The present study utilized inducible, cardiomyocyte-specific dominant negative O-GlcNAcase (dnOGAh) transgenic mice, as described previously (7). Briefly, to generate dnOGAh mice, TRE-EGFP-dnOGA (tetracycline-responsive element, enhanced green fluorescent protein, rat OGA splice variant) mice were bred with αMHC-rtTA (α-myosin heavy chain promoter driven codon optimized reverse tetracycline transactivator) mice. To induce the transgene, 10 week old male dnOGAh mice were injected with 100 μg doxycycline (DOX) (MilliporeSigma, Burlington, MA, USA) in 0.9% NaCl solution and immediately switched to 1 g/kg DOX-supplemented NIH31 diet (Envigo, Madison, WI, USA). αMHC-rtTA mice treated with DOX for the same duration served as controls (CON). This study also utilized male cardiomyocyte-specific BMAL1 knockout (CBK) mice; littermate floxed mice served as controls (CON), as described previously (14). All mice were housed at the Center for Comparative Medicine at the University of Alabama at Birmingham (UAB), under temperature-, humidity-, and light- controlled conditions. A strict 12 h light/12 h dark cycle regime was enforced [lights on at 6 a.m.; zeitgeber time (ZT) 0]; the light/dark cycle was maintained throughout these studies, facilitating investigation of diurnal variations (as opposed to circadian rhythms). Mice were housed either in standard micro-isolator cages or CLAMS (Comprehensive Laboratory Animal Monitoring System) cages, and received food and water in an ad libitum fashion. CLAMS cages allowed the non-invasive assessment of energy expenditure, food intake and physical activity in a continuous manner, as described previously (17). All mice were on the C57BL/6 background strain, were male, and were 12 weeks old at the time of assessments and/or tissue collection. All animal experiments were approved by the Institutional Animal Care and Use Committee of the University of Alabama at Birmingham.

RNA was extracted from biventricular samples using standard procedures. Candidate gene expression analysis was performed by quantitative RT-PCR, using previously described methods (18, 19). For quantitative RT-PCR, custom-designed Taqman assay were utilized; sequences for custom-designed assays have been reported previously (14, 20). All quantitative RT-PCR data are presented as fold change from an indicated control group.

Transcriptomic analysis was performed using RNA sequencing (RNA-seq) in the UAB Genomics Core facility. Following initial testing of RNA samples using an Agilent BioAnalyzer, RNA with RIN values greater than 7.0 were subsequently utilized for library preparation (after DNAse treatment). RNA-Sequencing libraries were next generated using the NEBNext Ultra II RNA kit (NEB, Ipswich MA); resulting libraries were sequenced on the Illumina NextSeq 500 (Illumina, Inc. San Diego CA) using paired end 75 bp sequencing reads, per standard methods. Sequencing data can be accessed in GEO.

Qualitative analysis of protein expression was performed via standard western blotting procedures, as described previously (8). Briefly, 15–30 μg protein lysate was separated on polyacrylamide gels and transferred to PVDF membranes. Membranes were probed for with anti- BMAL1 (Cell Signaling Tech #14020; 1:2000), REVERBα (Cell Signaling Tech #13418; 1:2000), PER2 (Alpha Diagnostics Inc #PER21-A; 1:2500), and E4BP4 (Cell Signaling Tech #14312; 1:2000) antibodies, or with the CTD110.6 antibody (UAB Epitope Recognition Core) for detection of O-GlcNAcylation. Rabbit and mouse HRP-conjugated secondary antibodies (Cell Signaling 7074 and 7076, respectively; 1:2000–5000) were used for chemiluminescent detection with Luminata Forte Western Blotting substrate (Millipore, WBLUF0100). All densitometry data were normalized to amido black staining. Importantly, to minimize the contribution that position on the gel might have on outcomes, samples were randomized on gels; all original gels are included in Supplemental files.

Cross sections from the middle region of the left ventricle were taken immediately upon removal of the heart, and were fixed in formalin for 24 h (followed by storage in 70% ethanol at 4°C prior to paraffin embedding and sectioning). Assessment of left ventricular interstitial fibrosis employed either Picrosirius Red staining (CBK hearts) or Masson Trichrome staining (dnOGAh hearts), followed by semi-quantitative analysis using Image-J software (NIH), as described previously (21). All original images are included in Supplemental files.

