To begin to understand the factors upstream of STAT6 needed for atherosclerosis resolution, we examined the requirements for the canonical activators of STAT6, namely IL-4 and IL-13. Mice deficient in both Il4 and Il13 (Il4^-/-Il13^-/-) were injected with a liver-specific AAV expressing Pcsk9 to result in LDLr deficiency and fed Western diet (WD) to promote hypercholesterolemia in order to develop advanced atherosclerotic lesions by 20 weeks (Bjørklund et al., 2014; Roche-Molina et al., 2015; Peled et al., 2017). At this time, a baseline (BL) group was harvested to examine the size and composition of the established plaques. To decrease plasma cholesterol thereafter and induce resolution of atherosclerosis (Res), mice were injected with an anti-sense oligonucleotide (ASO) to apolipoproteinB (ApoB) and switched to chow diet for 3 weeks, as described (Lin et al., 2019; Sharma et al., 2020; Figure 1A).

Treatment with the Pcsk9 AAV and WD caused a significant increase in total plasma cholesterol (~500 mg/dL), whereas subsequent ApoB ASO injections decreased plasma cholesterol to normal levels (~60 mg/dL, Figure 1B). We previously showed, in an identical experimental design, that ApoB ASO administration to Pcsk9 AAV and WD-treated WT mice for 3 weeks resulted in decreased plaque size and macrophage (CD68+ cells) content in aortic roots (Sharma et al., 2020). In the present study, following cholesterol lowering, there were no significant changes in aortic root plaque size (Figure 1C), macrophage content (Figure 1D–F), or necrotic core area (Figure 1G) in mice deficient in IL-4 and IL-13. Collectively, these data indicate that IL-4/13 are needed for atherosclerosis resolution.

We previously reported that atherosclerosis resolution requires recruitment of new Stat6-expressing monocytes to plaques (Rahman et al., 2017). Because we demonstrated that IL-4/13 are important for atherosclerosis resolution (Figure 1), we next examined whether the source of the IL-4/13 is from cells newly recruited in the resolution phase or from cells in the plaques prior to the induction of lipid lowering.

Thus, we performed aortic arch transplant studies (as described previously Rahman et al., 2017; Chereshnev et al., 2003), in which aortic arches from hyperlipidemic Ldlr^-/- donors fed WD for 20 weeks (BL group) were transplanted into normolipidemic WT or Il4^-/-Il13^-/- mice. Four weeks post transplant, the grafted aortic segments were evaluated for plaque burden and macrophage content (see Experimental design; Figure 2A). As expected, the arch donors (and BL group) had elevated total cholesterol levels (978 ± 141 mg/dL), whereas the transplant recipients were normocholesterolemic (79 ± 22 mg/dL and 73 ± 10 mg/dL for WT and Il4^-/-Il13^-/- mice, respectively; Figure 2B). Although plaque size did not change significantly between the BL and transplant recipients (Figure 2C), there was a significant and comparable reduction (~60%) in plaque macrophages in both the WT and Il4^-/-Il13^-/- recipient groups (Figure 2D, E). The necrotic core area significantly diminished as well in both recipient groups (Figure 2F). Thus, resolution of atherosclerosis was similar in WT and Il4^-/-Il13^-/- recipients (Figure 2D–G), indicating that IL-4/13 production by cells newly recruited is not essential for resolution of atherosclerosis. These results were confirmed in an additional aortic transplant experiment, in which plaque-bearing aortic arches from Apoe^-/- donors fed WD for 16 weeks were grafted into either WT or Il4^-/- recipients for 2 weeks (Figure 2—figure supplement 1A), resulting in plaque size and macrophage content changes consistent with those in Il4^-/-Il13^-/- transplant recipients (Figure 2—figure supplement 1B–G). Notably, there was a marked decrease in macrophage content of both WT and Il4^-/- arch recipients (Figure 2—figure supplement 1D–E, G).

Our data thus far indicate that IL-4/13 are needed for atherosclerosis resolution, and the source of these cytokines is not from cells that are newly recruited in the resolution phase, but from cells accumulating during the disease progression phase.

We next evaluated whether IL-4/13 content is increased during atherosclerosis resolution. To measure cytokine abundances in plaques, Ldlr^-/- mice fed WD for 20 weeks (BL) and treated with ApoB ASO for an additional 3 weeks (Res) were examined. Upon harvest, plaques from aortic arches and brachiocephalic arteries were carefully excised and evaluated for CCL2, IL-4, and IL-13. CCL2 was used as a control since we previously showed that its levels are decreased in resolving plaques (Rahman et al., 2017; Feig et al., 2011b). Consistent with our previous reports (e.g., Rong et al., 2001), CCL2 abundance in plaques showed a marked decrease from progression to resolution (Figure 3A).

