Adult mice (C57Bl6 background) were bred and aged to 16 weeks of maturity for experiments and housed at 23°C–24°C with 45% humidity and maintained on a 12 h light/dark cycle with food and water provided ad libitum. This study was approved by the Animal Care Committee at the University of Guelph and all experiments were carried out in accordance with the guidelines from the Canadian Council on Animal Care.

Inducible CreLox transgenic mice expressing Epo LoxP (Zeigler et al., 2010) and Cre recombinase under the control of the cardiomyocyte-specific promoter, alpha-myosin heavy chain (αMHC-MerCreMer), were used in this study (Sohal et al., 2001) (Jackson Laboratory Strain # 005657, Supplementary Figure S1). Briefly, the 5′loxP site was inserted into intron 1 of the Epo gene (located 94 base pairs upstream from exon), and the 3’ loxP site was inserted into intron 4 (located 86 base pairs downstream of the exon 4), with the NEO cassette flanked by both loxP sites.

Experimental mice were bred using the following schemes: female EPO LoxP without Cre (EPO^fl/fl: αMHC-MerCreMer^−/−) crossed with male EPO LoxP mice expressing homozygous Cre Recombinase (EPO^fl/fl: αMHC-MerCreMer^+/+), which resulted in 100% heterozygosity; female EPO^fl/fl: αMHC-MerCreMer^+/+ crossed with male EPO^fl/fl: αMHC-MerCreMer^+/+ (100% homozygous), female EPO^fl/fl: αMHC-MerCreMer^+/− crossed with male EPO^fl/fl: αMHC-MerCreMer^+/− (expected Mendelian genetics were 25% homozygous, 50% heterozygous, 25% wildtype) and female EPO^fl/fl: αMHC-MerCreMer^+/− crossed with male EPO^fl/fl: αMHC-MerCreMer^+/+ (expected Mendelian genetics were 50% homozygous, 50% heterozygous). Genotyping used tail biopsies and PCR (REDExtract-N-Amp Tissue PCR Kit; Sigma-Aldrich, Oakville, ON, Canada) according to the manufacturer’s instructions. To confirm the presence of the floxed EPO alleles, the following primers sets were used: EPO-KO-F: 5′-AGT​GAA​GTT​TGG​CCG​AGA​AG-3’ (PCR Reaction A), EPO-KO-R: 5′- AGA​TCG​AAC​TTG​GCT​CCT​CA-3’ (PCR Reaction A), EPO-TAR-R: 5′-GTG​GGA​CGT​TCT​GGA​AGA​AA-3’ (PCR Reaction A). PCR Reaction–Stage 1: 1 Cycle, 95°C for 2 min; Stage 2: 40 Cycles, 95°C for 45 s; Annealing: 59°C for 1 min; Extension: 72°C for 1 min; Stage 3 Additional Extension: 1 Cycle, 72°C for 5 min; Hold at 4°C. Gel Electrophoresis–load 18.5 uL sample/well and run gel for 45 min at 90V (lights OFF). Expected Results–Homozygous EPO flox sites = one band at 344 bp; Heterozygote EPO flox sites = band at 228 bp and 344 bp; Wildtype (i.e., no flox) = one band at 228 bp. Representative agarose gel (1.5%) characterizing EPO floxing is presented in Supplementary Figure S1B.

For expression of Cre recombinase under the αMHC-MerCreMer promoter, the following primer sets and conditions were used: MerCreMer Common (Fwd): 5′-TCT​ATT​GCA​CAC​AGC​AAT​CCA-3’ (PCR Reaction A and B), MerCreMer Reverse: 5′-CCA​GCA​TTG​TGA​GAA​CAA​GG-3’ (PCR Reaction A), Wild Type Reverse: 5′-CCA​ACT​CTT​GTG​AGA​GGA​GCA-3’ (PCR Reaction B). PCR Protocol for Reaction A and B–Stage 1: 1 Cycle, 95°C for 2 min; Stage 2: 40 Cycles, 95°C for 30 s; Annealing: 60°C for 30 s; Extension: 72°C for 1 min; Stage 3 Additional Extension: 1 Cycle, 72°C for 5 min; Hold at 4°C. Gel Electrophoresis–load 16 uL sample/well and run gel for 35 min at 95 V (lights OFF). Expected Results–Transgenic allele (αMerCreMer) = ∼300 bp; Heterozygote allele = 295 bp and ∼300 bp; Wild type allele = 295 bp. Representative agarose gel (1.5%) characterizing αMerCreMer Cre recombinase is presented in Supplementary Figure S1C.

