Introduction

Atherosclerosis is a chronic inflammatory disease triggered by imbalanced lipid metabolism (Rocha & Libby, 2009; Weber & Noels, 2011). In atherosclerotic plaques, macrophages ingest lipoproteins and transform into lipid-laden foam cells. The foamy macrophages lose their ability to migrate away from the plaques, where they sustain a local state of sterile inflammation, which occurs in the absence of pathogens and is typically associated with the release of immune-recognizable cellular content from damaged or dying cells (Randolph, 2014). Cholesterol efflux and efferocytosis (the engulfment and clearance of apoptotic cells (AC)) by macrophages, on the other hand, help to resolve inflammation and contribute to plaque stability as counterbalancing mechanisms that oppose plaque rupture (Khera et al, 2011; Westerterp et al, 2013, 2018; Kojima et al, 2017; Yurdagul et al, 2017). Cholesterol efflux by plaque macrophages is the first step in a multistep process, referred to as “reverse cholesterol transport” (RCT), that reduces lipid accumulation in plaques. Macrophages efflux intracellular cholesterol using their plasma membrane cholesterol transporters (such as the ATP-binding cassette (ABC) transporters subfamily A member-1 (ABCA1) and subfamily G member-1 (ABCG1)), which in turn hand the exported cholesterol over to lipid-poor apolipoproteins, forming high-density lipoprotein (HDL) particles (Costet et al, 2000; Oram et al, 2000). The cholesterol efflux pathway is transcriptionally activated by the metabolic by-products of AC-derived cholesterol in the efferocytic macrophages. As such, efferocytosis and RCT synergize to reduce necrosis and resolve inflammation in plaques (Klucken et al, 2000; Joyce et al, 2002; Tall & Yvan-Charvet, 2015; Zimmer et al, 2016; Guo et al, 2018).

Transcriptional and post-transcriptional regulation play critical roles to set sterile inflammation in motion but also in its resolution and, eventually, plaque regression. Post-transcriptional regulatory steps often involve RNA-binding proteins (RBPs) that interact with mRNA to alter its processing, transport, translation, and degradation. Several RBPs have been shown to alter the mRNA stability and translation of cytokines to turn-off the inflammatory response and key molecular regulators of cholesterol homeostasis and some have been linked to atherosclerosis development (atherogenesis) (Chiu et al, 1997; Kang et al, 2011; Zhang et al, 2013; Ramírez et al, 2014; Mobin et al, 2016; Haneklaus et al, 2017). Fragile X Mental Retardation Protein (FMRP) is an RBP that has been widely studied in neurons due to its causal role in the Fragile X Mental Retardation Syndrome (FXS) (Hersh & Saul, 2011; Bagni et al, 2012; Hunter et al, 2014). In FXS, a hypermethylated CGG repeat expansion in the 5′ untranslated region (5′ UTR) of the FMR1 mRNA results in its transcriptional silencing. A subpopulation of the individuals afflicted with FXS and FMRP-deficient (Fmr1−/−) mice have lower cholesterol levels (Leboucher et al, 2019b). While a prior study has shown FMRP protein expression is induced in macrophage-enriched areas of the human atherosclerotic plaque (Hansmeier et al, 2018), FMRP’s contribution to the atherosclerotic process has not been investigated directly, and our knowledge of FMRP function has remained largely limited to studies in neurons (Darnell et al, 2011). A key serine phosphorylation (S500 in human; S499 in mouse) on FMRP triggers hierarchical phosphorylation of surrounding serines and threonines, while enhancing FMRP’s translation-repressing activity on many synaptic function-linked mRNAs bound by it. The identity of the kinase(s) that phosphorylates FMRP has remained subject to intense discussion (Ceman et al, 2003; Narayanan et al, 2008; Coffee et al, 2012; Niere et al, 2012; Bartley et al, 2014, 2016; Prieto et al, 2020).