In contrast to CCL2, IL-4 levels did not significantly change between progression (111 pg/mL) and resolution (123 pg/mL; Figure 3A) and were strikingly lower (over two orders of magnitude) than those commonly used in cell culture experiments (10–20 ng/mL) (Minutti et al., 2017; Bosurgi et al., 2017; Sanin et al., 2018). Interestingly, IL-13 was undetectable (<3 pg/mL) in both progressing and resolving plaques. To confirm that plaque IL-13 levels are negligible, we examined its expression in a recently published dataset of plaque immune cells pre- and post-atherosclerosis resolution (Sharma et al., 2020). In accordance with the undetectable protein levels, only one cell in the entire dataset showed Il13 expression (Figure 3—figure supplement 1). Hence, we focused on IL-4 in subsequent experiments.

Because IL-4 production is needed in cells that accumulate during atherosclerosis progression, but not in cells newly recruited during resolution (Figures 1 and 2), we next examined the abundance of IL-4-producing cells in plaques. IL-4-reporter mice were injected with Pcsk9-AAV and fed WD for 16 weeks to promote atherosclerosis formation. BL mice were harvested, and the disease resolution group was induced with ApoB ASO treatment and chow diet (Figure 3B). We confirmed that ApoB ASO treatment led to reduction in plaque macrophages and promoted resolution of atherosclerosis (Figure 3C). Aortic arches were digested and analyzed using flow cytometry (Figure 3D). Results showed no significant difference in the proportion of IL-4-producing leukocytes (CD45+ cells) when comparing mice from the BL (3.3%) and resolution (2.7%) groups (Figure 3E).

Our data suggest that small amounts of IL-4-producing cells accumulate in progressing plaques and do not increase further during the resolution phase. In the reparative environment, the IL-4 that accumulated is apparently sufficient to activate STAT6, which in turn polarizes plaque macrophages to pro-resolving cells.

Because IL-4 levels do not change in resolving plaques, we postulated that this environment contains other factors that license IL-4 activation of STAT6.

To explore this, we reviewed our transcriptome data from macrophages taken by laser-capture microdissection from progressing and resolving plaques (Ramsey et al., 2014). These data (from two independent mouse models (Feig et al., 2011a) suggested activation of the Wnt pathway (Ramsey et al., 2014). This was of particular interest since a mutation causing downregulation of this pathway led to early coronary artery disease in humans (Mani et al., 2007). Moreover, activation of Wnt signaling in macrophages was recently shown to inhibit atherosclerosis progression (Wang et al., 2018).

To confirm this activation in resolving plaques, single-cell RNA-sequencing (scRNA-seq) data from progressing and resolving plaque macrophages that we recently published (Lin et al., 2019) were analyzed for the expression of Wnt-related genes (GO term GO0060070 ‘Wnt signaling Pathway’). The expression of these genes was summed per cell, aggregated by Louvain clusters and plotted (Figure 4A). This analysis showed expression of Wnt-related genes in all the macrophage clusters (Figure 4A). Notably, comparison between progressing and resolving plaque macrophages demonstrated that in disease resolution there is an overall increase in the expression of Wnt-related genes across most clusters (Figure 4B). The relative expression levels of each Wnt-related gene from resolving versus progressing plaque macrophages are presented in Figure 4C. Among the Wnt regulated genes, Cttnb1, the gene encoding for β-catenin, the downstream canonical effector of Wnt, was the most upregulated across macrophage clusters (Figure 4C).

To extend our analysis of Wnt signaling to human plaque macrophages, we analyzed a recently published dataset characterizing the transcriptome of human endarterectomy leukocytes (Fernandez et al., 2019). First, Wnt-responsive gene expression data from mouse plaque macrophages (Figure 4C) were averaged across all clusters. The 10 highest differentially expressed genes in resolving plaques were identified and their expression examined in carotid plaque macrophages from symptomatic (Sym) versus asymptomatic (Asym) patients (Fernandez et al., 2019; Figure 4D). Results show that of the 10 genes 7 had lower expression levels in Sym patients’ plaque macrophages compared to Asym patients (Figure 4D). A permutation test resulted in a p-value=0.008, thereby supporting a bona fide inverse relationship between Wnt signaling and disease severity across species, consistent with the genetic loss-of-function association with premature CAD (coronary artery disease; Mani et al., 2007).