At 8 weeks of age, EPO^fl/fl: αMHC-MerCreMer^+/− mice were injected intraperitoneally (i.p.) with corn oil/tamoxifen mixture (25 mg/kg) once daily for 5 days to activate Cre recombinase and induce cardiomyocyte-specific deletion of Epo (denoted EPO^Δ/Δ). Experiments were performed at 8 weeks post-injection (16 weeks old). Age-matched control mice (EPO^fl/fl: αMHC-MerCreMer^−/−, EPO^fl/fl: αMHC-MerCreMer^+/+ and EPO^fl/fl: αMHC-MerCreMer^+/− without tamoxifen) demonstrated no differences and were subsequently combined (denoted EPO^fl/fl). Cre-null mice provided tamoxifen (EPO^fl/fl: αMHC-MerCreMer^−/− with tamoxifen) and mice without EPO floxed but with Cre (EPO^+/+: αMHC-MerCreMer^+/− with and without tamoxifen) showed no major phenotypical differences compared to wildtype mice (Supplementary Table S1).

EPO^fl/fl and EPO^Δ/Δ mice were subjected to a 3-day voluntary wheel running protocol. Mice were housed individually and allowed to run freely on an in-cage wheel (12 cm in diameter). Rotations were transmitted to a cycling computer (VDO M2.1 WR Cycling Computer) and distance, time, and pace were recorded. Data were presented as the averages from all 3 days of voluntary wheel running following a 24-h acclimation period.

The Comprehensive Laboratory Animal Monitoring System (CLAMS) metabolic caging apparatus (Columbus Instruments Oxymax) is a sealed indirect calorimeter used for the simultaneous measurement of multiple parameters, including oxygen consumption (VO₂), carbon dioxide production (VCO₂), and calculation of respiratory exchange ratio (RER) across 24-h. EPO^fl/fl and EPO^Δ/Δ mice were weighed and individually placed into the CLAMS caging for a 24-h acclimatization period followed by a subsequent 24-h data collection period. Mice were maintained on a 12-h light/dark cycle and provided food and water ad libitum. Data was recorded every 15 min for metabolic readings (VO₂, VCO₂) and total energy expenditure. The RER was calculated as the quotient of VCO₂/VO₂.

For the determination of hematocrit, blood was collected from the left ventricle and saphenous vein of EPO^fl/fl and EPO^Δ/Δ mice and centrifuged in heparinized microcapillary tubes (5000rpm at 23°C for 10 min). To calculate hematocrit, the length of red blood cells was divided by the length of total blood volume and expressed as a percentage. To measure hemoglobin (g/L), saphenous vein blood was collected by microcuvettes and measured by a HemoCue^® Hb 201 (cat# 111716 Life Supply).

Mice were anesthetized using an isoflurane/oxygen mix (2%/100%). Echocardiography was performed using the Vevo2100 system (VisualSonics Inc., Toronto, ON, Canada) with the 40 MHz MS550D ultrasound transducer. Mice were maintained at 37.5°C throughout data collection, as confirmed with a TH-5 rectal probe thermometer (Physiotemp Instruments LLC, Clifton, NJ, United States). B-Mode images were captured from the parasternal long axis mid-papillary region to measure the left ventricle endocardial length (i.e., LV chamber length) from the aortic annulus to the apex during in diastole, as previously described (Vinhas et al., 2013). M-Mode images were collected from the parasternal long axis mid-papillary region and analyzed using the left ventricle trace function from the cardiac package (VisualSonics Inc., Toronto, ON, Canada) as previously described (Platt et al., 2017). Measurements represent the average data collected over three consecutive cardiac cycles. Data collection and analyses were performed in a blinded manner.

Cardiac function was investigated in vivo by invasive hemodynamics. Mice were anaesthetized using an isoflurane/oxygen mix (2%:100%). Animals were maintained at 37.5°C throughout data collection using a heated water pad. A 1.2F catheter (FTS-1211B-0018; Transonic Scisense Inc.) was inserted into the right carotid artery and advanced into the left ventricle to collect hemodynamics measurements. Hemodynamics were collected for 15 min. All hemodynamic signals were digitized at a sampling rate of 2000 Hz and recorded by computer using iWorx^® analytic software (Labscribe2, Dover, NH, USA). Data collection and analyses were performed in a blinded manner. Tissues were collected, weighed, and randomly assigned to either histology, qPCR, or western blotting.

