Caloric Restriction Resolves Atherosclerosis

stCR in obese mice promotes atherosclerosis resolution. We previously showed in multiple mouse models with hypercholesterolemia that substantial lipid lowering results in the resolution of atherosclerosis, as judged by decreased content and inflammatory properties of macrophages (e.g., 24–26). To isolate the effects of weight loss in obesity on established atherosclerosis, a model in which cholesterol levels are not dramatically affected was required. Another consideration in study design was that switching the feeding of a high-fat diet (HFD) to normal chow results in a severe reduction in food intake (27) (mimicking long-term fasting), as well as in epigenetic changes in macrophages and their precursors related to differences in diet composition (e.g., 15, 16). Thus, we adapted a protocol of mild stCR (28), keeping the diet composition the same, to investigate the role of a clinically relevant level of reduced caloric intake and subsequent weight loss in inflammation resolution in atherosclerosis, independent of cholesterol lowering.

Thus, WT mice were treated with Ldlr antisense oligonucleotide to induce LDL receptor (LDLr) deficiency (as we described previously) (29) and fed a HFHC diet ad libitum for 24 weeks to develop obesity and advanced atherosclerotic plaques. Tissues from the baseline (BL) group (i.e., mice after 24 weeks of HFHC diet) were harvested. To examine early changes induced by weight loss, the stCR group was switched to daily feeding of 70% of their ad libitum consumption of the same HFHC diet for an additional 2 weeks (Figure 1A). The data show that after 24 weeks of treatment, mice were obese, and after 2 weeks of stCR, they lost 14.3% of their weight (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/JCI172198DS1). Upon harvest, several tissues were weighed and a reduction in eWAT mass was observed (Supplemental Figure 1B) with no significant changes to the masses of inguinal white adipose tissue (iWAT), brown adipose tissue (BAT), liver, or kidney (Supplemental Figure 1, C–F).

Examination of metabolic parameters showed marked improvements with stCR, including reduced fasting glucose (Supplemental Figure 1G), lower HOMA-IR (a measure of insulin resistance; Supplemental Figure 1H), and improved glucose tolerance (Supplemental Figure 1, I and J). Moreover, plasma cholesterol levels remained elevated after stCR, with nonsignificant changes between the 2 groups (Figure 1B). To investigate the lipoprotein distribution of plasma cholesterol, a subset of plasma samples was fractioned by fast protein liquid chromatography (FPLC), and cholesterol was measured in each fraction. The results showed no differences in the cholesterol levels in LDL or VLDL fractions between groups. Though HDL cholesterol (HDL-c) was higher in the stCR group by FPLC (Figure 1C), direct measurements of HDL-c in additional plasma samples showed no difference between BL and stCR mice (Supplemental Figure 1K).

For the evaluation of atherosclerosis, aortic roots were sectioned and investigated for plaque size and composition. While plaque area was comparable between the groups (Supplemental Figure 1L), the stCR group had fewer macrophages, observed both as a decrease in the area of CD68+ cells (Figure 1D) and their proportion of the total plaque area (Figure 1E). In an independent analysis, consistent findings were observed from digested aortic arches that were analyzed using flow cytometry. These results showed fewer macrophages in aortic arches of stCR mice compared with the BL group (Supplemental Figure 1M). To further establish that the changes in the macrophage content of atherosclerotic plaques were independent of plasma cholesterol levels, we investigated whether these parameters were correlated. Statistical analysis shown in Figure 1F demonstrated no correlation.

The change in plaque cellular composition without a decrease in area is reminiscent of several of our previous studies (e.g., 26, 30, 31), in which inflammation-resolving plaque properties were the predominant feature, with size less significantly affected as the decrease in plaque macrophages was counterbalanced by collagen enrichment, presumably because the content of matrix metalloprotease–producing (inflammatory) macrophages declined. In human plaques, such depletion of macrophages and enrichment in collagen are taken as signs of increased stability (e.g., 32). To quantify changes in the collagen content of plaques, aortic root sections were stained with picrosirius red, and the positive areas were quantified from polarized light images, which represent collagen I, the most common type in atherosclerotic plaques. Indeed, consistent with our previous data and concurrent with decreased plaque macrophages, there was increased collagen content following stCR (Figure 1G). Representative images of aortic roots stained for CD68 and picrosirius red are presented in Figure 1H. As alluded to above, these compositional changes to plaques are increasingly appreciated as more clinically relevant than plaque size in terms of the risk of plaque rupture and myocardial infarction (33).