To independently support the clinical relevance of the inverse relationship between Wnt signaling and disease severity, we next examined whether the Wnt pathway is associated with atherosclerosis resolution in humans. A recently described cohort of statin-naïve patients that were treated with high-dose atorvastatin following presentation with an acute coronary syndrome (Alkhalil et al., 2018) was used. Using advanced imaging techniques, the authors evaluated carotid artery plaque volume at presentation and 3 months later. We investigated the expression of seven Wnt-associated genes (Axin2, Ddx3x, Wls, Senp2, Ctnnb1, Csnk1a1, and Amfr) in peripheral blood mononuclear cells (PBMCs) obtained at presentation and follow-up. The fold-change of each gene expression at the follow-up compared to presentation was calculated per patient and summed to create a ‘Wnt score’. These scores at the 3-month follow-up were significantly increased relative to disease presentation, indicated by the increased Wnt score (Figure 4E). Notably, the expression of Axin2, the hallmark Wnt target gene (Jho et al., 2002; Lustig et al., 2002), significantly correlated with a decrease in plaque volume (R² = 0.408, p=0.002; Figure 4F). These data provide further support that atherosclerosis resolution is associated with upregulation of the Wnt pathway not only in mice, but also in humans.

Because Wnt signaling is upregulated in resolving plaque macrophages and IL-4 levels were similar in progressing and resolving plaques, we examined whether this pathway augments the effects of IL-4. Wnt3a is a wnt pathway activator and is most commonly used for studies in vitro. Thus, we treated bone marrow-derived macrophages (BMDMs) with recombinant Wnt3a protein and IL-4 (at its plaque concentration). First, we confirmed that treatment with Wnt3a activated the Wnt pathway, indicated by the upregulation of Axin2 (Figure 5—figure supplement 1A) and the accumulation of the transcription co-activator β-catenin (Figure 5—figure supplement 1B). Examination of genes associated with pro-resolution (e.g., Arg1, Ccl17, and Socs1) revealed that Wnt3a treatment did not alter their expression. However, when cells were co-treated with Wnt3a and IL-4, these transcripts showed increased expression relative to IL-4 alone (Figure 5A–C).

To examine whether the enhanced expression of these pro-resolving genes is due to Wnt3a-induced production of IL-4/13, we repeated the in vitro experiments with Il4^-/-Il13^-/- BMDMs. The data show that the expression of the pro-resolving genes was enhanced in cells co-treated with Wnt3a and IL-4 (Figure 5D–F), indicating that the increased gene expression cannot be explained by induced production of IL-4/13. Since the signal transducer of IL-4 is STAT6 (as noted above, previously shown to be required for plaque resolution Rahman et al., 2017), we studied its involvement in the enhanced pro-resolving gene expression by using BMDMs from Stat6^-/- mice. As shown in Figure 5G–I, Stat6 deficiency completely abrogated the expression of Arg1, Ccl17, and Socs1 and the enhancement observed in cells co-treated with Wnt3a and IL-4. In sum, the data suggest that Wnt signaling cooperates with IL-4/STAT6 signaling to augment macrophage pro-resolving gene expression.

Our in vitro and in vivo data suggest a role for the Wnt signaling pathway in resolving atherosclerosis. To strengthen this suggestion, we performed an atherosclerosis resolution study. Reversa mice (Feig et al., 2011a) were fed WD for 16 weeks to establish advanced atherosclerotic plaques (baseline group; BL). LDL-C levels were then lowered and mice treated with an established Wnt inhibitor, LGK974 (Liu et al., 2013), or vehicle, for an additional 3 weeks (Figure 6A).

BL mice were hypercholesterolemic (~810 mg/dL), and mice from both lipid-lowering groups were normocholesterolemic (~140 mg/dL; Figure 6B). Mice treated with LGK974 had similar body weights (Figure 6—figure supplement 1A), lipid profiles (Figure 6B), hematological parameters (Figure 6—figure supplement 1B–E), and frequencies of Ly6C^hi and Ly6C^lo circulating monocytes (Figure 6—figure supplement 1F) compared to the control (Ctrl) group. To confirm that Wnt inhibition was successful in the LGK974-treated mice, plaques were stained for the active form of β-catenin (Figure 6—figure supplement 1G), which showed significantly less active β-catenin. To verify that LGK974 treatment inhibited Wnt signaling in plaque macrophages, we excised them using laser capture microdissection (LCM) and examined the expression of Wnt-responsive genes. The data show decreased expression of Lef1 and Axin2 from LGK974 compared to vehicle-treated mice (Figure 6—figure supplement 1H).

Plaque area and macrophage content were significantly decreased in the control group compared to BL (Figure 6C–F). However, mice that received the Wnt inhibitor did not have a significant decrease in plaque or macrophage area (Figure 6C–F), despite equally low cholesterol levels (Figure 6B). Furthermore, we previously reported that resolving plaques are enriched with collagen, presumably because the content of matrix metalloprotease-producing macrophages decline (Peled et al., 2017). We again observed this increase in plaque collagen in the control resolution group, but it was not detected in the Wnt inhibitor group (Figure 6F, G). As noted above, atherosclerosis resolution is characterized by the change in the inflammatory state of plaque macrophages from pro-inflammatory to pro-resolving (Rahman et al., 2017; Ramsey et al., 2014; Feig et al., 2012). Consistent with this, plaque immunostaining showed that in control resolution there was downregulation of the pro-inflammatory markers CCL2 and VCAM1, which were not attenuated in the LGK974 group. Additionally, compared to BL mice, in the control group there were increases in pro-resolving markers ARG1 and MRC1, which were not seen in the LGK974 group (Figure 6F, G). Overall, these data suggest that Wnt inhibition impairs atherosclerosis resolution and results in decreased accumulation of pro-resolving plaque macrophages.