The ex vivo Langendorff preparation provides the direct assessment of systolic and diastolic cardiac function (and its susceptibility to ischemia-reperfusion injury), independent of preload, afterload, and heart rate, without the influence of neurohormonal effects or humoral factors in the blood. Accounting for these variables allows for conclusions to be made about purely intrinsic cardiac function. Mice were heparinised (200 IU/kg of body weight) for ex vivo cardiac assessment. After 20 minutes, mice were anesthetized using isoflurane, followed by a midline thoracic incision made to rapidly excise the heart. The heart was rinsed in ice-cold phosphate buffered saline and the aorta was cannulated onto a 21-gauge needle to allow for retrograde perfusion. Hearts were perfused with carbogenated (95% O2: 5% CO2) Krebs Henseleit buffer (pH of 7.4) at 70–75 mmHg. The Krebs Henseleit buffer contained the following compounds (in mM/L): 118 mM NaCl, 4.7 mM KCl, 1.2 mM MgSO4, 1.2 mM KH2PO4, 0.5 mM C3H3NaO3, 0.05 mM EDTA, 11 mM glucose, and 2 mM CaCl2. The left atrium was removed to allow the insertion of a deflated balloon attached to a pressure catheter into the left ventricle. The balloon was inflated to achieve an end diastolic pressure of 5–8 mmHg. Hearts were paced using a Grass SD9 Stimulator at a frequency of 7 Hz. A stabilization period of 20 min was followed by recording of baseline measurements, followed by 25 min of global no-flow ischemia. Afterwards, the perfusate line was reopened and hearts were re-perfused for 45 min. To assess cytoprotective function ex vivo, percent recovery of left ventricular pressure, and rate of change in pressure (dP/dt_max–index of contractility/inotropy, and dP/dt_min–index of relaxation) were calculated after 45 min of reperfusion. Data collection and analyses were performed in a blinded manner.

Blood was collected from the left ventricle via cardiac puncture and allowed to clot for 2 h on ice, then centrifuged in 1.5 mL Eppendorf tubes at (4,000 rpm at 4°C for 20 min). The supernatant was removed, snap frozen in liquid nitrogen, and stored at −80°C for quantification of serum EPO levels. EPO protein concentration was quantified from serum using a Quantikine Mouse EPO ELISA (MEP00B, R&D Systems) according to the manufacturer’s instructions.

Tissues selected for qPCR were snap frozen in liquid nitrogen and kept at −80°C (n = 8 per group for EPO^fl/fl and EPO^Δ/Δ, n = 7 EPO^fl/- Cre^+/−). Total RNA was isolated from the free left ventricle and kidneys using TRIzol (Invitrogen, Burlington, ON, Canada) with the Qiagen RNeasy kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. RNA samples were treated with DNase (Qiagen), according to manufacturer’s instructions. Prior to cDNA synthesis, RNA concentrations were quantified (NanoDrop, ND1000; Thermo Fisher Scientific, Waltham, Massachusetts, USA). Protein contamination was assessed by measuring absorbance at 280 nm. Generation of cDNA was completed using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems by Thermo Fisher Scientific, Waltham, Massachusetts, USA) according to the manufacturer’s instructions, using 2000 ng of RNA per sample. Target gene RNA was quantified using the Platinum SYBR Green qPCR SuperMix-UDG with ROX (Invitrogen, Burlington, ON, Canada) using the primers listed in Table 1. Primers were designed to span an exon-exon region to eliminate any possibility of priming genomic contamination. qPCR was performed using 7,500 Real Time PCR detection system (Applied Biosystems, Foster City, CA, USA) with the following protocol: 1 cycle at 50°C for 2 min, 1 cycle at 95°C for 5 min, then 40 cycles at 95°C for 15 s, 1 min at 60°C for all genes (excluding Epo, where the annealing temperature was 58°C), followed by a dissociation curve to assess specificity of the reaction (Supplementary Figure S2). Samples were run in duplicate 25 uL reactions. Undetectable Epo RNA was assigned the value of the limit of detection of the assay (CT = 40). Results were analyzed according to the delta-delta CT method using reference genes (Table 1) and normalized to the EPO^fl/fl group.

The hearts were excised and block fixed in 10% buffered formalin for 24 h, followed by transfer into 70% ethanol for storage before processing and embedding in paraffin wax. Tissue microarrays were created using the tissue micro array (TMA) Master II instrument from 3D Histech Ltd. Regions of interest from the left ventricle were selected, cored, and added into a host paraffin block to create an array of 84 tissue cores, each measuring 2.0 mm in diameter (biological replicates n = 3).