In this study, we present evidence that ER stress and hypercholesterolemia in mice induces the phosphorylation of macrophage FMRP on S500 by the inositol-requiring enzyme-1 (IRE1), a conserved endoplasmic reticulum (ER) stress-sensing kinase/endoribonuclease (RNase). Metazoans have two IRE1 paralogues, IRE1α (referred to as IRE1 in this paper) and IRE1β. While IRE1α is ubiquitously expressed, IRE1β expression is restricted to gastrointestinal epithelium (Cloots et al, 2021). To date, IRE1 has been described to trans-autophosphorylate as a first step in the activation of its RNase modality, which initiates a nonconventional RNA-splicing reaction and the production of the transcription factor known as spliced X box protein-1 (XBP1s), which is one of the key drivers of the unfolded protein response (UPR) (Walter & Ron, 2011; Çimen et al, 2016a; Robblee et al, 2016). IRE1 senses both protein folding stress induced by an accumulation of unfolded proteins in the ER lumen and ER membrane lipid bilayer stress induced by an accumulation of cholesterol or saturated fatty acids (SFA) (Seimon et al, 2010; Sukhorukov et al, 2020). While previous studies by us and others have shown that ER stress and subsequent IRE1 activation are causally associated with atherosclerosis progression, the mechanism by which IRE1 contributes to the disease pathogenesis has remained elusive (Erbay et al, 2009; Zhou & Tabas, 2013; Tufanli et al, 2017). Here, we show that IRE1 phosphorylates FMRP on S500, which in turn leads to post-transcriptional suppression of cholesterol efflux and efferocytosis by macrophages. FMRP deficiency and IRE1 kinase inhibition both enhance RCT and efferocytosis in vivo, reducing foam cell formation and atherosclerosis progression in mice. These findings reveal a novel role for FMRP in macrophages in the regulation of cholesterol homeostasis and efferocytosis and provide mechanistic insight into IRE1-driven atherosclerotic processes during hypercholesterolemia.

Results Lipids induce FMRP phosphorylation in IRE1 kinase-dependent manner

Previously published mass spectrometry-based IRE1 interactome data revealed multiple FMRP-interacting proteins potentially associating with IRE1 (Fig 1A) (Acosta-Alvear et al, 2018). These proteins share significant sequence homology with FMRP and are usually found in homo/heteromeric complexes with FMRP (Zhang et al, 1995). Based on these observations, we reasoned that FMRP might also physically interact with IRE1. Indeed, FMRP co-immunoprecipitated with IRE1 in non-stress and ER stress conditions induced by thapsigargin (TG; an inhibitor of the ER Ca2+ pump) and tunicamycin (TM; an inhibitor of N-linked glycosylation) from human embryonic kidney cell line (HEK293T) cells that were transiently transfected with plasmids encoding both proteins (Fig 1B).