To investigate the molecular mechanisms underlying the enhanced response of BMDMs to Wnt3a and IL-4, we performed RNA-seq of cells treated for 16 hr with media (untreated) or with media containing Wnt3a, IL-4, or Wnt3a+IL-4. Principal component analysis (PCA) shows that the untreated BMDMs cluster with the IL-4-treated cells (Figure 7A), indicating a high similarity of transcriptional profiles (Figure 7—figure supplement 1A) before Wnt-mediated augmentation. Wnt3a-treated (−/+ IL-4) cells clustered together were distinct from the control and IL-4-treated cells (Figure 7A) and exhibited extensive transcriptional changes (Figure 7—figure supplement 1A). The data suggested a possible engagement of the JAK-STAT pathway in Wnt3a-treated cells as both Wnt3a and Wnt3a+IL-4-treated cells had the KEGG pathway term ‘JAK-STAT signaling pathway’ in the top seven most significantly enriched pathways (Figure 7—figure supplement 1B). Of the members of the JAK-STAT pathway, we focused on STAT3 because of its known anti-inflammatory functions (reviewed in Hillmer et al., 2016) and its being implicated in Wnt3a-induced enhancement of IL-4 responses of macrophages (Feng et al., 2018). Moreover, the STAT3 activating cytokines IL-6 and IL-10 have been reported to collaborate with IL-4 to enhance Arg1 expression in macrophages (Makita et al., 2015; Mauer et al., 2014), as we also found (Figure 7B, C).

We next tested whether STAT3 is activated (as reflected by phosphorylation) by Wnt3a in BMDMs. The data show that STAT3 is not phosphorylated after a short exposure to Wnt3a (Figure 7—figure supplement 1C); however, after 16 hr of treatment, there was substantial STAT3 phosphorylation (Figure 7D). Furthermore, the STAT3 target gene Socs3 was upregulated following Wnt3a treatment (Figure 7E). To directly examine the involvement of STAT3 in the enhanced expression of pro-resolving genes following Wnt3a treatment, we knocked-down (KD) Stat3 mRNA in BMDMs using siRNA and treated these cells with Wnt3a, IL-4, or Wnt3a+IL-4. The siRNA decreased Stat3 transcript levels by ~80% (Figure 7—figure supplement 1D). Results show that Stat3 KD attenuated Arg1 expression in response to Wnt3a+IL-4 (Figure 7F), suggesting that STAT3 is required for the enhancement of IL-4 responses by Wnt3a.

To explore whether Wnt inhibition affects STAT3 activity in vivo, aortic root sections from control and LGK974-treated mice (Figure 6A) were stained for STAT3, and its nuclear localization was determined. Images and their analyses show less nuclear STAT3 localization in plaques of mice treated with LGK974 compared to the control group (Figure 7G). Collectively, these data suggest that Wnt enhances the response to IL-4 via STAT3 activation.

We next aimed at understanding how Wnt signaling promotes STAT3 activation. Since STAT3 is most commonly activated downstream of cytokine receptors, together with our observation that Wnt3a promotes STAT3 phosphorylation 16 hr post treatment, but not in early time points, we postulated that Wnt3a induces the production and secretion of a factor or factors that accumulate in the conditioned media (CM) and activate STAT3 in an autocrine/paracrine manner. To test this hypothesis, CM were generated from BMDMs treated with Wnt3a for 16 hr. Naive BMDMs were then treated with the CM or Wnt3a itself for 15, 30 and 60 min, and STAT3 phosphorylation was examined. The data show that the CM caused STAT3 phosphorylation after 15 min, while Wnt3a itself failed to do so (Figure 7—figure supplement 1E). Furthermore, to examine whether Wnt CM augments pro-resolving gene expression, BMDMs were treated with IL-4 for 2.5 hr either after 16 hr of Wnt3a treatment or in combination with Wnt CM. Similar to our previous results, co-treatment of BMDMs with Wnt3a, as well as Wnt CM, in combination with IL-4 promoted enhanced Arg1 expression compared with IL-4-only treatment (Figure 7—figure supplement 1F). To confirm that the augmentation in Arg1 expression is not due to residual Wnt3a, CM was treated with a Wnt3a-blocking antibody prior to its exposure to naïve BMDMs. The data show that the CM retained its IL-4 augmentation ability (Figure 7—figure supplement 1F), suggesting that Wnt3a stimulated the production and/or secretion of a factor/s other than itself that, in turn, activated STAT3.

