We recently showed that dietary β-carotene delays atherosclerosis progression in Ldlr^-/- mice (Zhou et al., 2020), which prompted us to examine whether β-carotene also impacts the resolution of inflammation in complex atherosclerotic lesions. To achieve these lesions, we utilized two distinct mouse models of atherosclerosis in combination with WD-VAD. In our first model, we established LDLR deficiency in wild-type mice by injecting anti-sense oligonucleotide targeting the low-density lipoprotein receptor (ASO-LDLR) weekly for a period of 16 wk. The characteristics of the plaques before undergoing resolution were established by sacrificing a subset of mice after 16 wk on diet (baseline). The remaining mice were injected once with SO-LDLR to promote atherosclerosis resolution and divided into two groups fed either WD-VAD (resolution–control) or WD-β-carotene (resolution–β-carotene). Mice were harvested 3 wk after SO-LDLR injections (Figure 1A).

Male mice gained more weight than female mice, although we did not observe differences between the three experimental groups (Figure 1—figure supplement 1A). For the remaining outcomes (plasma lipid profile, retinoid levels, and plaque composition), we did not observe sex differences in any of our experiments, and therefore, we combined results from both sexes. Both resolution groups, independently of β-carotene, showed drastic reduction in total plasma cholesterol and triglyceride levels in comparison to the baseline group. HDL-C levels remained constant between groups (Figure 1B). Using pooled samples, we compared plasma cholesterol and triglyceride partitioning between groups by fast performance liquid chromatography (FPLC). Very low-density lipoprotein (VLDL) and LDL fractions in both resolution groups showed a comparable reduction in total cholesterol and triglyceride content in comparison to baseline mice (Figure 1C and D). Lipid normalization was accompanied by the upregulation of hepatic LDLR expression at the mRNA and protein levels (Figure 1E and F).

Despite mice being fed a WD-VAD for several weeks, HPLC measurements of circulating vitamin A and hepatic retinoid stores ruled out vitamin A deficiency in any of the experimental groups (Figure 1G and H). However, mice fed WD-β-carotene showed an increase in hepatic vitamin A content in comparison to those fed WD-VAD, which was mediated by the conversion of β-carotene to vitamin A (Figure 1H). Hepatic Cyp26a1 expression, which is commonly utilized as a surrogate marker of vitamin A and retinoic acid production (Emoto et al., 2005; Abu-Abed et al., 2002), appeared upregulated in response to WD-β-carotene (Figure 1—figure supplement 1B).

To determine the effect of β-carotene supplementation on atherosclerosis resolution, we examined lesion size and composition at the level of the aortic root (Figure 1I). Lesion area remained constant among the three experimental groups as previously reported in mice treated with ASO-LDLR under similar experimental conditions (Figure 1J; Basu et al., 2018). We next examined plaque composition by focusing on two parameters commonly utilized as a surrogate indicators of atherosclerosis resolution: CD68⁺ area, a myeloid (macrophage) marker that represents plaque inflammation, and collagen area, a marker of plaque stability in humans (Libby and Aikawa, 2002; Soehnlein and Libby, 2021). In comparison to baseline mice, the lesions of resolution–control and resolution–β-carotene groups presented a reduction in CD68 content of approximately 36 and 45%, respectively (Figure 1K). The collagen content in lesions increased in both resolution groups in comparison to baseline, reaching statistical significance only in the β-carotene-fed mice. In this group, we observed 215 and 60% increase in collagen content compared to the baseline and the resolution–control groups, respectively (Figure 1L). Lastly, we plotted the relative CD68 and collagen contents in the lesion to perform a descriptive discriminant analysis. Individual samples from three distinct experimental groups clustered together, highlighting significant differences between the three experimental groups for all the comparisons (Figure 1M).

To validate the results obtained in our ASO-LDLR reversible model of atherosclerosis, we utilized Ldlr^-/- mice fed WD-VAD subjected to a dietary switch strategy to lower cholesterol in plasma and promote atherosclerosis resolution (Amengual et al., 2021; Yu et al., 2017). Baseline Ldlr^-/- mice were harvested after 12 wk on WD-VAD, while the remaining animals were switched to either a standard diet without vitamin A (resolution–control) or the same diet supplemented with β-carotene (resolution–β-carotene) for four more weeks. To prevent changes in food intake due to consistency and hardness of the feed, we provided standard diets as powder. This approach prevented a reduction in body weight, which could have resulted in confounding alterations in atherosclerosis resolution (Figure 2—figure supplement 1A).

