Cardiomyocyte Pdk4 response is associated with metabolic maladaptation in aging

Abstract Ischemic heart disease (IHD) is the leading cause of death, with age range being the primary factor for development. The mechanisms by which aging increases vulnerability to ischemic insult are not well understood. We aim to use single‐cell RNA sequencing to discover transcriptional differences in various cell types between aged and young mice, which may contribute to aged‐related vulnerability to ischemic insult. Utilizing 10× Genomics Single‐Cell RNA sequencing, we were able to complete bioinformatic analysis to identity novel differential gene expression. During the analysis of our collected samples, we detected Pyruvate Dehydrogenase Kinase 4 (Pdk4) expression to be remarkably differentially expressed. Particularly in cardiomyocyte cell populations, Pdk4 was found to be significantly upregulated in the young mouse population compared to the aged mice under ischemic/reperfusion conditions. Pdk4 is responsible for inhibiting the enzyme pyruvate dehydrogenase, resulting in the regulation of glucose metabolism. Due to decreased Pdk4 expression in aged cardiomyocytes, there may be an increased reliance on glucose oxidization for energy. Through biochemical metabolomics analysis, it was observed that there is a greater abundance of pyruvate in young hearts in contrast to their aged counterparts, indicating less glycolytic activity. We believe that Pdk4 response provides valuable insight towards mechanisms that allow for the young heart to handle ischemic insult stress more effectively than the aged heart.


| INTRODUC TI ON
Cardiovascular disease (CVD) remains as the worldwide leading cause of death, particularly, ischemic heart disease (IHD) or coronary artery disease (CAD) is the most prevalent (Tsao et al., 2022).
IHD is best described as a chronic plaque buildup in the coronary artery eventually causing ischemic conditions in the heart, leading to tissue death and metabolic alterations (Jiang et al., 2021). This chronic buildup of plaque leads to greater IHD incidence and death increasing with age (Khan et al., 2020). However, the mechanisms that protect the young heart from ischemic stress are not well understood. Various studies have indicated that the aged heart is not only more vulnerable to IHD but also less equipped to handle the stress of ischemic conditions (Ren et al., 2022).
Single-cell RNA sequencing (scRNA-seq) has become an increasingly popular method of genetic sequencing since its founding (Hwang et al., 2018). The value of scRNA-seq comes in the manner of classifying the sequenced data. In contrast to bulk RNA sequencing, scRNA-seq allows us to observe differences in cell types versus blanket transcriptional differences (Hwang et al., 2018). In our study, we aim to utilize modern sequencing technology to find transcriptional differences between young and aged hearts under healthy physiology and under ischemia/reperfusion (I/R) conditions. These transcriptional differences in the heart may provide insight to differing adaptive responses caused by aging in the form of gene regulation. We will investigate four common cell types found in cardiac tissue: cardiomyocyte, endothelial cells, fibroblast, and macrophage.
We aim to discover novel gene regulation under ischemic conditions in the various cell types that may provide clues to understanding the effects of aging.