As a first step in the investigation of the Wnt3a-stimulated secreted factor that activates STAT3, we mined our RNA-seq dataset (Figure 7A) for soluble proteins with known links to STAT3 activation that showed differential expression following Wnt3a treatment. This showed differential expression of transforming growth factor β (TGFβ) and C-X-C motif chemokine receptor 3 (CXCR3) in Wnt3a-treated cells. Next, TGFβ and CXCR3, as well as the main receptors upstream of STAT3 activation (glycoprotein 130 [gp130], IL-10 receptor [IL10R], and hepatocyte growth factor receptor [HGFR]), were inhibited using specific blocking antibodies, concomitant with treatment of the BMDMs with IL-4 alone or a combination of IL-4 and Wnt3a. Arg1 expression was not attenuated by blocking any of the aforementioned proteins in Wnt3a+IL-4-treated cells (Figure 7—figure supplement 1G), indicating that these factors do not play a role in the enhanced pro-resolving gene expression following Wnt3a treatment.

Because we could not find a potential secreted protein to activate STAT3 following Wnt3a treatment, together with experiments showing that boiled Wnt CM retained its ability to augment Arg1 expression (data not shown), we postulated that the Wnt3a-stimulated secreted factor is a lipid. Thus, targeted liquid chromatography-tandem mass spectrometry (LC-MS)/MS metabololipidomics analysis of Wnt3a CM was performed. We were particularly interested in inflammation-resolving lipid mediators (representative MS/MS fragmentation spectra are shown in Figure 8—figure supplement 1A, B). We found that several lipid mediators were upregulated in CM of Wnt3a-treated BMDMs (Figure 8A), notably at the time when CM enhances the response to IL-4. Specifically, there was a global increase of arachidonic acid metabolites, such as TXB₂, PGE₂, PGD₂, PGF_2a, peaking 16 hr post-Wnt3a treatment (respective fold changes of 18, 58, 101, and 5, Figure 8A).

Of the upregulated metabolites, we focused on PGE₂ because of its function as an amplifier of IL-4 responses (Sanin et al., 2018; Gao et al., 2016; Luan et al., 2015; Heffron et al., 2021; Figure 8—figure supplement 1C) and its role as a STAT3 activator in various cell types (Jw et al., 2019; Chun et al., 2010; Wang et al., 2015; Halpern et al., 2019). For example, Sanin et al. recently examined the relationship between PGE₂ and IL-4 to regulate pro-resolving gene expression and have performed RNA-seq of BMDMs co-treated with IL-4 and PGE₂ (Sanin et al., 2018). Thus, we compared those data to our RNA-seq data of BMDMs co-treated with Wnt3a and IL-4. The data show a high degree of correlation in gene expression between BMDMs treated with Wnt3a+IL-4 and PGE₂+IL-4, with r² = 0.481 and p-value<2.34⁻⁷⁵ (Figure 8B), despite marked differences in the IL-4 concentrations used in these experiments (0.1 and 20 ng/mL, Sanin et al., 2018, respectively).

Because the canonical Wnt pathway was previously shown (e.g., Araki et al., 2003; Nunez et al., 2011) to regulate Ptgs2 expression (the enzyme that metabolizes arachidonic acid to PGH₂, the PGE₂ precursor), we examined its levels following Wnt3a stimulation in BMDMs. The results show increased Ptgs2 expression upon Wnt3a treatment, as well as upregulation of the next enzyme in the PGE₂ production pathway, Ptges (Figure 8C). In addition to the increase in these enzymes that produce PGE₂, Wnt3a also decreased the expression of Hpgd, which inactivates PGE₂ and serves as a negative regulator of its levels (Figure 8C).

To directly assess whether PGE₂ is the secreted factor in Wnt3a CM that cooperates with IL-4, we employed two strategies. First, Wnt CM was treated with a PGE₂-blocking antibody and then added to naive BMDMs together with IL-4 for 2.5 hr. While blocking of Wnt3a itself in Wnt CM did not alter the augmented Arg1 expression (Figure 7—figure supplement 1F), PGE₂ blockade completely abrogated the enhancement in gene expression (Figure 8D). In a complementary experiment, we inhibited the PGE₂ receptors EP2 and EP4 prior to exposure of BMDMs to Wnt CM and IL-4. The results showed that control BMDMs had increased Arg1 expression upon treatment with Wnt3a+IL-4 or CM+IL-4 (compared with IL-4 alone), with similar responses in BMDMs pre-treated with either EP2 or EP4 inhibitors (Figure 8E). However, BMDMs in which both receptors were inhibited simultaneously did not show enhanced Arg1 expression upon Wnt3a+IL-4 or CM+IL-4 treatment, but rather responded similarly to IL-4-only-treated cells (Figure 8E).

