Using standard techniques (19), in-vivo terminal invasive pressure catheterization was performed on both groups (n_Control = 8; n_Aging = 5). Open-chest hemodynamic measurements were performed under anesthesia induced via inhalation of isoflurane, while the animals were placed on a heated table (37°C) and monitored using a rectal probe. Pressure waveforms were then acquired using a conductance catheter and analyzed for common hemodynamic metrics of RV function. Due to lack of cuvette calibration of catheters for conversion of volume measurements from relative-volume-units to absolute measurements, only pressure-based hemodynamic data are reported and compared in the current work. RV heart rate was calculated by evaluating the periodicity of the waveforms (peak-to-peak time). Peak pressures were characterized as the maximum pressure experienced by the RV during a cardiac cycle (P_max), while end-diastolic pressures were identified as pressures at the point of the maximum second derivative of the waveforms, (d2pdt2)max. Load-dependent measures of RV contractility and relaxation were, respectively, calculated by evaluating the maximum and minimum of the time derivatives of pressure waveforms (dpdtmax and dpdtmin). Contractility index was then obtained using the ratio of dpdtmax over the maximum pressure experienced by the RV over a cardiac cycle (RV peak pressure; P_max). Additionally, the time-constant of RV relaxation (tau) was calculated as (26):

Where P represents the RV pressure waveform beginning at dpdtmin until the minimum RV pressure experienced during a cardiac cycle (P_min), B is an intercept, t represents the time during a cardiac cycle, and τ (tau) is the time-constant of RV relaxation, measured via linear regression of equation 1 to the acquired pressure waveforms in MATLAB (Mathworks, Natick, MA). Hemodynamic parameters for each animal were evaluated using average measurements from at least three consecutive cycles, except for tau, which was obtained from a single beat (due to limitations of our custom hemodynamic analysis subroutines for estimation of tau).

Following hemodynamic measurements, the heart was harvested and arrested by placement in cardioplegic solution (27). Subsequently, the RV free wall (RVFW) was dissected and measurements were acquired for the Fulton index [ratio of RV weight to weight of the left ventricle (LV) + intraventricular septum] and RVFW thickness (n_Control = 9, n_Aging = 6). All measurements were performed in air, at room temperature (23°C), using a scale (0.1 mg readability; Mettler-Toledo International Inc., Columbus, OH) and a thickness gauge (0.025 mm precision; L. S. Starrett Company, Athol, MA).

Following morphological measurements, square specimens with a circumferential-longitudinal orientation (Figure 1) were harvested from the RVFW to undergo biaxial mechanical testing (n_Control = 6, n_Aging = 6). Specimens were mounted on a BioTester testing device (CellScale, Waterloo, ON, Canada), using a suture and pulley mechanism for minimal shear loading (28). Samples were then submerged in modified Krebs solution with 2,3-Butanedione monoxime and oxygen to ensure tissue viability (27). Our previous work (19) has shown that this media bath can effectively maintain tissue viability up to 90 min, via passive diffusion. All measurements in this study were concluded within 45–80 min of harvesting the heart.

RVFW mechanical properties were characterized using multi-protocol displacement-controlled biaxial loading scenarios (1:1, 1:2, 2:1, 1:4, 4:1, 1:6, and 6:1 displacement ratios). Previous work has demonstrated that this loading protocol can effectively capture the biaxial RVFW properties under a wide range of possible strains (18, 27, 29), generating adequate data for parameter characterization of constitutive models. Each specimen underwent 15 cycles of 1:1 displacement-controlled preconditioning, before the start of data acquisition. Four visual tracking markers were placed on the epicardial surface of the RVFW specimens and marker displacements (recorded using a CCD camera) were post-processed via standard techniques (27, 29, 30) to obtain the deformation gradient tensor (F), using a four-node finite-element approximation (31). Components of the Green–Lagrange strain tensor (E) were then calculated as E=12(FTF-I), where I is the identity tensor. Biaxial force measurements and initial specimen dimensions were used to obtain the 1st Piola-Kirchhoff stress tensor (P) by calculating the ratio of forces in the deformed configuration over the cross-sectional area in the reference configuration. The 2nd Piola-Kirchhoff stress tensor (S) was then evaluated as S = F⁻¹P. Stress-strain data was post-processed under a plane-stress approximation, using a finite deformation analysis framework in Mathcad (PTC, Needham, MA).