To study Epo gene mRNA expression and localization from cardiac cell types, the microarray sections were stained using the following fluorophores: Epo with FITC (channel 2, accession # NM_007942.2) using RNAScope 2.5 LS Probe Mm-Epo-C2 Cat. No. 315508-C2, and with Cy5 (channel 1, accession # NM_007942.2) using RNAScope 2.5 LS Probe Mm-Epo-01 Cat. No. 444948. Then, kinase insert domain receptor (Kdr), an endothelial cell marker, was stained with Cy5 (channel 1, accession # NM_01612.2) using RNAScope 2.5 LS Probe-Mm-Kdr Cat. No. 414818. Myosin heavy chain 6 (Myh6), a cardiomyocyte marker, was stained with TRITC (channel 3, accession # NM_001164171.1) using RNAScope 2.5 LS Probe-Mm-Myh6 Cat. No. 506258. Finally, 4′,6-diamidino-2-phenylindole (Dapi) was used to stain nuclei. In-situ hybridization duplex staining was performed using commercially available assays and the ACD Bio program for the Leica Bond RX immunostainer. The slides were scanned using an Olympus VS 120 automated slide scanner, and images of 1 mm x 1 mm were acquired at ×40 magnification. The acquired images were analyzed using the HALO image analysis software from Indica Labs (v3.2.1851.229).

Mice were exsanguinated and 10 mL of 1x PBS, 10 mL of 0.5 mol/L KCl, and 10 mL of 10% buffered formalin (VWR, Mississauga, ON, Canada) were perfused through the right carotid artery to fix cardiomyocytes in diastole. Tissues were harvested, stored in 10% buffered formalin for 24 h, and then transferred into 70% ethanol. Hearts were processed and embedded in paraffin wax. Cross sections of the heart (5 μm) were mounted onto charged 1.2 mm Superfrost slides (Fisher Scientific). Paraffin embedded sections were then stained with either hematoxylin and eosin or Picrosirius Red for the determination of cardiomyocyte cross-sectional area (CSA) or percent fibrosis, respectively, within the left ventricle. Bright field images were acquired using an Olympus FSX 100 light microscope and analyzed blinded using ImageJ (for CSA) and cellSens (for interstitial fibrosis).

Cobalt chloride (CoCl₂) is used as a hypoxia mimetic for its ability to chemically induce HIF-α stabilization and upregulate downstream HIF target genes under normoxic conditions (reviewed and discussed in detail (Muñoz-Sánchez and Chánez-Cárdenas, 2019; Packer, 2020)). The effects of CoCl₂ are widespread, resulting in HIF-2α stabilization across multiple cell populations and tissues, including the myocardium (Wiesener et al., 2003). Mice were injected i. p. with either CoCl₂ (30 mg/kg) dissolved in 1 mL sterile PBS (to simulate hypoxia) or vehicle control (PBS). Hearts were collected for HIF-2α western blotting at the following endpoints: 45 min, 1.5 h, or 3 h after injection.

For probing of HIF-2α, left ventricular samples were homogenized in a buffer with a phosphatase (PhosSTOP, cat. # 4906845001, Sigma) and protease inhibitor cocktail (cat. #P8340, Sigma), and separation of nuclear from cytoplasmic extracts was performed using the NE-PER kit as per the manufacturer’s instructions (cat. # 78833, Life Technologies). Nuclear protein extract concentrations were measured by bicinchoninic acid assay (cat. # 23277, Fisher Scientific). Samples were equally loaded (20ug/well) and separated by 10% SDS-PAGE, followed by immunoblotting. Nitrocellulose membranes were rinsed in ddH2O and then incubated in reversible Ponceau Stain for 7 min to confirm equal protein transfer. The stain was stripped using 200uM NaOH for 1 min and rinsed in ddH2O for 5 min x 3. Membranes were blocked (5% skim milk in 1x TBST (0.1% tween)) and incubated in a primary HIF-2α antibody (Sun et al., 2015) (1:1000; cat. # NB100-122SS, Bio-Techne; 5% bovine serum albumin in 1x TBST) overnight at 4°C. Membranes were washed for 5 min x 3 in TBST and then incubated with a goat anti-rabbit IgG horseradish peroxidase–conjugated secondary antibody (1:1000, cat. # HAF008, Bio-Techne; 1% skim milk for 1 h at 22°C). Membranes were washed for 5 min x 3 in 1x TBST. Signal was detected and quantified via enhanced chemiluminescence (cat. # 1705060, Bio-Rad) using a FluorChem HD imaging system (Alpha Innotech, Santa Clara, CA, USA). Values were obtained by measuring the target band (normalized to Ponceau) relative to the EPO^fl/fl group.

