Left ventricular hypertrophy is one of the main risk factors for cardiovascular mortality and morbidity. It has been proposed that hypertrophic stimuli act in great part by increasing the size of cardiomyocytes, and that the latter characteristic is a necessary condition to differentiate left ventricular hypertrophy from other benign forms of cardiac enlargement. To test whether the same genetic loci control the size of cardiomyocytes and left ventricular mass, we performed whole genome linkage analyses in a panel of 24 recombinant inbred AXB/BXA mouse strains. Whereas one major locus was linked to left ventricular mass in both males and females, loci linked to the size of cardiomyocytes were clearly distinct and showed sex-specific linkage. Moreover, the parental origin of chromosome Y had strong effects on the size of cardiomyocytes in male mice but did not affect left ventricular mass. In addition to showing that genetic loci that increase the size of cardiomyocytes are not necessarily linked to increased left ventricular mass, our findings have important consequences in evaluating cardiac phenotypes when performing genetic manipulations in mice, and in determining the cause of sex-specific differences when using models derived from C57BL/6J mice.
- cardiac hypertrophy
- cardiovascular genomics
- linkage analysis
- quantitative trait locus
numerous clinical studies have repeatedly and convincingly shown that increased left ventricular (LV) mass (LVM) is highly predictive of cardiovascular morbidity and mortality (4, 10, 15, 30, 41). Inappropriately high LVM is known as left ventricular hypertrophy (LVH). However, it is difficult to set a threshold that differentiates benign cardiac enlargement from maladaptive LVH, because cardiovascular risk increases continuously and proportionally with LVM over a wide range of LVM values, even those extending below what had often been considered as the “upper normal” limit (10, 41). Another difficulty relates to the fact that changes in LVM may result from differences in the number (hyperplasia) or size (hypertrophy) of cardiomyocytes (CMs), in the number of non-CM cells (primarily fibroblasts), and/or in the composition of the extracellular matrix (17). Consequently, there is no simple definition of what LVH corresponds to at the cellular level. The fact that most signaling pathways that lead to LVH have been identified on the basis of their effect on the size of neonatal rat cultured CMs suggests that increased size of CMs is an important feature of LVH (23). Consequently, it has been proposed that any phenotypic characterization of LVH should test whether CMs are enlarged, and that increased size of CMs is one essential feature that distinguishes LVH from other benign forms of cardiac enlargement (17, 21). A natural extension of this postulate is that genes and/or pathways leading to LVH are the same as those that affect CM size. This has led in part to the common practice of measuring CM size to determine whether increased LVM in genetically modified mice truly corresponds to LVH (17).
Genetic crosses between normotensive rodents (to date, mostly rats) have made it possible to identify loci and/or genes linked to LVM independently of blood pressure (5, 13, 25, 31, 42, 46). However, the extent to which genes controlling LVM overlap with those controlling the size of CMs in such crosses has never been verified systematically. This is difficult to test in genetic inter- or backcrosses, because precise measurement of the size of CMs requires their isolation from adult hearts (20) and this procedure is not compatible with measurements of LVM. An alternative tool for linkage studies is that provided by panels of recombinant inbred (RI) strains, where each strain contains a particular combination of genes from the two parental strains consolidated by inbreeding (11). This makes it possible to perform multiple phenotyping, since all individuals within a particular strain share the same genotype. For instance, the panel of AXB/BXA RI strains has been constructed from reciprocal crosses between A/J and C57BL/6J mice (34). Although both strains are normotensive, it has been reported that LV from C57BL/6J mice harbor an eccentric hypertrophy compared with A/J counterparts (24). We therefore used this panel to perform linkage analyses and compare the genetic pathways that control LVM and the size of CMs.