Finally, to determine whether Wnt inhibition in vivo may affect PGE₂ production, we examined the expression of Ptgs2, Ptges, and Hpgd in macrophages LCM-excised from plaques of control or LGK974-treated mice (as in Figure 6—figure supplement 1I). Although expression of Ptgs2 and Hpgd did not differ between the two groups, the expression of Ptges was reduced by 58% in the LGK974 group compared to controls (Figure 8F). Taken together, our data suggest that macrophages activated by the Wnt pathway produce and secrete PGE₂, activating STAT3 in an autocrine/paracrine fashion, which subsequently cooperates with IL-4/STAT6 (Figure 8G). In atherosclerosis, this mechanism primes the response of plaque macrophages to IL-4 that accumulates in plaques during disease progression and together induce resolution of atherosclerosis.

This study aimed at determining molecular mechanisms upstream of STAT6 that promote atherosclerosis resolution. Mouse studies of atherosclerosis resolution were performed on the basis of LDLr deficiency (Ldlr^-/-, Pcsk9-AAV (Peled et al., 2017; Bjørklund et al., 2014; Roche-Molina et al., 2015) and Reversa Feig et al., 2011a; Lieu et al., 2003), with WD (100244; Dyets Inc) feeding for 16–20 weeks, followed by aggressive cholesterol lowering (using ApoB-ASO for Ldlr^-/- and Pcsk9-AAV or poly I:C injections for Reversa; Feig et al., 2011a; Lieu et al., 2003) with chow diet feeding, according to the protocol (number IA16-00494) approved by the Institutional Animal Care and Use Committee of the New York University School of Medicine. In the aortic transplant model (Rahman et al., 2017; Chereshnev et al., 2003), aortic arches from atherosclerotic (Ldlr^-/-/Apoe^-/-) mice were intrapositioned with the abdominal aorta in recipient mice (Il4^-/-Il3^-/-/Il4^-/-, maintained on a chow diet), and blood flow was directed through the graft. For in vivo Wnt inhibition studies, mice were administered 3 mg/kg LGK974 (Liu et al., 2013) (S7143; Celleckchem) three times/week by oral gavage. For all studies, mice were randomized into the groups at the BL time point. Researchers were not blinded to the identity of the study groups, with blinding prior to data analysis. Based on a power analysis (performed as described on https://stats.idre.ucla.edu/other/mult-pkg/seminars/intro-power/), we typically enroll 15 mice to have at minimum 10 mice/group; this is estimated to give 90% power to detect differences greater than 20% in atherosclerotic plaque parameters with ≈15% standard deviation between wild type and test mice based on two-sided two-sample t-test with α = 0.05. In some cases, fewer mice were used in order to have enough tissues for all of the sample types needed.

C57BL/6, Il4^-/-, Stat6^-/-, Apoe^-/-, Ldlr^-/-, and Reversa mice were purchased from The Jackson Laboratory. Il4^-/-Il13^-/-mice (C57BL6/J background) were provided by Dr. Thomas Wynn (then at the National Institutes of Health). IL-4/eGFP mice were provided by Dr. Richard Locksley (University of California San Francisco). Male and female mice at ages 6–10 weeks were used.

This new atherosclerosis resolution model was recently described (Peled et al., 2017; Bjørklund et al., 2014; Roche-Molina et al., 2015; Sharma et al., 2020). Briefly, mice were injected intraperitoneally with an AAV that expresses a hyperactive form of Pcsk9 (AAV.8TBGmPCSK9D377Y, Penn Vector Core), under the regulation of a liver-specific promoter, at 1 × 10¹² viral particles/mouse. After 16–20 weeks of WD feeding, lipid lowering was achieved by discontinuing the WD feeding and bi-weekly i.p. injections of 50 mg/kg ASO to ApoB, kindly provided by Ionis Pharmaceuticals. Mice were sacrificed after 3 weeks of ApoB ASO treatment and their tissues collected.

For the data displayed in Figure 4D, they were based on the initial analysis of human atherosclerosis samples as previously reported (Fernandez et al., 2019). Subjects gave informed written consent, following the approval by the Institutional Review Board of the Icahn School of Medicine at Mount Sinai (IRB 11-01427). For the data displayed in Figure 4E, F, the examination of atherosclerosis resolution features in humans was described previously (Alkhalil et al., 2018). Ethical approval was obtained from National Research Ethics Services (NRES) and Oxford R and D committee prior to commencement of the study. All patients provided written informed consent prior to participation in this study, which included the collection of peripheral blood cells and their analysis.