Figure 1. FMRP is a novel IRE1 kinase substrate

STING analysis of published IRE1 interactome proteins in relation to FMRP (Acosta-Alvear et al, 2018). HEK293T cells were co-transfected with IRE1 and FMRP plasmids and stimulated with TG (600 nM) or TM (1 mg/ml) for 2 h. Protein lysates were immunoprecipitated (IP) with anti-IRE1 or IgG (control) antibodies and analyzed by Western blotting using specific antibodies for FMRP and IRE1 (n = 3 biological replicates). RAW 264.7 mouse macrophages were treated with either oxLDL (50 µg/ml) or TG (300 nM) for 6 h. Protein lysates were treated with λ Phosphatase (PPase) for 30 min and analyzed by Western blotting using specific antibodies for pFMRP, FMRP, pIRE1, IRE1, and β-Actin. pFMRP/FMRP fold induction is depicted above the blots (n = 6 biological replicates). Apoe−/− mice were fed with chow diet (CD) or western diet (WD) for 16 weeks followed by peritoneal macrophage (PM) isolation. Protein lysates were analyzed by Western blotting using specific antibodies for pFMRP, FMRP, pIRE1, IRE1, and β-Actin. pFMRP/FMRP-fold induction is depicted above the blots (n = 5 mice per group). Control- or IRE1-siRNA transfected HEK293T cells were stimulated by either PA (500 µM) or TG (600 nM) for 4 h. Protein lysates were analyzed by Western blotting using specific antibodies for pFMRP, FMRP, pIRE1, IRE1, and β-Actin. pFMRP/FMRP fold induction is depicted above the blots (n = 4 biological replicates). Protein lysates of thioglycolate-elicited PM from IRE1α+/+ and IRE1α−/− mice (after 16 weeks on WD) were analyzed by Western blotting using specific antibodies for pFMRP, FMRP, pIRE1, IRE1, and β-Actin. pFMRP/FMRP fold induction is depicted above the blots (n = 4 mice per group). MEF cells were transfected with either empty vector, EGFP-FMRP or 3xFLAG-IRE1 plasmids then pre-treated either with vehicle (dimethyl sulfoxide, DMSO) or AMG-18 (25 µM; 1 h) followed by TG (600 nM) stimulation for 4 h. Protein lysates were analyzed by Western blotting using specific antibodies for pFMRP, FMRP, pIRE1, IRE1, and β-Actin. pFMRP/FMRP fold induction is depicted above the blots (n = 4 biological replicates). C57BL/6 were injected either with DMSO or AMG-18 (30 mg/kg; 8 h), followed by TM injection (1 mg/kg; 8 h). Protein lysates of thioglycolate-elicited PM were analyzed by Western blotting using antibodies for pFMRP, FMRP, pIRE1, IRE1, and β-Actin. pFMRP/FMRP fold induction is depicted above the blots (n = 4 mice per group). HEK293T cells were transfected with either empty vector (EV), IRE1-WT, or IRE1–KD plasmids and stimulated by TG (600 nM; 1 h). Protein lysates from each transfection were separately immunoprecipitated (IP) with anti-IRE1 antibody and subjected to a kinase reaction with purified hFMRP protein and ATP-γ-S (100 µM) in kinase buffer. The IP protein were analyzed by Western blotting using specific antibodies for thiophosphate esters (ThioP), IRE1, and FMRP (n = 3 biological replicates). Purified FMRP and IRE1 kinase (activated) proteins were subjected to kinase assay and analyzed by Western blotting using specific antibodies for ThioP, IRE1, and FMRP (n = 3 biological replicates) and with LC-MS/MS. Identified IRE1 kinase-mediated FMRP phosphorylation sites (bottom). Fmr1−/− mouse embryonic fibroblasts (MEF) were transfected either with EV, WT-FMRP, SA-FMRP, or STSA-FMRP plasmids followed by PA treatment (500 µM; 6 h). Protein lysates were analyzed by Western blotting using specific antibodies for FMRP, pFMRP, pIRE1, and β-Actin (n = 3 biological replicates).

Data information: A representative blot is shown. In D, E, G, and H data are cumulative results of two independent experiments. Data are mean ± SEM. Unpaired t-test with Welch’s correction or paired t-test.

Source data are available online for this figure.

Since FMRP phosphorylation is critical for translation suppression and the association with IRE1 juxtaposes it to a kinase whose substrate is unclear, we wondered whether ER stress alters FMRP phosphorylation state. To assess this possibility, we treated cultured macrophages with known ER stressors, such as TG and oxidized low-density lipoprotein (oxLDL, another inducer of ER stress). Using specific antibodies that recognize S724 phosphorylation on IRE1 and S500 phosphorylation on FMRP (Reynolds et al, 2015; Wang et al, 2021), we found that TG and oxLDL significantly induced IRE1 autophosphorylation and FMRP phosphorylation but did not affect levels of FMRP protein or Fmr1 mRNA (Fig 1C and Appendix Fig S1A and B). Phosphatase treatment of the samples partially reversed the ER stress-induced increase in the pFMRP/FMRP ratio (Fig 1C, Appendix Fig S1B), implying FMRP phosphorylation is enhanced by these ER stressors.

