High-Resolution Longitudinal MRI of the Transition to Heart Failure

Abstract

The development of heart failure (HF) is an evolving process that entails both structural and functional changes through time. While the physiological state of cardiac pathologies has been well characterized, less is known about the transition from a normal to a maladaptive state. Magnetic resonance imaging (MRI) is a noninvasive technique that facilitates longitudinal experiments to follow the progression of cardiac structural and physiological disorders over time. Transgenic murine models of cardiac disease, such as the muscle LIM protein-deficient strain used in this study, offer populations of a reproducible phenotype that readily lend themselves to serial studies. In this longitudinal study, high spatial and temporal resolution time-course MR images revealed an abrupt and brief phase of major anatomical restructuring during which the ventricular chamber dilated and the wall thinned. The ability of MRI to acquire spatially and temporally resolved images enabled the 3D estimation of cavity volume and wall mass changes with time. It was concluded that, using an imaging protocol of high temporal resolution, MRI has the adequate spatial and temporal imaging resolution to allow for the detection and quantification of rapidly occurring transitional phases in a single mouse heart as it progresses toward failure.

Chronic pathophysiologies, such as heart disease, evolve through time and are marked by both structural and functional changes (1). In particular, left ventricular (LV) chamber dilatation and wall thinning are established characteristics of the transition toward overt failure (2). By characterizing these geometric alterations and the resulting changes in systolic performance, and correlating them in time, we can obtain a fundamental understanding of the disease progression and identify targets of possible therapeutic intervention.

It is difficult to quantify complex chronic syndromes, such as heart failure (HF), because of their diverse etiologies and time courses. The recent development of genetically engineered mice has provided clinically relevant models of cardiovascular disease that offer a number of advantages to facilitate biomechanical analyses (3). The HF model used in this study was a genetically modified strain lacking a cytoskeletal muscle LIM-only protein. Many of the clinical characteristics associated with congestive HF have been shown in adult failing HF mice (4). However, younger knockouts have been found to exhibit normal characteristics for most functional parameters (5). These results suggest the presence of a definite transitional period, which makes HF mice an ideal model for longitudinal studies.

Intensive work has been done on several specific lesions that are known to lead to cardiac disease, including myocardial infarction (MI) (6) and gene overexpression (7). The majority of these studies involve terminal experiments, and thus are limited to focusing on selected time points that precede and follow the intervention, be it surgical or pharmacological. The difficulty of resolving a time course that quantifies the remodeling process throughout the transition from a normal to a depressed state therefore poses logistical problems for terminal experimental techniques, and suggests the use of a nondestructive approach.

In the present study we conducted a longitudinal experiment with (to our knowledge) the highest temporal resolution to date in a genetically modified mouse model of HF, with strong clinical similarities to human cardiomyopathy (4), in an attempt to describe the time course of remodeling during the transition toward failure. We hypothesized that a distinct and detectable time course of remodeling exists, and that MRI has the unique qualities necessary to characterize that time course in high spatial and temporal detail. The HF mice and wild-type (WT) controls were imaged with a weekly protocol from 2 to 32 weeks of age. From these images we derived high-resolution time courses that described the time-varying changes that occurred in chamber volume and myocardial mass as the mice progressed toward overt failure. Our results have demonstrated the ability of MRI to quantify changes in anatomical structure over short intervals of a lifespan and accurately describe a period of rapid ventricular dilatation and wall thinning. A highly resolved temporal imaging protocol may therefore be necessary to detect substantial but quickly occurring transitional phases during cardiac remodeling.

MATERIALS AND METHODS

Mouse Model

HF mice (8) were obtained from a University of California–San Diego (UCSD) colony. The homozygous HF mice were inbred, so WT littermates were not available. The background strain to the HF was a hybrid cross of 129/Sv and C57BL/6 strains. A genetic marker analysis showed that after 20 generations of intercrossing there was an indistinguishable background between the HF and 129/Sv genotypes, which made age-matched 129/Sv mice (Charles River) an appropriate WT control (5). All protocols were performed according to the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the UCSD Animal Subjects Committee.