Using previously established techniques (19, 32), equibiaxial strain-controlled responses of RVFW specimens were interpolated from the acquired multi-protocol displacement-controlled experimental data via biharmonic spline interpolations in MATLAB (Supplementary Figure 2). As previously discussed, equibiaxial strain-controlled responses are accompanied by unique tissue kinematics with no fiber rotations (33) and, therefore, could be used to estimate fiber-level mechanical properties from tissue-level measurements, independent of fiber orientation and splay (19, 21, 32). Effective fiber-ensemble (EFE) stresses, representing the fiber-level response of combined collagen and myofiber bundles, were then estimated from tissue-level measurements as (19, 32):

Here, S_EFE represents the EFE stress of the combined collagen-myofiber bundles, and (_{S11)Equibiaxial} and (_{S22)Equibiaxial} are the interpolated biaxial tissue-level 2nd Piola-Kirchhoff stresses under equibiaxial strains, respectively, in the circumferential and longitudinal directions. We assumed the initial nearly-linear, low-strain portion of the EFE stress-strain responses to be mostly dominated by myofibers, while collagen fibers dominated the high-strain response following recruitment (18, 34) (Supplementary Figure 3). To categorize the data before and after collagen recruitment, equation 2 was differentiated with respect to EFE strain (E_EFE), to evaluate the changes in EFE stiffness (TMEFE=∂SEFE∂EEFE; where TM_EFE is the EFE stiffness). For specimens in both groups, we observed a relatively constant-stiffness region (relatively linear stress-strain behavior, dominated by myofibers), followed by beginning of collagen recruitment and an abrupt increase in EFE stiffness (Supplementary Figure 4). The strain at which collagen fibers begin recruitment was defined as the point where there is a significant elevation in EFE stiffness compared to the stiffness trends prior to that point. This was quantified as the point where TM_EFE (EFE stiffness) was significantly elevated outside of the Z = 4.417 confidence interval of the distribution of TM_EFE measurements before that point. A Z-value of 4.417 (99.999% confidence interval) was chosen as a threshold for maximal confidence in the detected increase in stiffness, avoiding false detection of collagen recruitment strain due to potential fluctuations in the low-strain data, resulting from data acquisition noise. The EFE strain at the n + 1th point of the EFE stiffness-strain plot (Supplementary Figure 4) was defined as the collagen recruitment strain, if:

Here, (TM_EFE)_n+1 is the EFE stiffness at the n + 1th point of the EFE stiffness-strain data (Supplementary Figure 4). The right-hand side of the inequality represents the upper bound of the TM_EFE confidence interval based on the TM_EFE data up to the n^th point. The beginning of collagen recruitment was defined as the first point where the inequality in equation 3 is satisfied. The EFE stress-strain data before collagen recruitment was then used for myofiber stiffness estimations, using a rule-of-mixtures approach (18, 19, 35):

Where (T_{MEFE)Before Collagen Recruitment} is the slope of the line fitted to the initial low-strain portion of the EFE stress-strain curve (Supplementary Figure 5), ϕ_Myofiber and ϕ_Collagen represent the myofiber and collagen area fractions in RVFW specimens (measures of tissue content; acquired from histological measurements), and TM_Myofiber and TM_Collagen are the effective myofiber and collagen stiffnesses, respectively. Assuming the initial portion of the stress-strain data to be dominated by myofibers (minimal collagen recruitment, TM_Collagen = 0), effective myofiber stiffness for each specimen was estimated as:

Additionally, a non-linear anisotropic constitutive model (36) was used to model the response of the RVFW specimens in each cohort:

Here, W is the strain energy density, E₁₁ and E₂₂, respectively, represent the circumferential and longitudinal (apex-to-base) Green-Lagrange strains, B₀ is a scaling factor and b₁, b₂ and b₃ are metrics for the circumferential, longitudinal and in-plane coupling stiffness of the RVFW, respectively (19). 2nd Piola–Kirchhoff stress components were obtained by differentiating equation 6 with respect to Green-Lagrange strain:

Where (S₁₁)_{Model−Predicted} and (S₂₂)_{Model−Predicted} are the model-predicted stress components in the circumferential and longitudinal directions, respectively. Using equation 7 and the acquired multi-protocol experimental stress-strain data, model parameters were estimated for each specimen using a trust-region-reflective non-linear least-squares optimization algorithm in MATLAB, to minimize the difference between model-predicted and experimentally acquired data. A R² measure was used to evaluate the goodness of fit. Cohort-specific strain energy maps in the low-strain and high-strain regions were then generated by taking the average of all strain energy distributions in the circumferential-longitudinal strain space for specimens in each cohort, facilitating holistic model-based evaluation of RVFW biomechanical properties over a wide range of loading scenarios.