Previous studies showed that hyperlipidemia induces ER stress in plaque macrophages in vivo (Moore et al, 2013; Kim et al, 2016). We next investigated whether hyperlipidemia also has an effect on FMRP phosphorylation on S499 (mouse S499 corresponds to human S500). To this end, we used mice deficient in apolipoprotein E (Apoe−/−) as this is a protein found in plasma lipoprotein particles and facilitates cholesterol clearance from the circulation (Davignon et al, 1999). In agreement with previous reports, we observed that ER stress (as monitored by IRE1 autophosphorylation) was induced in the peritoneal macrophages (PM) obtained from hyperlipidemic, Apoe−/− mice that were fed with a Western diet (WD) for 16 weeks when compared to Apoe−/− mice fed with chow diet (CD) (Fig 1D). We found that FMRP S499 phosphorylation, which leads to FMRP-mediated translational suppression, was 1.4-fold elevated by a chronic exposure to hypercholesterolemia, whereas FMRP protein and Fmr1 mRNA expression levels remained unchanged (Fig 1D; Appendix Fig S1C and D).

We next asked what role IRE1 plays in ER stress-induced FMRP phosphorylation. To address this question, we transfected IRE1-specific siRNA to suppress IRE1 expression in a human cell line, followed by ER stress induction by a saturated fatty acid, palmitate (PA), that is known to induce ER stress, or TG. While both ER stressors induced the phosphorylation of IRE1 and FMRP, this was prevented in IRE1 knock-down cells (Fig 1E and Appendix Fig S1E). In addition, we analyzed hypercholesterolemia-induced FMRP phosphorylation in PMs obtained from Apoe−/− mice with a genetic deletion of IRE1α in the myeloid lineage (IRE1−/−) (see Methods) after feeding with WD for 16 weeks. FMRP phosphorylation, but not FMRP protein or Fmr1 mRNA, was reduced in PM isolated from IRE1−/− mice when compared to those isolated from IRE1+/+ mice (Fig 1F, Appendix Fig S1F and G). While these results show a clear reduction in FMRP phosphorylation upon siRNA treatment or gene knock-out, we observed residual signal in both cases (Fig 1E and F). As we confirm below (Fig 1K), the antibody used in these experiments is phosphorylation specific. We, therefore, surmise that partial phosphorylation on S500/S499 may also be mediated by the other kinases that are known to phosphorylate this residue on FMRP (Narayanan et al, 2008; Bartley et al, 2016), in addition to IRE1 as we show here. Collectively, our in vitro and in vivo results strongly support that ER stress-induced activation of IRE1 kinase leads to enhanced phosphorylation of FMRP on S500/S499.

IRE1 phosphorylates FMRP

To begin investigating the role of IRE1’s kinase activity in ER stress-induced FMRP phosphorylation, we expressed a 3xFLAG-tagged IRE1 (FLAG-IRE1) and/or EGFP-FMRP in wild-type mouse embryonic fibroblasts (MEFs). Endogenous FMRP and IRE1 migrated faster in SDS–PAGE gel than the epitope tagged EGFP-FMRP and FLAG-IRE1, respectively. In all conditions, the MEFs were TG-treated (to stimulate IRE1 kinase activity) in the absence or presence of an IRE1 kinase-specific inhibitor (AMG-18) (Papandreou et al, 2011; Ghosh et al, 2014; Tufanli et al, 2017; Harnoss et al, 2019). In the absence of AMG-18, both IRE1 and FMRP were phosphorylated. AMG-18 treatment prevented IRE1 phosphorylation while clearly reducing FMRP phosphorylation (Fig 1G and Appendix Fig S1H). While these data further support the notion that IRE1 phosphorylates FMRP during ER stress, alternative kinase(s) appears to mediate the same reaction. This is consistent the published data that several kinases can phosphorylate the same residue on FMRP (Narayanan et al, 2008; Bartley et al, 2016). Furthermore, in an in vivo setting where mice were injected with TM to induce IRE1 kinase activity, treatment with AMG-18 inhibited IRE1 kinase activity and reduced FMRP phosphorylation in PM (Fig 1H and Appendix Fig S1I). Taken together, our findings support the notion that IRE1 kinase activity makes an important contribution to ER stress-induced FMRP S499 phosphorylation in mouse macrophages and MEFs as well as S500 in HEK293T cells. Our data also support that other known or unknown FMRP kinase(s) are responsible for basal FMRP phosphorylation that is observed in non-stress conditions.