MRI Study

Animal Preparation and Monitoring

Four HF and four WT mice were imaged longitudinally at weekly intervals from 2 to 32 weeks of age. These eight mice represented a subset of data presented by Costandi et al. (9) on the physiological significance of the methods described here. Anesthesia was chamber-induced with 5 Vol-% inhaled isoflurane in 100% O₂, and then sustained under free-breathing conditions in a custom-built restraint with 1.5 Vol-% isoflurane at 1.5l/min for the remainder of the imaging session. A previous validation study demonstrated that heart rates under isoflurane anesthesia are independent of the mouse strain employed (HF: 432 ± 9.5, WT: 459 ± 14.6, P = 0.11). Physiological monitoring was performed using a custom-designed component-based system integrated with a data acquisition network (LabVIEW 6.1; National Instruments). External copper foil or subcutaneous needle ECG leads were placed on the front fore-limbs for ECG triggering (HSB-T; Rapid Biomedical). The raw ECG traces were preprocessed to remove artifactual currents induced by gradient switching. Triggering thresholds were manually adjusted with gain and low- and high-pass filtering of the signal to accurately detect the ascending limb of the QRS complex. Temperature was tightly controlled by heated airflow through the magnet bore. Core body temperature was monitored with a rectal thermocouple probe, and the bore and near animal temperatures were measured with thermocouples (four probes) distributed on the RF coil. The mice were maintained and observed at 450–550 beats per minute (BPM) and 36–38°C for the duration of the imaging protocol.

MRI

In vivo NMR cardiac imaging was performed on a 7.05T (300 MHz) horizontal-bore MR scanner (Varian, Palo Alto, CA, USA) equipped with a shielded 12-cm bore gradient system capable of 22G/cm gradient strength and a 300-μs maximum rise time (Magnex Scientific, Oxford, UK). A 1.9-cm custom-built dual quadrature-driven transverse electromagnetic mode (TEM) volume coil was used for transmission and reception of the RF signal for all mice in the study (7.3–35.4 g). The coil configuration was designed to allow for external tuning so that the coil could be adjusted while it was loaded with the mouse and monitoring leads, to maximize signal strength.

High-resolution bright-blood MRI experiments were conducted using an ECG-triggered fast low-angle shot (FLASH) gradient-echo (GE) pulse sequence tailored for murine imaging, as previously described (10). The scanning parameters were optimized for the signal-to-noise ratio (SNR) as follows: flip angle = 90°, echo time (TE) = 1.8 ms, repetition time (TR) = ~R-R interval, RF pulse width = 1.0 ms, and number of averages = 5. Since one phase-encoding step was sampled per heartbeat, and sufficient time was available for full T₁ relaxation, the flip angle was set at 90° to maximize the contrast-to-noise ratio (CNR) between blood and myocardium. Fractional k-space acquisition of the echo (62.5%) and the receiver bandwidth (32 kHz) were empirically optimized to reduce the TE while minimizing phase errors and SNR loss, respectively. A custom-designed time-series averaging scheme was employed as follows: for the five averages at each time point, the portion of k-space collected was averaged first, and then the complex conjugate estimation was calculated to fill the remaining uncollected complex points. All images were acquired with a field of view (FOV) of 25 mm² and data matrix of 128² to yield an in-plane resolution of 195 μm².

Each imaging protocol resulted in seven to nine 1-mm-thick short-axis images spanning the whole heart from apex to base with no gap between slices. A prospective gating scheme was used such that images were acquired at a range of time delays relative to the ECG trigger with a temporal resolution of 10 ms. The equatorial frame that contained the largest chamber diameter was selected as the end-diastolic phase of the cardiac cycle. All data presented in this study are from this end-diastolic state. Figure 1 illustrates a representative axial series of HF (h–n) and age-matched control (a–g) mice at about 28 weeks of age. At the base (a and h), the outflow tracts and atria are visible, while toward the apex (g and n) the contiguous texture of the diaphragm can be seen over most of the cross section.

FIG. 1.