Transmural histological staining was performed on a sub-group of specimens (n_Control = 3, n_Aging = 3) to quantify the effects of aging on RV fiber architecture. Sample sizes were chosen based on our previous work showing minimal between-sample variabilities in RV content and fiber architecture (18, 19). Specimen fixation was carried out using 10% neutral buffered formalin followed by staining of RVFW specimens using Masson's trichrome, resulting in collagen fibers stained in blue and myofibers in red/pink. A total of 11–17 sections were obtained for each specimen, from epi to endocardium, at 50–75 μm increments. Transmural area fractions of collagen and myofibers were then quantified via manual RGB-based thresholding of the histological images (collagen: blue, myofibers: red/pink) to analyze the effects of aging on RVFW composition. Area fractions were calculated as the ratio of the area occupied by respective blue/red pixels, divided by the total area within the region of interest. In addition, cardiomyocyte width was measured from the histological data to investigate the role of aging in RV hypertrophy (40 measurements performed on each specimen at different sites along the myofibers). Furthermore, similar to previous work (18), the orientation of RVFW collagen and myofibers and the coherency of collagen fiber distributions were quantified transmurally, using gradient-based image analysis techniques (37). Following segmentation of histological sections based on the appropriate RGB threshold, local image gradients at each section were used to construct the structure tensor of the gradient map in order to analyze the transmural orientation of RVFW collagen and myofibers (18, 19, 37):

Here, T is the symmetric positive-definite structure tensor, R(x,y) is a gaussian weighting function which specifies the integration region of interest (37), and I_x and I_y are the partial spatial derivatives of the histological image (I), respectively, in x and y directions. The 1st eigen vector of T indicates the dominant fiber orientation at each histological section (19, 37). For all data presented in this work, 0° corresponds to the circumferential direction, while +90° points toward the apex-to-base (longitudinal) direction. Moreover, using the eigen values of the structure tensor in equation 8, collagen fiber coherency was evaluated as:

where λ₁ and λ₂ correspond to the 1st and 2nd eigen values of the structure tensor T (37). 0% collagen fiber coherency corresponds to a sparse (non-coherent), randomly distributed fiber architecture, while 100% coherency indicates a highly-aligned, tightly packed, continuous (coherent) collagen fiber distribution (38).

A total of 66 histological sections were analyzed for the control and aging groups. We performed linear interpolations to report the histological data on an equally-spaced grid, against normalized tissue thickness (0–100% thickness). In case of data categorization (Epi, Mid and Endo groups), the data between 0 and 20% thickness were used for the epicardium, while the data between 80 and 100% thickness correspond to the endocardium. Orientation analysis and image segmentation were performed using the OrientationJ toolbox (37, 39) in ImageJ (imagej.nih.gov).

Data are presented with mean ± standard error of the mean. Sample normality and homogeneity of variances were assessed using the Shapiro–Wilk test and Bartlett's test of homoscedasticity. Circular statistics was employed for fiber orientation analysis, using the Watson–Williams test in the CircStat toolbox (40) in MATLAB. For all other data, in case of normality and homoscedasticity, a two-sided unpaired student's t-test was used for statistical comparisons, while non-normal distributions were compared using Mann–Whitney U-tests. For all purposes, p < 0.05 was considered statistically significant. Statistical analyses were performed in the R software package (R Foundation for Statistical Computing, Vienna, Austria).

Healthy aging did not show an effect on the heart rate (Figure 2A; 271.5 ± 11.7 vs. 292.3 ± 14.1 BPM for Aging-vs.-Control; p = 0.326). Aging resulted in increased RV peak pressures (Figure 2B; 26.8 ± 0.9 vs. 23.0 ± 0.9 mmHg for Aging-vs.-Control; p = 0.017), while showing a modest non-significant effect on end-diastolic pressures (Figure 2C; 1.9 ± 0.4 vs. 1.3 ± 0.1 mmHg for Aging-vs.-Control; p = 0.085). Effects of aging on the load-dependent measures of RV contractility and relaxation are shown in Figure 2D. Aging significantly increased dpdtmax (1,611.7 ± 90.5 vs. 1,063.8 ± 101.7 mmHg/s for Aging-vs.-Control; p = 0.004) but did not demonstrate any effects on dpdtmin (−823.9 ± 60.4 vs. −814.7 ± 85.5 mmHg/s for Aging-vs.-Control; p = 0.940). Increased contractility index was observed for the aging group (Figure 2E; 60.1 ± 2.2 vs. 45.8 ± 3.5 1/s for Aging-vs.-Control; p = 0.012), while the time-constant of RV relaxation (tau) remained unchanged (Figure 2F; 10.7 ± 1.6 vs. 9.9 ± 0.8 ms for Aging-vs.-Control; p = 0.595).

