Cytokine Response to Diet and Exercise Affects Atheromatous Matrix Metalloproteinase-2/9 Activity in Mice

Abstract

Background: The aim of this study is to identify the principal circulating factors that modulate atheromatous matrix metalloproteinase (MMP) activity in response to diet and exercise.

Methods and Results: Apolipoprotein-E knock-out (ApoE^−/−) mice (n=56) with pre-existing plaque, fed either a Western diet (WD) or normal diet (ND), underwent either 10 weeks of treadmill exercise or had no treatment. Atheromatous MMP activity was visualized using molecular imaging with a MMP-2/9 activatable near-infrared fluorescent (NIRF) probe. Exercise did not significantly reduce body weight, visceral fat, and plaque size in either WD-fed animals or ND-fed animals. However, atheromatous MMP-activity was different; ND animals that did or did not exercise had similarly low MMP activities, WD animals that did not exercise had high MMP activity, and WD animals that did exercise had reduced levels of MMP activity, close to the levels of ND animals. Factor analysis and path analysis showed that soluble vascular cell adhesion molecule (sVCAM)-1 was directly positively correlated to atheromatous MMP activity. Adiponectin was indirectly negatively related to atheromatous MMP activity by way of sVCAM-1. Resistin was indirectly positively related to atheromatous MMP activity by way of sVCAM-1. Visceral fat amount was indirectly positively associated with atheromatous MMP activity, by way of adiponectin reduction and resistin elevation. MMP-2/9 imaging of additional mice (n=18) supported the diet/exercise-related anti-atherosclerotic roles for sVCAM-1.

Conclusions: Diet and exercise affect atheromatous MMP activity by modulating the systemic inflammatory milieu, with sVCAM-1, resistin, and adiponectin closely interacting with each other and with visceral fat.

Atheromatous matrix metalloproteinase (MMP) activity is related to plaque vulnerability and rupture, the final common pathway for the thromboembolic complications of atherosclerosis, such as ischemic stroke or myocardial infarction.^1–4 Identifying the principal factors that affect atheromatous MMP activity may help predict and prevent these events.

Using an MMP-activatable protease reporter, we previously described that 10-weeks of treadmill exercise reduced in vivo atheromatous MMP-2/9 activity in Apolipoprotein-E (ApoE) knock-out (ApoE^−/−) mice with pre-existing plaque, even if the animals remained on a high fat diet and plaque growth was not attenuated.⁵ The study suggested that circulating adiponectin and soluble vascular cell adhesion molecule-1 (sVCAM-1) might be related to the reduced MMP activity in atheromas, but the mechanisms for this effect were not known. The present study aims to address the mechanisms leading to this observation, relating the MMP-modulating effect of diet and exercise to cytokine homeostasis. We also considered body mass and visceral fat amount, which could be affected by diet and exercise, and ultimately contribute to obesity-related re-configuration of the systemic cytokine network.

We hypothesized that there is a cytokine pathway leading to reduced or enhanced MMP activity, and that we could identify the principal components and most important effectors of this cascade, in our model system.

Cohorts of ApoE^−/− mice with pre-existing plaque, fed either a high-fat diet or a normal chow diet, underwent either 10 weeks of exercise training or no exercise, forming 4 experimental groups. We measured atherosclerosis-related cytokines using bead-based multiplex assays, lipid levels, and visceral fat, as determined by micro-computed tomography (microCT), and correlated these outcome measures to ex vivo atheromatous MMP activity. The goal of the experiments was to determine the cytokine mechanism by which diet and exercise affect plaque-destabilizing protease activity. Factor analysis (see Supplementary Introduction), an unbiased statistical method for dimension reduction,⁶ was performed on the serological variables, which were measured to extract the principal components that could explain the observed effects of diet and exercise. Path analysis (see Supplementary Introduction) was then performed to devise a pathway of interactive cytokines that contribute to the altered MMP-2/9 activity in atheromas.

Methods

Further details are available in the Supplementary Methods.

Animals and Experimental Groups

After 1 week of adaptation, 8-week-old ApoE^−/− mice (~22–28 g) were fed either a normal chow diet (ND) (n=28) or a Western diet (WD) (n=28) for a total of 18 weeks. Our previous study⁵ established that 8 weeks of WD in this strain was sufficient to induce atherosclerotic plaques in the aorta or branch arteries. Starting at the end of these 8 weeks, for the following 10 weeks, half of each diet group of mice were trained to run on a treadmill. Thus, there were a total of 4 experimental groups (n=14 per group): ND without exercise training (ND−ET), ND with exercise training (ND+ET), WD without exercise training (WD−ET), and WD with exercise training (WD+ET).

In two separate sets of experiments using WD-fed ApoE^−/− mice with pre-existing plaque (n=28), in vivo or ex vivo MMP imaging of carotid atheromas and quantitative histological studies was performed after either a short-term (2 weeks) period of exercise training or no training. The timing for imaging and histology was determined so as to observe the early evolution to better investigate the potential causality for the statistical link between pro-/anti-atherosclerotic mechanisms and plaque MMP activity in the longer-term main experiments.