Ldlr^-/- mice in both resolution groups presented a comparable reduction in plasma cholesterol and triglyceride levels in comparison to the baseline group. These changes were not accompanied by alterations in HDL-C levels between groups (Figure 2A). Systemic and hepatic vitamin A levels failed to show indications of vitamin A deficiency, and hepatic vitamin A stores increased in β-carotene-fed mice (Figure 2—figure supplement 1B).

We next characterized atherosclerotic lesions at the level of the aortic root (Figure 2B). Lesion size area was comparable between experimental groups (Figure 2C), although CD68 content decreased to the same extent in both resolution groups in comparison to baseline mice (Figure 2D). In comparison to the baseline group, collagen accumulation in the lesion increased 100% in the resolution–control and over 200% in the resolution–β-carotene group, respectively (Figure 2E). Descriptive discriminant analysis showed comparable results to those observed for our ASO-LDLR model, where differences between the three experimental groups reached statistical significance (Figure 2F).

Together, these data show that β-carotene supplementation during atherosclerosis resolution results in the improvement of lesion composition in two independent mouse models.

To establish the contribution of BCO1 on the effect of β-carotene on atherosclerosis resolution, we utilized LDLR-ASO to promote atherogenesis in Bco1^-/- mice following the same experimental approach described for wild-type mice (Figure 1A). Consistent with our results in our two resolution models (Figures 1B and 2A), plasma cholesterol levels decreased in both resolution groups in comparison to baseline mice (Figure 3—figure supplement 1A). Similarly, HPLC measurements of plasma and hepatic vitamin A ruled out vitamin A deficiency in Bco1^-/- mice baseline and resolution mice (Figure 3—figure supplement 1B and C). The accumulation of β-carotene in tissues and plasma is characteristic in Bco1^-/- mice (Hessel et al., 2007). Indeed, HPLC quantification of β-carotene in plasma and liver of Bco1^-/- resolution– β-carotene mice presented 4-fold and 400-fold greater β-carotene levels found in wild-type resolution–β-carotene, respectively (Figure 3A and B).

The characterization of the atherosclerotic lesions showed no alterations in lesion size between the three experimental groups (Figure 3C and D). When compared to baseline mice, both resolution groups displayed lower CD68 and higher collagen contents, although we did not observe differences in CD68 and collagen contents between both resolution groups (Figure 3E and F). Descriptive discriminant analysis highlighted differences between baseline Bco1^-/- mice and their resolution littermates, independently of the presence of β-carotene in the diet (Figure 3G).

Our data show that β-carotene favors atherosclerosis resolution in two independent mouse models (Figures 1M and 2F), and results in Bco1^-/- mice directly implicate vitamin A formation in this process (Figure 3G). Retinoic acid is the transcriptionally active form of vitamin A and promotes Treg differentiation by upregulating FoxP3 expression in various experimental models (Gundra et al., 2017; Ouimet et al., 2015; Gundra et al., 2014; Girgis et al., 2014; Girgis et al., 2013; Broadhurst et al., 2012). Treg number decreases in developing lesions and increases during atherosclerosis resolution (Sharma et al., 2020; Foks et al., 2015; Ait-Oufella et al., 2006), but whether the dietary manipulation of β-carotene or vitamin A affects Treg number during atherosclerosis remains unanswered. To investigate whether the Tregs are responsible for the effect of dietary β-carotene on atherosclerosis resolution, we induced atherosclerosis in Foxp3^EGFP mice by injecting ASO-LDLR as described above (Figures 1 and 3). Foxp3^EGFP mice co-express EGFP and FoxP3 under the regulation of the endogenous FoxP3 promoter (Haribhai et al., 2007). After 16 wk on diet, mice undergoing resolution were treated twice with either the PC61 anti-CD25 monoclonal antibody (anti-CD25) or an isotype control (IgG) (Figure 4A; Sharma et al., 2020). We did not observe changes in body weight among groups, and the three groups undergoing resolution showed a normalization in plasma lipids in comparison to baseline controls (data not shown).