| Integration and cell types
Our comparative analysis allowed us to observe transcriptional differences in young to aged C57BL/6J wildtype mice. Four RNA samples were received for analysis that allowed for our integrated bioinformatic analysis (Figure 1). Observing the effects of ischemic insult, we compared young I/R to young sham groups and aged I/R to aged sham groups. Additionally, comparison of aged sham to young sham and aged I/R to young I/R were done to investigate the effects of aging ( Figure 2a). Following the integration vignette from the Satija Laboratory (Tim Stuart et al., 2019), the maximum overlap of each sample was achieved (Figure 2b). Vital to our study, integration ensured when each cluster was annotated with cell-type, it contained cells from both samples that form each dataset.
Determining that manually annotating cell types led to the most accurate and consistent data in our datasets, multiple marker genes were found that allowed for annotations (Figures 1f and 2c).
Integration established largely consistent marker genes for each celltype throughout our datasets. Cardiomyocyte marker genes Myl4 and Myl7 were observed in all four datasets, likewise Nppa/Nppb were profound datasets except for the aged and young sham comparison. The aged and young sham dataset expressed two unique cardiomyocyte marker genes, Trdn and Mylk3, were strongly expressed and used for annotation for this comparison (Figure 2e). Across all groups, the most consistent marker gene for endothelial cells was Cdh5; notably, each dataset expressed unique endothelial marker genes. Exclusive to the young I/R and sham comparison, Rgcc was significantly expressed and used for endothelial cell annotation ( Figure 2c). Furthermore, Nrp1 was found significantly expressed in the young I/R and sham comparison along with the aged and young sham comparison and used as a marker gene (Figure 2c,e).
Cd94 and Cd34 were used as endothelial cell markers in the aged I/R and sham, aged and young sham, and aged and young I/R comparisons (Figure 2d Annotating fibroblast cells also revealed generally consistent marker genes with some exceptions. Pdgfra was remarkably expressed in all datasets and widely regarded as a strong fibroblast marker gene. Collagen-type genes: Col1a1, Col1a2, Col3a1, and Col6a2 were also found throughout the datasets and were used to annotate fibroblast cells (Figure 2c-f). In the young I/R and sham comparison, Prg4 was uniquely identified as significantly expressed as used as a fibroblast marker gene ( Figure 2c).
Finally, annotating macrophage cell populations required further analysis. The marker genes identified throughout all datasets were Cd74 and Ctss/Ctsb, while S100a8 and Mrc1 were found in most datasets and used to annotate macrophage cells (Figure 2c

| Differential expression and enrichment
Integration established stable cell types between samples allowing for analyzing differentially expressed genes (DEGs) in specific cell types between two samples-for example, comparing gene regulation in fibroblast cells between I/R and sham conditions. In differential expression testing, cardiomyocyte cell populations were targeted to compare gene (feature) regulation between conditions ( Figure 3a). The data from this test was used to generate our Gene Ontology (GO) biological process enrichment data.
Our enrichment data are split into our I/R and sham data ( Figure 3b) and our aged and young data ( Figure 3c). Under I/R stress, many biological cardiac processes are downregulated regardless of age including heart and muscle contraction and heart development. Notably, upregulation of fatty acid oxidation (FAO) processes and downregulation of glycolytic processes under I/R stress were observed in young mice ( Figure 3b). Interestingly, while there is downregulation of ventricular cardiac tissue morphogenesis and upregulation of striated muscle hypertrophy in aged mice, the opposite was observed in the young mice.
Aged mice also exhibited biological processes, such as wound healing, that were not seen in the young mice enrichment (Figure 3b).
In our aged and young dataset (Figure 3c), we recognized that many similar processes that were downregulated in the previous dataset ( Figure 3b) were also downregulated in this comparison. The enrichment data retrieved from our differential expression testing revealed that not only were FAO processes downregulated under ischemic stress conditions compared to sham conditions (Figure 3b) but were also downregulated in aged mice compared to young mice under ischemic stress ( Figure 3c). This encouraged us to attempt to reveal individual DEGs that may aide in these metabolic changes associated with our differential expression data.

| Identification of pyruvate dehydrogenase kinase 4
During our study, we identified Pyruvate Dehydrogenase Kinase 4 (Pdk4) as a gene that had significant changes in regulation during our differential expression testing (Figure 3a). Pdk4 is predominantly expressed in cardiomyocytes (Figure 4a,b). While this is expected, as Pdk4 is well regarded as a strong cardiomyocyte marker gene, the differing expression levels are quite notable. In contrast to sham conditions, Pdk4 is expressed remarkably after inducing ischemic stress in both aged and young mouse cardiomyocytes (Figure 4c). The relative expression of Pdk4 in cardiomyocytes is over three times greater in young mice than in aged mice when comparing them to their healthy F I G U R E 1 Graphical synopsis of experimental design. We isolated left ventricle tissue from young (3-5 months) and aged (24-26 months) C57BL/6J mice under sham operations or 45 min of ischemia/24-h of reperfusion (I/R) conditions. Once single-cell suspensions were achieved, cells were prepared for sequencing using the 10× Chromium system.