Synthesis of the MMP-2/9 Activatable Molecular Imaging Probe

A polymeric nanoparticle-based MMP-2/9 activatable probe was synthesized, as reported previously.^5,7–9

Experimental Procedures

The experimental procedures used were similar to those of a previously published study.⁵ However, a group (ND+ET) was added as an additional control in the current study, and abdominal microCT was performed to measure the amount of visceral fat. At the end of the study period, 4 µmoL MMP-2/9 probe in 150 µL phosphate-buffered saline (PBS) was intravenously injected.

Four hours after fasting and 24 h after tail vein injection of the probe, abdominal microCT imaging was performed. After blood was collected from the retro-orbital plexus, the animals were euthanized, and the aortas were carefully harvested. Ex vivo MMP-2/9 imaging of the aortas was performed using a Cy5.5 NIRF reflectance imaging machine. To measure plaque size, Oil-Red-O staining was performed for the entire aorta (n=10 randomly selected mice per group).

After centrifugation of the blood samples, a Luminex bead-based multiplexing assay (Millipore, Billerica, MA, USA) was performed to measure serum MMP-9 and 9 cytokines: sVCAM-1, interleukin (IL)-1β, IL-6, IL-10, adiponectin, resistin, leptin, monocyte chemoattractant protein (MCP)-1, and tumor necrosis factor (TNF)-α. In addition, a commercial assay kit (EHDL-100 and DIGL-200; BioAssay-Systems, Hayward, CA, USA) was used to measure low-density lipoprotein+very low-density lipoprotein (LDL) cholesterol and high-density lipoprotein (HDL) cholesterol levels.

Exercise Training

The animals in the ND+ET group or WD+ET group were trained, as previously published,⁵ to run on a treadmill with a rubber belt driven at a controlled speed (DJ-344; Daejong, Korea) for 30 min/day, 5 days/week.

Abdominal microCT Imaging and Quantification of Visceral Fat Amount

microCT imaging was performed, as previously published.¹⁰

Near-Infrared Fluorescent Imaging and Quantification of MMP Activity

Ex vivo NIRF imaging and lesion quantification were performed, as reported previously.^5,7,8,11

Measurement of Plaque Size

Atherosclerotic lesion size was determined using the En Face Method.^5,8

Statistical Methods

Data are presented as mean±standard error. All statistical analyses were conducted using the SPSS software package (SPSS 18.0; SPSS Inc., Chicago, IL, USA) or the Amos software package (Amos 21.0; SPSS Inc.). An explorative factor analysis⁶ was conducted to identify principal components (principal component analysis, PCA¹²) in the serological datasets. Based on the PCA results, path analysis was performed to estimate direct and indirect contributions of principal factors to atheromatous MMP-2/9 activity. The selection of independent variables (i.e., individual cytokines) for path analysis was based upon their relationships with the dependent variable (i.e., atheromatous MMP activity), as well as their relative importance within each of the PCA-derived factors (i.e., clusters of cytokines).

A Second Set of Experiments for In Vivo Imaging and Histopathological Assessment of Carotid Atheroma, Performed After Short-Term Exercise Training

Eight-week-old ApoE^−/− mice (n=18) were additionally used. After 1 week of adaptation, mice were fed a WD for a total of 10 weeks. Starting at the end of the 8 weeks, for the following 2 weeks, half of the mice were trained to run on a treadmill. Thus, there were a total of 2 experimental groups (n=9 per group): shorter-term WD−ET and shorter-term WD+ET.

In vivo MMP imaging was performed at the 8^th (baseline) and 10^th week (follow up), as previously reported,⁹ with some modifications. After the follow-up in vivo imaging, the carotid artery was carefully excised and stored for histology.

Key inflammatory components in carotid atheromas were quantified after immunohistochemical staining for macrophages, MMP-9, and VCAM-1. A priori selection of these immunomarkers was based on statistical analyses of data from the first set of experiments. In addition, staining for von Willebrand factor (vWF) was performed to delineate an intact endothelial lining in a section adjacent to each VCAM-1-stained section.

Immunopositive areas were quantified, as previously described.^5,8,11 The extent of endothelial VCAM-1 expression is reported as a VCAM-1/vWF area percentage.

A Third Set of Shorter-Term Exercise Training Experiments for Western Blot (for M1 vs. M2 Macrophages) and Masson’s Trichrome Staining (for Collagen Fibers) Assessment of Carotid Atheromas

Eight-week-old ApoE^−/− mice (n=10) were additionally studied to see if exercise training, performed as in the second set of experiments, would be associated with less pro-inflammatory M1 (vs. anti-inflammatory M2) macrophages and more collagen fibers in carotid atheromas, when compared with WD only. After shorter-term exercise training, the left carotid artery was homogenized with a homogenizer (Kimble Chase, Rockwood, TN, USA), using a protein extraction kit (PRO-PREP^TM Protein Extraction Solution; INtRON Biotechnology, Sungnam, Korea) supplemented with a cocktail of protease inhibitors. The homogenates were centrifuged at 13,000 rpm (about 20,000 g) for 15 min at 4℃, and the supernatants were collected. Proteins extracted from the left carotid artery tissue were quantitated using a Pierce^TM BCA Protein Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA), separated by 8% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), and analyzed by using a Western blot, with an enhanced chemiluminescence detection kit (Thermo Fisher Scientific), for F4/80-positive macrophages (Proteintech, Chicago, IL, USA), M1 macrophages (inducible nitric oxide synthase, iNOS; Abcam, Cambridge, UK) and M2 macrophages (arginase-1; Abcam).

