Inter-individual differences in response to medication are well-established clinical issues, the study of which is a large field of research in itself. Yet, pharmacogenetic studies in humans are hampered by several important concerns such as compliance issues, environmental confounders and limitations to non-invasive measurements. In this context laboratory mice, which have already played a key role for dissecting the genetic basis of disease susceptibility and providing key support for many current concepts of disease pathogenesis and treatment strategies, are also likely to become instrumental for complementing and furthering human pharmacogenetics 
. Indeed, several recent studies already addressed the genetic basis of strain-specific susceptibility to carcinogens 
and analgesics 
or the metabolism of warfarin, testosterone and irinotecan 
in inbred mice.
Here we undertook a similar approach focusing on the cardiovascular system. Our large survey assessed the suitability of using a large panel of inbred mouse strains to investigate physiological responses to β-adrenergic treatments. To the best of our knowledge, it provides the first large-scale standardised characterisation of heart rate, systolic blood pressure, ECG and cardiac weight indices in response to sustained infusion of atenolol and isoproterenol in such a panel.
The robustness of our experimental standards is supported by several essential validation steps: first, mean strain values measured in ctr
condition correlated well with independent datasets of the Mouse Phenome Database. In agreement with the notion that morphological parameters are highly heritable and least affected by environment, traits related to body and cardiac weight were the most consistent with independent studies. Even though conservation across independent datasets was much reduced for SBP and HR, the number of significant between-projects correlations was higher than expected by chance for both traits. As discussed by others 
, this confirms that despite marked susceptibility to environmental factors, a significant portion of HR and SBP variance observed in inbred mouse strains is genetically determined. Second, phenotypic variances were generally smaller within than between strains, indicative of elevated trait heritabilities (i.e.
values were indeed significantly higher than in human populations, in particular for SBP, HR and ECG intervals (see for instance 
), where H2
estimates typically vary around 0.6 in twin studies and around 0.25 in nuclear families. This further emphasises the potential of using our mouse model instead of traditional human cohorts for downstream genome-wide association scans. Third and regardless of the treatment conditions, strain means for all phenotypes had uni-modal distributions, consistent with the idea that they are complex traits under the control of multiple genetic loci.
We found that cardiovascular responses to drug exposure were trait-, drug-, and dose-specific. In order to obtain a comprehensive and unbiased overview of the structure of the phenotypic data with respect to the relationships across strains, treatments, and phenotypes, we performed a comprehensive comparative investigation. Our unsupervised bi-clustering analyses facilitated processing the large phenotypic data in order to extract key information regarding strain proximity and drug-dependent phenotypic perturbations. For instance our clustered matrices of significance score made immediately apparent that phenotypic changes were usually more significant under β-stimulation than β-blockade (consistent with iso and ate being administered to healthy animals housed in a reduced stress environment). When considering all experimental conditions, HR was the most, and SBP and QTc the least affected phenotypes in response to both pharmacological compounds. Strain responses of phenotypes related to body and heart weight (i.e. HW, AW, AWI, VW, VWI, BWE, BWG) were well correlated with drug-induced changes of heart rate and ECG intervals (in particular PR, Pdur, QT and ST), but not with the responses of systolic blood pressure and ECG wave amplitudes or areas. Trait-specific patterns of sensitivity to iso ranged from maximal response under treatment with iso1 for HR to negligible effects under either concentrations of iso for SBP. We detected compartmental and strain-specific cardiac sensitivity to iso, with atria responding at lower concentrations than ventricles in the majority of the strains. At this stage, the biological mechanisms underlying differential sensitivities to chronic β-stimulation are not known but might reflect distinct and strain-specific distributions of atrial and ventricular β-adrenergic receptors and/or differential downstream signalling pathways. Altogether, these data suggest that responses to β-adrenergic drugs by themselves are determined by complex genetic architectures. Beyond the specific context of the present study, it is worth mentioning that the unsupervised bi-clustering analyses we employed in this study could also be useful for dissecting other large datasets of complex phenotypic traits, which are likely to become more and more frequent.
Due to their broad use to generate transgenic or knock-out models, the lines C57BL/6J and 129S1/SvImJ are of particular interest. Our comprehensive analyses as well as other reports 
have clearly highlighted their differential behaviours for baseline phenotypes such as SBP, atrial and ventricular size, heart rate, heart rate variability and cardiac metabolism. Recently, Barrick et al
. further emphasised on the specificity of 129S1/SvImJ and C57BL/6J responses to prolonged trans-aortic constriction (TAC), another model of left ventricular hypertrophy associated with increased hemodynamic load and sustained β
-adrenergic stimulation 
. Upon aortic banding, C57BL/6J mice had an earlier onset and more pronounced impairment in contractile function, with corresponding left and right ventricular dilatation, fibrosis, change in expression of hypertrophy markers, and increased liver weights at five weeks post-constriction. In contrast, 129S1/SvImJ mice had delayed transition to decompensated heart failure, with relatively mild alterations in histology and markers of hypertrophy at five weeks post-TAC and preserved systolic function until eight weeks post-TAC, suggesting that 129S1/SvImJ genetic modifiers might protect against the earlier and more severe pathological changes seen in C57BL/6J mice 
. In contrast, in our study, iso10
-mediated increase of VWI was marked in strain 129S1/SvImJ while C57BL/6J mice were significantly more resistant. This indicates that different signalling pathways may lead to cardiac hypertrophy, depending on the type of upstream trigger (i.e.
pressure overload vs
adrenergic activation). These observations put an additional emphasis on the need for detailed characterising of these and other strains, both in terms of baseline cardiovascular phenotypes and in response to pressure overload or β
Future work will aim at associating the rich phenotypic data of this study with genetic markers. Indeed the vast majority of the strains of our CV-PGX panel has been extensively genotyped 
. Early studies already suggested that the genome of laboratory mice is a mosaic of regions of distinct but limited sub-specific origins 
. More recent resequencing efforts have refined this picture by cataloguing over 8 million SNP alleles across 16 inbred strains 
and genetic maps of similar densities were further imputed in 49 strains 
. These invaluable resources allow for inferring the ancestry of most of the genome with good confidence and provide mapping resolution usually higher than in other model organisms. Thus, it is now becoming feasible to map trait variation in inbred mouse strains by testing association to loci of inferred ancestry 
. Here we already used genomic SNP profiles to show that overall genetic similarity of the strains exhibits little concordance with their phenotypic relatedness in terms of the collection of cardio-vascular traits we measured. This indicates that cardiovascular phenotypes are unlikely to segregate according to global phylogeny, but rather be governed by smaller, local differences in the genomes of the various strains. Thus, given the significant heritability of many of these traits, we are confident that association studies using our phenotypic resource have good chances to reveal new candidate loci related to differential cardiovascular responses under treatments.