Representative axial series for HF and WT controls. Axial images spanning the heart from base to apex in control (a–g) and HF (h–n) mice were used to estimate chamber volume and myocardial mass at end-diastole. At the base the introduction of the atria and outflow tracts is evident, while toward the apex the diaphragm can be seen over most of the cross section.

Image Analysis

Four 2D contours were manually planimetered for each heart at end-diastole (left ventricular (and right ventricular) epicardium and endocardium) from axial sections with the use of image analysis software (ImageJ, NIH). Contour areas were calculated from the calibrated units, multiplied by the slice thickness, and summed together to approximate chamber volumes. LV mass was determined according to the following equation:

$$\text{LV}\text{mass}={\gamma }_{\mathit{myoc}}\mathrm{\sum }(V_{\mathit{epi}}-V_{\mathit{endo}})\times \text{ST}$$

where the specific gravity (γ) of myocardial tissue is 1.055 g/ml, ST is the slice thickness (1 mm), and V_epi and V_endo are the epicardial and endocardial volumes, respectively.

A total of 142 MRI experiments were conducted. For figure clarity, the HF data were binned so that each data point is the average of the nearest two time points (i.e., 3–4 weeks, 5–6 weeks, etc.).

Statistical Analysis

Data are expressed as the means ± SEM. All statistical comparisons were made by means of two-way analysis of variance (ANOVA) followed by Fisher’s protected least significant difference (PLSD) post-hoc analysis over time and genotype. When the ANOVA genotype, time, or interaction terms were significant, a post-hoc analysis was performed on data binned at eight time points: 6, 9, 12, 15, 21, 24, 27, and 31 weeks. Statistical differences of P < 0.05 were considered significant for individual pairs.

RESULTS

A weekly imaging protocol tracked a genetic mouse model of HF and its WT controls from about 2 to 32 weeks of age. Diastolic anatomy was assessed by reconstructing chamber and myocardial volumes from axial images spanning the LV from apex to base. Though the figures presented here are from a single representative HF mouse and an age-matched WT control, the statistical analyses are based on all mice imaged in the longitudinal study (N = 4 HF, N = 4 WT) as described by Costandi et al. (9).

High-resolution bright-blood imaging allowed for the accurate delineation of endocardial and epicardial surfaces. Representative images of failing and normal hearts at end-diastole are shown in Fig. 2. Two time points (an early age of about 11 weeks and a late age of around 28 weeks) are shown for comparison. The WT mice are comparable at both early and late ages, as would be expected for normal hearts at maturity. In contrast to younger hearts, the older HF hearts are dilated. The HF mice have dilated ventricles at both time points, and more myocardial mass compared to controls. This demonstrates the ability of MRI to detect changes in anatomical structure both between groups and within the same group over time.

FIG. 2.

Representative axial images of HF and WT hearts. Note the more-dilated ventricles and larger wall mass in HF vs. WT. Chamber dilatation continues from early (~11 weeks) to late (~28 weeks) ages in HF hearts.

A comprehensive validation study was conducted to assess the ability of MRI to accurately estimate myocardial mass from axial images. Since an accepted standard for approximating chamber volume in the murine model is not available, we used a mass analysis with gravimetry as an appropriate alternative for technique validation. LV wet weights (WW) measured in excised HF and WT hearts at four time points (8, 15, 22, and 31 weeks) were compared with the myocardial masses estimated from MR images. The results are summarized in Table 1. A linear regression analysis showed a good correlation between the measured weights and MR calculations [MR_mass,mg = WW_mass,mg(0.93) + 1.6, R² = 0.89]. In addition to the mass validation, a directly measured geometrical dimension was compared between the excised hearts and MRI. The largest epicardial cross-sectional dimension of the LV was measured from the excised hearts and MR axial images. Linear regression analysis again confirmed the good correlation between measurements [MR_x-sec,mm = Excised_x-sec,mm(1.02) − 0.41, R² = 0.99]. This demonstrates the ability of MRI to accurately predict myocardial tissue volume and geometric dimensions in the rapidly beating mouse heart both within the same group and between groups over time.

Table 1.