Finally, we stained the right carotid artery (predefined equidistant axial sections, six 4-µm-thick sections with 100 μm spacing/animal for Masson’s trichrome) and quantified blue-stained areas to calculate the relative proportion of collagen within plaques of the right carotid artery.

Results

There were a total of 4 experimental groups (n=14 per group): ND−ET, ND+ET, WD−ET, and WD+ET. Table S1 summarizes the results of cytokine assays, lipid panel tests, and ex vivo atheromatous MMP-2/9 near-infrared fluorescent (NIRF) imaging, visceral fat quantification by abdominal microCT, and plaque size.

At baseline, body weight did not differ across the 4 groups. At the end of the study period, body weight was higher in WD-fed animals than in ND-fed animals (both _adjP<0.01; Figure 1A). In these ApoE^−/− mice with pre-existing atheromas, 10 weeks of treadmill exercise appeared to slightly reduce the body weight in animals on a WD, which, however, was not statistically significant. In ND-fed animals, exercise training did not change the body weight, either. Likewise, exercise did not reduce visceral fat (Figure 1B) or plaque size (Figure 1D) in either ND-fed mice or WD-fed mice. However, exercise did reduce the atheromatous MMP activity ex vivo in WD-fed animals (_adjP<0.01), which was not the case in ND-fed animals (Figure 1C). Visceral fat amount (all corrected P<0.01) and plaque size (all _adjP=0.006–0.015) were higher in the 2 WD groups than in the 2 ND groups (Figure 1B,D). However, plaque MMP activity was lower not only in the ND groups but also in the WD+ET group, compared with the WD−ET group (all _adjP<0.01) (Figure 1C).

Figure 1.

Effects of diet and exercise on weight, visceral fat amount, and aortic atheromas in apolipoprotein-E knock-out (ApoE^−/−) mice. Bar graphs represent mean±SE. Following Kruskal-Wallis tests, adjusted P values (*<0.05, **<0.01) were obtained by using post-hoc Mann-Whitney U-tests with Holm-Bonferroni correction to adjust for multiple comparisons: 6 possible pairwise comparisons among the 4 groups.

In the correlation matrix of 12 circulating factors including MMP-9, cytokines, HDL, and LDL (Table S2), the determinant (value=0.001) was greater than the necessity value to avoid multicollinearity (0.00001). The Kaiser-Meyer-Olkin test (value=0.62) to measure sampling adequacy indicated that the data were appropriate for PCA. Bartlett’s test was highly significant (P<0.01), which also indicated that the data were appropriate for PCA, because there were relationships between the variables. Using the 12 variables, PCA was performed to extract factors with the eigenvalues greater than 1. As a non-orthogonal rotation method, direct oblimin rotation was used after considering that the extracted factors might correlate with each other. Thus, 4 principal components were identified (Table).

Table. Principal Component Analysis of Atherosclerosis-Related Serologic Factors in Apolipoprotein-E Knock-Out Mice on a Normal Diet or Western Diet With/Without Exercise

	Components
	I	II	III	IV
TNF-α	0.944
IL-6	0.917
MCP-1	0.782
Resistin	0.745
MMP-9	0.700
VCAM-1		0.753
Adiponectin		−0.711
HDL		−0.643
IL-1β			0.917
IL-10			0.874
Leptin				0.907
LDL				0.652

The component matrix, a matrix of the factor loadings (scores) for each variable onto each factor, was calculated after a non-orthogonal rotation method (direct Oblimin rotation). TNF-α, tumor necrosis factor-α; IL, interleukin; MCP-1, monocyte chemoattractant protein-1; MMP-9, matrix metalloproteinase-9; VCAM-1, vascular cell adhesion molecule-1; HDL, high-density lipoprotein; LDL, low-density lipoprotein.

The First Principal Component Was Composed of Five Pro-Inflammatory Cytokines, Including Resistin, Which Correlated With Atheromatous MMP-2/9 Activity

TNF-α levels were lower in the ND+ET group compared with the 2 WD groups (Figure S1A): the WD+ET group and the WD−ET group (both _adjP<0.05). Similarly, resistin levels were lower in the ND+ET group than in the 2 WD groups (both _adjP<0.01; Figure 2A). In addition, the ND–ET group had lower resistin levels than the WD+ET group (_adjP<0.01).

Figure 2.

Effects of diet and exercise on principal serologic factors, which were correlated with atheromatous MMP-2/9 activity in apolipoprotein-E knock-out mice. (A–C) Bar graphs represent mean±SE. Following Kruskal-Wallis tests, adjusted P values (*<0.05, **<0.01) were obtained by using post-hoc Mann-Whitney U-tests with Holm-Bonferroni correction to adjust for multiple comparisons: 6 possible pairwise comparisons among the 4 groups. (D–F) Scatter plots with a line of best fit (r, Pearson coefficient). sVCAM-1, soluble vascular cell adhesion molecule-1; MMP-2/9, matrix metalloproteinase-2/9; NIRF, near infrared fluorescence; A.U., arbitrary unit.

Monocyte chemoattractant protein (MCP)-1 levels differed across the groups with a marginal significance (P=0.05); and there were no inter-group differences (Figure S1B). Interleukin (IL)-6 (Figure S1C) and MMP-9 (Figure S1D) levels did not significantly differ across the groups, although IL-6 levels tended to be lower in the ND-fed animals compared with the WD-fed animals.