MATERIALS AND METHODS
All procedures on animals were approved by the Institut de Recherches Cliniques de Montréal (IRCM) Institutional Animal Care Committee and conducted according to guidelines issued by the Canadian Council on Animal Care. Breeding trios for the parental A/J and C57BL/6J strains, the chromosome-substitution (consomic) A/J-chrYC57BL/6J/NaJ and C57BL/6J-chrYA/J/NaJ strains, and the RI AXB/PgnJ and BXA/PgnJ strains were obtained from the Jackson Laboratory (Bar Harbor, ME). For simplicity, all RI strains and consomics are identified without the PgnJ or NaJ suffix, respectively. Because females from some RI strains (BXA1, BXA2, BXA12, and BXA26) are known to be poor mothers, they were substituted with ovariectomized C3SnSmn.CB17-Prkdcscid/J females implanted with ovaries from the corresponding RI strains by the Jackson Laboratory technical services and mated to males from the cognate RI strain. All animals were then bred in the IRCM pathogen-free animal facility. To test for parent-of-origin effects, reciprocal F1 progenies were generated by mating either A/J males with C57BL/6J females (B6AF1) or C57BL/6J males with A/J females (AB6F1). According to a published note from the Jackson Laboratory (2), some nominally independent AXB/BXA RI strains are near-duplicates. Thus we chose only one sister strain per duplicate group: AXB13 [related to AXB14 (now called AXB13a)] and AXB18 [now called AXB19a, related to AXB19 and AXB20 (now called AXB19b)]. A total of 13 AXB (AXB1, 2, 4, 5, 6, 8, 10, 12, 13, 15, 18, 23, and 24) and 11 BXA (BXA1, 2, 4, 11, 12, 13, 14, 16, 17, 24, and 25) were used in the present study. The BXA26 strain was discarded because of breeding problems. All animals were used at 12 wk of age for phenotyping procedures.
Tissues were collected from 7–11 male and female individuals for each strain (parental, consomic, or RI) or F1 cross. On the day of tissue collection, each mouse was weighed for determination of whole body weight (BW) and killed by cervical dislocation, and several tissues (heart, thymus, adrenal glands, and kidneys) were collected. Heart ventricles were dissected into right ventricle (RV, without the septal wall) and LV (including the septal wall). Each tissue was blotted dry and weighed individually. Relative tissue weights were then calculated for LV (LVW/BW), RV (RVW/BW), thymus (TW/BW), both adrenal glands (AW/BW), and both kidneys (KW/BW).
Cardiomyocyte isolation and videomicroscopy.
CMs were isolated from four or five male and female individuals for each strain (parental, consomic, or RI) or F1 cross. After anesthesia was induced the heart was quickly removed from each mouse and mounted for cardiac retrograde aortic perfusion, and CMs were isolated as described previously (47). After the last Ca2+ restoration, cells were fixed for 10 min in 80 mM phosphate buffer containing 1.5% glutaraldehyde, a solution shown to preserve all dimensions of isolated CMs (22). After fixed cells were pelleted by gravity for ∼10 min, they were resuspended in 100 mM phosphate buffer and examined with an Axiovert S100TV microscope (Zeiss) with slight modifications of our previously published procedure (14). In short, the profiles of cells were visualized by autofluorescence (excitation 540 nm, emission 620 nm), and their image was captured with a Retiga EXi monochrome camera (Qimaging, Surrey, BC, Canada). The area, length, width, and width-to-length ratio of ∼100 CMs per heart were measured with the public domain Java image processing program ImageJ (40).
Blood pressure measurements.
To measure diastolic blood pressure (DBP), systolic blood pressure (SBP), mean arterial pressure (MAP), pulse pressure (PP) and heart rate (HR) in conscious unrestrained mice, three male and female individuals from each strain (parental or RI) or F1 cross were implanted under isoflurane anesthesia with telemetry devices (Data Sciences International, St. Paul, MN). The catheter of the implant was inserted into the aortic branch through the left carotid and maintained by two ligatures, and the body of the transmitter was inserted into a subcutaneous pocket at the right inguinal level. Animals were allowed 1 wk to recover from surgery, and all measured values were then recorded for 48 h. Males from the BXA14 and BXA24 strains were discarded because of consistent postimplantation behavioral problems.
Genotype data [either single nucleotide polymorphisms (SNPs) or simple sequence length polymorphisms (SSLPs)] were obtained from GeneNetwork (www.genenetwork.org) and edited to remove unused strains and loci with identical strain distribution patterns. Undefined genotypes in sister strains were edited with original data from the Wellcome Trust Center for Human Genetics (www.well.ox.ac.uk/mouse) and the Mouse Genome Informatics (www.informatics.jax.org). A total of 965 autosomal and 29 X-specific polymorphic markers were used. None of the quantitative trait values was transformed, because they all conformed to the normal distribution according to the D'Agostino test for normality. Genomewide scans were performed separately in males (with only autosomal markers) and females (with both autosomal and X-linked markers).