All data are presented as mean ± standard error of the mean (SEM). For non-omics data (i.e., not RNA-seq data), comparisons among groups with two variables were analyzed by two-way analysis of variance (ANOVA) through use of Prism statistical software to investigate main effects of genotype and time. Normality of data was assessed through use of the Shapiro–Wilks test, followed by either parametric (t-tests for only two experimental groups or Sidak's post-hoc test for multiple pairwise comparisons) or non-parametric (Mann–Whitney for only two experimental groups or Kruskal–Wallis with Dunn's correction for multiple pairwise comparisons). Cosinor analyses were performed to determine whether 24 h time series data significantly fit a cosine curve; if they did, then mesor (daily average value), amplitude (peak-to-mesor difference), and acrophase (timing of the peak) were calculated and compared between experimental groups, as described previously (22). For RNA-seq data, JTK cycle was performed using the online tool NiteCap. For the JTK cycle analysis, transcripts were considered to have a 24 h oscillation if they significantly fit a cosine curve (q < 0.05); amplitude and acrophase were calculated within NiteCap, which also determined significant differences between genotypes for these cosine parameters. NiteCap was further utilized for heat map generation and principal component analysis (PCA). In addition, the online tool PANTHER was utilized for pathway enrichment analysis. In all analyses, the null hypothesis of no model effects was rejected at p < 0.05.

The primary outcomes of this study included gene expression, protein expression, and fibrosis; one or more of these outcomes are influenced by time-of-day, sex, age, diet, and genetic background. Consistent with the stated purpose of the study, two main variables were investigated: time-of-day and genotype (i.e., CBK, dnOGAh, and littermate controls). This resulted in a total of 28 experimental groups: 16 experimental groups for the CBK model and 12 experimental groups for the dnOGAh model. To the extent that was possible based on the study design, other biological variables remained constant. Male 12 week old mice on the C57BL/6 background were chosen, allowing direct comparison of RNA-seq data with previously published transcriptomic data obtained from CBK and littermate control hearts (14). Our recent studies investigating male and female dnOGAh mice at this age revealed comparable phenotypes (7). One variable that was not identical between CBK and dnOGAh models was diet; dnOGAh mice were fed a doxycycline containing diet (to induce the transgene), whereas CBK mice were fed a normal rodent chow.

To identify roles of the cardiomyocyte circadian clock, we previously generated a cardiomyocyte-specific BMAL1 knockout (CBK) mouse model (14). To confirm that the cardiac clock is disrupted in CBK mice, hearts were collected from CBK and littermate control mice at 3 h intervals over a 24 h period, followed by assessment of circadian clock components at mRNA and protein levels. At the transcript level, 10 core circadian clock components were investigated; bmal1, clock, npas2, reverbα, reverbβ, per1, per2, per3, cry1, and cry2 mRNA. A 2-way ANOVA revealed significant time main effects for 9 out of 10 core clock component mRNAs (Figure 1A; Supplementary Table 1); only cry2 mRNA did not exhibit a significant time main effect (p = 0.187). Significant genotype main effects were also observed for 9 out of 10 core clock component mRNAs (Figure 1A; Supplementary Table 1); per2 mRNA did not exhibit a significant genotype main effect (p = 0.183). Of those transcripts that exhibited significant genotype main effects, bmal1, reverbα, reverbβ, per1, and per3 were repressed, while clock, npas2, cry1, and cry2 were induced (Figure 1A; Supplementary Table 1). Cosinor analysis indicated that all 10 core clock component mRNAs exhibit 24hr oscillations in CON hearts; these oscillations were either abolished or significantly attenuated in CBK hearts for 8 out of 10 core clock component mRNAs (cry2 and per2 mRNA 24 h oscillations persisted in CBK hearts; Supplementary Table 2). At the protein level, 3 core circadian clock components were investigated; BMAL1, REVERBα, and PER2. A 2-way ANOVA revealed significant time and genotype main effects for all 3 core clock proteins; both BMAL1 and REVERBα protein levels were decreased in CBK hearts, while PER2 levels were increased (Figure 1B; Supplementary Table 1). Cosinor analysis indicated that all 3 core clock component proteins exhibit 24 h oscillations in CON hearts; REVERBα oscillations were significantly attenuated in CBK hearts (Supplementary Table 2). Consistent with the cardiomyocyte-specific nature of the murine model, CBK mice exhibited 24 h oscillations in energy expenditure, food intake, and physical activity that were comparable to littermate control mice (Supplementary Figure 1A, as well as Supplementary Tables 1, 2). Similarly, we have previously reported that distinct circadian clock components are not altered in livers of CBK mice (14). Next, we investigated protein O-GlcNAcylation levels in CON and CBK hearts isolated during the light phase (ZT9); protein O-GlcNAcylation levels were significantly increased in CBK hearts (2.26-fold; Figure 1C). Collectively, these data are consistent with the concept that cardiomyocyte BMAL1 deletion disrupts the clock in the heart, and this is associated with augmentation of protein O-GlcNAcylation.