Among these variables in the first component, both TNF-α levels and MCP-1 levels correlated with plaque size (r=0.453, P<0.01 and r=0.438, P<0.01, respectively). However, only resistin levels showed a significant correlation with both plaque size (Figure S2) and atheromatous MMP-2/9 activity (Figure 2D) (r=0.48, P<0.01 and r=0.291, P<0.05, respectively). Furthermore, all the other variables in this component correlated with resistin (Table S2). Therefore, only resistin was included as a variable for modeling cytokine paths to atheromatous MMP-2/9 activity.

The Second Principal Component Was Composed of Anti-Atherosclerotic Adiponectin and HDL, as Well as Pro-Atherosclerotic sVCAM-1, Which Correlated With Atheromatous MMP-2/9 Activity

Adiponectin levels (Figure 2B) but not HDL levels (Figure S3A) differed across the groups (P<0.01 and 0.484, respectively). The ND−ET group and the ND+ET group had higher adiponectin levels than the WD−ET group (_adjP=0.035 and <0.01, respectively). In addition, the WD+ET group tended to have higher adiponectin levels compared with the WD−ET group (_adjP=0.076). Adiponectin levels were inversely correlated with body weight (r=−0.412, P<0.01) and visceral fat amount (r=−0.396, P<0.01) (Figure S3B). Adiponectin levels and HDL levels, both of which correlated with each other (Figure S3C), did not correlate with atheromatous MMP-2/9 activity (Figures 2E and S3D, respectively).

The sVCAM-1 levels were lower in the ND−ET group (_adj P<0.05) and ND+ET group (_adjP<0.05) than in the WD−ET group (Figure 2C). However, between the 2 ND groups and the WD+ET group, sVCAM-1 levels did not differ significantly. The WD+ET group had relatively low sVCAM-1 levels compared with the WD−ET group, which, however, did not reach a statistical significance (unadjusted P=0.108). sVCAM-1 levels correlated with atheromatous MMP activity (r=0.383, P<0.01; Figure 2F) as well as plaque size (r=0.416, P<0.01; Figure S4). Therefore, the positive-signed sVCAM-1 was included as a variable for modeling cytokine paths to atheromatous MMP-2/9 activity. Between the 2 variables with a negative sign, adiponectin was selected because the adipokine levels better reflected the effects of diet and exercise than HDL levels, as described above.

The Third Principal Component Was Composed of Anti-Inflammatory IL-10 and Pro-Inflammatory IL-1β, Both of Which Did Not Correlate With Atheromatous MMP-2/9 Activity

Please refer to the Supplementary Results and Figure S5 for more information.

The Fourth Principal Component was Composed of LDL and Leptin; Only the Latter Tended to Correlate With Atheromatous MMP-2/9 Activity

Please refer to the Supplementary Results and Figure S6 for more information.

Models of Cytokine Paths to Atheromatous MMP-2/9 Activity

Based on the aforementioned variable selection procedures and the results of the bivariate correlations, an initial hypothetical model was formulated using the 4 variables (resistin, adiponectin, sVCAM-1, and MMP-2/9 activity), with reciprocal paths between them (Figure 3A). A revised model without reciprocal paths was generated (Figure 3B) by conducting a ‘specification searches’ function in the Amos software and examining how well each of the various models fit the data. This explorative procedure assists in finding the best theoretical model. The revised model fit the data well (χ²=0.430, d.f.=2, P=0.807, NFI=0.977, CFI=1.000, RMSEA=0.000) and accounted for 19% of the variance in atheromatous MMP-2/9 activity.

Figure 3.

Path analyses to estimate direct and indirect contributions of principal serologic factors to atheromatous matrix metalloproteinase (MMP)-2/9 activity in apolipoprotein-E knock-out mice on a normal diet or Western diet with/without exercise. (A) An initial hypothetical model. (B) A revised model. (C) Another model that incorporates the variable, ‘visceral fat amount’. (D) A more comprehensive model incorporating 2 latent variables (anti-atherosclerotic vs. pro-atherosclerotic). Black lines, hypothetical path; red lines, significant (P<0.05) inhibitory path; blue lines, significant (P<0.05) excitatory path; light blue lines, (P<0.1) excitatory path; gray lines, non-significant (P>0.05) path. Numbers are standardized regression coefficients for each path. sVCAM-1, soluble vascular cell adhesion molecule-1; HDL, high-density lipoprotein; IL, interleukin.

Three of four selected hypothesized paths were significant (all P<0.05), which is displayed as red (inhibitory path) and blue (excitatory path) lines with corresponding standardized regression coefficients (Figure 3B). sVCAM-1 was directly positively related to atheromatous MMP activity. However, adiponectin was indirectly negatively related to atheromatous MMP-2/9 activity by way of sVCAM-1. Resistin was indirectly positively related to atheromatous MMP activity by way of sVCAM-1. In addition, resistin also had a non-significant trend (P=0.098) toward being directly positively related to atheromatous MMP-2/9 activity, which was displayed as a light blue line with a corresponding standardized regression coefficient in the path diagram (Figure 3B).