Healthy aging led to increased RVFW thickness (0.90 ± 0.05 vs. 0.65 ± 0.05 mm for Aging-vs.-Control; p = 0.001), while not affecting the Fulton index (0.26 ± 0.03 vs. 0.27 ± 0.01 mg/mg for Aging-vs.-Control; p = 0.140). Moreover, aging was associated with decreased RV and LV weight normalized to body weight (RV WeightBody Weight: 0.05 ± 0.007% vs. 0.06 ± 0.003% for Aging-vs.-Control, p = 0.026; LV WeightBody Weight: 0.18 ± 0.007% vs. 0.23 ± 0.009% for Aging-vs.-Control, p = 0.0005). Specimen-specific RVFW thickness and Fulton index measurements are reported in Supplementary Table 1.

Aging demonstrated a bimodal effect on the RVFW biaxial properties by resulting in increased circumferential and longitudinal stiffness under lower strains, while progressing to decreased biaxial stiffness at higher strains (Figure 3A). A similar effect was observed on the EFE (effective fiber-ensemble) stress-strain properties of combined RVFW collagen-myofiber bundles (Figure 3B). Using a rule-of-mixtures approach, this translated into increased effective myofiber stiffness (Figure 3C; 159.5 ± 23.6 vs. 66.2 ± 5.2 kPa for Aging-vs.-Control; p = 0.003), while no significant effects were observed on collagen recruitment strain (Figure 3D; 11.9 ± 0.7% vs. 10.4 ± 0.9% for Aging-vs.-Control; p = 0.197). Specimen-specific constitutive model parameters for each group are shown in Supplementary Table 2. Overall, the employed model showed an acceptable fit quality (R²) to our experimental data (R² = 0.95 ± 0.01 and 0.96 ± 0.01 for Aging and Control, respectively). Age-specific strain energy maps, representing the combined effects of all model parameters, are demonstrated in Figures 3E,F for each cohort at the low-strain and high-strain regions.

Representative histological sections for each group are demonstrated in Figure 4A. Aging resulted in increased cardiomyocyte width (Figure 4B; 25.42 ± 0.34 vs. 14.94 ± 0.64 μm for Aging-vs.-Control; p = 0.0001). Quantifying the transmural orientation of RVFW fibers revealed myofiber (Figure 4C) and collagen (Figure 4D) reorientation toward the longitudinal direction at sub-endocardial levels. Overall, myofibers showed similar orientations to collagen fibers. Aging significantly shifted the overall orientation of myofibers (circular mean of transmural fiber angles, dotted lines in Figure 4C) by 14.6° toward the longitudinal direction (Figure 4E; p = 0.017). Similarly, the overall orientation of collagen fibers was shifted by 16.4° (p = 0.013). Aging also resulted in cardiomyocyte loss and decreased myofiber area fractions at both epicardium (Figure 4F; 90.8 ± 0.3% vs. 95.3 ± 0.7% for Aging-vs.-Control; p = 0.004) and endocardium (Figure 4F; 82.4 ± 1.5% vs. 95.3 ± 1.9% for Aging-vs.-Control; p = 0.007). Furthermore, aging led to RVFW fibrosis and increased collagen area fractions at epicardium (Figure 4G; 5.3 ± 0.4% vs. 3.4 ± 0.3% for Aging-vs.-Control; p = 0.015) and the mid-ventricular region (Figure 4G; 5.0 ± 0.4% vs. 3.4 ± 0.3% for Aging-vs.-Control; p = 0.037). Analyzing the coherency of collagen architectures revealed decreased coherency at the endocardium (Figure 4H; 10.4 ± 1.1% vs. 19.7 ± 1.1% for Aging-vs.-Control; p = 0.003), while showing no effects on the other regions.