Flow cytometry analyses demonstrated the effectiveness of the anti-CD25 treatment by reducing the ratio of circulating and splenic CD25⁺FoxP3⁺ (CD25⁺ Tregs) in comparison to the other experimental groups (Figure 4B–D). We also observed a depletion of CD25⁺ Tregs and CD25⁺FoxP3^- T cells in the lesion and lymph nodes (Figure 4E and F), in agreement with previous reports showing that anti-CD25 fails to completely deplete CD25^-FoxP3⁺ Tregs (CD25^- Tregs), which retain strong anti-inflammatory properties (Couper et al., 2007). Hence, we quantified the number of total Tregs independently of the presence of CD25 by counting the number of GFP/FoxP3/DAPI triple-positive cells (Figure 4G). All the groups undergoing resolution presented a greater total Treg number than baseline mice, although the results did not reach statistical significance between baseline and resolution–control group (p=0.10). Resolution–β-carotene injected with IgG presented the highest Treg cell number among all the other experimental groups, suggesting that dietary β-carotene favors Treg expansion in the plaque (Figure 4H).

Circulating total Treg number remained constant, but we observed a slight decrease in the number of splenic total Tregs in mice undergoing resolution in comparison to baseline mice. This reduction was more pronounced in the resolution–β-carotene mice+anti-CD25 (Figure 4—figure supplement 1A and B). The number of circulating CD25^- Tregs increased in the resolution–β-carotene mice+anti-CD25 group in comparison to baseline mice, remaining constant in the spleen among all groups (Figure 4—figure supplement 1C and D).

Tregs possess strong anti-inflammatory properties independently of the presence of CD25 (Couper et al., 2007; Hayes et al., 2020) and play an important role in plaque remodeling during atherosclerosis regression (Sharma et al., 2020). Hence, we evaluated plaque composition in our Foxp3^EGFP mice (Figure 5A). As expected, all experimental groups showed comparable lesion size at the level of the aortic root (Figure 5B). Mice in the resolution–β-carotene+IgG group displayed a reduction in CD68 content in the lesion of approximately 50% in comparison to baseline mice. Resolution–control+IgG and resolution–β-carotene+anti-CD25 groups showed a 30% reduction in comparison to baseline mice, although only the former reached statistical significance (Figure 5C). Collagen content in the lesion of resolution–β-carotene+IgG mice was significantly higher in comparison to any other experimental group. Resolution control+IgG and resolution–β-carotene+anti-CD25 showed comparable results, while baseline mice had the lowest collagen content among all the groups (Figure 5D).

We next performed a descriptive discriminant analysis to examine pairwise comparisons between our four experimental groups. The baseline group was significantly different in comparison to all the resolution counterparts. Resolution–β-carotene+IgG mice were different from their resolution–control+IgG, as we observed in our two previous resolution experiments (Figures 1M and 2F). We did not observe statistical differences when we compared resolution–β-carotene+anti-CD25 to resolution–control+IgG (FDR = 0.38) or resolution–β-carotene+IgG to (FDR = 0.10) (Figure 5E).

The presence of anti-inflammatory macrophages in the lesions, together with a net egress of pro-inflammatory macrophages, are key features of atherosclerosis resolution (Moore et al., 2018). We quantified arginase 1 content in the lesion, a key anti-inflammatory marker in mice that contributes to collagen synthesis (Szondi et al., 2021), and is synergistically upregulated in anti-inflammatory macrophages exposed to retinoic acid (Yurdagul et al., 2020; Pinos et al., 2023; Figure 5F). Only those mice fed β-carotene, independent of the injection with IgG or anti-CD25, showed an upregulation of arginase 1 in the lesion (Figure 5G). We next evaluated retinoic acid signaling in the atherosclerotic lesion by quantifying the levels of the retinoic acid-sensitive marker Cyp26b1 in the plaque (Figure 5H; Ross and Zolfaghari, 2011). We observed an upregulation in Cyp26b1 expression in mice fed β-carotene, independently of receiving IgG or anti-CD25 (Figure 5I). We also evaluated macrophage egress by injecting a single dose of EdU a week before sacrificing the baseline mice (see ‘Methods’ for details; Figure 5J). Only those mice fed β-carotene, independent of the antibody treatment, displayed a greater egress in comparison to baseline mice (Figure 5K). We did not observe changes in monocyte recruitment evaluated by injecting fluorescently labeled beads 24 hr before the sacrifice (Figure 5L and M), nor changes in Ki67⁺CD68⁺ cells, as an indicator of macrophage proliferation in the lesion (Figure 5—figure supplement 1A and B).