Axitinib is an FDA-approved selective inhibitor of cellular phosphorylation of VEGF receptor tyrosine kinases (VEGFR-1, VEGFR-2, VEGFR-3) for the treatment of advanced renal cell carcinoma (Tyler and FCSHP, 2012; INLYTA, 2023). Axitinib (cat. #S1005, Selleck Chemicals) was prepared as previously described (Ma and Waxman, 2009). Briefly, axitinib was suspended at 5 mg/mL in polyethylene glycol 400 (cat. # PX1286B-2, Sigma) and sonicated at room temperature for 30 min until dissolved. Using 0.1N HCl, the pH was adjusted to 2.5, followed by a second round of 10-min sonication. To achieve a final ratio of 3:7 (v/v) of polyethylene glycol 400 to water, acidified water (pH 2.5) was added. The solution was prepared fresh and stored in the dark at 4°C. Axitinib was administered once a day through i. p. injection at a dose of 25 mg/kg body weight in a volume of 5 μL/g body weight (Ma and Waxman, 2009) for 4 days x 2 cycles (with 2 days rest). Data (invasive hemodynamics, saphenous vein hemoglobin levels, and tissues for qPCR) were collected 24 h after the last injection.

A one-step TUNEL in situ apoptosis kit (cat. # E-CK-A320, Elabscience Biotechnology Co.) was used to quantify differences in cell apoptosis between EPO^fl/fl and EPO^Δ/Δ hearts according to the manufacturer’s instructions. Fluorescent images were acquired using an Olympus FSX 100 microscope and the amount of green fluorescence of TUNEL-positive cells was analyzed blinded using ImageJ.

Graphing and statistical analyses of the data presented was performed using Prism version 9 software developed by (GraphPad Inc., La Jolla, CA, USA). Power calculations were used to determine the number of mice needed to detect a significant effect. Results were reported as mean ± SD (morphometrics, serum, hematocrit, qPCR, western blotting) or mean ± SEM (CLAMS, echocardiography, histology, invasive hemodynamics, Langendorff). To confirm whether data was normally distributed, a Shapiro-Wilk test was used. If the data was normally distributed, either a one-way ANOVA followed by Dunnett’s or Tukey’s post-hoc test or an unpaired Student’s t-test was performed. If the data was not normally distributed, a Kruskal–Wallis test with Dunn’s post-hoc test or Mann-Whitney U test was performed. Simple linear regression correlation analyses were run comparing Epo RNA expression to dP/dt_max, dP/dt_min, and dP/dt@LVP40 values and goodness-of-fit values were provided. p-value <0.05 was considered significant.

In the developing fetus, mice null for Epo and the Epor die by E13.5 due to impaired cardiogenesis and anemia (Wu et al., 1999; Kertesz et al., 2004). Our previous work using the Mlc2v promoter established cardiomyocyte specific Epo deletion during embryogenesis in mice induces a phenotype with less cardiomyocyte hypoplasia, compensatory cellular hypertrophy, and upregulation of Epo in the heart by the endothelial cell persisting into adulthood (Allwood et al., 2024). Therefore, the presence of cardiac EPO is critical for proper development of the heart, yet endogenous cardiac Epo expression, regulation, and physiological significance in the adult heart remains unknown. For this reason, in our follow up study, we hypothesized that the demand for compensatory cardiac EPO would be reduced in a fully, and normally, developed adult heart. Upon induced deletion of adult cardiomyocyte Epo in mice after an ordinary course of cardiogenesis, we expected whole-heart morphology to be normal and cardiac Epo production to be low. To verify this, mice were subjected to morphometric and quantitative tissue PCR analysis. In line with our previous reports (Allwood et al., 2024), the morphological measurements revealed no gross differences in body weight, heart weight, left ventricular weight, or heart weight/tibial length (HW/TL) ratio between groups (Supplementary Figure S3C; Table 2). Surprisingly, after tamoxifen induced MerCreMer deletion of Epo expression, there was a significant upregulation in cardiac Epo in EPO^Δ/Δ mice compared to the EPO^fl/fl group by qPCR (Figure 1A). This data was confirmed by a mildly graded response in mice with only one floxed EPO allele (i.e., EPO^fl/- Cre^+/−) (Figure 1A), suggesting this phenomenon does not exist in an all-or-nothing manner. Further, we sought to confirm our data was not specific to male mice. Indeed, female EPO^Δ/Δ mice also showed a significant increase in whole-heart Epo upon targeted cardiomyocyte Epo deletion (Figure 1B) with no apparent sex effect (Supplementary Figure S4). To identify the specific cell type(s) responsible for Epo expression under normoxic conditions in EPO^fl/fl and EPO^Δ/Δ mice, we used RNA fluorescent in situ hybridization. As expected, we observed low basal Epo mRNA production by the cardiomyocyte (Figure 1Ci) and endothelial cell (Figure 1Cii) in EPO^fl/fl mice. By contrast, in the EPO^Δ/Δ mice, an over-abundance of Epo signal (shown using two regions of interest, Figure 1Ciii, iv) was co-localized with the endothelial cells, indicating that upon successful cardiomyocyte-Epo knockout, the endothelium compensated for the loss by increasing its own Epo expression. Therefore, not only have we verified this phenomenon in the adult mouse using a second independent CreLox line to support our earlier work (Allwood et al., 2024), but we also showed hyper-compensated cardiac Epo occurs regardless of floxed allele homozygosity (EPO^Δ/Δ) or heterozygosity (EPO^fl/-: Cre^+/−), and in both sexes. Importantly, despite marked Epo overexpression in the adult EPO^Δ/Δ mouse, on a whole-body and organ level, there were no visual abnormalities (Supplementary Figure S3A). There was also no change in renal Epo expression (Figure 2A), serum EPO (Figure 2B), or hematocrit (Figure 2C), suggesting that there would be no altered physiological consequences from compensatory cardiac endothelial overexpression.

