Pulmonary arterial remodelling by deficiency of peroxisome proliferator-activated receptor-γ in murine vascular smooth muscle cells occurs independently of obesity-related pulmonary hypertension

SmPparγ^−/− mice were generated using transgenic mice expressing Cre-recombinase under the control of the smooth muscle protein 22-alpha promotor (The Jackson Laboratory, Bar Harbor, Maine) and homozygously floxed PPARγ mice [22]. Eight to 12 week old littermate male WT (Cre-) and SmPparγ^−/− (Cre+) mice on a mixed genetic background (C57BL/6, SJL, DBA/2J, C3H) were randomized as follows: WT and SmPparγ^−/− fed ad libitum a low-fat diet (LFD) (10% kcal from fat; Brogaarden, Gentofte, Denmark; D12450J) and WT and SmPparγ^−/− fed ad libitum a high-fat diet (HFD) (60% kcal from fat; Brogaarden; D12492) for 24 weeks. Mice were housed at room temperature with a 12 h light/dark cycle. After sacrificing, organs were excised, weighed, shock-frozen in liquid nitrogen and stored at − 80 °C until further investigation.

Body weight was measured twice weekly throughout the study. Food and water intake were measured for up to 24 h in a LabMaster (TSE Systems; Bad Homburg, Germany) after 12 weeks and at the end of the experiment (24 weeks). At the beginning and after 12 and 24 weeks, intraperitoneal (i.p.) insulin tolerance tests (ITT) using a dose of 0.5 U/kg insulin (Insuman® Rapid, Sanofi Aventis, Berlin, Germany) and i.p. glucose tolerance tests (GTT) with 1 g/kg glucose (Glucosteril, Fresenius, Bad Homburg, Germany) were carried out in ~ 4 h and ~ 12 h fasted mice, respectively. Right ventricular systolic pressure (RVSP), a measure for pulmonary arterial systolic pressure (PASP), was recorded with a Millar® microtip catheter (SPR1000), which was inserted into the right ventricle (RV) through the jugular vein. Systemic arterial pressure (SAP) was measured in the carotid artery. The catheter signal was amplified using a PowerLab® amplifier and converted to corresponding pressure curves using LabChart7® software (AD Instruments, Sydney, Australia). To determine the individual SAP and RVSP three to five tracings at different time points were used.

Lungs were perfused with ice cold phosphate buffered saline (PBS) and subsequently with 1% paraformaldehyde (PFA), extracted and fixed in 4% PFA. After dehydration, lungs were embedded in paraffin using standard procedures. To quantify the degree of muscularization of pulmonary arteries, 3 μm tissue sections were stained for α-smooth muscle actin (Sigma-Aldrich, #2547) and von Willebrand factor (Dako, A0082) using standard immunohistochemical protocols. The degree of muscularization was defined by α-smooth muscle actin positive parts as percentage of the total pulmonary artery cross section: non-muscularized: < 25%, muscularized: ≥ 25%. Data are shown as ratio of muscularized to non-muscularized arteries. Furthermore, the medial wall thickness was determined in vessels with a diameter of 20–70 μm as well as the lumen area (defined as the area within the lamina elastic interna). Osteopontin was visualized by standard immunohistochemistry using anti-osteopontin antibodies (polyclonal anti-osteopontin, #AF808, R&D Systems) and immunoreactive area was quantified in five animals per group.

The pulmonary artery was dissected and perfused with ice-cold PBS with 1% penicillin/streptomycin (P/S). The artery was incubated in an enzyme mixture (collagenase type I, elastase, trypsin inhibitor) in DMEM supplemented with 20% fetal bovine serum (FBS) for 15 min at 37 °C. To terminate the digestion procedure the arteries were washed with PBS with P/S. The surrounding adventitia and endothelium was mechanically removed, and the artery was cut into pieces and incubated in the enzyme mixture for additional 90 min at 37 °C. Afterwards, primary PASMCs were extracted by centrifugation and transferred to cell culture plates. PASMCs were cultured in DMEM containing 20% FBS and P/S at 37 °C and 5% CO₂.

Standard immunoblotting was performed using PAGE after protein isolation from aortae. The following primary antibodies were used for protein expression analyses: monoclonal anti-PPARγ (#2443, Cell Signaling), anti-GAPDH (#MAB374, Millipore).

Isolated PASMCs were incubated on glass tissue slides and fixed with 4% PFA. Slides were washed three times in ice-cold PBS prior to permeabilization with Triton-X. Following one additional washing step, cells were blocked with 5% BSA and incubated with antibodies against α-smooth muscle actin (#2547, Sigma Aldrich) and PPARγ (#2443, Cell Signaling). Slides were washed, and incubated with secondary antibodies (anti-rabbit Alexa Fluor 488 and Cy3-conjugated anti-rabbit IgG, Jackson Immuno Research) and counterstained with 4′,6-diamidino-2-phenylindole (DAPI) for visualization of cell nuclei.