Given the known relationship between visceral fat and resistin and adiponectin levels,^13,14 we generated another model (Figure 3C) to include the variable ‘visceral fat amount’, which was indirectly positively associated with atheromatous MMP-2/9 activity, by way of sVCAM-1 (positively), adiponectin (negatively), and resistin (positively). This model also fit the data well (χ²=3.796, d.f.=3, P=0.284, NFI=0.911, CFI=0.971, RMSEA=0.069). It was estimated that when the visceral fat amount increased by 1 cm³, adiponectin decreased by 1.1 μg/mL, resistin increased by 0.5 ng/mL, sVCAM-1 decreased by 0.3 μg/mL, and thus atheromatous MMP-2/9 activity increased by 3.5 arbitrary units. Conventional regression analyses re-confirmed that sVCAM-1 was an independent predictor of atheroma MMP activity after adjusting for resistin, adiponectin, or visceral fat amount (Table S3).

Based on the results of the aforementioned path analyses, a more comprehensive model incorporating 2 latent variables (anti-atherosclerotic vs. pro-atherosclerotic) was formulated for future studies with a larger sample size (Figure 3D). In the second set of experiments, in vivo imaging and histological studies supported the final model, which had been derived from the prior ex vivo imaging and serological data, by demonstrating: (1) attenuation of the increase of in vivo atheromatous MMP activity after exercise vs. baseline; and (2) smaller immunopositive areas for endothelial VCAM-1 expression, plaque macrophages, and MMP-9 in mice after exercise vs. WD only (Figure S7). As expected, there was a strong positive correlation between atheromatous MMP-2/9 activity (NIRF signal intensities) and immuno-positive areas for plaque macrophages (r=0.683, P=0.007). In the third set of experiments, Western blotting showed that shorter-term exercise training (vs. WD only) was associated with a marginally significant decrease of F4/80-positive macrophages (P=0.068) and a significant decrease of the M1/M2 ratio (P<0.01) in carotid atheromas (Figure 4). Masson’s trichrome staining in Figure S8 shows a relatively high proportion of collagen fibers in carotid atheromas after shorter-term exercise training vs. WD only, although there was no statistical significance in grouped quantification data for all animals (respectively, 52.4±20.4% vs. 45.2±18.0%, P=0.564 by t-test).

Figure 4.

Effects of exercise, assessed by Western blotting for F4/80-positive macrophages, M1 (inducible nitric oxide synthase, iNOS) and M2 macrophages (arginase-1). Ten 8-week-old apolipoprotein-E knock-out mice were on a Western diet (WD) with (n=5) vs. without (n=5) exercise training (ET) for 2 weeks (respectively WD+ET and WD−ET). Note that the M1/M2 (iNOS/arginase) ratio is significantly lower in the WD+ET group than in the WD−ET group. The expression of proteins was quantified by densitometric analysis and normalized to β-actin (with a factor of 100; i.e., expressed in percentage). P values from t-tests. Bars throughout denote mean±SE.

Discussion

Statistical analyses of the complex interplay between 12 circulating cytokines identified 4 principal components contributing directly or indirectly to the diet- and exercise-related modulation of atheromatous MMP-2/9 activity in ApoE^−/− mice with pre-existing plaque. These components are: (1) pro-inflammatory cytokines including resistin, TNF-α, IL-6, MCP-1, and MMP-9; (2) anti-atherosclerotic adiponectin/HDL and pro-atherosclerotic sVCAM-1; (3) anti-inflammatory IL-10 and pro-inflammatory IL-1β; and (4) LDL and leptin. Among these serological variables , resistin, adiponectin, and sVCAM-1 were determined to be the main effectors.

Atheromatous MMP-2/9 protease activity was increased possibly due, in part, to circulating sVCAM-1, a factor that is known to increase the process of endothelial adhesion and trans-endothelial migration of monocytes in atherosclerosis. sVCAM-1 is thought to derive from proteolytic cleavage and shedding from endothelial cells,¹⁵ although further studies should clarify other cell types of origin, and mechanisms of generation and clearance of the molecules.¹⁶ Our study showed that sVCAM-1 levels were affected by 2 adipokines; resistin increased sVCAM-1 levels, whereas adiponectin decreased them. Resistin levels were positively correlated with the amount of visceral fat, whereas adiponectin levels were negatively correlated with it.

sVCAM-1 has shown inconsistent associations with atherosclerosis or cardiovascular events.¹⁷ In the present study, only sVCAM-1 but not the other 11 atherosclerosis-related serologic markers were independently associated with atheromatous MMP activity, which was assessed by using cleavage-activatable molecular imaging agents. The distinct association that was observed in mice with pre-existing plaque supports the notion that sVCAM-1 is a strong predictor of the extent and severity of established atherosclerosis.^18,19 Thus, VCAM-1-mediated recruitment of inflammatory monocytes into established atheromas may be a dominant factor that determines the net proteolytic activity⁸ within the atheromas. A continued activation of the vascular endothelium and persistent recruitment of inflammatory cells could keep inflammation high within atheromas^9,11 and thus lead to plaque destabilization. Eventually, sufficient proteolytic activity and inflammation will thin out and rupture the fibrous plaque cap, exposing the lipid core to the circulation, activating the blood and platelet coagulation cascades, and thus lead to acute thromboembolic vascular events.