Exposing mice to either a physiological or pathological challenge could reveal additional insight to the innate compensatory mechanisms. We therefore subjected mice to an exercise test, and considering there was no change in hematocrit, we hypothesized no difference would be observed between groups. Yet, voluntary wheel running across three consecutive days revealed EPO^Δ/Δ mice had increased running performance compared to control animals (Figure 2D). Since the difference in exercise capacity was not the result of erythropoiesis, we determined it could be linked to reported effects of EPO as a regulator of energy homeostasis as it increases metabolic activity, cellular respiratory capacity, and oxygen utilization shown previously in transgenic mice (Wang et al., 2013). However, using whole-body indirect calorimetry (CLAMS setup), our data showed there were no differences in VO₂ (Figures 2E, F), VCO₂ (Figures 2G, H), RER (Figures 2I, J), or total energy expenditure between groups (Figure 2K, L). We cannot exclude the possibility for differences in myocardial oxygen consumption at the level of the skeletal muscle by these data alone. It is plausible that EPO^Δ/Δ mice were more efficient at extracting oxygen for mitochondrial cellular respiration (i.e., ATP production). Indeed, positive hypertrophic cardiac remodeling might account for the heightened exercise tolerance due to increased cardiac output.

Prior work showed that constitutive deletion of cardiomyocyte-specific EPO during embryogenesis causes early hypoplasia, and eventually, cardiac Epo over-expression led to hypertrophy, with no change in overall cardiac mass in adult hearts (Allwood et al., 2024). We sought to explore the physiological significance of cardiomyocyte deleted Epo in adult mice, independent of cardiogenesis. By echocardiography, there was a significant decrease in end systolic dimension (ESD) and a trending reduction in end diastolic dimension in EPO^Δ/Δ mice (EDD, Figures 3A, B; Table 3). The left ventricle chamber length was decreased (Figures 3A, C) and posterior wall thickness was increased (Figures 3A, D). While heart rate (Figure 3E) and cardiac output (Figure 3F) were unchanged, EPO^Δ/Δ mice had greater ejection fractions (Figure 3G), suggesting systolic function was being augmented. Together, these parameters indicated a concentric hypertrophy phenotype in the EPO^Δ/Δ mice. To confirm these findings, we evaluated histologically cardiomyocytes in perfusion-fixed hearts (Figure 3H), identifying a significant increase in EPO^Δ/Δ cross-sectional area compared to control (Figure 3I). These data combined suggest that as cells of the left ventricle wall widened, the length of the hearts shortened, resulting in no difference in global heart mass.

Both pathological (e.g., hypertension) and physiological (e.g., exercise) conditions stimulate cellular and gross ventricular hypertrophy with inverse relationships to long-term health (Schoenfeld et al., 1998). However, a hallmark feature of pathological remodeling is the re-expression of certain fetal genes (Schoenfeld et al., 1998). Therefore, we quantified relative RNA levels of Anp, Bnp, α-Mhc, ß-Mhc, and the ratio of ß-Mhc/α-Mhc between groups by qPCR. No differences amongst Anp, Bnp, α-Mhc expression, or the ratio of ß-Mhc/α-Mhc (Figure 3J) were seen though ß-Mhc RNA was increased in EPO^Δ/Δ mice. An incipient cause of ß-Mhc re-expression during cardiac hypertrophy and normal aging is linked to elevated fibrosis (Pandya et al., 2006). Accordingly, we quantified interstitial fibrosis of the left ventricle between groups (Figures 3Ki,ii), finding no difference. EPO^Δ/Δ mice had compensatory cardiac function (i.e., increased ejection fraction) and the EPO^Δ/Δ phenotype is inconsistent with pathological remodeling. Nonetheless, we reconciled this line of inquiry with a comprehensive assessment of cardiac function by invasive hemodynamics.

