Analysis of Cre-mediated genetic deletion of Gdf11 in cardiomyocytes of young mice

Abstract

Administration of active growth differentiation factor 11 (GDF11) to aged mice can reduce cardiac hypertrophy, and low serum levels of GDF11 measured together with the related protein, myostatin (also known as GDF8), predict future morbidity and mortality in coronary heart patients. Using mice with a loxP-flanked (“floxed”) allele of Gdf11 and Myh6-driven expression of Cre recombinase to delete Gdf11 in cardiomyocytes, we tested the hypothesis that cardiac-specific Gdf11 deficiency might lead to cardiac hypertrophy in young adulthood. We observed that targeted deletion of Gdf11 in cardiomyocytes does not cause cardiac hypertrophy but rather leads to left ventricular dilation when compared with control mice carrying only the Myh6-cre or Gdf11-floxed alleles, suggesting a possible etiology for dilated cardiomyopathy. However, the mechanism underlying this finding remains unclear because of multiple confounding effects associated with the selected model. First, whole heart Gdf11 expression did not decrease in Myh6-cre; Gdf11-floxed mice, possibly because of upregulation of Gdf11 in noncardiomyocytes in the heart. Second, we observed Cre-associated toxicity, with lower body weights and increased global fibrosis, in Cre-only control male mice compared with flox-only controls, making it challenging to infer which changes in Myh6-cre;Gdf11-floxed mice were the result of Cre toxicity versus deletion of Gdf11. Third, we observed differential expression of cre mRNA in Cre-only controls compared with the cardiomyocyte-specific knockout mice, also making comparison between these two groups difficult. Thus, targeted Gdf11 deletion in cardiomyocytes may lead to left ventricular dilation without hypertrophy, but alternative animal models are necessary to understand the mechanism for these findings.

NEW & NOTEWORTHY We observed that targeted deletion of growth differentiation factor 11 in cardiomyocytes does not cause cardiac hypertrophy but rather leads to left ventricular dilation compared with control mice carrying only the Myh6-cre or growth differentiation factor 11-floxed alleles. However, the mechanism underlying this finding remains unclear because of multiple confounding effects associated with the selected mouse model.

INTRODUCTION

Growth differentiation factor 11 (GDF11) is a member of the transforming growth factor-β (TGF-β) superfamily, best known for its morphogenic roles during development (14). The role of GDF11 in cardiac aging has been controversial (9, 21, 24, 25). We initially demonstrated that GDF11 is a circulating factor that, when administered to aged mice, can decrease age-related cardiac hypertrophy as indicated by a decrease in heart weight-to-tibia length ratios (12). Others have shown that GDF11 administration may be beneficial following myocardial infarction (6). However, following our initial report, contrasting work indicated that GDF11 administration to aged mice does not affect heart weight-to-body weight ratios (24) or that GDF11 supplementation impairs cardiac function in concert with reducing cardiomyocyte size and cardiac mass (21). Furthermore, additional controversy has been raised because of a lack of specificity of antibody and aptamer reagents used to detect and quantify GDF11 separate from the closely related protein myostatin (also known as GDF8 and encoded by the gene Mstn). Because GDF11 and GDF8 share ~90% identity in their active domains, it has been challenging to quantify levels of each protein independently (7).

Human clinical data indicate that low serum levels of the GDF11 + myostatin pool at study entry predict increased mortality in adult patients with heart disease over the subsequent eight years (16). These human data point to low levels of the circulating pool of GDF11 + myostatin as a potential pathogenic factor in human heart disease. However, conflicting human data have also been published suggesting that higher levels of GDF11 in adults with severe aortic stenosis are associated with an increased risk of adverse events following valve replacement surgery (22). Thus, further studies are needed to understand how perturbations to the GDF11 and myostatin system might contribute to cardiovascular risk.

The effect of GDF11 deficiency on the heart also remains unclear, but prior studies have evaluated the effect of myostatin deficiency on the heart. Mstn^−/− mice are viable and have increased skeletal muscle weight as well as increased heart weight and heart weight to tail length, but they show no increase in heart weight-to-body weight ratio (10). Mstn^−/− mice also have mildly increased left ventricular volumes and mildly reduced systolic function at baseline but have an enhanced response to isoproterenol stress with a higher percent change in fractional shortening compared with wild-type mice (10). Inducible cardiomyocyte-specific deletion of Mstn leads to transient chamber dilation and impairment of systolic function; however, upregulation of Mstn in noncardiomyocytes occurs, leading to an unexpected overall increase in Mstn expression in whole heart extracts of knockout mice (3). Conversely, overexpression of Mstn leads to decreased heart weights in males but not females (20). Despite similarities in sequence between mature myostatin and GDF11, there are structural differences between the two proteins that could lead to distinct effects with deletion of Gdf11 in cardiomyocytes that have not previously been described with cardiomyocyte-specific Mstn deletion (25).

We sought to test the effects of reduced levels of GDF11 in the heart during young adulthood in mice. Because systemic germline deletion in the mouse (Gdf11^−/−) results in a number of developmental defects and is lethal by 1 day of age, likely because of renal agenesis (14), we generated a mouse model to specifically delete Gdf11 only within cardiomyocytes. We chose to delete Gdf11 from cardiomyocytes for our initial study as opposed to other cardiac cell types given the demonstrated phenotype seen with cardiomyocyte deletion of Mstn and challenges of deleting Gdf11 only in the heart if targeting other cell types. By crossing a cardiomyocyte-specific Myh6-cre allele with a Gdf11-floxed allele (13), we were able to induce genomic excision of the regions encoding mature GDF11 protein exclusively in cardiomyocytes. Our experiment was designed to test the hypothesis that postnatal Gdf11 deficiency in cardiomyocytes would promote cardiac hypertrophy. We found instead that targeted cardiomyocyte deletion of Gdf11 during young adulthood, using the Myh6-cre system, does not result in cardiac hypertrophy and rather leads to progressive left ventricular dilation that is apparent in both females and males by 6 mo of age. However, because of multiple adverse effects from Cre recombinase itself on the heart and potential differential expression of the cre gene across genotypes, we are unable to define the molecular mechanism of the dilated cardiomyopathy phenotype that develops when the Gdf11 gene is removed from cardiomyocytes using this mouse model.

MATERIALS AND METHODS

Animals: constitutively active Cre.