All procedures were approved by the Institutional Animal Care and Use Committees of the University of Illinois at Urbana Champaign. For all our studies, we utilized comparable number of male and female wild-type, Ldlr^-/- (#002207, Jackson Labs, Bar Harbor, ME), Bco1^-/- (Hessel et al., 2007), and Foxp3^EGFP mice (#006772, Jackson Labs). All mice were in C57BL/6J background. Mice were kept under controlled temperature and humidity conditions with a 12 hr light/dark cycle and free access to food and water. Mice were weaned at 3 wk of age onto a breeder diet (Teklad global 18% protein diet: Envigo, Indianapolis, IN) in groups of 3–4 mice.

Control and β-carotene diets contained either placebo beadlets or β-carotene beadlets at a final concentration of 50 mg β-carotene/kg diet. For reference, the content of β-carotene in carrots is 80 mg/kg (USDA Database). Beadlets were a generous gift from DSM Nutritional Products (Sisseln, Switzerland). All diets were prepared by Research Diets (New Brunswick, NJ) by cold extrusion. The exact composition of all the diets used for this study is provided in Supplementary file 1. For all the experiments, we stimulated the development of atherosclerosis with a Western diet deficient in vitamin A (WD-VAD), as done in the past (Zhou et al., 2020; Amengual et al., 2020).

Before tissue harvesting, mice were deeply anesthetized by intraperitoneal injection with a mixture of ketamine and xylazine at 80 and 8 mg/kg body weight, respectively. We collected blood by cardiac puncture using ethylenediaminetetraacetic acid (EDTA)-coated syringes. We then perfused the mice with 10% sucrose in 0.9% sodium chloride (NaCl)-saline solution prior to tissue harvesting. Tissues were immediately snap-frozen in liquid nitrogen and kept at –80°C.

Aortic roots were collected after removing fat under a binocular microscope, embedded in optimum cutting temperature compound (OCT, Tissue-Tek, Sakura, Torrance, CA) and kept at –80°C.

To transiently deplete LDLR expression in wild-type, Bco1^-/-, and Foxp3^EGFP mice, we administered weekly an intraperitoneal injection containing 5 mg/kg body weight of ASO-LDLR for a period of 16 wk. At the end of the treatment, we injected a single dose of 20 mg/kg of the sense oligonucleotide (SO-LDLR) antidote, which binds and deactivates the ASO-LDLR (Basu et al., 2018). All oligonucleotide treatments were generously provided by Ionis Pharmaceuticals (Carlsbad, CA).

Nonpolar compounds were extracted from 100 μl of plasma or 30 mg of liver under a dim yellow safety light using methanol, acetone, and hexanes. The extracted organic layers were then pooled and dried in a SpeedVac (Eppendorf, Hamburg, Germany). We performed the HPLC with a normal phase Zobax Sil (5 μm, 4.6 × 150 mm) column (Agilent, Santa Clara, CA). Isocratic chromatographic separation was achieved with 10% ethyl acetate/hexane at a flow rate of 1.4 ml/min. For molar quantification of β-carotene and retinoids, we scaled the HPLC with a standard curve using the parent compound.

We measured plasma total cholesterol, cholesterol in the high-density lipoprotein fraction (HDL-C), and triglyceride levels using commercially available kits (FUJIFILM Wako Diagnostic, Mountain View, CA), according to the manufacturer’s instructions.