MR Validation of Myocardial Mass^*

	Week 8 (mg)	Week 15 (mg)	Week 22 (mg)	Week 31 (mg)
HF
MRI	69.7 ± 1.7 (N = 10)	86.4 ± 3.4 (N = 8)	86.9 ± 4.0 (N = 8)	83.3 ± 3.0 (N = 6)
Gravimetry	70.4 ± 2.1 (N = 8)	89.6 ± 2.4 (N = 8)	87.4 ± 1.6 (N = 10)	85.8 ± 2.8 (N = 13)
WT
MRI	60.0 ± 2.3 (N = 8)	69.3 ± 1.6 (N = 8)	71.7 ± 2.5 (N = 8)	74.7 ± 2.7 (N = 8)
Gravimetry	63.5 ± 2.5 (N = 8)	72.7 ± 2.9 (N = 12)	76.2 ± 3.8 (N = 14)	86.7 ± 4.8 (N = 8)

^*Values are mean ± S.E.M.

Chamber dilatation and wall thinning are hallmarks of the transition toward HF, yet their time course remains unknown. Time-course plots derived from longitudinal MRI revealed a distinct dilatory phase beginning around week 15 and continuing to about week 27. End-diastolic volume (EDV) was significantly increased in HF compared to WT mice at each time point (HF vs. WT: P < 0.01), and ventricular chamber volume was seen to increase by about 75% by week 31 (Fig. 3). Within the HF group the EDV initially increased from 6 (47.01 ± 0.93 μl) to 9 (54.25 ± 1.41 μl) weeks (6 wk vs. 9 wk HF: P < 0.005), most likely as a function of normal growth. The EDV plateaued between 9 and 15 weeks (54.63 ± 1.86 μl), at which point no further increase was observed (9 wk vs. 15 wk HF: P = 0.89). However, it abruptly increased again from 15 to 21 weeks (70.49 ± 3.17 μl; 15 wk vs. 21 wk HF: P < 0.0001), marking a distinct dilatory phase. A second plateau was reached and maintained for the remainder of the study period, and no further increase was measured (31 wk HF: 72.87 ± 3.35μl; 21 wk vs. 31 wk HF: P = 0.49). During this dilatation, wall mass remained constant (Fig. 4), indicating that the endocardial surface was increasing relative to the epicardium, and the wall was thinning. Though HF hearts had more myocardial mass at each time point vs. WT (HF vs. WT: P < 0.05), the only statistically significant increase in mass within the HF group between consecutive time points occurred from 6 (82.97 ± 2.03 μl) and 9 (93.55 ± 2.27 μl) weeks (6 wk vs. 9 wk HF: P < 0.005), again most likely as a function of physiological growth. After 9 weeks, the only difference in HF heart mass measured between any two time points was vs. the 31-week point, supporting the conclusion that mass was constant throughout the dilatation phase.

FIG. 3.

EDV for HF vs. WT mice. Shaded and open symbols are the average of the nearest two time points for each animal at each time point. Note the abrupt dilatation phase in HF hearts, as well as the consistently larger chamber at all time points. Statistical significance refers to the difference in volume between the earliest and latest time points within each bracket for the entire HF data set (N = 4). HF vs. time: * P < 0.005. NS: not statistically significant.

FIG. 4.

Ventricular wall mass for HF vs. WT mice. Shaded and open symbols are the average of the nearest two time points for each animal at each time point. Note the larger HF hearts and the peak in growth at around week 12. Statistical significance refers to the difference in mass between the earliest and latest time points within each bracket for the entire HF data set (N = 4). HF vs. time: * P < 0.005. NS: not statistically significant.