Mice expressing constitutively active Cre recombinase driven by the Myh6 promoter [B6.FVB-Tg(Myh6-cre)2182Mds/J] were obtained from The Jackson Laboratory. Mice containing a floxed (flanking loxP) Gdf11 allele with loxP sites flanking exons 2 and 3 of Gdf11 were generously provided by Dr. Se-Jin Lee (Johns Hopkins University) (13). Mice were bred to obtain three genotypes used in this study: Myh6^cre/wt;Gdf11^fl/fl (experimental genotype), Myh6^cre/wt;Gdf11^wt/wt (Cre-only control genotype, referred to as Myh6^cre/wt), and Myh6^wt/wt;Gdf11^fl/fl (flox-only control genotype and littermate control, referred to as Gdf11^fl/fl). All mice were on a mixed C57Bl/6J and 6N background.

Young adult male and female mice of all three genotypes were weighed weekly starting at 2 mo of age. Echocardiograms were performed in all mice at 1–2 and 6 mo of age in a blinded manner. Additional echocardiogram time points at 3 or 4 mo of age were also performed in a subset of mice. Mice were observed for up to 6 mo, the animals were euthanized, and serum and tissues were harvested for further analysis. We also determined tibia length as a normalizing parameter because GDF11 administration has been shown to decrease body weight (17). Tissues for downstream DNA, RNA, and protein analysis were flash-frozen in liquid nitrogen and stored at −70°C until further processing was performed. All experiments were conducted according to the Guide for the Use and Care of Laboratory Animals and approved by the Institutional Animal Care and Use Committee of Harvard University Faculty of Arts and Sciences.

Animals: tamoxifen-inducible Cre.

Mice expressing tamoxifen-inducible Cre recombinase (Mer-Cre-Mer, abbreviated MCM) driven by the Myh6 promoter [B6.FVB(129)-A1cf^{Tg(Myh6-cre/Esr1*)1Jmk}/J] were obtained from The Jackson Laboratory (backcrossed to a C57Bl/6J agouti background per vendor). Mice were bred with Gdf11^fl/fl mice to obtain two genotypes used in this experiment: Myh6^MCM/wt;Gdf11^fl/fl (experimental genotype) and Myh6^wt/wt;Gdf11^fl/fl (flox control genotype). Male and female mice were stratified based on gender and then randomized at 5 mo of age to receive injections of either 4-hydroxytamoxifen (4OH-tamoxifen, 75 mg/kg) or vehicle (10% ethanol-90% sunflower oil) via intraperitoneal injection for 5 days. Echocardiography was performed at baseline (at 5 mo of age) and 4 wk later (at 6 mo of age) before harvest to evaluate left ventricular size and function. Tissues were harvested and processed for histology or quantitative PCR (qPCR).

Echocardiography.

Mice were sedated with 0.1–0.5% inhaled isoflurane for echocardiography, with the dose titrated to maintain heart rates of >500 beats/min for acquired images. Mice were placed on a heating pad, and echocardiograms were obtained with the Vevo770 (Visualsonics, Toronto, Ontario, Canada). M-mode was used to measure left ventricular interventricular septal wall thickness, left ventricular posterior wall thickness, and left ventricular internal diameter during both systole and diastole. Fractional shortening (%), left ventricular mass, and left ventricular volumes were calculated with the Visualsonics software package.

Cardiomyocyte and noncardiomyocyte isolation from adult hearts.

Cardiomyocytes and noncardiomyocytes were isolated from adult male and female mice using a Langendorff-free method as previously described (1). We modified the original protocol to include blebbistatin (5 µM; Sigma) instead of 2–3-butanedione monoxime in the culture media. In addition, we used a peristaltic pump rather than hand injection to better control the flow rate of the perfused solutions. Briefly, mice were anesthetized with isoflurane, and then the chest cavity was opened. The descending aorta and inferior vena cava were cut followed by perfusion of 7 ml of EDTA buffer in the apex of the right ventricle. The aorta was then clamped and cut distal to the clamp, and the heart was removed. EDTA buffer (10 ml) was then perfused in the apex of the left ventricle followed by injection of 3 ml of perfusion buffer and then 30–40 ml of collagenase buffer delivered in the left ventricular apex. The clamp was removed, and the heart was manually dissociated. Stop buffer (perfusion buffer + 5% fetal bovine serum) was added, cells were passed through a 300-µm strainer, and then cardiomyocytes were allowed to gravity settle for 20 min. The supernatant containing noncardiomyocytes and debris was plated in an uncoated tissue culture plate in DMEM–F-12–10% FBS for 3 h. The cardiomyocyte fraction in the pellet then underwent sequential gravity settling with low-speed centrifugation (12 g, 3 min) with calcium reintroduction followed by plating in laminin-coated plates for 3 h. After 3 h, both cardiomyocytes and noncardiomyocytes were washed and harvested in TRIzol. Samples were frozen at −70°C until RNA extraction by QIAcube.

PCR.

DNA was extracted from harvested tissue using the REDExtract-N-Amp Tissue PCR Kit (Sigma). Table 1 shows the primers that were used to amplify Myh6, Myh6-cre, and Gdf11 PCR products and their expected product sizes. PCR products were run on a 2% agarose gel in TAE buffer at 100 V for 45–60 min, and gels were imaged in a Gel Doc EZ system (Bio-Rad).

Table 1.

Primers for genotyping by PCR

Product	Primers	Product Size, bp
Gdf11 wild type	F (GDF11b): 5′-AAGGCTTGGGAAGCAGGCAAG-3′	359
	R (GDF11c): 5′-AGGTATGGTTAGGGTGTGGAG-3′
Gdf11 flox	F (GDF11a): 5′-ATGCAGATGGTAATACTTGGG-3′	393
	R (GDF11c): 5′-AGGTATGGTTAGGGTGTGGAG-3′
Gdf11 Δ2–3 (postrecombination)	F (GDF11b): 5′-AAGGCTTGGGAAGCAGGCAAG-3′	300
	R (GDF11c): 5′-AGGTATGGTTAGGGTGTGGAG-3′
Myh6 wild type	F (MYH6–1): 5′-ATGACAGACAGATCCCTCCTATCTCC-3′	895
	R (MYH6–2): 5′-AGAGGTGGTGGCTCTTAGCA-3′
Myh6-cre	F (MYH6–1): 5′-ATGACAGACAGATCCCTCCTATCTCC-3′	300
	R (MYH6–3): 5′-CTCATCACTCGTTGCATCATCGAC-3′
Myh6 MCM	F (Myh6MCM F): 5′-CGTCCTCCTGCTGGTATAG-3′	405
	R (Myh6MCM R): 5′-GTCTGACTAGGTGTCCTTCT-3′

Growth differentiation factor 11 (Gdf11) primers from McPherron et al. (13). F, forward; R, reverse; MCM, Mer-Cre-Mer.