We, and others, have reported that several circadian clock components are O-GlcNAc modified, impacting their cellular localization, DNA binding, and/or protein stability (8, 9, 12, 13). Moreover, disruption of the cardiomyocyte circadian clock impacts protein O-GlcNAcylation levels (Figure 1C), consistent with an interrelationship between the clock and O-GlcNAcylation (13). We therefore hypothesized that cardiomyocyte dnOGA induction would impact the cardiac circadian clock and protein O-GlcNAcylation in a manner similar to that observed in CBK hearts. To test this hypothesis, hearts were isolated from dnOGAh and littermate CON mice at 4 h intervals over the 24 h day, 2 weeks after doxycycline treatment initiation; this duration of transgene induction is not associated with cardiac dysfunction (7). Following RNA isolation, RT-PCR was performed for the same 10 core circadian clock components investigated in CBK hearts. A 2-way ANOVA revealed significant time main effects for all 10 core clock component mRNAs (Figure 2A; Supplementary Table 3). Significant genotype main effects were observed for 7 out of 10 core clock component mRNAs (Figure 2A; Supplementary Table 3); clock, reverbα, and per1 mRNA did not exhibit significant genotype main effects (p-values of 0.054, 0.821, and 0.217, respectively). Of those transcripts that exhibited significant genotype main effects, reverbβ, per3, cry1, and cry2 were repressed, while bmal1, npas2, and per2 were induced (Figure 2A; Supplementary Table 3). Cosinor analysis indicated that all 10 core clock component mRNAs exhibit 24 h oscillations in CON hearts; these oscillations were significantly altered in dnOGAh hearts for only npas2 mRNA (Supplementary Table 4). At the protein level, 3 core circadian clock components were investigated; BMAL1, REVERBα, and PER2. A 2-way ANOVA revealed significant time main effects for all 3 core clock proteins, and significant genotype main effects for BMAL1 and PER2 (Figure 2B; Supplementary Table 3); REVERBα did not exhibit a significant genotype main effect (p = 0.916). Both BMAL1 and PER2 protein levels were increased in dnOGAh hearts (Figure 2B; Supplementary Table 3). Cosinor analysis indicated that all 3 core clock component proteins exhibit 24 h oscillations in CON hearts; PER2 oscillations were significantly augmented in dnOGAh hearts (Supplementary Table 4). Consistent with the cardiomyocyte-specific nature of the murine model, dnOGA mice exhibited 24 h oscillations in energy expenditure, food intake, and physical activity that were comparable to littermate control mice (Supplementary Figure 1B, as well as Supplementary Tables 3, 4). Similarly, 24 h oscillations in bmal1, reverbα and dbp mRNA levels were not different between CON and dnOGAh livers (Supplementary Figure 2). Next, we investigated protein O-GlcNAcylation levels in CON and dnOGA hearts isolated during the light phase (ZT8); protein O-GlcNAcylation levels were significantly increased in dnOGAh hearts (Figure 2C). Collectively, these data are consistent with the concept that increasing protein O-GlcNAcylation in the heart impacts the circadian clock, in a manner that is distinct from CBK mice.

Core circadian clock components directly regulate the expression of numerous genes that do not feedback on the clock mechanism, but instead impact cellular functions (10). These clock output genes include the PAS transcription factors, DBP and TEF, which are direct BMAL1/CLOCK targets; DNA binding activity of DBP and TEF is directly antagonized by the REV-ERBα/β target E4BP4 (23, 24). As such, dbp, tef, and e4bp4 mRNA levels were next assessed in both CBK and dnOGAh mouse models. A 2-way ANOVA revealed significant time and genotype main effects for all 3 clock output mRNAs in both CBK and dnOGAh hearts (Figures 3A,B, as well as Supplementary Tables 1, 3). Transcripts for dbp and tef were significantly decreased in CBK and dnOGAh hearts, while e4bp4 mRNA was significantly increased in hearts of both mouse models (Figures 3A,B, as well as Supplementary Tables 1, 3). Cosine analysis revealed that 24 h oscillations of these 3 output mRNAs were either abolished or significantly attenuated in CBK hearts (Supplementary Table 2). Similarly, 24 h oscillations were significantly attenuated for dbp mRNA in dnOGAh hearts (Supplementary Table 4). E4BP4 protein levels were next investigated in both mouse models. A 2-way ANOVA revealed significant time and genotype main effects for E4BP4 protein in both CBK and dnOGAh; E4BP4 protein was increased in both models (Figure 3C, as well as Supplementary Tables 1, 3). Although 24 h oscillations in E4BP4 protein were abolished in CBK hearts, these oscillations were not significantly altered in dnOGAh hearts (Supplementary Tables 2–4).