To ask whether IRE1 can phosphorylate FMRP directly, we employed in vitro assays. To this end, we immunoprecipitated wild type (WT) or kinase dead (KD) mutant IRE1 from HEK293T cells (after treating with ER stressor) and incubated the immunoprecipitates with purified, recombinant human FMRP protein in a kinase reaction. The reaction included ATP-γ-S instead of ATP, which allows kinases to thio-phosphorylate their substrates. The resultant kinase reaction was analyzed by Western blotting using an anti-thiophosphate ester antibody. In the reaction containing IRE1-WT, both proteins were thio-phosphorylated, but not in the reaction containing the IRE1-KD mutant (Fig 1I).

To determine the specific amino acids phosphorylated by IRE1, we performed the kinase reaction using purified, recombinant human IRE1-kinase/RNase domains (amino acids 468–977) and human FMRP proteins. Both IRE1 and FMRP were phosphorylated in this reaction in an AMG-18-sensitive manner (Fig 1J). Liquid chromatography-mass spectrometry (LC-MS/MS)-based analysis of the phosphorylated FMRP residues in this kinase reaction revealed eight IRE1-induced phosphorylation sites on FMRP at serine (S362, S494, S497, S500, S504) and threonine (T502, T592, T594) (Fig 1J, Appendix Fig S1J and K). Importantly, S500 in human FMRP is the previously identified FMRP phosphorylation site that was shown to enhance FMRP-mediated translational suppression (Ceman et al, 2003; Narayanan et al, 2008; Coffee et al, 2012; Niere et al, 2012; Bartley et al, 2014, 2016; Prieto et al, 2020).

Using site-directed mutagenesis, we engineered two mutant versions of FMRP, in which we either changed S500 to alanine (SA mutant) or S500, T502, and S504 to alanine (STSA triple mutant) to block phosphorylation at S500 and amino acids in its proximity. We then induced ER stress with PA in Fmr1−/− MEFs transiently transfected with WT FMRP and FMRP mutants (SA and STSA). PA induced FMRP phosphorylation in WT FMRP-reconstituted cells but failed to do so in cells expressing the SA and STSA mutants of FMRP (Fig 1J). These data confirm the specificity of the FMRP antibody for phosphorylated S500 and is consistent with the notion that IRE1 kinase directly phosphorylates human FMRP protein on S500.

IRE1-FMRP signaling induces foam cell formation while suppressing RCT

We next wondered what role FMRP plays in macrophage biology that is relevant to atherosclerotic plaque development. We performed an in vivo macrophage foam cell formation assay in the peritoneum of Fmr1−/− and Fmr1+/+ mice, using a well-established method in which we induced hyperlipidemia using a combination of adenoviral-delivery of proprotein convertase subtilisin kexin 9 (AAV_PCSK9), a protein that directs hepatic low-density lipoprotein (LDL) receptors for degradation, and feeding with WD for 16 weeks (Li et al, 2007; Tsimikas et al, 2011; Peled et al, 2017). FMRP deficiency significantly reduced foam cell formation in vivo (Fig 2A). Next, we fed Apoe−/− mice with a WD (12 weeks) and injected them daily with the IRE1 kinase inhibitor, AMG-18, or vehicle for the last 4 weeks. AMG-18 reduced foam cell formation in the peritoneum in vivo (Fig 2B). This result indicates that IRE1-FMRP signaling axis enhances foam cell formation.