Quantitative PCR.

RiboZol reagent (VWR) and the E.Z.N.A. Total RNA I kit (Omega) were used to isolate RNA from homogenized whole organ tissue, followed by the High Capacity cDNA Reverse Transcription kit (Thermo Fisher Scientific) to reverse transcribe mRNA to cDNA according to the manufacturers’ instructions. qPCR was performed using TaqMan probes for natriuretic peptide B (Nppb, Mm01255770_g1), Gdf11 (Mm01159973_m1, spanning exons 1–2), Gdf15 (Mm00442228_m1), inhibin subunit-βA (Inhba, activin A, Mm00434339_m1), Mstn (Mm01254559_m1), and transforming growth factor-β receptor 1 (Tgfbr1, Mm00436964_m1), with TATA-binding protein (Tbp) used as the housekeeping gene (Mm01277042_m1) using a Bio-Rad CFX384 Real-Time System. Because of low yields of isolated cardiomyocytes and noncardiomyocytes from adult hearts, RNA was extracted using the Qiagen QIACube from Trizol (Thermo) followed by reverse transcription to cDNA using the SuperScript VILO cDNA synthesis kit (Thermo) per the manufacturer’s instructions. The isolated cDNA (10–20 ng) was then preamplified using the Taqman PreAmp Master Mix (Thermo) for 14 cycles. The preamplified cDNA was then diluted 1:20 for quantification of Gdf11 expression in isolated cardiomyocytes and noncardiomyocytes by qPCR on the QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems).

Flow cytometry.

Flow cytometry was performed to quantify the percentage of cardiomyocytes in the cell population obtained immediately after the final step of the extraction procedure. In brief, the cardiomyocyte fraction was fixed with 1 ml of 70% ethanol and then stored at −20°C. The noncardiomyocyte fraction was plated into a six-well plate and allowed to expand for 7 days in DMEM–F-12–10% fetal bovine serum + 1% penicillin-streptomycin. Cells were dissociated with trypsin and then fixed in 70% ethanol. The fixative was removed via low-speed centrifugation (20 g × 3 min), and the cells were washed and then permeabilized in 0.1% Triton in PBS. Cells were labeled overnight at 4°C using an antibody to cardiac troponin T [ab8295, diluted 1:250 in phosphate-buffered saline (PBS) supplemented with 10% goat serum; Abcam] to detect cardiomyocytes, and an antibody to vimentin (ab92547, diluted 1:200 in PBS supplemented with 10% goat serum; Abcam), which is highly expressed in fibroblasts. After washing steps and incubation with the corresponding AlexaFluor 568-conjugated secondary antibody for 2 h at room temperature (mouse IgG1 and rabbit IgG, respectively; Thermo), cells were analyzed by flow cytometry (MoFlo Astrios, using the nozzle of 200 µm; Beckman Coulter). Data were analyzed with FlowJo software (version 10.0.8).

Histology.

Harvested tissues were fixed in 4% paraformaldehyde for 24 h and then exchanged for 70% ethanol for 3 days before embedding in paraffin. Sections were stained with Masson’s trichrome staining as previously described (8). Global fibrosis was quantified using a Python script written to quantify the percentage of blue pixels out of the total pixels from a cross-sectional image of the heart (https://github.com/jgarbern/global-fibrosis). Average cardiomyocyte cross-sectional area for each mouse was quantified in a blinded manner with a total of 40 cardiomyocytes measured from multiple sections from each heart.

Quantitative mass spectrometry.

Blood was obtained by retro-orbital collection at the time of harvest and transferred to serum separator tubes. Tubes were spun at 2,000 g for 5 min, and serum was transferred to clean low-binding microcentrifuge tubes and stored at −70°C until further processing. Samples (100 µl) were submitted to the Brigham Research Assay Core at Brigham and Women’s Hospital for quantitative mass spectrometry. Mouse serum was denatured, reduced, and alkylated, followed by pH-based fractionation using cation ion exchange-solid phase extraction; appropriate elution fraction was digested with trypsin. After desalting and concentrating of tryptic digest, the peptide mixture was separated and eluted by liquid chromatography followed by mass spectrometric analysis operated in positive electrospray ionization mode. The most intensive and unique proteotypic peptides from GDF11 and myostatin as surrogated peptides along with heavy-labeled unique peptides as internal standards were used for quantitative determination of GDF11 and myostatin.

Statistical analysis.

Data are expressed as means ± SE unless noted otherwise. Data were evaluated with the D’Agostino and Pearson omnibus normality test; normally distributed data were evaluated with Student’s t-test or two-way ANOVA with Tukey’s post hoc analysis, whereas data that were not normally distributed were analyzed with Mann-Whitney or Kruskal-Wallis with Dunn’s multiple-comparisons test as indicated. Kaplan-Meier survival curves were evaluated with the Mantel-Cox log-tank test. Body weight trends with time were analyzed with a two-way repeated-measures ANOVA with Tukey’s multiple-comparisons test. A P value < 0.05 was considered statistically significant.

RESULTS

Cardiomyocyte-specific deletion of Gdf11 leads to left ventricular dilation that progresses with age.