Leukocyte subpopulations in plaques and eWAT dramatically change with stCR. To investigate at the molecular level how stCR influences the immune compartment in atherosclerotic plaques, first, single-cell suspensions were obtained from aortic arches harvested from mice in both experimental groups. CD45+ cells (i.e., all leukocytes) that were viable were sorted, and transcripts of individual cells were sequenced using the 10× Genomics platform (following the method described) (34). Because we have also obtained adipose tissue CD45+ single-cell transcriptomic data from the same mice (13), the gene expression profiles from both tissues were merged to identify common subpopulations. Quality control and data filtering are displayed in Supplemental Figure 2A.

Unbiased clustering of the single cells found 23 distinct clusters (Figure 2A and Supplemental Figure 2, B–D). To annotate the different clusters, we used a published meta-analysis of plaque single-cell transcriptomes as a reference dataset (35) (Supplemental Figure 2, C and E). Representative top differentially expressed genes (DEGs) in each cluster are presented in Supplemental Figure 2D and Supplemental Table 1. Many of the clusters in our dataset corresponded to previously published work (35); however, some clusters not found previously were identified as well. For these, we used our previously published dataset from the eWAT CD45+ cells (Figure 2A) as the reference dataset (13). Most notable was the appearance of Fcgr4+ macrophages uniquely in our dataset, which was identified due to their preferential accumulation in stCR conditions. Cell proportions were plotted for each tissue in BL and stCR conditions (Figure 2B). Note that while all clusters are shared across eWAT and plaques, their distributions considerably differed in both the obese and stCR conditions.

We also investigated whether obesity and stCR drove similar gene expression in both eWAT and plaques, as well as in distinct leukocyte clusters within each tissue. The expression of each DEG in eWAT was plotted per leukocyte cluster, and its corresponding expression in plaques is shown in Figure 2C. We classified all DEGs (columns) as either BL biased, with log2 fold change ≥ 1 higher expression in plaque BL (blue), or stCR biased, with higher log2 fold change ≥ 1 expression in stCR (red) across all clusters (rows). Many columns (i.e., DEGs) show a signal in multiple rows (cell clusters), indicating that several clusters differentially express the same genes within each tissue, and often in the same direction (either BL or stCR biased). To look further into this, we plotted the number of clusters that shared DEGs in each tissue (agnostic of whether they are BL or stCR biased). Supplemental Figure 2F shows that most changes in gene expression are restricted to a single cell cluster. Numerous genes, however, were differentially expressed in multiple clusters, with some changing coordinately in >15 (Supplemental Figure 2F and Supplemental Tables 2 and 3).

We next investigated whether genes change concordantly (i.e., undergo expression changes in the same direction in response to stCR treatment) across tissues (Figure 2C). For example, macrophages may have changes resulting from their well-known plasticity in different tissue environments, but there are also likely to be similarities related to the common origin of these cells from circulating monocytes or their response to treatment. When looking at individual genes (columns) across tissues (top and bottom panels), several appear to be similarly changing in both eWAT and plaques (as reflected by showing signals in the same columns in the top and bottom plots). This suggests that a core set of genes (Supplemental Tables 2 and 3) is regulated similarly not only between clusters, but also across tissues (e.g., same column in Figure 2C). Most genes, however, appear to be uniquely differentially expressed in 1 tissue or the other (i.e., not showing concordant signal in Figure 2C), consistent with previous studies (36). Supplemental Table 2 summarizes all DEGs shared by 5 or more clusters, further indicating if the expression is BL or stCR biased and in which tissue.