Figure 2. FMRP deficiency enhances RCT while reducing foam cell formation in vivo

A. Fmr1+/+ and Fmr1−/− mice were injected with AAV_PCSK9 and fed with 16 weeks of WD. Residential PM were stained with Oil Red O (ORO) and imaged (n = 7 mice per group; Scale bar = 50 µm). B. Apoe−/− mice were fed with WD (12 weeks) and injected with vehicle (DMSO) or AMG-18 (30 mg/kg/day) in the last 4 weeks of WD. Residential PM were stained with ORO and imaged (n = 5 mice per group; Scale bar = 50 µm). C–E. Flow cytometry analysis of BMDMs after dil-ac-LDL (25 µg/ml) loading for 24 h; (C) control- or Fmr1-siRNA transfected BMDM (n = 4 biological replicates), (D) Fmr1+/+ and Fmr1−/− BMDM (n = 4 biological replicates), (E) BMDM pre-treated with either vehicle (DMSO) or AMG-18 (5 µM; 1 h) (n = 6 biological replicates). F, G. Macrophages were pre-loaded with fluorescently labeled cholesterol (16 h) followed by incubation in efflux medium including APOA1 (25 µg/ml) or HDL (50 µg/ml) as acceptors for 6 h. % Efflux was calculated as cholesterol signal in medium/cholesterol signal in medium and cell: Cholesterol efflux in (F) Fmr1+/+ and Fmr1−/− BMDM (n = 4 biological replicates) and in (G) BMDM that were pre-treated either with DMSO or AMG-18 (5 µM; 1 h) (n = 4 biological replicates). H–K. RCT experiment: (H) Schematic representation of C57BL6 mice were injected with [3H]-cholesterol-loaded foamy Fmr1+/+ and Fmr1−/− BMDM, (I) plasma cholesterol levels after 24 and 48 h, (J) liver cholesterol levels after 48 h, and (K) feces cholesterol levels after 48 h (n = 12 mice per group).

Data information: Data are mean ± SEM. Unpaired t-test with Welch’s correction.

Source data are available online for this figure.

We also transfected wild type BMDMs with either Fmr1-specific or control siRNA and followed by loading the cells with 3,3′-dioctadecylindocarbocyanine (dil)-labeled acetylated LDL (Ac-LDL) (Carotti et al, 2021). Flow cytometry analysis revealed that Fmr1 silencing significantly reduced % dil-acLDL internalized in macrophages (Fig 2C, Appendix Fig S2A). Likewise, Fmr1−/− BMDM displayed reduced % dil-acLDL internalization when compared to Fmr1+/+ BMDM (Fig 2D). A similar reduction in foam cell formation was observed with the IRE1 kinase inhibitor (Fig 2E). These results indicate that both the inhibition of IRE1 kinase activity and the genetic deletion of its proposed substrate, FMRP, reduce foam cell formation in vitro and in vivo. Reduced foam cell formation could be explained with less cholesterol uptake, however, neither FMRP knock down nor IRE1 kinase inhibition altered cholesterol uptake in macrophages (Appendix Fig S2B). We reasoned that this observation is most likely related to an increase in cholesterol export (RCT; due to increased translation of cholesterol exporters) from Fmr1−/− macrophages. Indeed, FMRP deficiency led to an increase in cholesterol efflux coupled with its loading onto the cholesterol carriers, apolipoprotein-A1 (APOA1) and HDL (Fig 2F), and, likewise, the IRE1 kinase inhibitor enhanced cholesterol efflux (Fig 2G). Thus, ER stress-associated reduction in cholesterol efflux is dependent on both IRE1 kinase activity and FMRP.