| Alterations in Pdk4 protein levels
Taking advantage of immunoblotting, we measured Pdk4 protein content. We found that protein levels of Pdk4 matched trends observed in transcription. Pdk4 protein level was significantly in-

| Metabolomic analysis
The metabolomics data allowed for testing of key glucose metabolites and their comparative abundance. Pdk4 is known to inhibit pyruvate dehydrogenase (PDH) when expressed. We compared the abundance of pyruvate as well as its downstream and upstream metabolites in I/R versus Sham and Aged versus Young conditions ( Figure 6b,c).
Following the results obtained from differential expression testing (Figure 4d), we found a greater presence of pyruvate under I/R stress compared to sham conditions in both aged and young mice ( Figure 6b). This indicates that with the increased expression of Pdk4, pyruvate accumulates in the cytoplasm as it is no longer being metabolized. We observed that most glycolytic upstream metabolites before phosphoenolpyruvate are more abundant in sham conditions in both aged and young mice except for glyceraldehyde-3-phosphate which was present in greater quantities under I/R conditions in aged mice ( Figure 6b).
We discovered that aged mice contain more pyruvate in their myocytes than young mice under sham conditions ( Figure 6c). We could expect this result as transcriptional Pdk4 expression is slightly upregulated leading to inhibition of PDH causing accumulation of pyruvate. It is improbable that the marginal upregulation of Pdk4 is the primary cause for the results observed here and further research is needed. We detected that pyruvate is greater in young mice than aged mice under I/R conditions ( Figure 6c). Based on our transcriptional data indicating major Pdk4 expression in young mice ( Figure 5d), we expected greater inhibition of PDH in young mice than in aged mice under ischemic conditions. These results strongly propose that Pdk4 is regulating glucose metabolism in response to stress such as ischemia conditions but is also impacted during the aging process, advancing research in metabolic alterations due to aging.

| Glucose oxidation dependency
Finally, a mitochondrial fuel dependency test was performed to characterize the role of Pdk4 in glucose oxidation of cardiomyocytes under physiological and pathological conditions. The Seahorse XF analyzer paired with a glucose oxidation dependency test was used on isolated cardiomyocytes from both aged and young hearts under sham and I/R conditions. Furthermore, we tested all groups with and without Pdk4 inhibition. The results show a decrease in glucose oxidation dependency after ischemic insult in both aged and young cardiomyocytes (Figure 6d,e).
There were not significant alterations in young cardiomyocytes under sham or I/R conditions (Figure 6e), while a significant difference occurred in aged cardiomyocytes under I/R versus sham ( Figure 6e). In addition, we found a significant increase of glucose oxidation dependency in young cardiomyocyte from I/R versus sham with Pdk4 inhibitor (Figure 6e), indicating a critical role of F I G U R E 2 (a) UMAP dimensional projection of each integrated dataset, clusters colored by cell-type manually identified by marker genes. (b) Matching UMAP dimensional projections, clusters colored by sample age. (c-f) Corresponding dot plots for each integrated sample depicting key marker genes used for manual cell-type identification. Displays cluster gene expression level compared to global expression in assay (color) and gene expression level within each specific cell-type cluster (size). (g) Proportion of each cell-type observed in each integrated dataset. Pdk4 inhibitor treatment increases glucose oxidation dependency in young cardiomyocytes but not in aged cardiomyocytes from I/R groups ( Figure 6e). Correspondingly, this implies that Pdk4 activation in response to I/R stress as a metabolic adaptive response is impaired in aged hearts as compared to young hearts that could lead to cardiac vulnerability in aging to ischemic insults caused by ischemia and reperfusion.

| DISCUSS ION
Understanding the effects of aging remains to be an important objective in clinical medicine, especially pertaining to IHD. The mechanisms by which cardiac cells adjust and communicate during injury are still unknown. To uncover underlying transcriptional differences in aged and young mouse cells, we employed scRNA-seq.