Plasma lipoproteins were fractionated using FPLC with two Superose 6 10/300 GL columns (G.E. Healthcare, Boston, MA) on a Shimadzu HPLC system (Columbia, MD), as done in the past (Zhou et al., 2020; Amengual et al., 2021). Pooled samples (five mice/group) were used for analysis, and lipoprotein fractions were identified by quantifying cholesterol and triglyceride levels in each fraction.

We isolated the total RNA using TRIzol reagent (Thermo Fisher Scientific, Waltham, MA) and Direct-zol RNA Miniprep kit (Zymo Research, Irvine, CA) according to the manufacturer’s instructions. RNA was reverse-transcribed to cDNA using high-capacity cDNA reverse transcription kit (Applied Biosystems, Foster City, CA). Then we used TaqMan Master Mix (Applied Biosystems) to perform RT-PCR using the StepOnePlus RT-PCR system (ABI 7700, Applied Biosystems). We calculated the relative gene expression using the ΔCt method and normalized the data to β-actin (Actb). Probes (Applied Biosystems) include mouse Ldlr (Mm00440169_m1), mouse cytochrome P450 26a1 (Cyp26a1, Mm00514486_m1), and mouse Actb (Mm02619580_g1).

For the determination of hepatic levels of LDLR, proteins were extracted from liver lysates in RIPA buffer (50 mM Tris pH 7.4, 150 mM NaCl, 0.25% sodium deoxycholate, 1% Nonidet P-40) in the presence of protease inhibitors. Total protein amounts were quantified using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). A total of 80 µg of protein homogenates were separated by SDS-PAGE and transferred onto PVDF membranes (Bio-Rad, Hercules, CA). Membranes were blocked with fat-free milk powder (5% w/v) dissolved in Tris-buffered saline (15 mM NaCl and 10 mM Tris/HCl, pH 7.5) containing 0.01% Tween 100 (TBS-T), washed, and incubated overnight at 4°C with mouse anti-LDLR (Santa Cruz Biotechnologies, Dallas, TX) and mouse anti-GAPDH (Thermo Fisher Scientific) as a housekeeping control. Infrared fluorescent-labeled secondary antibodies were prepared at 1:15,000 dilution in TBS-T with 5% fat-free milk powder and incubated for 1 hr at room temperature.

To deplete Tregs, mice were injected twice with 250 µg of either the isotype control IgG (Ultra-LEAFPurified Rat IgG1, λ Isotype Ctrl antibody no. 401916, BioLegend, CA) or PC61 anti-CD25 monoclonal antibody (Ultra-LEAF Purified anti-mouse CD25 antibody no. 102040, BioLegend, CA). The first treatment took place a day after the administration of SO-LDLR, and the second injection 2 wk after, following established protocols (Sharma et al., 2020).

One week before harvesting the baseline group, we injected the mice intraperitoneally with 4 mg/ml of 5-ethynyl-2′-deoxyuridine (EdU) (Invitrogen, Waltham, MA) with the goal of labeling circulating monocytes to assess macrophage retention among groups. To compare monocyte recruitment among groups, we labeled circulating monocytes by injecting the mice retro-orbitally with 1 µm diameter Flouresbrite flash red plain microspheres beads (Polysciences Inc, Warrington, PA) diluted in sterile PBS 24 hr before harvesting, regardless of the experimental group. We assessed the efficiency of EdU and bead labeling by flow cytometry 48 hr and 24 hr after injection by tail bleeding, respectively (Weinstock and Fisher, 2019).

To assess the number of monocytes labeled with either EdU or fluorescent beads, we incubated blood samples with red blood cell lysis buffer (Thermo Fisher Scientific) and blocked unspecific bindings using anti-mouse CD16/CD32 (Mouse BD Fc Block, BD Biosciences, Franklin Lakes, NJ). Cells were then stained with FITC-conjugated anti-mouse CD45 (BioLegend), PE-conjugated anti-mouse CD115 (BioLegend), and PerCP-conjugated anti-mouse Ly6C/G (BioLegend, San Diego, CA) for 30 min on ice. Cells were fixed, permeabilized, and stained for EdU using Click-iT EdU Pacific Blue Flow Cytometry Assay Kit (Invitrogen) following the manufacturer’s instructions. Beads were detected with the 640 nm red laser using a 660/20 bandpass filter. Monocyte recruitment and egress were estimated based on monocyte labeling efficiency, following established protocols (Basu et al., 2018).