To gain a deeper understanding of the functional role of cardiac Epo overexpression and concentric cellular hypertrophy on cardiac function in our model, invasive hemodynamic analyses were performed (Figure 4A; Table 3). There were no differences in heart rate (Figure 4B). In terms of systolic function, EPO^Δ/Δ mice demonstrated elevated left ventricular systolic pressures (LVP, Figure 4C) and an increase in cardiac contractility (dP/dt_max, dP/dt@LVP40, Figures 4D, E). To assess diastolic function, a multi-parameter approach was used (Ogilvie et al., 2020). For this reason, we reported indices of both relaxation and compliance: dP/dt_min (active relaxation phase), Tau Logistic (active relaxation phase), and end diastolic pressure (EDP, passive filling phase). dP/dt_min was significantly improved (Figure 4D), though this was not reflected by differences in EDP (Figure 4F) or Tau Logistic (Figure 4G). Taken together, both systolic and diastolic function were superior in the EPO^Δ/Δ mice, suggesting endogenous cardiac EPO production was associated with better cardiac function. Supraphysiological levels of exogenous rhEPO confer inotropic and lusitropic effects ex vivo, therefore we were interested in the correlation between the levels of upregulated endogenous Epo RNA and these key hemodynamic parameters recorded in vivo. Simple linear regression analyses demonstrated a strong positive relationship between Epo RNA expression and dP/dt_max, dP/dt_min, and dP/dt@LVP40, with goodness-of-fits of R² = 0.49, 0.48, and 0.67, respectively (Figures 4H–J). Therefore, our results indicate that in mice with higher endogenous cardiac Epo expression, inotropic and lusitropic function are greater.

The in vivo assessment of cardiac function carries limitations–the data collected are dependent on preload, afterload, and heart rate. Conversely, the Langendorff preparation (i.e., ex vivo) allows for the isolated evaluation of systolic and diastolic cardiac function at a healthy baseline or after ischemic insult under controlled conditions without the influence of systemic neurohormonal or immunological factors. Since exogenous rhEPO has shown cardioprotective effects in vivo, investigating the physiological autocrine and/or paracrine effects of endogenous cardiac EPO on the heart was of particular interest. The EPO^Δ/Δ group demonstrated a significant increase in baseline LVP (Figures 5A, B (top left)), developed pressure (Figure 5C), and dP/dt_max and dP/dt_min (Figure 5A (bottom left), D). While there were no significant differences in EDP or paced heart rate as they were manually set and controlled, intrinsic heart rate prior to pacing was elevated in EPO^Δ/Δ mice (Table 4). After baseline recordings, mice were subjected to 25 min of global, no-flow ischemia, followed by 45 min of reperfusion (Figure 5A (right)). EPO^Δ/Δ mice displayed better recovery post-ischemia: 71% of LVP (Figure 5E), 76% of dP/dt_max, (Figure 5F), and 61% of dP/dt_min (Figure 5G), whereas the EPO^fl/fl mice recovered only 27% of LVP (Figure 5E), 28% of dP/dt_max (Figure 5F), and 25% of dP/dt_min (Figure 5G). Therefore, independent of heart rate, afterload, preload, and whole-body neurohormonal influence, hearts overexpressing endogenous cardiac EPO also displayed positive inotropic, lusitropic, chronotropic, and cardioprotective qualities. Having established the physiological impact of hyper-compensated cardiac Epo, we aimed to clarify some of the factors involved in regulating this phenomenon.

HIF-1α and HIF-2α are both primary transcription factors for regulating the hypoxic response. However, in terms of the transactivation of the Epo promoter, HIF-2α, and not HIF-1α, is the primary regulator in the adult kidney, liver, and brain (Warnecke et al., 2004; Gruber et al., 2007; Rankin et al., 2007; Yamashita et al., 2008; Yeo et al., 2008; Wang et al., 2015; Urrutia et al., 2016; Ghosh et al., 2021). Although there was no overt indication of hypoxia in this model (i.e., no differences in hematocrit or renal EPO expression), we aimed to rule out the canonical PHD2/VHL/HIF2α axis as the mechanism for mediating cardiac overexpression of Epo in EPO^Δ/Δ mice. By immunoblotting, we showed no difference in the HIF-2α protein levels in left ventricular nuclear extracts of EPO^fl/fl and EPO^Δ/Δ mice (Figures 6A–C). This finding confirmed the overexpression of left ventricular Epo in EPO^Δ/Δ mice was HIF2α-independent. Considering the unexpected nature of our findings, we wanted to conclusively exclude hypoxia as a stimulus for cardiac overexpression of Epo by investigating HIF1α-specific downstream target genes.