Identification of novel DEGs provide insight towards inadequately
understood shifts in adaptive responses to ischemic stress between the aged and young heart. Targeting the genes responsible for metabolic deficiency in the aged heart may facilitate therapies and potential treatments for the injured heart.
To best investigate the pathophysiology of IHD, a murine model for myocardial infarction was utilized. The 45-min ischemia, 24-h reperfusion period prior to sequencing causes physiological symptoms analogous to myocardial infarction and is an accepted method for simulating short-term I/R injury (Xu et al., 2014). Our previous research using this experimental method has shown that infarct area significantly increases in aged I/R mice compared to young I/R mice (Wang et al., 2018). Aged murine hearts after I/R conditions exhibit critically lower ejection fraction and fractional shortening compared to their young counterparts (Wang et al., 2018). The investigation of resulting adverse downstream effects was motive for this study.
Single-cell RNA sequencing was employed to accomplish the study of transcriptional differences in specific cell types. Utilizing the integration vignette in Seurat, we completed a comparative analysis between I/R and sham conditions as well as between aged and young mice. After identifying cell types, we enriched our differential expression data to provide insight into adverse biological processes.
We found FAO-related processes to be upregulated in young mice and downregulated in aged mice under ischemic stress.
Following the enrichment data, we sought to discover individual significant DEGs. This process uncovered notable changes in the expression of Pdk4 in cardiomyocytes which critically regulates pyruvate dehydrogenase (PDH) (Chambers et al., 2011;Shuai Zhang et al., 2014). Metabolic remodeling is a well-known consequence increased stress such as ischemia in the heart (Khan et al., 2020;Wu et al., 2017). In contrast, the young heart retains a stronger ability to recover from ischemic insult (Zhang et al., 2021). Based on our data, we propose that the aged heart under ischemic stress relies on inferior glucose oxidation as the heart's main energy source versus other energy-rich metabolic pathways.
Immunoblotting studies to observe the protein content of Pdk4 were performed to corroborate the RNA-seq findings. The results from immunoblotting support the increased expression of Pdk4 after ischemic insult in young mice. While we observed an upregulation of Pdk4 expression in aged mice, that result was not found to be significant. Notably, there was a significant increase of Pdk4 expression between young and aged mice following an ischemic insult. There was no significant difference in protein level between young and aged sham mice, the transcriptional Pdk4 upregulation in aged sham mice was modest, and thus, a significant change in the protein content was not expected. Though there was an average increased protein level in aged I/R compared to aged sham mice, this finding was also not considered significant. We consider due to the diminished transcriptional expression of Pdk4 in aged I/R mice, the resulting protein level expression would not be great enough to reach significance when compared to sham conditions. Encountering this change in an essential metabolite regulator on both transcriptional and protein level, we decided to use biochemical metabolomic data to test the abundance of key glycolytic me-

b) Enriched terms from Aged and Young I/R versus Sham cardiomyocyte integrated datasets. Testing variable is I/R condition (i.e., heart contraction under I/R conditions is downregulated in aged and young samples compared to sham operations). (c)
Enriched terms from I/R and Sham Aged versus Young cardiomyocyte integrated datasets. Testing variable is Aged condition (i.e., cardiac muscle tissue morphogenesis in the aged sample is downregulated under I/R conditions and upregulated under Sham operations compared to young sample). All enriched terms received an adjusted p-value of >0.05 from EnrichR and plotted with −log 10 for visualization.  The expression pattern in our single-cell data indicates that