Hearts were removed after perfusion at physiological pressure of saline containing 10% sucrose, embedded in OCT, and frozen. Hearts were sectioned through the aortic root (6 μm) and stained for CD68 (MCA1957; AbD Serotec) to detect macrophages, active-β-catenin (8814; Cell Signaling Technology), STAT3 (8768; Cell Signaling Technology), MRC1 (MCA2235; AbD Serotec), ARG1 (sc-20150; Santa Cruz), VCAM1 (ab134047; Abcam), and CCL2 (505908; BioLegend). Picrosirius red was used for collagen staining, and imaging done using polarizing light microscopy. Image analysis was performed with ImagePro Plus 7.0 software (Media Cybernetics) and ImageJ (National Institute of Health Schneider et al., 2012).

Aortic arches were digested as previously described (Tang et al., 2015), after their dissection following perfusion with PBS. Single-cell suspensions were added a live/dead cell staining (L23105; Invitrogen) and PE-Cy7 anti-CD45 antibody (103132; Biolegend) and analyzed using the LSRII cytometer (BD Biosciences).

Plaques from Ldlr^-/- mice fed WD for 20 weeks and treated or not with ApoB ASO for 3 weeks were carefully scraped out from the aortic wall, placed in a tube containing 100 μL PBS and protease/phosphatase inhibitor (A32961, Thermo Fisher Scientific) and weighed. Plaques were manually homogenized until no pieces were apparent (~2 min). Concentrations of CCL2, IL-4, and IL-13 in suspensions were measured using a custom mouse LEGENDplex assay (Biolegend).

Bone marrow was harvested from the femur and tibia of 6–20-week-old mice and treated with red-blood cell lysis buffer (R7757; Sigma-Aldrich) for 5 min. After washing, cells were seeded in DMEM media supplemented with 10% FBS, 1% penicillin/streptomycin, and 10 ng/mL macrophage colony-stimulating factor (m-CSF) (315-02; PeproTech). Cells were supplemented with media every other day and used for experiments between day 5 and 10 post harvest. Cells were typically treated with 100 ng/mL Wnt3a (1324-WN-010/CF; R&D Systems), 100 pg/mL IL-4, 5 ng/mL IL-6, and 2.5 ng/mL IL-10 (214-14, 216-16, and 210-10, respectively; Peprotech) for 16 hr, unless indicated otherwise. Experiments with Wnt CM or 100 pg/mL PGE₂ (14010; Cyaman Chemical) include treatment with IL-4 for 2.5 hr. For experiments with siRNA, BMDMs were transfected overnight with 25 nM scrambled (control) or STAT3 siRNA (L-040794-01; Dharmacon) using Lipofectamine RNAiMax Transfection reagent (13778075; Thermo Fisher Scientific), and started on Wnt3a and IL-4 treatments 48 hr after transfection. Blocking antibodies and protein inhibitors were used in the following concentrations: 0.5 µg/mL anti-TGFβ (521707; Biolegend), 0.1 µg/mL anti-gp130 from (MAB4682; R&D Systems), 1 µg/mL anti-HGFR (AF527; R&D Systems), 10 µg/mL anti-IL10R (112707; Biolegend), 0.1 µM AMG487 CXCR3 inhibitor (4487; Tocris Bioscience), 1 µg/mL anti-Wnt3a (MAB9025; R&D Systems), 100 nM anti-PGE₂ (10009814; Cayman Chemical), 30 µM AH6809 EP2 inhibitor (14050; Cayman Chemical), and 30 µM GW627368X EP4 inhibitor (10009162; Cayman Chemical).

Under strict RNAse-free conditions, aortic root sections were stained with hematoxylin-eosin and captured. RNA was then isolated using the PicoPure Kit (KIT0204; Thermo Fisher Scientific), and quality and quantity were determined using an Agilent 2100 Bioanalyzer (5067-1513; Agilent Technologies). RNA was converted to cDNA and amplified using the WT-Ovation Pico RNA Amplification Kit (3302-12; NuGEN) (Ramsey et al., 2014; Feig et al., 2012; Trogan et al., 2002).

Total RNA was isolated using TRIzol reagent (15596-018; Life Technologies) and Direct-zol RNA MiniPrep columns (R2052; Zymo Research) and quantified using Nanodrop 2000 (Nanodrop Products). Quantitative real-time PCR was performed using FAST SYBR Green Master Mix (4385612; Applied Biosystems) on the ABI PRISM 7300 sequence detection system (Applied Biosystems). The mRNA levels were normalized to Gapdh or Hprt as reference genes. Primer sequences are available in Supplementary file 1.

RNA was isolated from PBMCs collected from patients in the study in reference (Alkhalil et al., 2018), and the expression of Wnt target genes (Axin2, Ddx3x, Wls, Senp2, Ctnnb1, Csnk1a1, and Amfr) was measured by RT-PCR, with 18S RNA as a reference. Imaging of carotid atherosclerotic plaques was performed as described in Alkhalil et al., 2018.