Because of known sexual dimorphism in body weight and heart size in mice, we analyzed body weight, heart weight, and heart weight-to-body weight ratio data from males and females separately. Myh6^cre/wt;Gdf11^fl/fl mice had progressive left ventricular dilation with a significant increase in left ventricular end-diastolic volume [1–2 mo (Fig. 1A), 3–4 mo (Fig. 1B), and 6 mo (Fig. 1C)], a significant decrease in septal thickness (Fig. 1J), and a nonsignificant decrease in left ventricular posterior wall thickness (Fig. 1L) by echocardiography that was apparent in both genders by 6 mo of age. Male Myh6^cre/wt;Gdf11^fl/fl mice also had significantly increased left ventricular internal diameters at 6 mo of age compared with sex-matched Myh6^cre/wt mice (Fig. 1K). Cardiomyocyte-specific deletion of Gdf11 also led to a decrease in left ventricular function with decreased fractional shortening in Myh6^cre/wt;Gdf11^fl/fl females compared with sex-matched Gdf11^fl/fl controls and in Myh6^cre/wt;Gdf11^fl/fl males compared with sex-matched Myh6^cre/wt controls (Fig. 1G). The chamber dilation and functional decline were not associated with differences in either estimated left ventricular mass by echocardiography or heart weight compared with either the Cre-only control genotype (Myh6^cre/wt) or the flox-only control genotype (Gdf11^fl/fl) at 6 mo of age (Fig. 1, H, I, and F, respectively). Heart weight-to-body weight ratios were not different across groups at any age [1–2 mo (Fig. 1D), 3–4 mo (Fig. 1E), and 6 mo (Fig. 1F)]. There were no significant differences in tibia lengths or heart weight-to-tibia length ratios at 6 mo of age (Fig. 1, N and O, respectively).

Fig. 1.

Targeted cardiac growth differentiation factor 11 (Gdf11) deletion leads to left ventricular dilation. A–C: left ventricular end-diastolic volume by echocardiography at 1–2 (A), 3–4 (B), and 6 (C) mo of age. D–F: heart weight-to-body weight ratio at 1–2 (D), 3–4 (E), and 6 (F) mo of age. G: fractional shortening by echocardiography at 6 mo of age. H: estimated left ventricular mass by echocardiography at 6 mo of age. I: heart weight at 6 mo of age. J: interventricular septal thickness during diastole at 6 mo of age. K: left ventricular internal diameter during diastole at 6 mo of age. L: left ventricular posterior wall thickness during diastole at 6 mo of age. M: body weight at 6 mo of age (just before harvest). N: tibia length at 6 mo of age. O: heart weight-to-tibia length ratio at 6 mo of age. Sample sizes at 1–2 mo: Myh6^cre/wt (n = 5 females, n = 9 males), Gdf11^fl/fl (n = 11 females, n = 21 males), and Myh6^cre/wt;Gdf11^fl/fl (n = 12 females, n = 15 males) mice; n = 4/group for heart weight-to-body weight ratio parameter. Sample sizes at 3–4 mo: Myh6^cre/wt (n = 9 females, n = 12 males), Gdf11^fl/fl (n = 9 females, n = 13 males), and Myh6^cre/wt;Gdf11^fl/fl (n = 13 females, n = 14 males) mice for echocardiography; n = 4/group for heart weight-to-body weight ratio parameter. Sample sizes at 6 mo: Myh6^cre/wt (n = 6 females, n = 16 males), Gdf11^fl/fl (n = 3 females, n = 5 males), and Myh6^cre/wt;Gdf11^fl/fl (n = 6 females, n = 9 males). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by Kruskal-Wallis test with Tukey’s post hoc analysis.

We also used a tamoxifen-inducible cardiomyocyte deletion model with administration of 4OH-tamoxifen or vehicle to 6-mo-old Gdf11^fl/fl or Myh6^MCM/wt;Gdf11^fl/fl mice for 5 days at 75 mg/kg delivered intraperitoneally but saw no significant differences in left ventricular end-diastolic volume, estimated left ventricular mass, body weight, heart weight, or heart weight-to-body weight ratio with this treatment regimen (Fig. 2, E, F, G, I, and J, respectively) despite evidence of recombination by PCR (Fig. 2C). We did not see significant changes in Gdf11 mRNA expression following this treatment regimen by qPCR (Fig. 2D). However, we observed significant differences in survival across all groups in males (Fig. 2B) but not females (Fig. 2A). The etiology for the survival difference in 4OH-tamoxifen-treated males is unclear although we presume it reflects tamoxifen toxicity given the acute nature of the deaths. Survival of only the “healthiest” mice may also confound our long-term results. 4OH-tamoxifen significantly increased tibia length; therefore, we did not evaluate heart weight-to-tibia length ratios in this model (Fig. 2H); the effect of tamoxifen on long bone length has been reported previously (26). We also found significantly increased activin A (Inbha) mRNA expression within the spleens of 4OH-tamoxifen-treated Myh6^MCM/wt;Gdf11^fl/fl male mice, and a similar nonsignificant trend in other tamoxifen -treated groups (Fig. 2K), consistent with prior reports that tamoxifen may activate TGF-β signaling (4, 5, 15). We did not study this model further in depth because of the absence of a cardiac phenotype at the age selected (6 mo of age).

Fig. 2.

Survival curves, echocardiographic, heart weight, and body weight parameters in mice with tamoxifen-inducible Myh6-driven Cre expression [Myh6-Mer-Cre-Mer (MCM)]. A and B: Kaplan-Meier survival analysis in growth differentiation factor 11 (Gdf11)^fl/fl (gray) and Myh6^MCM/wt;Gdf11^fl/fl (black) female (A) and male (B) mice. Male mice have significant differences between groups with P < 0.01 by the Mantel-Cox test; n = 12–16/group at the start of study. C: recombination of Gdf11 gene seen in the heart but not liver, kidney, spleen, or skeletal muscle following administration of 4-hydroxytamoxifen (4OH-tamoxifen, 75 mg/kg ip × 5 days) by polymerase chain reaction (PCR). D: Gdf11 expression in RNA extracted from whole heart 1 mo after 4-hydroxytamoxifen administration in female and male mice by quantitative PCR (qPCR), n = 3/group. Data are normalized to TATA-binding protein (Tbp) and then to female Gdf11^fl/fl vehicle control. For echo and harvest data in E–J, n = 7–16/group; data were collected 1 mo after initiation of 4OH-tamoxifen injection (6 mo of age). E: left ventricular end-diastolic volume (µl) by echocardiography. F: estimated left ventricular mass (mg) by echocardiography. G: body weight (g) at harvest. H: tibia length. **P < 0.01 and ***P < 0.001 by 2-way ANOVA. I and J: heart weight (I) and heart weight-to-body weight ratio (J). K: activin A mRNA expression in spleen is significantly increased in male Myh6^MCM/wt;Gdf11^fl/fl mice 1 mo after initiation of 4OH-tamoxifen injection (6 mo of age), n = 4–6/group. *P < 0.05 by Kruskal-Wallis with Dunn’s multiple-comparisons test. Data are normalized to Tbp and then to female Gdf11^fl/fl vehicle control.