A previous clinical study showed that a 12-week bicycle exercise, 30-min of training 5 days per week, reduced circulating sVCAM-1 levels in patients with chronic heart failure.²⁰ In the present study animals, sVCAM-1 levels were lower in the 2 ND groups than in the 2 WD groups. Unlike the diet-related changes in sVCAM-1 levels, treadmill running did not reduce the cytokine levels, while the exercise training significantly attenuated the WD-induced increase of atheromatous MMP activity.

Adiponectin was reported to significantly suppress the VCAM-1 expression in atherosclerotic lesions of ApoE^−/− mice.²¹ In addition, exercise increases the secretion of adiponectin from adipocytes.^22,23 Indeed, exercise training in our study, which was not accompanied by a significant reduction in either body weight or visceral fat amount, could almost offset the WD-related decrease in circulating adiponectin levels. However, the exercise-mediated restoration of adiponectin levels was not associated with a significant reduction in circulating sVCAM-1 levels. A recent study²⁴ showed that adenovirus-expressing adiponectin DNA could increase serum adiponectin and decrease serum MMP-9 in ApoE^−/− mice, when the intervention was initiated relatively early (12 weeks) along with high-fat diet. In our experimental setting, path analysis revealed that adiponectin exerted no direct effect on atheromatous MMP-2/9 activity; only a weak indirect effect through affecting sVCAM-1 levels was observed. Therefore, further studies are required to discover direct or major indirect pathways for the exercise- and/or adiponectin-mediated reduction in atheromatous MMP-2/9 activity in WD-fed animals with pre-existing plaque.

Regardless of exercise training, the amount of visceral fat was higher in mice on a WD than in mice on a ND. Thus, WD may have increased atheromatous MMP activity, partly by affecting the levels of adipokines from the visceral fat. Path analysis demonstrated that when visceral fat amount increased, adiponectin decreased and resistin and sVCAM-1 increased, which was related to the net increase of atheromatous MMP-2/9 activity.

Resistin was reported to stimulate monocyte-endothelial cell adhesion by upregulating VCAM-1 expression in endothelial cells.^25,26 Because upregulation of endothelial VCAM-1 could be accompanied by the release of the soluble fraction, sVCAM-1, into the blood stream,^19,27,28 resistin could have increased sVCAM-1 and thus increased atheromatous MMP-2/9 activity in the present study, probably via recruitment of more monocytes. As was indicated in the path analysis, resistin could also increase the in vivo protease activity without taking the indirect route via sVCAM-1. Additional mechanisms that link resistin and plaque MMP activity should be explored in future studies.

Resistin levels were lower in the ND+ET group than in the 2 WD groups. Exercise training with WD did not reduce the resistin levels but non-significantly elevated them. Thus, resistin levels in the WD+ET group became significantly higher compared with the 2 ND groups. In addition, ND decreased the serum levels of leptin, which plays a key role in regulating energy intake and expenditure.²⁹ In contrast, WD increased leptin levels, which appeared to be non-significantly attenuated by exercise. Taken together, these results suggest the importance of dietary intervention or combining both dietary intervention and exercise training in achieving optimal levels of resistin and leptin.

Neither plaque size nor serum MMP-9 levels predicted atheromatous MMP-2/9 activity. Three observed variables (sVCAM-1, resistin, and adiponectin) in our path analysis model explain ~19% of the protease activity in atheromas. Further studies with a larger sample size should be performed to formulate a better model incorporating more of the key mediators. However, the low explanation power might suggest that the protease-mediated molecular optical imaging technique cannot be completely replaced by multiplex cytokine assays, traditional anatomy-based imaging, or a combination of blood assays and structural imaging. Even a large population-based study on carotid intima media thickness demonstrated that only 16% of the variance in the anatomical imaging data could be explained by a final multivariable model containing adiponectin and 9 traditional risk factors such as body mass index and smoking.³⁰

There are several limitations to be considered in interpreting our results (please refer to the Supplementary Discussion for more information). Further studies with a larger sample size are required to confirm if exercise training could actually decrease WD-mediated plaque vulnerability to rupture by attenuating matrix disorganization that includes degradation of collagen, through the mechanism proposed in this study (i.e., exercise-mediated anti-atherosclerotic modulation of the adiponectin/resistin/VCAM-1-related systemic cytokine network, culminating in the reduction of monocytes/macrophage recruitment, M1/M2 ratio, and MMP activity in atheromatous arteries).

In conclusion, diet and exercise could affect atheromatous MMP activity in mice with pre-existing plaque by modulating the systemic inflammatory milieu, where visceral fat and several principal cytokines such as sVCAM-1, resistin, and adiponectin closely interacted with each other by taking direct and/or indirect routes ‘towards’ or ‘away from’ the inflammatory protease activity (Figure 5) that may reflect rupture-prone plaque vulnerability. When translated into the clinic, our data on the cytokine network could serve as a useful biomarker panel. Furthermore, our results will contribute to the development of new anti-atherosclerotic therapeutics that could directly modulate cytokines or their networks, which likely play critical roles in a symptomatic manifestation of pre-existing plaques (i.e., ischemic stroke and myocardial infarction).

Figure 5.

Cytokine response to diet and exercise affecting atheromatous MMP-2/9 activity in mice. EC, endothelial cell; VSMC, vascular smooth muscle cell; MMP, matrix metalloproteinase. Green-colored arrow, excitatory path; red-colored line, inhibitory path.