For Treg quantifications, spleens were mashed and filtered through a 100 µm sterile cell strainer (Thermo Fisher Scientific). Blood and spleen homogenates were incubated with red blood cell lysis buffer (Thermo Fisher Scientific) and subsequently blocked with anti-mouse CD16/CD32 (BD Biosciences). We stained the cells with Fixable Viability Dye eFluor 780 (eBioscience, San Diego, CA) and eFluor 506-conjugated anti-mouse CD45 (BioLegend), PE-conjugated anti-mouse CD3e (BioLegend), FITC-conjugated anti-mouse CD4 (BioLegend), and BV421-conjugated anti-mouse CD25 (BioLegend) for 30 min on ice. We washed the cells twice before fixing and permeabilizing cell membranes using the eBioscience FoxP3/Transcription Factor Staining Buffer Set (Invitrogen) for 1 hr at room temperature, followed by incubation with Alexa Fluor 647-conjugated anti-mouse FoxP3 antibody (BioLegend) for 30 min at room temperature. All samples were measured on a BD LSR II analyzer (BD Biosciences), and results were analyzed with the FCS express 5 software (De Novo Software, Pasadena, CA).

Six-micrometer-thick sections were fixed and permeabilized with ice-cold acetone, blocked, and stained with rat anti-mouse CD68 primary antibody (clone FA-11; Bio-Rad) followed by biotinylated rabbit anti-rat IgG secondary antibody (Vector Laboratories, Burlingame, CA). CD68⁺ area was visualized using a Vectastain ABC kit (Vector Laboratories). Sections were then counterstained with hematoxylin, dehydrated in an ethanol gradient, xylene, and mounted with Permount medium (Thermo Fisher Scientific). Images were acquired using Axioskop 40 microscope (Carl Zeiss, Jena, Germany). For collagen content, frozen sections were fixed and stained using picrosirius red (Polysciences). Sections were scanned using the Axioscan.Z1 microscope (Carl Zeiss) using both bright field and polarized light. Total lesion area, CD68⁺, and collagen-positive areas were quantified using ImageJ software (NIH).

Aortic root sections were fixed in 10% formalin, permeabilized with 1% Triton X-100 in phosphate buffer, and stained using anti-FoxP3 (clone MF-14; BioLegend), anti-GFP (Cat# ab290; Abcam), anti-CD25 (clone IL2R.1; Thermo Fisher Scientific), anti-arginase 1 (Cat# 16001-1-AP; Thermo Fisher Scientific), anti-Cyp26b1b (Cat# 21555-1-AP; Thermo Fisher Scientific), and anti-Ki67 (clone SP6; Abcam). EdU+ cells in the lesion were visualized using Click-IT EdU Imaging Kit (Invitrogen). Secondary antibodies conjugated to Alexa fluorophores were purchased from Molecular Probes (Life Technologies). Images were taken using a Zeiss Axioscan.Z1 slide scanner (Carl Zeiss). Quantifications were performed using the ImageJ software (NIH).

Data represented as bar charts are expressed as mean ± standard error of the mean (SEM). Data were analyzed using GraphPad Prism software (GraphPad Software Inc, San Diego, CA) by one-way ANOVA, followed by Tukey’s multiple comparisons test. Differences between groups were considered significant with an adjusted p-value<0.05. Sample sizes were estimated based on previous experiments performed in our lab using comparable experimental approaches (Zhou et al., 2020; Amengual et al., 2021; Yu et al., 2017). Experimental samples were codified to prevent bias during analyses. The relationship between the two dependent factors, CD68 and collagen contents, and group differences was evaluated by descriptive discriminant analysis. Briefly, models were fitted under null and alternative hypotheses and goodness of fit was evaluated using the likelihood ratio test with randomized inference (bootstrap; n = 1000 simulations). Simultaneous hypothesis testing was corrected using the Benjamini–Hochberg procedure (Benjamini et al., 2001). The resulting false discovery rate (FDR) was considered significant with a cut-off <0.05.