We also aimed to infer cellular communications between the 23 cell clusters. Ligand–receptor interaction analysis was performed (see Methods), and the number of interacting pairs in plaques and eWAT (Supplemental Figure 2G) was plotted. The color represents the number of ligand–receptor pairs found between clusters in the x and y axes. Notably, in both plaques and eWAT, the cells with the highest number of significant ligand–receptor interactions among the leukocytes in both BL and stCR groups are the macrophages (such as Foamy-Trem2 macrophages and activated macrophages), which mostly communicate with other macrophages (Supplemental Figure 2G). In addition, in eWAT of both BL and stCR groups, there were also predicted interactions of inflammatory macrophages with T cells (CD8+, Treg, and NKT).

To further explore the responses of each macrophage cluster to stCR, the log fold change (LFC) values of DEGs compared with BL were hierarchically clustered and examined for pathway enrichment (Figure 2D). First, we performed differential expression analysis (see Methods) between stCR and BL within each macrophage cluster and calculated LFC values for those genes in all other macrophage clusters. We then performed hierarchical clustering on the LFC values across genes (rows) and macrophage clusters (columns).

There were 8 well-defined gene clusters capturing distinct patterns of differential gene expression across the macrophage clusters. Gene Ontology and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses to assess pathway enrichment in each hierarchical cluster were then performed, and the resulting terms are shown in Figure 2D. Interestingly, we have previously observed that Fcgr4+ macrophages accumulate in the eWAT in obese mice following stCR, and they express many genes associated with phagocytosis (13), a process that would be expected to have important functional effects in both tissues. Analysis of plaque macrophage transcriptome changes with stCR indicated that pathways upregulated in Fcgr4+ macrophages include “Fc-gamma receptor-mediated phagocytosis” and “regulation of lipolysis in adipocytes” (Figure 2D). Since Fcgr4+ macrophages were enriched in both plaques and eWAT with stCR, we chose to investigate these cells further and examine whether they play a role in stCR-induced atherosclerosis resolution.

Fcgr4+ macrophages accumulate with weight loss and promote beneficial changes in atherosclerotic plaques. As just noted, in both plaques and eWAT, stCR increased Fcgr4+ macrophages (Figure 2B and Figure 3A). These transcriptomic results were verified at the protein level using immunofluorescence staining of aortic roots and eWAT sections for macrophages (CD68 and F4/80, respectively) and FCGR4 (Figure 3B).

We hypothesized that the functional consequence of the enrichment would be enhanced tissue repair (specifically, inflammation resolution and favorable tissue remodeling), as suggested by Fc-receptors being potent mediators of phagocytosis (17). Moreover, these macrophages were found to be enriched in other phagocytosis-related genes (13). There is strong rationale for this hypothesis from the recognition in the atherosclerosis field that efferocytosis, or the phagocytosis of dying cells, is an inflammation-resolving process that limits the size of plaque necrotic cores (18). Thus, the fact that Fcgr4 is increased in stCR conditions made us interested to determine whether this had a functional consequence on efferocytosis.

To mimic this increase in vitro, we measured the efferocytotic activity of macrophages overexpressing the mRNA of the human homolog of Fcgr4, Fcgr3a. Control cells were similarly treated with a scrambled mRNA sequence. The cells were used in a standard assay (37) in which macrophages are incubated with fluorescent apoptotic cells, with efferocytotic activity quantified by counting the frequencies of macrophages that consume apoptotic cells. Fcgr3a-overexpressing macrophages had enhanced efferocytotic capacity compared with control cells (Figure 3, C and D), suggesting that the basal level of expression was limiting. To examine whether stCR induces efferocytosis in vivo, plaque necrotic core size was assessed in aortic root images, which, as alluded to above, has been shown to inversely correlate with macrophage efferocytotic activity (37, 38). Indeed, the data showed smaller necrotic cores in the stCR group (Figure 3, E–G).