We note that the different findings observed between tamoxifen-inducible Myh6-MCM versus constitutively active Myh6-cre models may be because of the different ages at which Gdf11 knockdown occurs in the two models. In addition, application of stress such as with pressure overload could be considered to enhance any potential phenotype caused with tamoxifen-induced Myh6-MCM expression. In the context of increased mortality in mice injected with 4OH- tamoxifen even without surgery, and concerns for confounding effects from 4OH- tamoxifen administration on TGF-β signaling, we chose not to apply pressure overload stress (e.g., transverse aortic constriction model) or repeat this experiment at a younger age in the tamoxifen-inducible cardiomyocyte deletion model and instead focused on the constitutively active Myh6-cre model for the remainder of this study.

Cardiomyocyte-specific deletion of Gdf11 is not sufficient to decrease Gdf11 mRNA expression in the whole heart.

Myh6 Cre-induced genomic recombination of Gdf11 was evident in the heart in neonatal mice (Fig. 3A). Gdf11 recombination was observed only in the heart by PCR and, as expected, was not seen in the lung, liver, kidney, spleen, or skeletal muscle at 6 mo of age (Fig. 3B). However, despite evidence of DNA recombination, we did not observe differences in Gdf11 mRNA expression in the whole heart at 6 mo of age (Fig. 3D), and in fact saw a significant increase in Gdf11 mRNA expression in Myh6^cre/wt;Gdf11^fl/fl male mice at 3 mo of age (Fig. 3C).

Fig. 3.

Targeted cardiomyocyte deletion of growth differentiation factor 11 (Gdf11) does not decrease total Gdf11 mRNA expression in mouse hearts. A: representative agarose gel image depicting Myh6 (wild-type band at 894 bp, Myh6-cre band at 300 bp) and Gdf11 [wild-type band at 359 bp, flox band at 393 bp, and ∆2–3 (postrecombination) band at 300 bp] alleles seen in the heart on day of life 0–1 by polymerase chain reaction (PCR) in Myh6^cre/wt, Gdf11^fl/fl, and Myh6^cre/wt;Gdf11^fl/fl pups. B: recombination of Gdf11 at 6 mo of age in females and males in the heart but not in lung, liver, kidney, spleen, or skeletal muscle by PCR. C and D: Gdf11 expression in RNA extracted from whole heart in 3 (n = 4/group, C)- or 6 (n = 3–4/group, D)-mo-old female and male mice by quantitative PCR (qPCR). Data are normalized to TATA-binding protein (Tbp) and then to female Myh6^cre/wt control. E and F: Gdf11 expression in RNA extracted from isolated adult (>2 mo old) cardiomyocytes (E) or noncardiomyocytes (F), with each data point (shown with open symbols) representing isolated cells from a single mouse of the same genotype, n = 3/group (2 males, 1 female). Data were normalized to Tbp and then Myh6^cre/wt control. Cardiomyocytes (E), P value = not significant (ns, 0.05); noncardiomyocytes (F), P value = ns (0.3) by Kruskal-Wallis test. G and H: serum levels of GDF11 (G) and myostatin (H) as determined by quantitative mass spectrometry in 6-mo-old mice; n = 3–4/group. *P < 0.05 by Kruskal-Wallis test followed by Tukey’s multiple-comparisons test.

Because approximately one-half of the whole heart is composed of noncardiomyocytes (including endothelial, fibroblast, and smooth muscle cells) (2), we isolated cardiomyocytes and noncardiomyocytes from adult (>2 mo) mice to evaluate Gdf11 mRNA expression in the two cell populations. In a representative batch, the cardiomyocyte population consisted of 84% cTnT⁺ cardiomyocytes, whereas the noncardiomyocyte population consisted of 95% vimentin⁺ fibroblasts, as quantified by flow cytometry (data not shown). Gdf11 expression shows substantial variability in Myh6^cre/wt mice in this analysis; however, we detected a decreased signal in cardiomyocytes from Myh6^cre/wt;Gdf11^fl/fl mice and a trend toward differences in Gdf11 expression in cardiomyocytes across genotypes (P = 0.05 by Kruskal-Wallis test) (Fig. 3E).

The high variability in Gdf11 expression in Myh6^cre/wt mice suggested that this genotype may be abnormal, which became evident in subsequent analysis. The Gdf11 signal in Myh6^cre/wt;Gdf11^fl/fl mice was greater than zero likely because of contamination from noncardiomyocytes in the cardiomyocyte population, although it is possible that this reflects incomplete recombination of the floxed alleles. We did not observe any differences in Gdf11 expression in noncardiomyocytes across genotypes (Fig. 3F). Direct comparison across cell types is challenging because of different amounts of RNA per cell in different cell types. If starting with an equal amount of RNA, there was a trend toward increased Gdf11 expression in noncardiomyocytes compared with cardiomyocytes in all genotypes (data not shown). However, we obtained much lower total RNA after isolation from noncardiomyocytes (predominantly cardiac fibroblasts) compared with an equal number of cardiomyocytes, and therefore comparing Gdf11 expression from the same amount of total RNA may be suboptimal.

Because of challenges in quantifying protein levels of GDF11 with antibody-based approaches (7), we measured GDF11 and myostatin levels in serum by quantitative mass spectrometry and found that Gdf11^fl/fl females had significantly lower circulating levels of GDF11 than Myh6^cre/wt females; however, there were no significant differences in GDF11 serum concentrations between Myh6^cre/wt;Gdf11^fl/fl mice and their sex-matched Cre- or flox-only control groups (Fig. 3G). Although the major source(s) of circulating GDF11 remains unclear, prior work from our group demonstrated that the spleen has the highest mRNA levels among organs studied (12); thus, it is not unexpected that serum levels of GDF11 were not altered in Myh6^cre/wt;Gdf11^fl/fl mice, since only cardiomyocyte Gdf11 was targeted for deletion in these animals. Circulating myostatin levels did not differ across groups (Fig. 3H).

Mechanism of left ventricular dilation confounded by adverse effects from Myh6-driven Cre recombinase activity.

