ser256 phosphorylation promoting cytosolic accumulation in cardiomyocytes⁴⁶. We detected no differences in the phosphorylation of FoxO1 at Ser256 or Thr24 between genotypes (Supp Fig. 5), suggesting that FoxO1 is unlikely to have contributed to the gene expression differences we observed in hearts of B55α cKO mice.