Measures of cardiac volume and mass are typically represented as a ratio that describes the balance between cavity size and myocardial tissue. The chamber volume to chamber plus wall mass ratio, a 3D analog to a radius to wall thickness measurement, is illustrated in Fig. 5. As the volume of the LV increases relative to the epicardial surface, the ratio will approach unity, indicating wall thinning in addition to chamber dilatation. Of particular interest is the abrupt onset and rate of increase in this index. Prior to week 15, no difference in this geometrical ratio was measured between groups (HF vs. WT: P > 0.22); however, after that time the difference was marked (HF vs. WT: P < 0.0002). Within the HF group, this index remained unchanged through week 15 (6 wk HF: 0.375 ± 0.005 μl; 15 wk HF: 0.370 ± 0.005μl; 6 wk vs. 15 wk HF: P = 0.61), but showed a significant increase from weeks 15–24 (24 wk HF: 0.473 ± 0.007 μl; 15 wk vs. 21 wk HF: P < 0.0001; 21 wk vs. 24 wk HF: P < 0.03). A plateau was reached after week 24 and no further increase was seen (31 wk HF: 0.485 ± 0.003μl; 24 wk vs. 31 wk HF: P = 0.34). These findings reveal an abruptly-occurring, brief remodeling phase during which the ratio increased by nearly 40% in the HF group in a 4-week span (from 17 to 21 weeks).

FIG. 5.

Chamber volume to wall mass plus chamber volume ratio for HF vs. WT mice. (Symbol definitions are the same as in Figs. 3 and 4.) In HF mice this ratio is comparable to that in controls up to about week 17, but it quickly increases as a result of chamber dilatation and wall thinning. This marks a clear remodeling phase during which ventricular geometry is altered. Note the brief duration of the increase in this ratio over a <7-week span (weeks 15–22). Statistical significance refers to the difference in μl/(mg + μl) between the earliest and latest time points within each bracket for the entire HF data set (N = 4). HF vs. time: * P < 0.0001. NS: not statistically significant.

DISCUSSION AND CONCLUSIONS

In this study a highly resolved time course of anatomical remodeling was measured in a mouse model of HF. A weekly imaging program revealed an abrupt restructuring phase during which the ventricular chamber dilated and the wall thinned, supporting the hypothesis that MRI can be used to determine the time course of rapidly occurring phases of cardiac remodeling.

The application of noninvasive MR methods to study cardiovascular disease in the murine model has expanded with recent advancements in disease models. Much work has been done to validate the ability of MRI to quantify cardiac anatomy and function in the LV (11) and right ventricle (RV) (12) during both normal growth (13) and hypertrophy (14). However, many of these studies follow traditional experimental designs that employ a minimal number of time points, and thus do not take advantage of the unique ability of MRI to serially track the progression of a disease state through time. Ross et al. (15) conducted such a longitudinal study in an MI model with moderately high temporal resolution (weekly for 2 weeks, and then biweekly for the remainder of 6 months). Although most of the remodeling that resulted from infarction occurred soon after the lesion, the power of serial imaging was clearly demonstrated.

The focus of the present work was to demonstrate the necessity of longitudinal studies for characterizing the evolution of cardiac remodeling. Longitudinal experiments require extensive development; dedicated expertise in MR physics, hardware, and software; specialized resources; and many hours of expensive scanner time. The proposition that such experiments are required to address certain scientific questions must then be justified. Current applications have yet to establish that MR can quantify cardiac remodeling over time in disease progression at high resolution. The necessity of this application is then still an open question in current mouse cardiac MRI and the data presented here may suggest that it is a warranted approach. This work complements a previous study by Costandi et al. (9), who investigated the role of anatomical remodeling in altering chamber compliance and mediating the progression of dilated cardiomyopathy. While that contribution applied MR as a tool to study a specific physiological question, the current article justifies the capability and necessity, of performing longitudinal experiments in the study of rapidly occurring remodeling processes.

The application of MR as a tool to measure such changes in the mouse has two major advantages: it provides greater accuracy compared to other experimental techniques, and makes it feasible to draw statistically significant conclusions.

An alternative to obtaining noninvasive measurements via ultrasound methods or MRI is to conduct ex vivo experiments on excised organs (16). Histological techniques have been extensively used; however, they are hampered by a number of substantial sources of experimental error (17), mainly attributed to variability in loading pressures at the time of cardiac arrest and tissue deformation during preparation and sectioning (18). Nondestructive modes of analysis are therefore preferred. Echocardiography has commonly been applied to the murine model, but is hindered by limitations stemming primarily from low spatial resolution and inaccurate probe placement (19), and 1D or 2D projections that require analytical model assumptions of the LV while making RV measurements problematic. These approximations may be sufficient in normal hearts, but may fail in diseased hearts that present with spatially heterogeneous changes in structure.