Next, we determined whether the absence of FMRP in BMDMs also enhances RCT in vivo. To this end, we pre-loaded Fmr1+/+ and Fmr1−/− BMDMs with [3H]-cholesterol and injected the cells subcutaneously into WT mice (Fig 2H). FMRP deficiency in macrophages significantly increased radioactivity counts ([3H]-cholesterol) in the plasma, liver, and feces of the recipient mice, when compared to recipient mice that received Fmr1+/+ macrophages (Fig 2I–K), demonstrating that FMRP deficiency in macrophages enhances RCT in mice.

FMRP regulates macrophage efferocytosis

We next investigated the impact of FMRP deficiency on efferocytosis, a primary process that promotes atherosclerotic plaque regression by removing apoptotic macrophages in the lesion area. To this end, we transfected BMDM (red fluorescent stained) with Fmr1-specific or control siRNA and incubated them with green fluorescent-labeled apoptotic cells (ACs), in which apoptosis was induced by ultraviolet (UV) irradiation. FMRP knock-down increased efferocytosis of ACs (as measured by colocalization of the fluorescent markers), when compared to control (Fig 3A). FMRP-deficient BMDMs also increased efferocytosis when compared to wild type BMDMs under both no-stress and PA-induced ER stress conditions (Fig 3B). Next, we induced hyperlipidemia in Fmr1−/− and Fmr1+/+ mice as described above using a combination of AAV_PCSK9 injection and feeding with a WD. We then injected the mice intraperitoneally with green fluorescent-labeled ACs and harvested PM macrophages. FMRP-deficient PMs displayed enhanced efferocytosis compared to WT PMs (Fig 3C).

Figure 3. FMRP deficiency increases efferocytosis in vivo

A–E. In vitro and in vivo efferocytosis experiments, where percentage of macrophages F4/80+ (red) that ingested apoptotic cells (AC) labeled with carboxyfluorescein succinimidyl ester (CFSE)+ (green) were reported as % efferocytosis. (A) BMDMs were transfected with Fmr1- or control-siRNA and incubated CFSE-labeled AC for the indicated hours (n = 4 biological replicates). (B) Fmr1+/+ and Fmr1−/− BMDMs were treated with PA (500 µM) for 6 h and then incubated with CFSE-labeled ACs for 4 h (n = 4 biological replicates). (C) Fmr1+/+ and Fmr1−/− mice were fed WD (16 weeks) and injected intraperitoneally with CFSE-labeled AC (1.5 h), followed by PM elicitation (n = 4–5 mice per group). (D) BMDM were pre-treated either with vehicle (DMSO) or AMG-18 (5 µM) for 1 h then incubated with CFSE-labeled Acs for 4 h (n = 3 biological replicates). (E) C57BL/6 mice were injected with AMG-18 (30 mg/kg) or vehicle (DMSO) for 8 h, followed by intraperitoneal injection with CFSE-labeled ACs for 1.5 h and PM elicitation (n = 4 mice per group). F, G. In vitro continuous efferocytosis experiments, where macrophages were stained for F4/80+ (red), AC were labeled with CFSE (AC-1; green) or Violet (AC2; violet). % continuous efferocytosis was determined by the ratio of F4/80+, CFSE+, and Violet+ (triple positive) cells to total F4/80+ and CFSE+ (double positive) cells. (F) Fmr1+/+ and Fmr1−/− BMDM were incubated with AC-1 for 2 h, and after 2 h interval, incubated with AC-2 for 2 more hours (n = 4–3 biological replicates). (G) BMDM were pre-treated either with vehicle (DMSO) or AMG-18 (5 µM) for 1 h, incubated with CFSE-labeled AC-1 for 2 h, followed by incubation with Violet-labeled AC-2 for 2 h and PM collection (n = 4 biological replicates).

Data information: For all images scale bar = 50 µm; Red: Macrophages, Green: AC/AC-1, Violet: AC-2. Data are mean ± SEM. Unpaired t-test with Welch’s correction.