Myh6^cre/wt male mice were significantly smaller than both Gdf11^fl/fl and Myh6^cre/wt;Gdf11^fl/fl mice by 6 mo of age (Fig. 1M), with a plateau in the body weight curve in Myh6^cre/wt mice appearing at around 4 mo of age in both genders (Fig. 4, C and D). In addition, there was a nonsignificant trend toward decreased survival in Myh6^cre/wt mice, with death in several mice at around 4 mo of age (Fig. 4, A and B).

Fig. 4.

Cre recombinase has adverse effects on survival and body weight. A and B: Kaplan-Meier survival analysis in female (A) and male (B) Myh6^cre/wt (dark gray, dashed line, n = 12 females, n = 24 males), growth differentiation factor 11 (Gdf11)^fl/fl (light gray, dotted line, n = 5 females, n = 11 males), and Myh6^cre/wt;Gdf11^fl/fl (black solid line, n = 12 females, n = 15 males) mice. C and D: body weight versus age in weeks in female (C) and male (D) Myh6^cre/wt (gray circles with dashed line, n = 5 females, n = 13 males), Gdf11^fl/fl (gray squares with dotted line, n = 5 females, n = 11 males), and Myh6^cre/wt;Gdf11^fl/fl (black tringles with solid line, n = 12 females, n = 15 males) mice. *P < 0.05 and **P < 0.01 by 2-way repeated-measures ANOVA with Tukey’s multiple-comparisons test.

There was a significant increase in cardiac mRNA expression of Mstn in Myh6^cre/wt;Gdf11^fl/fl mice compared with Myh6^cre/wt mice at 2 (Fig. 5A) but not 6 (Fig. 5E) mo of age. In contrast, at 6 but not 2 mo of age, Nppb, which encodes B-type natriuretic peptide, a clinically used marker of heart failure, had significantly higher expression in hearts of Myh6^cre/wt mice compared with Gdf11^fl/fl mice, with the experimental genotype (Myh6^cre/wt;Gdf11^fl/fl) having an intermediate expression profile [2 (Fig. 5B) and 6 (Fig. 5F) mo]. Expression of Tgfbr1, which encodes one of the receptors for GDF11, was also significantly lower in hearts of Gdf11^fl/fl mice compared with both Myh6^cre/wt and Myh6^cre/wt;Gdf11^fl/fl mice at 6 (Fig. 5G) but not 2 (Fig. 5C) mo of age. Finally, expression of Gdf15, which was reported to be upregulated following administration of supraphysiological doses of GDF11 (11), was significantly increased in hearts of Myh6^cre/wt mice compared with both Gdf11^fl/fl and Myh6^cre/wt;Gdf11^fl/fl mice at 6 (Fig. 5H) but not 2 (Fig. 5D) mo of age. These results suggest that cardiomyocyte deletion of Gdf11 has an early effect on Mstn expression that precedes left ventricular dilation and is followed later by differential regulation on other proteins involved in TGF-β signaling. However, the differences between the two control genotypes particularly at older ages suggest that the presence of Cre recombinase has adverse effects on the heart and make it difficult to identify molecular mechanisms to explain our echocardiographic findings in the experimental genotype.

Fig. 5.

Gene expression analysis from whole hearts of 2 (A–D)- and 6 (E–H)-mo-old mice of Mstn (A and E), natriuretic peptide B (Nppb, B and F), transforming growth factor-β receptor 1 (Tgfbr1, C and G), and growth differentiation factor 15 (Gdf15, D and H) by quantitative PCR (qPCR). Data are normalized to TATA-binding protein (Tbp) expression and then to Myh6^cre/wt control. No significant differences were observed between genders; therefore, data represent combined male and female data; n = 7–8/group. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by Kruskal-Wallis analysis with Dunn’s multiple-comparisons test.

Myh6-driven Cre recombinase expression leads to myocardial fibrosis and increased cardiomyocyte size in males.

Cardiotoxicity has been previously described by Pugach et al. with prolonged Myh6-driven Cre expression in this strain (18), and our results are consistent with these prior findings. We observed a significant increase in global myocardial fibrosis in males compared with females in the Cre-only control genotype (Myh6^cre/wt) at 6 mo of age, with males more sensitive than females to adverse effects from Cre, consistent with Pugach et al. (18) (Fig. 6, A and B). There was a similar nonsignificant trend observed in Myh6^cre/wt;Gdf11^fl/fl mice. In addition, there was a nonsignificant trend of increased global myocardial fibrosis in male mice expressing Cre recombinase compared with the flox-only control (Gdf11^fl/fl). Furthermore, the cardiomyocyte cross-sectional area was significantly increased in Myh6^cre/wt male mice compared with either Gdf11^fl/fl or Myh6^cre/wt;Gdf11^fl/fl mice (Fig. 6C).

Fig. 6.

Cre mRNA expression is higher in Myh6^cre/wt compared with Myh6^cre/wt; growth differentiation factor 11 (Gdf11)^fl/fl mice and is associated with myocardial fibrosis and increased cardiomyocyte size. A: representative histology sections of male and female Myh6^cre/wt, Gdf11^fl/fl, and Myh6^cre/wt;Gdf11^fl/fl mice stained with Masson’s trichrome. B: global fibrosis (%blue pixels) shows increased fibrosis in male (n = 5–9/group) Myh6^cre/wt mice compared with females (n = 3–5/group). C: cardiomyocyte cross-sectional area is increased in male Myh6^cre/wt mice compared with male Gdf11^fl/fl and Myh6^cre/wt;Gdf11^fl/fl mice. Cross-sectional area from 40 cardiomyocytes/mouse were averaged for each mouse, and then average cardiomyocyte cross-sectional area by mouse data was analyzed with n = 3–6 female mice and 5–10 male mice. *P < 0.05 and **P < 0.01 by 2-way ANOVA with Tukey’s post hoc analysis. D: Cre expression by quantitative PCR (qPCR) from male and female 3- to 6-mo-old mice. No significant differences were detected between genders; therefore, data depict combined male and female data. Data are normalized to TATA-binding protein (Tbp) and then to Myh6^cre/wt control; n = 16/group, *P < 0.05 by Mann-Whitney test. E: Cre protein detected by Western blot analysis in Myh6^cre/wt, Gdf11^fl/fl, and Myh6^cre/wt;Gdf11^fl/fl 4-mo-old male mice. Band densitometry analysis comparing Myh6^cre/wt mice with Myh6^cre/wt;Gdf11^fl/fl (not shown) is not significant with P value 0.2 by Mann-Whitney test.