F I G U R E 4
Pdk4 has greater expression in aged mice under sham conditions. Furthermore, we found that pyruvate is more abundance in aged mice than young mice under sham conditions. While this supports the scRNA-seq data and that Pdk4 is a major glycolytic regulator in the heart , it is unlikely that the slight transcriptional upregulation of Pdk4 under sham conditions lead to the results observed by metabolomics. We believe other signaling pathways may be involved as aging leads to impaired mitochondrial function that decreases oxidative phosphorylation and increases the production of reactive oxygen species (Chen et al., 2020;Lesnefsky et al., 2016).
It is intriguing to speculate that the increased transcriptional expression of Pdk4 in the aged heart in the baseline state may represent an adaptive response to the age-induced mitochondrial dysfunction.
Notably, recent attempts to downregulate Pdk4 to mitigate the effects of cardiac aging have been successful in baseline conditions signifying the link between Pdk4 and aging in normal physiology (Zhang et al., 2022). Under I/R stress, there is a larger abundance of pyruvate in the young mice than in the aged mice, consistent with transcriptional expression of Pdk4 being higher in young mice.
It is understood that increased glucose oxidation can be beneficial during ischemic conditions (Lopaschuk & Stanley, 1997;Tran & Wang, 2019); a chronic deficiency of other metabolic pathways is generally detrimental (Sithara & Drosatos, 2021). The expression pattern of Pdk4 observed suggests that the aged heart no longer strongly expresses the gene after ischemic insult. Seemingly, the aged heart suffers from the loss of flexibility to utilize other sources of energy compared to the young heart. Our enrichment data indicate that the aged heart has a decreased capacity for FAO under stress while the young heart upregulates FAO, consistent with findings of Pdk4 expression recognized as marker for increased FAO (Pettersen et al., 2019). We propose that Pdk4 is linked to this deficiency in metabolic programming that occurs with aging. Our results reveal that Pdk4 expression is linked to energy sources that alter with age in the heart after ischemic insult. Not only does vulnerability to IHD increase with age but the inability to adapt to the stress of ischemia is an additional consequence of aging (Zhang et al., 2021). Deficiency of Pdk4 in the aged heart has been observed in the past (Hyyti et al., 2010); its role following I/R indicates metabolic alterations that may contribute to increased injury compared to the young heart.
In the aging heart, mitochondria suffer additional damage with ischemia that is superimposed upon age-induced defects, enhancing myocardial injury during the early reperfusion period (Lesnefsky et al., 2016(Lesnefsky et al., , 2017Mohsin et al., 2019). It is possible that the baseline increase in Pdk4 expression in the adult heart represents a compensatory attempt to regulate the metabolism of these damaged mitochondria that are injurious to the heart. In response to I/R stress, Pdk4 is less robustly expressed in the aged heart. Suggesting that during reperfusion, metabolism in mitochondria that contain both aging and ischemic defects is restrained, potentially representing a mechanism that is a result of greater injury that occurs in the aged heart following I/R. Increased FAO results in higher stress on mitochondria (Menendez-Montes et al., 2021), these damaged cardiomyocytes are likely unable to handle the increased stress like their young counterparts. Further investigation regarding the young heart inhibiting PDH after ischemic conditions is required to better understand our results. We plan to further investigate Pdk4 and the relationship it has with age-related stress and possible therapeutic strategies impacting the metabolic remodeling in the aged heart.
Although we identified Pdk4 as a significantly differentially ex- expression of Pdk4 will provide considerable insight towards the heart's recovery from ischemic insult. Additionally, observing expression over time would permit us to discover other genes that follow the Pdk4 expression pattern revealed in our study.
In our study, we performed metabolomic analysis, immunoblotting, and in vitro metabolism testing. In future studies, it would be beneficial to include transgenic mice that have Pdk4 knocked out in cardiomyocytes and mice with Pdk4 overexpressed. While global knockout Pdk4 transgenic mice have been observed to increase glucose oxidation , we believe investigating cardiomyocyte-specific knockout in vivo would be the most beneficial. Cardiomyocyte-specific Pdk4 knockout mice have been generally found to have favorable outcomes under normal physiology (Cardoso et al., 2020), additional investigation of these mice under I/R conditions is essential to future studies. Furthermore, we plan to obtain Pdk4 overexpressed transgenic mice to observe effects on metabolism and heart function after ischemic insult as previous research indicates overexpression leads to increased fatty acid catabolism and without detriment (Chambers et al., 2011;Zhao et al., 2008).
While this study used a mouse model, it would be beneficial to test the expression of Pdk4 in the human heart to provide important insight into metabolic alterations present in humans with aging.
Additionally, though cardiomyocytes are vital cells that are impacted by ischemic stress, we aim to target more differentially expressed genes in other cell types in the future. In our scRNA-seq data, we identified cardiomyocytes, endothelial cells, fibroblasts, and macrophages. Observing transcriptional differences in these other cell types has the potential to further uncover increased adaptive responses in the young heart to ischemic stress.