Acknowledgments

This work was supported by Global Research Lab (GRL) program (NRF-2015K1A1A2028228) of the National Research Foundation, funded by the Korean government and the Korea Healthcare Technology R&D Project, Ministry of Health and Welfare (HI12C1847, HI12C0066).

Disclosures

The authors have no conflicts of interest.

Supplementary Files

Supplementary File 1

Supplementary Methods

Supplementary Results

Supplementary Discussion

Figure S1. Effects of diet and exercise on the atherosclerosis-related serological factors in the first component extracted by a principal component analysis (see the main text) in apolipoprotein-E knock-out mice.

Figure S2. Correlation between resistin and plaque size (area % on Oil-Red-O staining) in apolipoprotein-E knock-out mice on a normal diet or Western diet with/without exercise.

Figure S3. Effects of diet and exercise on the atherosclerosis-related serological factors in the second component extracted by a principal component analysis (see the main text) in apolipoprotein-E knock-out mice.

Figure S4. Correlation between soluble vascular cell adhesion molecule (sVCAM)-1 and plaque size (area % on Oil-Red-O staining) in apolipoprotein-E knock-out mice on a normal diet or Western diet with/without exercise.

Figure S5. Effects of diet and exercise on the atherosclerosis-related serological factors in the third component extracted by a principal component analysis (see the main text) in apolipoprotein-E knock-out mice.

Figure S6. Effects of diet and exercise on the atherosclerosis-related serological factors in the fourth component extracted by a principal component analysis (see the main text) in apolipoprotein-E knock-out mice.

Figure S7. Effects of exercise on the carotid atheromas, assessed by in vivo near-infrared fluorescence (NIRF) imaging and histological studies.

Figure S8. Representative animals that show exercise-mediated inhibition of collagen degradation in carotid atheromas, presumably due to reduced macrophage (F4/80) recruitment and matrix metalloproteinase (MMP)-2/9 activity.

Table S1. Atherosclerosis-related measures in apolipoprotein-E knock-out mice on a ND or WD with (+)/without (−) 10 weeks of exercise training (ET)

Table S2. Bivariate correlations between atherosclerosis-related serologic factors in 56 apolipoprotein-E knock-out mice

Table S3. Multiple regression analyses to predict atheromatous MMP-2/9 activity in apolipoprotein-E knock-out mice on a normal diet or Western diet with/without exercise

Please find supplementary file(s);