Cre mRNA expression is higher in Myh6^cre/wt mice compared with Myh6^cre/wt;Gdf11^fl/fl mice.

We evaluated whether Cre is differentially expressed across genotypes as a possible explanation for why Myh6^cre/wt mice have a more pronounced phenotype in terms of body weight, myocardial fibrosis, and expression of selected genes associated with heart failure, such as Nppb. We found that cre mRNA expression is significantly greater in Myh6^cre/wt mice compared with Myh6^cre/wt;Gdf11^fl/fl mice (Fig. 6D). In contrast to RNA levels, there were no differences noted in Cre protein expression in Myh6^cre/wt mice compared with Myh6^cre/wt;Gdf11^fl/fl mice (Fig. 6E). No cre protein expression was detected in Gdf11^fl/fl mice.

DISCUSSION

We observed a significant increase in left ventricular end-diastolic volume in mice with Cre-mediated genetic deletion of Gdf11 in cardiomyocytes. This finding was consistent in both males and females, with left ventricular end-diastolic volume significantly increased in Myh6^cre/wt;Gdf11^fl/fl mice compared with both Myh6^cre/wt and Gdf11^fl/fl control mice. Left ventricular dilation was associated with a decrease in left ventricular systolic function, suggesting a possible role for GDF11 signaling in dilated cardiomyopathy. We observed that Mstn and Gdf11 both transiently increase in RNA extracted from whole hearts of young adult Myh6^cre/wt;Gdf11^fl/fl mice before the onset of ventricular dilation, which suggests that there is a developmental component to the phenotype observed. However, our experimental design met numerous unanticipated challenges that prohibit clear interpretation of the underlying molecular mechanisms to explain these findings. First, we did not observe a significant decrease in Gdf11 expression in whole heart extracts from Myh6^cre/wt;Gdf11^fl/fl mice, suggesting that counterregulation in nonmyocytes may buffer cardiomyocyte-specific loss of Gdf11 in the heart. Second, mice expressing Cre recombinase had a different phenotype from mice not expressing Cre, with decreased body weights and increased global myocardial fibrosis in males, an effect only made clear because we used two control genotypes. Third, cre mRNA expression was different between the Cre control genotype and the Cre-containing experimental genotype. Although Cre protein levels were not different by Western blot analysis, in the context of Cre-associated toxicity, further study of protein levels at different ages should be performed in future work to elucidate which perturbations to the cardiac system should be attributed to differences in Cre levels versus Gdf11 deletion.

Although we observed a nonsignificant trend toward decreased Gdf11 expression in isolated adult cardiomyocytes of Myh6^cre/wt;Gdf11^fl/fl mice (complete absence of Gdf11 could not be shown likely because of contamination from noncardiomyocytes), we did not observe a significant decrease in Gdf11 expression in the whole heart despite constitutively active Myh6-driven Cre expression and PCR evidence of Gdf11 recombination. In fact, we observed a transient increase in Gdf11 expression in whole heart extracts from male Myh6^cre/wt;Gdf11^fl/fl mice compared with male Gdf11^fl/fl control mice at 3 mo of age and a similar nonsignificant trend compared with male Myh6^cre/wt control mice. A similar effect was previously reported when Mstn was targeted for deletion in cardiomyocytes using mice with tamoxifen-inducible Myh6-driven Cre expression, where Mstn mRNA expression increased in whole heart extracts because of upregulation in nonmyocytes despite cardiomyocyte deficiency of myostatin (3). Because of different amounts of total RNA in cardiomyocytes versus noncardiomyocytes, and low expression levels of Gdf11 in cardiomyocytes requiring preamplification of cDNA to detect a reliable signal by qPCR, we were unable to directly compare expression of Gdf11 in cardiomyocytes with noncardiomyocytes. Nonetheless, in noncardiomyocytes, we did not observe a difference in Gdf11 expression levels across genotypes. Taken together, this suggests that Gdf11 may act locally in cardiomyocytes, given the presence of a dilated phenotype even in the absence of expression differences at the organ level. In addition, we observed a significant increase in Mstn mRNA expression in whole heart extracts of 2-mo-old mice. This suggests that Cre-mediated cardiomyocyte deletion of Gdf11 leads to downstream signaling effects during development that precede the observed phenotype of left ventricular dilation starting at 3–4 mo of age. We also did not examine other organs given the lack of serum differences in GDF11 at 6 mo. Alternative models used in future work should focus on early downstream signaling changes in the heart and other organs (such as the spleen, which has higher baseline mRNA expression) as well as measuring circulating GDF11 levels at younger ages to better understand how these changes affect cardiac phenotype at later time points.

Myh6^cre/wt mice have previously been shown to develop progressive myocardial fibrosis and inflammation because of DNA damage at endogenous “loxP-like” or “pseudo-loxP” sites in the myocardium (18). In that study, the authors identified 227 loxP-like sites within genes that could be potentially recognized by Cre recombinase, when tolerating ≤4 mismatches in the canonical loxP sequence. Of these 227 degenerate loxP sites, 55 are expressed in the heart, leading to numerous off-target effects from Cre recombinase, including myocardial fibrosis and upregulation of apoptotic markers such as p53 and Bax (18). This previous study also found that males appear to be more sensitive to the off-target effects of Cre expression in cardiomyocytes, with an increased heart weight-to-body weight ratio in Myh6^cre/wt compared with wild-type control (C57Bl/6J) male mice at 6 mo of age (18). Although we did not have wild-type mice as a control in our study, we also observed that Myh6^cre/wt males had significantly increased global fibrosis compared with females, and Myh6^cre/wt males had significantly increased cardiomyocyte cross-sectional area compared with Gdf11^fl/fl control mice and Myh6^cre/wt;Gdf11^fl/fl knockout mice. It remains unclear whether the potency of Cre toxicity is identical in the presence or absence of loxP, or whether in the presence of loxP, there might be fewer off-target effects because of the stoichiometry of Cre:loxP versus Cre:loxP-like or pseudo-loxP sites. These results underscore the importance of inclusion of Cre control mice and evaluation of both genders when using Cre-lox technology.