Analyses were done in three stages as previously described (43), using the R/qtl package (6). First, main quantitative trait locus (QTL) effects were detected by performing single-QTL genome scans. Significance thresholds were set at the P < 0.63 (suggestive) and P < 0.05 (significant) levels as suggested previously (28, 33). Thresholds were calculated on the basis of permutation tests (1,000 permutations) performed on the data sets themselves (16). Confidence intervals (CIs) were determined by calculating the 1.5-log of odds ratio (LOD) support interval, as recommended recently (32). On the basis of the values of CM size obtained in reciprocal F1 and consomic animals (Fig. 1), the genome scans for CM size components were performed in males by defining the parental origin of chromosome Y (chr Y) as a covariate in either additive or interactive models. The interactive model was used when it yielded significant QTLs whose LOD score differed from the scores calculated by the additive model by a value >2.0 (27). The second step consisted of performing a two-QTL, two-dimensional genome scan for all traits to detect potential epistatic interactions. Because single-QTL analysis is based on its marginal effect and does not take into account other loci, the third step used a multiple regression model to reevaluate the statistical linkage of each locus to the phenotype of interest in the context of all other loci (identified as either QTLs or suggestive loci by the single-locus scans). With this analysis, terms that fail to meet significance thresholds under the new criteria are eliminated one at a time, and the analysis is repeated until all remaining loci are significant. This method also calculates the contribution of each significant locus to the total variance, based on adjusted (type III) sums of squares, and the percentage of the total variance is accounted for by all combined loci.
Differences between groups were evaluated by t-tests or one-way ANOVA followed by Keuls post hoc analysis. Correlations between phenotypes were evaluated by calculating r2 values on the basis of linear regression. All calculations were performed with GraphPad Prism (GraphPad Software, San Diego, CA).
In parental strains, LVW/BW (but not BW or RVW/BW) was higher in C57BL/6J mice than in A/J counterparts; likewise, the surface area of CMs from C57BL/6J was greater than that of A/J counterparts (Table 1). Both strains were normonsive, with MAP and SBP being slightly more elevated (by ∼10 mmHg) in male (but not female) C57BL/6J mice. Phenotypic values for other traits are compiled in the supplemental data for this article (Tables S1 and S2).1
Analysis of the reciprocal AB6F1 and B6AF1 crosses revealed a strong parent-of-origin effect on the size of CMs in males, but not in females (Table 1 and Fig. 1A). To test the potential contribution of chr Y from each strain, we used reciprocal consomic strains C57BL/6J-chrYA/J [C57BL/6J mice carrying the nonpseudoautosomal region (NPAR) of chr Y from A/J] and A/J-chrYC57BL/6J (A/J mice carrying the NPAR of chr Y from C57BL/6J) and tested whether substituting the NPAR of chr Y between the C57BL/6J and A/J strains affected the size of CMs. We found that the presence of chr Y from C57BL/6J into pure (C57BL/6J or A/J) or mixed (F1) genetic backgrounds was associated with increased size of CMs compared with counterpart strains carrying chr Y from A/J (Fig. 1A). In contrast to CM size, LVW/BW did not differ between consomic strains and their corresponding host strain (data not shown), or between reciprocal F1 animals (Table 1).
Distributions of the trait values among all 24 RI strains are shown in the supplemental data for this article (Tables S1 and S2). All strains were normotensive, but there was a significant correlation of the LVW/BW values with those of telemetric MAP; however, the values of the correlation coefficients (r2 = 0.23, P < 0.05 in males; r2 = 0.27, P < 0.01 in females) indicated that MAP was not responsible for >23–27% of the variance of the LVW/BW values (Fig. 1B). In contrast, there was no correlation between CM size and MAP. Of note, CM surface area is a very robust and precise phenotype. Accordingly, the intrastrain coefficient of variability of CM surface area averaged ∼3% in either males or females, whereas CM surface area in strains with the largest CMs was up to 42% higher than in strains with the smallest CMs. Moreover, despite the fact that LVW/BW and CM size were both greater in C57BL/6J than in A/J, the two traits segregated independently when examined in all 24 RI strains (Fig. 1B).