To further examine macrophage efferocytotic capacity in vivo, plaques were also assessed as described previously (37) for macrophage-associated apoptotic cells (observed as TUNEL+). The results showed that compared with BL mice, plaques in the stCR group tended to have more macrophages associated with apoptotic cells (Figure 3H), again indicative of increased efferocytosis. Notably, we previously reported in eWAT that after stCR, the content of macrophages that had multiple nuclei increased, consistent with their efferocytosis of apoptotic cells (13). Here, we found by immunostaining that in both BL and stCR mice, a sizable proportion of the efferocytes in plaques were FCGR4+, with substantially more efferocytes expressing FCGR4 in the stCR group (33% in BL and 62% in stCR plaques; Figure 3, I and J). It should be remembered that each macrophage performing efferocytosis typically clears several apoptotic cells (37–39). Therefore, even a modest increase in macrophages that have enhanced efferocytotic ability can result in a large increase in dead cell removal, consistent with what we have observed with changes in the necrotic core.

We also investigated whether stCR influences the expression of inflammatory and pro-resolving genes by performing flow cytometric analysis of plaque macrophages (Supplemental Figure 3A). This analysis revealed that stCR increased the levels of proteins associated with pro-resolution (e.g., CD163 and CD206) and decreased levels of those associated with inflammation (e.g., Ly6C and CD14). We further tested the response of macrophages overexpressing the human Fcgr3a to inflammatory and anti-inflammatory stimuli. To confirm that the Fcgr3a overexpression was successful and specific, we measured the expression of the human (Fcgr3a) and mouse (Fcgr4) genes and found no difference in the expression of Fcgr4 but a marked increase in Fcgr3a (Figure 3K). Upon exposure to the inflammatory stimulus LPS, Fcgr3a-overexpressing macrophages showed decreased expression of pro-inflammatory genes, including Il6, Tnfa, and Nos2 (Figure 3K). The response to IL-4 was more diverse, as Fcgr3a-overexpressing macrophages enhanced the expression of the pro-resolving gene mannose receptor (Mrc1), but not Arginase 1 (Arg1), as shown in Figure 3K.

We sought to further characterize Fcgr4+ macrophages by performing bulk RNA-Seq on eWAT macrophages that were either FCGR4 positive or negative. Obese mice were subjected to stCR, after which single cells from eWAT were isolated. FCGR4 positive and negative macrophages were flow-sorted and their transcriptomes compared. DEGs between macrophages expressing FCGR4 and those not expressing it are presented in Supplemental Table 4 and Figure 3L. We found that many genes upregulated in FCGR4+ macrophages are known to be important in the efferocytosis process, such as Il10 (40) and Pparg (41), adding further evidence of their phagocytic/efferocytotic function. We also queried the significantly up- and downregulated genes (adjusted P value < 0.1) of FCGR4+ macrophages for KEGG pathway enrichment (Supplemental Figure 3B). Both “response to interferon γ” and “antigen processing and presentation” were enriched in the genes from FCGR4+ macrophages, while “extracellular matrix organization,” “regulation of angiogenesis,” and “collagen fibril organization” were enriched in the genes with lower expression in FCGR4+ macrophages. Interestingly, genes involved in “cell killing” were enriched in genes both up- and downregulated in FCGR4+ macrophages, suggesting that different components of these pathways are at play in FCGR4+ macrophages compared with FCGR4– macrophages.

Taken together, these data suggest that stCR induces a desirable environment in both plaques and eWAT that is associated with decreased expression of inflammatory genes, increased expression of pro-resolving genes, and enrichment of Fcgr4+ macrophages, with these cells promoting clearance of apoptotic cells.

eWAT-derived Fcgr4+ macrophages contribute to the reduction in plaque necrotic core. The data above suggest that cells in the Fcgr4+ macrophage cluster promote a reduction in plaque necrotic core upon stCR-induced weight loss. Despite being relatively enriched to similar levels in both plaques and eWAT following stCR (Figure 3A), Fcgr4+ macrophages, in terms of cell frequency among all leukocytes, constituted a much larger population in eWAT (Figure 2B) (13). Thus, we hypothesized that Fcgr4+ macrophages in eWAT may contribute to resolving atherosclerotic inflammation by interorgan mechanisms.