Given similarities between CBK and dnOGAh hearts, in terms of impact on established clock output genes (i.e., dbp, tef, e4bp4), we next investigated transcripts involved in metabolism (dgat2, nampt) and signaling (pik3r1, rhobtb1) that we have previously reported to be regulated by the cardiomyocyte circadian clock (14, 25). A 2-way ANOVA revealed significant time main effects for dgat2, pik3r1, and rhobtb1 (Figures 4A,B, as well as Supplementary Tables 1, 3). In addition, genotype main effects were found for all 4 clock-controlled mRNAs, in both CBK and dnOGAh hearts; these 4 transcripts were decreased in both models (Figures 4A,B, as well as Supplementary Tables 1, 3. Cosine analysis revealed significant 24 h oscillations of these 4 transcripts in CON hearts (except for CBK littermate CON hearts, for which 24 h oscillations did not achieve statistical significance; Supplementary Tables 2–4). 24 h oscillations in these transcripts were abolished in CBK and dnOGA hearts (except for pik3r1 in CBK hearts and rhobtb1 in dnOGAh hearts; Supplementary Table 2–4). Collectively, these data are consistent with attenuated circadian clock output in the heart following increased protein O-GlcNAcylation.

Previous transcriptomic analyses indicate that the cardiomyocyte circadian clock orchestrates diurnal variations of approximately 5%–10% of the cardiac transcriptome (14, 26). Given that our candidate gene approach suggested that circadian clock output is attenuated in dnOGAh hearts (Figures 3B,4B), we next evaluated the diurnal transcriptome at a global level. More specifically, RNAseq was performed on RNA isolated from dnOGAh and littermate CON hearts collected at 4 h intervals over the 24 h day (2 weeks after doxycycline treatment initiation). JTK cycle identified 528 transcripts that significantly oscillate with a periodicity of 24 h in CON hearts; the number of oscillating transcripts was only 45 in dnOGA hearts (Supplementary Tables 5, 6, respectively). Further analysis revealed that 496 transcripts oscillate only in CON hearts, 13 transcripts oscillate only in dnOGA hearts, while 32 transcripts oscillate in both CON and dnOGA hearts (Supplementary Tables 7–9, respectively). Of the transcripts oscillating in both CON and dnOGA hearts, 5 had an amplitude that was significantly reduced in dnOGA hearts (Supplementary Table 9). As such, a total of 501 oscillating transcripts in CON hearts were either abolished or significantly attenuated in dnOGA hearts, representing 95% of all oscillating transcripts in CON hearts (Figure 5A). The impact of dnOGA induction on the diurnal transcriptome was further illustrated in a PCA plot; in addition to marked differential expression at the level of genotype, dnOGA hearts do not exhibit the same time-of-day-dependent pattern observed for CON hearts (Figure 5B). Comparing our previously published microarray analysis of CBK hearts (14) with the newly generated RNAseq analysis of dnOGAh hearts, we identified 125 transcripts exhibiting attenuated/abolished 24 h oscillations in both CBK and dnOGA hearts (Figure 5C; Supplementary Table 10). Pathway analysis of these 125 transcripts indicated notable enrichment in collagen and extracellular matrix (Figure 5D). Collectively, these data indicate that the diurnal transcriptome is markedly impaired in dnOGAh hearts.

As indicated in Figure 5D, multiple transcripts involved in the extracellular matrix exhibited attenuated/abolished 24hr oscillations in both CBK and dnOGAh hearts; these transcripts included several collagen isoforms (Supplementary Table 10). RT-PCR was next performed to validate 24 h patterns of col1a1, col3a1, and col4a1 mRNA in CBK, dnOGAh, and littermate control hearts. A 2-way ANOVA revealed significant time main effects for col1a1, col3a1, and col4a1 mRNA (Figure 6A). In addition, genotype main effects were observed for all 3 collagen isoforms, in both CBK and dnOGAh hearts; these 3 transcripts were increased in both models (Figures 6A,B). Importantly, consistent with increased collagen isoform expression, interstitial fibrosis was augmented in both CBK and dnOGAh hearts (Figure 6C). These observations suggest that increased protein O-GlcNAcylation and attenuated circadian governance are associated with cardiac fibrosis.