After post mortem isolation of the heart, the right ventricle (RV) was separated from the left ventricle (LV) and ventricular septum (S), and the wet weights were determined. RV hypertrophy was calculated using the Fulton’s index: RV/(LV + S) [23]. Additionally, 3 μm paraffin sections of the RV and LV were stained with hematoxylin and eosin. Cardiomyocyte cross-sectional area (CSA) was determined in the RV and LV as measures of cardiac remodelling. An area of ~ 50 cardiomyocytes in each ventricle from n = 4–5 animals per group were evaluated. Microscopic images were analyzed in an observer-blinded manner applying Cell D Imaging Software.

RNA was isolated utilizing the RNeasy Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instruction for purification from cells and heart tissue. Synthesis of cDNA was performed using the High Capacity RNA-to-cDNA™ Kit (Applied Biosystems, Darmstadt, Germany). Gene expression analysis by quantitative real-time polymerase chain reaction (qPCR) (SybrGreen, Applied Biosystems) was performed in duplicate with an Mx3000P cycler (Stratagene; Agilent Technologies) and normalized to the housekeeping gene 18S. The following primer sequences (at final concentrations of 100 nM) were used: 18S rRNA (Rn18s) 5′-GGACTCTTTCGAGGCCCTGTA-3′, 5′-CACCAGACTTGCCCTCCAAT-3′; Osteopontin (SPP1) 5′-CTCCAATCGTCCCTACAGTCG-3′, 5′-AGGTCCTCATCTGTGGCATC-3′, Egr-1 (EGR1) 5′-CCGAGCGAACAACCCTATGA-3′, 5′-CGAGTCGTTTGGCTGGGATA-3′; BNP (NPPB) 5′-GGTCCAGCAGAGACCTCAAAA-3′, 5′-GCCAGGAGGTCTTCCTACAAC-3′ (forward and reverse, respectively).

Data were expressed as means ± SEM or medians and ranges in case of data without normal distribution. We analyzed a full two-factor linear model with repeated measurements for the glucose tolerance test (GTT) and the insulin tolerance test (ITT) with SPSS V25 (IBM). The model was constant + diet + genotype + diet * genotype with measurement time series GTT/ITT as inner subject factor. Difference to baseline was chosen as within-subject contrast. The assumption of normal data distribution was checked for 2 × 2 subgroups by a Kolmogorov-Smirnov test with Lilliefors correction. We performed a Box-test to check for homogeneity of covariances of different groups. The homogeneity assumption did not hold, but Pillai-Bartlett trace was found to be more robust than alternatives when the assumption of homogeneity was violated but sample sizes of the groups were equal [24]. If the maximum ratio of sample sizes was 1.5:1 or less, the F statistic is reported to be robust against violation of the homogeneity of variances assumption [25]. This condition holds since our subgroup sample sizes were nearly equal. We report significance of multivariate analysis based on exact F-statistics and Pillai trace. For further testing the with-in subject effects we tested for sphericity by a Mauchly test and used an appropriate correction (e.g. Huynh Feldt) for the within-subject effects. Homogeneity of variance between within-subject contrasts was asserted by a Levene test in GTT and ITT. Significance of contrasts was assessed by an F-test. Model effects (main effects and interactions) were compared with Tukey correction. A full model univariate two-factor ANOVA was performed with other parameters to test for factor (diet, genotype) effects and interaction between factors. Tests for heteroskedasticity were considered. In case of deviation from normal distribution a logarithmic transformation of the parameter was performed. Transformed parameters did not significantly deviate from normal distribution. Additionally, non-parametric tests with original parameters (Mann-Whitney U, Kruskal Wallis) were performed. P-values are given as numbers with the exception of very small p-values (p < 0.001). Alpha< 0.05 was regarded as significant. P-values near alpha are referred to as borderline. Due to the small power type 2 errors are a concern and borderline significance may be neither an assertion nor a rejection of the null-hypothesis. Trends were assessed by correlation analysis.

As a prerequisite for evaluating the relevance of PPARγ in VSMCs in PAH, we analyzed expression levels in isolated PASMCs from WT and SmPparγ^−/− mice. Immunofluorescence of PASMCs revealed that expression of PPARγ was virtually lost in tissues derived from SmPparγ^−/−, whereas clearly detectable in WT mice (Fig. 1a). The VSMC-specific knockout of PPARγ was further validated in isolated aortae using immunoblotting procedures (Fig. 1b). Thus, SmPparγ^−/− mice served as a valid animal model for analyses of the impact of PPARγ in VSMCs for metabolic disturbances and pulmonary vascular changes.

WT and SmPparγ^−/− mice were subjected to either a low fat-diet (LFD) or a high fat-diet (HFD) for a total of 24 weeks. Both LFD- (~ 25% of body weight increase) and HFD-fed (~ 70% of body weight increase) mice displayed significant weight gain over time (Fig. 1c). Loss of PPARγ in VSMC (SmPparγ^−/−) in HFD mice further enhanced body weight.