We used both Myh6^cre/wt and Gdf11^fl/fl controls to attempt to account for adverse effects from Cre recombinase. Our results demonstrate that, without inclusion of both controls, our data would have been easily misinterpreted with potentially incorrect conclusions drawn, a lesson that was learned from experience after we failed to include the Myh6^MCM/wt control line in the tamoxifen-inducible study described in results. For example, without Gdf11^fl/fl controls, we may have incorrectly concluded that deletion of Gdf11 leads to a decrease in Nppb expression. Conversely, without Myh6^cre/wt controls, we may have incorrectly concluded that deletion of Gdf11 leads to an increase in Nppb expression. However, with inclusion of both control genotypes, we see that there is actually a confounding effect with significant differences in expression of Nppb between Gdf11^fl/fl and Myh6^cre/wt control mice. It is challenging to reconcile these differences among the different genotypes. It appears that the presence of Cre induces stress on the mouse (with increased mortality, decreased body weight, increased Nppb expression, and increased cardiomyocyte size). Comparing the Myh6^cre/wt control mice with Myh6^cre/wt;Gdf11^fl/fl mice, one might conclude that cardiomyocyte deletion of Gdf11 is actually cardioprotective (with increased survival, increased body weight, decreased Nppb expression, and decreased cardiomyocyte size). However, other explanations are possible as well, such as differential off-target effects in the presence or absence of true loxP sites or differential expression of Cre in the two genotypes. Given that comparison of Gdf11^fl/fl control mice with Myh6^cre/wt;Gdf11^fl/fl mice leads to different conclusions than when comparing Myh6^cre/wt control mice with Myh6^cre/wt;Gdf11^fl/fl mice, we are unable to definitively determine the mechanisms by which Gdf11 deletion in cardiomyocytes leads to left ventricular dilation. Inclusion of multiple control groups admittedly adds significant costs and time required to maintain a larger animal colony, but careful selection of control groups is necessary to obtain meaningful results.

We observed higher cre mRNA but not protein expression in Myh6^cre/wt compared with Myh6^cre/wt;Gdf11^fl/fl mice in this study. Variable cre RNA expression in different generations has been reported in other Cre lines (19, 23). For example, in an albumin-Cre model, despite transmission of albumin-cre in genomic DNA of successive generations, some mice did not express cre in the liver (23). The authors speculated that this could be because of multiple homologous recombination events leading to segregation of inactive copies of the transgene, binding of transcriptional inhibitors, or posttranscriptional silencing of cre (23). In addition, cytosine methylation of loxP sites following Cre recombination of a parent can lead to inhibition of Cre-mediated recombination in subsequent generations (19). Finally, although all mice were on a mixed C57Bl/6J and 6N background, there may be generational differences because of different degrees of backcrossing to the underlying genetic background. It is possible that there is a more complex interaction with Cre and the underlying genetic background that will be difficult to uncover. To try to address this confounding effect, we used littermate controls to compare Gdf11^fl/fl with Myh6^cre/wt;Gdf11^fl/fl mice. However, our breeding strategy paired Myh6^cre/wt with wild-type (C57Bl/6J) mice and Gdf11^fl/fl with Myh6^cre/wt;Gdf11^fl/fl; thus, Myh6^cre/wt were not littermates with Myh6^cre/wt;Gdf11^fl/fl mice and may have had varying degrees of methylation or varying genetic backgrounds in the breeders. Future work to understand how Cre mRNA and protein levels change with age in both genotypes is necessary to interpret whether variable Cre levels might be confounding the phenotype seen with cardiomyocyte Gdf11 deletion.

In conclusion, deletion of the Gdf11 gene from cardiomyocytes may lead to left ventricular dilation and decreased systolic function, consistent with a dilated cardiomyopathy phenotype. However, because of numerous confounding factors associated with the selected, and commonly used, Cre-lox system as well as challenges in working with Gdf11 itself and its complicated regulatory system, we are unable to attribute a mechanism to this phenotype. These data highlight the importance of using appropriate control groups when using the Cre recombinase system, since opposite conclusions could have been drawn had only one of the two control genotypes been used for comparison. Further work to develop alternative animal models that avoid Cre toxicity and investigate alternative proteins involved in regulation of GDF11 expression is warranted.

GRANTS

J. C. Garbern was supported by the John S. LaDue Memorial Fellowship in Cardiology from Harvard Medical School and National Institutes of Health (NIH) T32 Fellowship 5T32-HL-007572. V. Bassaneze was supported by the Lemann Foundation Cardiovascular Research Postdoctoral Fellowship through Harvard University/Brigham and Women’s Hospital. J. M. Goldstein was supported by NIH F32 Postdoctoral Fellowship F32-AG-050395. R. G. Walker was supported by NIH T32 Fellowship T32-HL-007208-39. This work was supported by NIH Grants AG-047131, HL-119230, AG-048917, and AG-057428 (to R. T. Lee) and AG-048917 and AG-057428 (to A. J. Wagers) and from the Glenn Foundation (to A. J. Wagers).

DISCLOSURES

Richard Lee, Ryan Walker, and Amy Wagers are co-founders of, members of the scientific advisory board for, and hold private equity in Elevian, Inc., a company that aims to develop medicines to restore regenerative capacity. Elevian also provides sponsored research support to the Lee Laboratory and Wagers Laboratory.

AUTHOR CONTRIBUTIONS

J.G., A.C.K., V.B., A.V., H.S., R.G.W., A.J.W., and R.T.L. conceived and designed research; J.G., A.C.K., V.B., A.V., H.S., R.S., L.P., and E.M.R.-B. performed experiments; J.G., A.C.K., V.B., A.V., H.S., R.S., L.P., and E.M.R.-B. analyzed data; J.G., A.C.K., V.B., A.V., H.S., J.M.G., R.G.W., A.J.W., and R.T.L. interpreted results of experiments; J.G. and V.B. prepared figures; J.G. and L.P. drafted manuscript; J.G., V.B., A.V., J.M.G., R.G.W., S.B., A.J.W., and R.T.L. edited and revised manuscript; J.G., A.C.K., V.B., A.V., H.S., R.S., L.P., E.M.R.-B., J.M.G., R.G.W., S.B., A.J.W., and R.T.L. approved final version of manuscript.