After the single-locus scan, we detected significant QTLs for LVW/BW and CM size components (Fig. 2, Table 2), but not for any of the other phenotypes measured, including blood pressure values and normalized weight of other tissues (data not shown). For LVW/BW, we detected one major QTL (named QTL Lvm1 for left ventricular mass QTL 1) showing linkage in both sexes (Figs. 2 and 3). Interestingly, the A/J allele at the closest marker to Lvm1 was linked to the highest phenotypic value, despite the fact that the A/J parental strain had a lower LVW/BW than C57BL/6J. QTLs linked to CM size components were different in males and females and clearly distinct from Lvm1 (Fig. 2). In females, one major QTL on chr 14 (named Cms1 for cardiac myocyte size QTL 1) was linked to both CM area and length (Figs. 2 and 3). A peak showing the same profile at the same position also showed suggestive linkage to CM width. This suggested that both length and width contributed to the variance of CM area, as confirmed by the multiple regression model (see below). In females, another QTL on chr 5 (named Cms2) showed significant linkage to CM width only. In males, given that the origin of chr Y had a strong effect on CMs from either reciprocal F1 or consomic strains (Fig. 1A), genome scans for CM size components were performed by defining the parental origin of chr Y as a covariate. With the additive model, one major QTL (named Cms3) showed significant linkage to both CM area and width (Fig. 2). Although found on chr 5, the profile and location of Cms3 in males was clearly distinct from that of Cms2 in females (Fig. 3).
The two-QTL, two-dimensional genome scan did not detect any epistatic interaction between the autosomal QTLs in both males and females. However, to test possible interactions of QTLs with chr Y in males, we repeated the single-locus scan by setting the origin of chr Y as an interactive covariate. This analysis revealed two additional QTLs that were not detected in the previous analysis: Cms4 on chr 10 (linked to CM area and width) and Cms5 (linked to CM length) (Fig. 4). All the above QTLs (along with their positions, LOD score, and phenotypic effects) are summarized in Table 2.
To reevaluate the importance of chr Y and the single-scan QTLs in the context of all other loci (either significant or suggestive), build the final linkage model, and calculate which percentage of the total variance is accounted for by all loci, all data were recalculated with the multiple regression model. The results (Table 3) show that 1) both in males and females, Lvm1 contributed to a large part of the variance (∼31%) of LVW/BW; 2) in females, Cms1 and Cms2 each contributed to a large part of the variance of CM size components; 3) some loci other than QTLs identified by single-locus scans contributed to part of the variance of LVW/BW (both in male and females) and CM size components (in females); 4) in males, chr Y (as well as its interactions with Cms4 and Cms5) contributed to a large part of the variance of CM size components (ranging from 15% to 32%); and 5) other QTLs contributing individually to CM size components in males were Cms1, Cms2, and Cms3.
The present study allowed us to identify for the first time natural genetic variants that control CM size in mice. Most of these loci were sex specific and, importantly, were different and distinct from the major QTL Lvm1 that controlled LVM in both sexes. The distinct influence of loci controlling LVM and CM size was reinforced when examining the effects of the origin of alleles on the phenotypes. Indeed, strains with AA genotypes for Lvm1 had greater LVM than their BB counterparts, whereas the CM strains with BB genotypes at all Cms loci had increased size compared with their AA counterparts. Interestingly, regulation of the width and the length of CMs has been suggested to correspond to different biological processes controlled by different regulation pathways, with increased width corresponding to the addition of sarcomeres in parallel, while increased length corresponds to the addition of sarcomeres in series (3, 38). This notion is compatible with our findings showing that (in addition to the effects of chr Y) 1) in females Cms1 was linked to both area and length of CMs, whereas Cms2 was linked only to CM width, and 2) in males Cms 3 and Cms4 were linked to both area and width of CMs, whereas Cms5 was linked only to CM length. Others had suggested that beyond absolute values, the width-to-length ratio of CMs was a very conserved and tightly regulated phenotype (20). However, in contrast to absolute values of CM size, we did not detect in the present study any QTL linked to this relative ratio.
None of the QTLs identified in the present study showed any obvious overlap with QTLs previously reported (either by us or by others) as being linked to LVM, either in mice or in syntenic regions in other species, nor did we identify within these QTLs any obvious candidate genes on the basis of their known function. However, the latter strategy suffers from limitations, as it has been shown that it may distract from identification of genes that are truly linked to a trait but could not have been suspected on the basis of their known functions (39). Identification of culprit genes will therefore need to wait until congenic lines further narrow the regions of interest.