Under hypoxic conditions, HIF-1α escapes oxygen-dependent degradation, translocates to the nucleus, and complexes with HIF-1β (Wang et al., 1995). Working in concert with hepatocyte nuclear factor 4 (HNF-4) (Galson et al., 1995; Huang et al., 1997; Jelkmann, 2007) and the transcriptional co-activators, p300 and cAMP response element (CREB)-binding protein (Firth et al., 1995; Bunn et al., 1998), the HIF-1 complex binds the hypoxia response element to initiate the transcription of >500 downstream target genes (Mole et al., 2009), including Vegfa (Forsythe et al., 1996), Glut1 (Iyer et al., 1998), Ho-1 (Lee et al., 1997), Pgk-1 (Semenzas et al., 1994; Li et al., 1996a), Cas 8 (Zhao et al., 2018). Therefore, to rule out hypoxia and the involvement of HIF-1α in regulating cardiac Epo overexpression, we examined the expression levels of select HIF1α-specific target genes (e.g., Vegfa, Glut1, Hmox-1, Pgk-1, Cas 8) by qPCR. We measured no differences in their gene expression, and even downregulation of Cas 8 in EPO^Δ/Δ mice (Figure 7). This indicated the upstream regulation of Epo in the EPO^Δ/Δ mice by hypoxia and HIF-1α was unlikely. Epo is also regulated by HIF-independent mechanisms in a tissue-specific manner (Yasuda et al., 1998; Tam et al., 2006). One study demonstrates potent inhibition of VEGF induces hepatic synthesis of Epo and subsequent erythropoiesis through a HIF1α-independent mechanism (Tam et al., 2006). Therefore, we investigated the hypoxia-independent VEGF isoform, Vegfb, and its receptor, Vegfr1, by qPCR analyses. Interestingly, both genes were significantly upregulated in EPO^Δ/Δ mice (Figure 7). Co-upregulation of endothelial-derived EPO and whole-heart Vegfb and Vegfr1 suggested a complex cellular interaction was governing this novel physiological concentric hypertrophy and cardioprotective phenotype. To identify a relationship between cardiac EPO and VEGF, we used the FDA-approved VEGF-specific tyrosine kinase inhibitor, axitinib, to interrupt VEGF signaling and observe the corresponding impact on Epo RNA levels in the heart.

Several independent labs show active systemic crosstalk between EPO-EPOR and VEGF-VEGFR signal transduction pathways. rhEPO increases Vegf expression (Ribatti et al., 1999; Westenbrink et al., 2008; Lu et al., 2012; Oztas et al., 2020) resulting in angiogenesis and attenuated interstitial fibrosis, while VEGF-VEGFR blockade induces non-renal Epo expression (Tam et al., 2006) and erythropoiesis (Tam et al., 2006; Johnson et al., 2017). Whether these pathways interact similarly in the heart by endogenous forms of EPO and VEGF, is unknown. Therefore, to improve our understanding of the interplay between VEGF-VEGFR and EPO-EPOR in the heart, we interrupted VEGF-VEGFR signaling by inhibition using axitinib (Ma and Waxman, 2009) in EPO^fl/fl mice. Preclinically, in a mouse model of prostate cancer, 24 days of axitinib treatment effectively arrests tumour growth by supressing tumour patency (i.e., blood perfusion) without disrupting blood perfusion to normal tissues (e.g., heart, kidney, liver, lung, muscle) (Ma and Waxman, 2009). In our study, axitinib was well-tolerated, and mice did not display any overt signs of distress during, or after, treatment. Short-term axitinib treatment did not elicit changes in erythropoiesis, as determined by no differences detected in hemoglobin levels measured between EPO^fl/fl and EPO^fl/fl+AXI groups (Figure 8A). When VEGF signaling was inhibited, cardiac expression of Epo RNA was significantly elevated by qPCR (Figure 8B). However, by invasive hemodynamics, we observed blunted heart rates and reduced left ventricle relaxation in axitinib-treated mice (Supplementary Table S2). These data were expected since tyrosine kinase inhibitors affect the cardiac conduction system, causing bradycardia and QTc prolongation (Kloth et al., 2015; Shopp et al., 2022), and are known to have cardiotoxic effects (Force and Kolaja, 2011; Dobbin et al., 2021). Our findings indicated cardiac Epo upregulation following VEGF inhibition does not rescue hearts from suspected axitinib-related bradycardia and cardiotoxicity. Importantly, however, we revealed the existence of a previously unrecognized link between VEGF and EPO in the heart. Future work is needed to fully define how the cardiac EPO-EPOR and VEGF-VEGFR axes are coordinated.