| Bioinformatic analysis: Clustering and celltype identification
In all our integrated datasets, we scaled the data and performed a PCA reduction with 10 principal components prior to UMAP dimensional reduction using all 10 principal components from PCA reduction. The FindClusters() function was used to identify clusters and the resolution was adjusted to generate ~50 clusters. Utilizing the FindAllMarkers() function in Seurat, we identified marker genes in each cluster to annotate cell types. We subset all clusters with remarkable cell-type specific marker genes (Log2FC-value < 2, pvalue > 0.05) to allow for cell-type annotation into four essential cell types: cardiomyocyte, endothelial, fibroblast, and macrophage. To verify the accuracy of our marker genes, we employed the Human Protein Atlas (Uhlén et al., 2015) as a genome directory to crossanalyze our markers to associated cell types in the atlas.

| Bioinformatic analysis: Differential expression and enrichment
Once all four comparative datasets had identified cell types, we followed the Seurat vignette for differential expression testing.
In our testing parameters, the two samples that form each dataset were compared with each cell-type isolated. Particularly, the RNA data assay is used to generate the data in comparison with using the integrated assay. The RNA assay in contrast to the integrated assay retains the original expression profiles of cells prior to "correction" that occurs during the integration protocol.
The original expression profiles of the cells were set as the identities in the FindMarkers() function with cell-type set as the subset identity. We omitted all genes that produced an adjusted p-value greater than 0.05 from all further downstream analysis. These data were used to generate pathway enrichment using the EnrichR R-package (Chen et al., 2013;Kuleshov et al., 2016) and Gene Ontology (GO) term biological processes (Ashburner et al., 2000;Gene Ontology, 2021). The statistical expression values (Log2FC) of Pdk4 were taken from the differential expression testing, and related plots were created using the ggplot2 R-package to visualize the expression data (Wickham, 2016).

| Metabolomic data analysis
Data analysis was completed using XCMS software for spectrum deconvolution and MetSign software to identify metabolites, peak alignment, normalization, and statistical analysis (Wei et al., 2011(Wei et al., , 2012(Wei et al., , 2014. We matched our data to our in-house generated database of ion m/z, MS/MS spectra, and metabolite retention time to identify most metabolites. Unmatched data were analyzed in Themo-Fisher Compound Discoverer Software 2.0 where metabolites with at least 40% similarity scores were selected. We targeted metabolites associated with glycolysis and plotted the abundance change (foldchange) data using the ggplot2 R-package (Wickham, 2016). The metabolomics datasets used in this study are available at the National Metabolomics Data Repository under accession number: 2842.

| Cardiomyocyte isolation
We tested isolated cardiomyocytes to best understand metabolism modulation in our testing groups. Heparin IV (Fresenius Kabi AG) was administered by intraperitoneal injection to prevent coagulation during perfusion. Once anesthetized, the mice heart was excised and cannulated by the aorta to be connected to the perfusion apparatus (Radnoti LCC). The heart was perfused at 37°C with perfusion buffer followed by digestion buffer. To isolate cardiomyocytes, the heart was removed from the apparatus and gently minced, the remaining solution was suspended and filtered through 100 μm filter to obtain isolated cardiomyocytes.

| Seahorse mitochondria testing
The Seahorse Bioscience XF24 Extracellular Flux Analyzer Kit  (Satija et al., 2015). We filtered out all genes containing an adjusted p-value greater than 0.05. Likewise, after using EnrichR, it produces a dataset containing all associated GO biological processes and corresponding adjusted p-values. EnrichR produces Benjamini-Hochberg adjusted p-values (Chen et al., 2013;Kuleshov et al., 2016). We filtered out all GO terms with an adjusted p-value of 0.05 and selected terms representing metabolic and development alterations in the heart.
In our metabolomics data analysis, the Grubbs' test was used to identify outlier data proceeded with a pairwise two-tail t-test. To determine significance, metabolites that were present in at least 75% of the pooled samples in each group as well as acquired a p-value of <0.05 from the two-tailed t-test were selected. The dependency data were validated by performing a twoway ANOVA statistical test with Šídák correction in Prism 9.4.1 (GraphPad Software). Statistical significance was determined by p-value < 0.05. All figures are shown with means ± standard error