All study groups were subjected to repetitive metabolic phenotyping. Blood glucose levels at baseline GTT (0 weeks) displayed a similar curve pattern in WT and SmPparγ^−/− mice, without evidence of a genotype-specific difference (Fig. 2a). After 12 weeks of diet (Fig. 2b), HFD-fed mice displayed significantly higher fasting glucose levels (p < 0.001) and higher serum glucose throughout the GTT until 150 min, also evidenced by quantification of the area under the curve (AUC; p < 0.001 LFD vs. HFD). Highest glucose levels were detected at 30 min in HFD-fed WT mice, while the maximum in HFD-fed SmPparγ^−/− mice was seen at 60 min, indicating glucose intolerance in these mice. After 24 weeks of diet (Fig. 2c), again glucose levels after fasting as well as after glucose challenge in both HFD-treated WT and SmPparγ^−/− mice were higher compared to their genotype-related LFD-treated WT and SmPparγ^−/− animals.

At baseline (0 weeks) SmPparγ^−/− displayed no major difference in insulin tolerance compared to their WT littermates (Fig. 3a). After 12 weeks (Fig. 3b) and 24 weeks (Fig. 3c) only LFD-fed WT mice showed preserved insulin sensitivity, while the other groups displayed insulin resistance. Indeed, this pattern became more evident at 24 weeks, now showing a more profound difference of insulin sensitivity in SmPparγ^−/− LFD-fed mice compared to their WT littermates. Thus, SmPparγ^−/− in LFD resembled the impact of HFD in WT mice. This underlines an impact of smooth muscle PPARγ for whole body insulin sensitivity. Noteworthy, HFD did not further significantly increase the level of insulin resistance in SmPparγ^−/− (Fig. 3c).

We conclude that both diet and genotype have an effect on glucose metabolism with a relevant interaction.

Food intake was equal between LFD wild-type and LFD SmPparγ^−/− mice (1.80 ± 0.34 g/18 h and 1.75 ± 0.38 g/18 h) and between HFD wild-type and HFD SmPparγ^−/− mice (1.42 ± 0.27 g/18 h and 1.44 ± 0.34 g/18 h).

Following sacrificing the animals after 24 weeks, organ measurements were performed. As shown in Table 1, liver and kidney weights were higher in HFD-treated mice compared to LFD-mice. Additionally, perirenal fat and epididymal fat tissue were higher in HFD- vs. LFD-fed mice. In contrast, brown adipose tissue (BAT) remained unchanged under HFD.

Right ventricular systolic pressure (RVSP), as a measure for pulmonary artery systolic pressure (PASP), was determined, along with systemic artery pressure (SAP) and heart rate. RVSP was significantly enhanced by HFD without interaction of genotype (Fig. 4a). Of note, SAP and heart rate were comparable in all investigated groups (Figs. 4b, 5 and 6c).

We performed additional statistical analyses for potential correlation between insulin sensitivity and RVSP. As shown in Fig. 4d, a significant correlation was calculated, underlining an association between metabolic disturbances and pulmonary vascular pressure.

Muscularization of pulmonary arteries is an early crucial morphological feature of pulmonary vascular remodelling. SmPparγ^−/− enhanced the proportion of muscularized to non-muscularized arteries in both LFD- and HFD-fed animals (Fig. 5a-b), indicating an impact of PPARγ in VSMCs for de novo muscularization of pulmonary arteries.

SmPparγ^−/− LFD-fed mice displayed increased expression of osteopontin (Fig. 5c), an extracellular matrix protein exhibiting both tissue remodelling and inflammatory properties, which further was established as contributing to PAH [26], indicating enhanced extracellular lung tissue remodelling in these mice. Similarly, HFD also enhanced osteopontin content in WT animals.

As a measure of RV hypertrophy, the Fulton’s index was determined post mortem. The Fulton’s index showed a median of 0.26 (95% confidence interval (CI) 0.23–0.28) in LFD WT animals and did not increase in HFD-fed WT mice (median = 0.27, 95% CI 0.24–0.29). There was no marked difference related to the genotype in both diets (Fig. 6a).

Changes in cardiomyocyte morphology are integral components of heart remodelling associated with PAH. Figure 6b depicts representative hematoxylin-eosin stainings and corresponding quantification. We evaluated the cross-sectional area (CSA) of cardiomyocytes as one parameter displaying heart tissue remodelling. Figure 6c depicts that neither HFD nor SmPparγ^−/− reached a significant effect on CSA in the RV and LV after 24 weeks treatment in the subgroups.

Transcript analyses were performed to detect potential differences in gene expression that had yet transferred to only minor morphological changes but, in particular, functional changes in the RV as demonstrated by increased RVSP. Brain natriuretic peptide (BNP) is an established and widely used clinical parameter for heart function and failure, and ventricular remodelling. While BNP was shown being a prognostic marker in PAH [27], others ruled out the NT-proBNP response to exercise as a disease severity assessment marker of PAH [28]. No significant upregulation of BNP was detectable in the RV due to the application of different diets (Fig. 7a). Osteopontin had previously been shown being enhanced in the RV in animal models of PH [29]. HFD upregulated osteopontin in the RV (Fig. 7b). Further, early growth response protein 1 (Egr-1), critically involved in vascular remodelling in PAH [30], was higher expressed in the RV in SmPparγ^−/− mice in both LFD- and HFD treated animals (Fig. 7c).