Of note, others had previously used the same panel of AXB/BXA RI mice to acquire cardiac phenotypes by M-mode echocardiography and estimate LVM, but these authors did not report on association of a QTL with LVM (36). However, we have verified that there was no correlation between the values of LVW/BW acquired by ourselves with the values of LVM estimated on the basis of noninvasive echocardiography. We postulate that, despite its utility in evaluating LV function and morphology in the very contrasted parental strains, this method does not have sufficient resolution to detect smaller differences in LVM in the RI strains that constitute this panel. In contrast to LVW/BW, the values we obtained for telemetric blood pressure were very similar to those published by others using the same method and the same parental C57BL/6J and A/J strains (35). However, we did not detect any QTL linked to any component of blood pressure, and the latter cannot therefore account for any of the phenotypic differences in LVW/BW or CM size.
It is well established that there are important differences between males and females in cardiac morphology, function, and disease progression, both in humans and in animal models (9, 12, 18, 29). From a genetic standpoint, sex-specific differences in linkage from various studies do not manifest themselves in the same fashion for all loci. For instance, in a previous study with rats, QTLs linked to LVM in males were different from those linked to the same phenotype in females (31). This is in contrast to the present study, where there was a perfect match of the linkage profile of Lvm1 in both male and female mice. From a phenotypic standpoint, sex-specific differences concern many aspects of cardiac physiology and are more the rule than the exception (9, 12, 18, 29). Although these differences may arise from multiple mechanisms (including differential effects of sex steroids), there has been evidence that chr Y carries genes that may affect cardiovascular outcome (8). Accordingly, chr Y variants have been linked to blood pressure in rats (19) and humans (44), as well as to cholesterol levels in the human population (7). However, the mechanism by which chr Y associated to these phenotypes has always been elusive and has never been linked to a phenotype to the level of a well-defined cell type. In our study, we found that chr Y accounted for a large part of the variance of CM size, both in the background of parental strains and in the mixed background of RI strains. The interactions of the responsible genes on chr Y with that of other QTLs (and how the latter differ from those linked to LVM) are summarized in part in Fig. 5. Outside of their effect on sex determination and germ cell function, the function of genes on chr Y are still poorly understood (26). It has been reported that chr Y from most mouse laboratory strains (including A/J and C57BL/6J) all originated from common male Mus musculus molossinus ancestors (37). However, it was reported recently that chr Y polymorphisms in A/J and C57BL/6J mice are capable of modifying autoimmune disease susceptibility (45), indicating that some genetic drift in chr Y must have occurred. Accordingly, recent data indicate that ∼10% of all 4,935 identified SNPs on chr Y are polymorphic between A/J and C57BL/6J (1). Further experiments on the mechanisms responsible for the effect of chr Y on CM size might provide an important adjunct for understanding how genes on chr Y affect cardiovascular outcome.
Importantly, there is an ever-growing list of examples where the effects of gene manipulations in mice (resulting from knockout, overexpression, or dual targeting) on cardiomyopathy phenotypes differ between male and females (18). These differences concern survival, the level of hypertrophy and/or fibrosis, as well as dysfunction (18). The mechanisms responsible for these sex-specific differences have often not been resolved, but C57BL/6J mice are the preferred genetic background for most genetic manipulations. We do not know yet whether chr Y (beyond its effect on CM size) also affects cardiac function, particularly in the context of disease and perturbed genetic regulation. However, our data suggest that, at the very least, sex-specific differences in the C57BL/6J background should not be generalized to all mouse strains.
In summary, our data show that genetic loci that increase the size of CMs do not necessarily lead to increased LVM, and that different loci can even influence both phenotypes independently from each other. Consequently, one cannot assume that genetic modifications in animals should automatically alter both phenotypes to the same extent. We also found that chr Y from C57BL/6J harbors allelic variants that have strong effects on the size of CMs. Since C57BL/6J mice are the preferred background for most genetic manipulations, the origin of chr Y should be assessed when genetic manipulations affect phenotypes in a sex-specific fashion in models derived from C57BL/6J mice.
This work has been supported by National Heart, Lung, and Blood Institute Grant HL-69122 and by Canadian Institutes for Health Research Grant MOP-64391.
↵1 The online version of this article contains supplemental material.
Address for reprint requests and other correspondence: C. F. Deschepper, Experimental Cardiovascular Biology Research Unit, Institut de Recherches Cliniques de Montréal (IRCM), 110 Pine Ave West, Montréal, QC, Canada H2W 1R7 (e-mail:).
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).
- Copyright © 2007 the American Physiological Society