http://dx.doi.org/10.1253/circj.CJ-16-1196

References

• 1.   Halpert I, Sires UI, Roby JD, Potter-Perigo S, Wight TN, Shapiro SD, et al. Matrilysin is expressed by lipid-laden macrophages at sites of potential rupture in atherosclerotic lesions and localizes to areas of versican deposition, a proteoglycan substrate for the enzyme. Proc Natl Acad Sci U S A 1996; 93: 9748–9753.
• 2.   Uzui H, Harpf A, Liu M, Doherty TM, Shukla A, Chai NN, et al. Increased expression of membrane type 3-matrix metalloproteinase in human atherosclerotic plaque: Role of activated macrophages and inflammatory cytokines. Circulation 2002; 106: 3024–3030.
• 3.   Sukhova GK, Schonbeck U, Rabkin E, Schoen FJ, Poole AR, Billinghurst RC, et al. Evidence for increased collagenolysis by interstitial collagenases-1 and -3 in vulnerable human atheromatous plaques. Circulation 1999; 99: 2503–2509.
• 4.   Sluijter JP, Pulskens WP, Schoneveld AH, Velema E, Strijder CF, Moll F, et al. Matrix metalloproteinase 2 is associated with stable and matrix metalloproteinases 8 and 9 with vulnerable carotid atherosclerotic lesions: A study in human endarterectomy specimen pointing to a role for different extracellular matrix metalloproteinase inducer glycosylation forms. Stroke 2006; 37: 235–239.
• 5.   Shon SM, Park JH, Nahrendorf M, Schellingerhout D, Kim JY, Kang BT, et al. Exercise attenuates matrix metalloproteinase activity in preexisting atherosclerotic plaque. Atherosclerosis 2011; 216: 67–73.
• 6.   Fabrigar LR, Wegener DT, MacCallum RC, Strahan EJ. Evaluating the use of exploratory factor analysis in psychological research. Psychological Methods 1999; 4: 272–299.
• 7.   Kim DE, Kim JY, Schellingerhout D, Kim EJ, Kim HK, Lee S, et al. Protease imaging of human atheromata captures molecular information of atherosclerosis, complementing anatomic imaging. Arterioscler Thromb Vasc Biol 2010; 30: 449–456.
• 8.   Kim DE, Kim JY, Schellingerhout D, Shon SM, Jeong SW, Kim EJ, et al. Molecular imaging of cathepsin B proteolytic enzyme activity reflects the inflammatory component of atherosclerotic pathology and can quantitatively demonstrate the antiatherosclerotic therapeutic effects of atorvastatin and glucosamine. Mol Imaging 2009; 8: 291–301.
• 9.   Shin IJ, Shon SM, Schellingerhout D, Park JY, Kim JY, Lee SK, et al. Characterization of partial ligation-induced carotid atherosclerosis model using dual-modality molecular imaging in ApoE knock-out mice. PLoS One 2013; 8: e73451.
• 10.   Kim DE, Kim JY, Sun IC, Schellingerhout D, Lee SK, Ahn CH, et al. Hyperacute direct thrombus imaging using computed tomography and gold nanoparticles. Ann Neurol 2013; 73: 617–625.
• 11.   Shon SM, Choi Y, Kim JY, Lee DK, Park JY, Schellingerhout D, et al. Photodynamic therapy using a protease-mediated theranostic agent reduces cathepsin-B activity in mouse atheromata in vivo. Arterioscler Thromb Vasc Biol 2013; 33: 1360–1365.
• 12.   Jolliffe IT. Principal component analysis. Springer Series in Statistics, 2nd Edition. New York: Springer-Verlag New York, 2002; pp. 1–9. doi:10.1007/b98835.
• 13.   Kwon K, Jung SH, Choi C, Park SH. Reciprocal association between visceral obesity and adiponectin: In healthy premenopausal women. Int J Cardiol 2005; 101: 385–390.
• 14.   Choi HY, Kim S, Yang SJ, Yoo HJ, Seo JA, Kim SG, et al. Association of adiponectin, resistin, and vascular inflammation: Analysis with 18F-fluorodeoxyglucose positron emission tomography. Arterioscler Thromb Vasc Biol 2011; 31: 944–949.
• 15.   Pigott R, Dillon LP, Hemingway IH, Gearing AJ. Soluble forms of E-selectin, ICAM-1 and VCAM-1 are present in the supernatants of cytokine activated cultured endothelial cells. Biochem Biophys Res Commun 1992; 187: 584–589.
• 16.   Ballantyne CM, Entman ML. Soluble adhesion molecules and the search for biomarkers for atherosclerosis. Circulation 2002; 106: 766–767.
• 17.   Rohatgi A, Owens AW, Khera A, Ayers CR, Banks K, Das SR, et al. Differential associations between soluble cellular adhesion molecules and atherosclerosis in the Dallas Heart Study: A distinct role for soluble endothelial cell-selective adhesion molecule. Arterioscler Thromb Vasc Biol 2009; 29: 1684–1690.
• 18.   Cybulsky MI, Iiyama K, Li H, Zhu S, Chen M, Iiyama M, et al. A major role for VCAM-1, but not ICAM-1, in early atherosclerosis. J Clin Invest 2001; 107: 1255–1262.
• 19.   Ley K, Huo Y. VCAM-1 is critical in atherosclerosis. J Clin Invest 2001; 107: 1209–1210.
• 20.   Adamopoulos S, Parissis J, Kroupis C, Georgiadis M, Karatzas D, Karavolias G, et al. Physical training reduces peripheral markers of inflammation in patients with chronic heart failure. Eur Heart J 2001; 22: 791–797.
• 21.   Okamoto Y, Kihara S, Ouchi N, Nishida M, Arita Y, Kumada M, et al. Adiponectin reduces atherosclerosis in apolipoprotein E-deficient mice. Circulation 2002; 106: 2767–2770.
• 22.   Bluher M, Bullen JW Jr, Lee JH, Kralisch S, Fasshauer M, Kloting N, et al. Circulating adiponectin and expression of adiponectin receptors in human skeletal muscle: Associations with metabolic parameters and insulin resistance and regulation by physical training. J Clin Endocrinol Metab 2006; 91: 2310–2316.
• 23.   Kriketos AD, Gan SK, Poynten AM, Furler SM, Chisholm DJ, Campbell LV. Exercise increases adiponectin levels and insulin sensitivity in humans. Diabetes Care 2004; 27: 629–630.
• 24.   Wang X, Chen Q, Pu H, Wei Q, Duan M, Zhang C, et al. Adiponectin improves NF-kappaB-mediated inflammation and abates atherosclerosis progression in apolipoprotein E-deficient mice. Lipids Health Dis 2016; 15: 33.
• 25.   Hsu WY, Chao YW, Tsai YL, Lien CC, Chang CF, Deng MC, et al. Resistin induces monocyte-endothelial cell adhesion by increasing ICAM-1 and VCAM-1 expression in endothelial cells via p38MAPK-dependent pathway. J Cell Physiol 2011; 226: 2181–2188.
• 26.   Verma S, Li SH, Wang CH, Fedak PW, Li RK, Weisel RD, et al. Resistin promotes endothelial cell activation: Further evidence of adipokine-endothelial interaction. Circulation 2003; 108: 736–740.
• 27.   Calabresi PA, Prat A, Biernacki K, Rollins J, Antel JP. T lymphocytes conditioned with Interferon beta induce membrane and soluble VCAM on human brain endothelial cells. J Neuroimmunol 2001; 115: 161–167.
• 28.   James WG, Bullard DC, Hickey MJ. Critical role of the alpha 4 integrin/VCAM-1 pathway in cerebral leukocyte trafficking in lupus-prone MRL/fas(lpr) mice. J Immunol 2003; 170: 520–527.
• 29.   Friedman JM, Halaas JL. Leptin and the regulation of body weight in mammals. Nature 1998; 395: 763–770.
• 30.   Rundek T, Blanton SH. Response to letter regarding article, “Traditional risk factors are not major contributors to the variance in carotid intima-media thickness”. Stroke 2013; 44: e236.