Although it is believed that genetically identical HF mice will progress similarly through the remodeling phase, biological variability will result in a population of mice with a distribution in absolute heart size. Quantifying changes in a geometrical ratio, such as that depicted in Fig. 5, with terminal experiments would quickly become unfeasible, since one would have to have a number of mice in order to gather data at many time points. Statistical power in longitudinal MR studies is instead derived from relative changes in a measure over time within the same animal rather than among different animals.

Whether biological processes are assessed by ex vivo terminal experiments or noninvasively in vivo, a fundamental concept persists that such processes can occur quickly and thus require high temporal resolution to characterize, particularly in rodents. Figure 5 clearly suggests that undersampling of the data in time can have significant consequences for attempts to accurately characterize the remodeling process (since most alterations apparently occur over a 6-week interval beginning at around week 17) and to isolate the time of onset in this transitional phase. Though an understanding of the physiology that underlies change may still be gained from the endpoints of such curves, often it is the rate at which changes occur that is of importance (20,21). The identification of these events by noninvasive means may then warrant further analysis at these now-established time points of interest.

In the present study, serial MRI was used to characterize a phase of rapid change in geometry within the LV. Figure 6 illustrates a schematic coronal section through a normal and a failing heart scaled to axial dimensions measured in a 24-week-old HF mouse and its age-matched control. Both ventricular cavities are enlarged, while both the septal and LV free wall are thinned in the HF heart. Chamber dilatation and wall thinning have a direct effect on the passive mechanics (22), chamber compliance (23), wall stress (24), and fiber strain (25) in the heart, and have been closely associated with a number of abnormalities leading to the clinical characterization of HF (26 –28). Animal models of cardiac disorder (particularly genetically alterable mouse models) that present with the structural characteristics consistent with those found in humans are therefore an indispensable tool for studying clinically relevant modes of cardiac failure. The imaging techniques available to conduct anatomical and functional cardiac studies must be scaled to these models both spatially and temporally, and must be capable of quantifying quickly-occurring transitional phases in time. The MR techniques applied in this study meet these requirements with 195-μm in-plane spatial and 10-ms temporal resolution employed in a weekly imaging protocol over an 8-month period.

FIG. 6.

Schematic coronal sections through WT (a) and HF (b) hearts scaled from MR images at 24 weeks. Chamber dilatation is evident in both ventricles, while wall thinning is apparent in the LV septal and free wall. The imaging plane identifies the location of the MR axial cross section, where measurements were taken to scale the drawings. LV, left ventricle; RV, right ventricle; LA, left atrium; RA, right atrium; MV, mitral valve; TV, tricuspid valve; AV, aortic valve; PV, pulmonary valve; SW, septal wall; FW, free wall; IP, imaging plane.

The limitations of the proposed methods are typical for fast imaging of the murine anatomy (29). The isolation of the end-diastolic state is a function of temporal resolution through the cardiac cycle, which was ±10 ms. It is often difficult to acquire an appropriate ECG trace from which to trigger the pulse sequence in mice, particularly with short TRs and high gradient slew rates (30), which further complicates the isolation of end-diastole. Also, since resolution in the longitudinal direction is limited by slice thickness, volume calculations may be affected. Lastly, although the endocardial surface is easily detectable due to the high contrast between blood and myocardium, the epicardial surface is more difficult to delineate, particularly where the anterior wall of the ventricle approaches the chest wall. However, the good correlation established by the validation studies reflects the small magnitude of these errors.

In summary, we were able to quantify a rapidly-occurring remodeling phase in a mouse model of HF using a high temporally and spatially resolved MRI protocol. The observed dilatation and concomitant wall thinning emphasize the presence of biological events that are of short duration but have a substantial impact on physiology. Furthermore, these data suggest that noninvasive tools capable of characterizing a 3D geometry are required to accurately describe these transitional phases during cardiac remodeling processes.