BMDMs were treated with media (untreated), Wnt3a, IL-4 ,and Wnt3a+IL-4 for 16 hr. RNA was isolated as described above. Single-read RNA sequencing was performed using the Illumina HiSeq 2500. All samples had high quality, when checked with FastQC v0.11.7 (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Illumina adapter sequences and poor-quality bases were trimmed using trimmomatic v0.36 (Bolger et al., 2014). Trimmed sequences were mapped to the mm10 mouse reference genome using STAR v2.6.0a (Dobin et al., 2013), indexed using samtools v1.9 (Li et al., 2009), then quantified for UCSC genes using HTSeq-count v0.11.1 (Anders et al., 2015). Comparative analysis between conditions was performed using DESeq2 v1.24.0 (Love et al., 2014) with default parameters. PCA was performed on the RNA (19,862 transcripts) dataset using PCA function in MATLAB (R2017a).

We summed the expression values of genes associated with the GO term GO:0016055 (Wnt signaling pathway), added the summed expression as a metadata feature using Seurat, and plotted the expression in UMAP using Seurat FeaturePlot. We also plotted the level of summed Wnt expression for each cell, summarized as a violin plot, and split by treatment group. Significance was assessed using the Wilcoxon rank-sum test within each cluster between groups.

For each of gene associated with GO:0016055, we calculated a log fold-change (logFC) value for resolution/progression for each cluster using the Seurat FindMarkers function. We then performed hierarchical clustering based on the gene logFC values and visualized them using pheatmap.

The sum and average of the logFC across macrophage clusters in the mouse dataset were analyzed for GO:0016055. The top 10 genes with the highest sum were the same as the top 10 genes with the highest average, and these genes were chosen for further analysis. LogFC was calculated for symptomatic versus asymptomatic patients. To determine the significance of the enrichment for negative logFC values (symptomatic/asymptomatic) in the subset of 10 Wnt responsive genes, we calculated logFC values (symptomatic/asymptomatic) for all genes in the dataset and ran a permutation test with 10,000 permutations. Specifically, for each of the permutations, we randomly selected 10 genes from the set of all genes and recorded the number of negative logFC values in that random selection. p-Values were calculated as the fraction of permutations that yielded seven or more negative logFC values. Human studies consent was obtained as described in Fernandez et al., 2019; Alkhalil et al., 2018.

BMDM lysates were run in 12% SDS-polyacrylamide gels under denaturing conditions and blots were stained with anti-pY-STAT3 (9131;Cell Signaling Technology), anti-total-STAT3 (8768; Cell Signaling Technology), β-catenin (sc-7963; Santa Cruz), and anti-GAPDH antibodies (G8795; Sigma-Aldrich). Immunoreactive bands were visualized using the Odyssey imaging system (LI-COR Biosciences) by using horseradish peroxidase-conjugated secondary antibodies (ab205718; Abcam and 626520; Thermo Fisher Scientific) and chemiluminescence.

Media of BMDMs treated with Wnt3a was immediately added to two volumes of ice-cold methanol and stored at −80°C (Dalli et al., 1730). Prior to solid phase extraction, 500 pg of deuterated synthetic standards (d5-RvD2, d5-LXA₄, d4-LTB₄, and d4-PGE₂; Cayman Chemical) were added to the samples. Samples were centrifuged (3000 rpm for 10 min at 4°C) and supernatants collected, acidified to pH 3.5, and extracted using C18 cartridges (Biotage). Finally, lipid mediators were eluted from the column by addition of methyl formate. The solvent was then evaporated under N₂ gas. Samples were resuspended in a 1:1 mixture of methanol and water and subjected to LC-MS/MS analysis using a Poroshell reverse-phase C18 column (100 mm × 4.6 mm × 2.7 µm; Agilent Technologies) equipped high-performance liquid chromatography system (Shimadzu) coupled to a QTrap 5500 mass-spectrometer (AB Sciex) operating in negative ionization mode. Scheduled multiple reaction monitoring (MRM) transitions coupled with information-dependent acquisition and enhanced product ion-scanning was used thereafter. Specific lipid mediators were identified by matching their retention time and at least six diagnostic MS/MS ions, compared with external standards (Cayman Chemical) analyzed in parallel. For quantification, the MRM peak was compared with synthetic standards. Extraction recovery was determined using the internal deuterated standards added to the samples prior to extraction.

GraphPad Prism 7 (GraphPad Software) was used for statistical analysis. Data are expressed as mean ± SEM. With three or more groups, statistical analysis was performed using one-way ANOVA, with Dunnett’s multiple comparisons testing and Gaussian distribution. Comparison of two parameters for two or more groups was performed using a two-way ANOVA, with Sidak’s multiple comparison testing. Data involving comparisons between two groups were analyzed using two-tailed t-test. p<0.05 was considered significant.