Source data are available online for this figure.

We next asked how IRE1 kinase activity impacts efferocytosis by macrophages. We observed that AMG-18 increased efferocytosis of ACs in BMDMs (Fig 3D). Next, we injected wild type mice with AMG-18 (for 8 h) followed by injection of green fluorescent-labeled ACs. AMG-18 also induced efferocytosis of ACs by PM in vivo (Fig 3E).

Continued clearance of ACs by macrophages prevents the accumulation of necrotic cells and is an important process that promotes atherosclerosis regression (Yurdagul et al, 2017). To determine whether Fmr1−/− macrophages can efficiently internalize multiple ACs over consecutive rounds of engulfment, we incubated macrophages with green fluorescent-labeled AC for 2 h, followed by second incubation with violet fluorescent-labeled AC for 2 more hours. FMRP-deficient BMDMs displayed an increase in continued efferocytosis when compared to wild type BMDMs (Fig 3F). Likewise, treatment of wild type BMDM with AMG-18 enhanced continued efferocytosis of ACs (Fig 3G). Collectively, these results demonstrate that the ablation of IRE1 kinase activity and its proposed substrate, FMRP, enhances efferocytosis in vitro and in vivo.

Translational suppression of cholesterol transporters and efferocytosis regulators by FMRP during ER stress

Consistent with our observations that macrophage FMRP plays a role in suppressing cholesterol efflux and efferocytosis, published data describing the FMRP-RNA interactome also suggest that FMRP interacts with mRNA-encoding proteins involved in cholesterol trafficking (such as Abca1 and Abcg1) and efferocytosis receptors (such as c-Mer tyrosine kinase (Mertk) and LDL receptor-related protein 1 (Lrp1)), suggesting FMRP may impair their translation (Darnell et al, 2011; Ascano et al, 2012). To assess this notion in macrophages, we next performed polyribosome profiling in Fmr1+/+ and Fmr1−/− BMDMs under PA-induced ER stress conditions (Fig 4A and B, Appendix Fig S3A). Indeed, the mRNA abundance for the cholesterol transporters, Abca1 and Abcg1, and the mRNA abundance for efferocytosis regulators, Mertk, Lrp1, and Cd36, were increased in translating polysome fractions and decreased in non-translating (NTR) fractions in the Fmr1−/− BMDMs when compared to Fmr1+/+ BMDMs (Fig 4B), while the abundance of these mRNAs in the total cell lysate was unchanged (Appendix Fig S3B).

Figure 4. FMRP targets in macrophages

A, B. RNA lysates from Fmr1+/+ and Fmr1−/− BMDM that were treated with PA (500 µM; 6 h) were fractionated using a 10–50% sucrose gradient and separated to polysome, monosome/NTR fractions. The absorbance (260 nm) of RNA was measured and plotted as a function of time (n = 3 biological replicates). (A) Representative profile for RNA distribution from genotypes based on UV absorbance readings after sucrose gradient fractionation. (B) The ratio of the Abca1, Abcg1, Mertk, Lrp1, Cd36, Cd47, and Rac1 mRNA in polysome to NTR fraction (n = 3 biological replicates). C. BMDM were isolated from Fmr1+/+ and Fmr1−/−, and protein lysates were analyzed by Western blotting using specific antibodies for ABCA1, ABCG1, MerTK, LRP1, FMRP, and β-Actin antibodies and fold inductions relative to β-Actin are depicted above the blots (n = 6 biological replicates). D. Fmr1−/− MEF cells were transfected with EV, WT-FMRP, or STSA-FMRP plasmids followed by PA treatment (500 µM; 6 h). Protein lysates were analyzed by Western blotting using specific antibodies for ABCA1, MerTK, LRP1, pFMRP, FMRP, and β-Actin and fold inductions relative to β-Actin are depicted above the blots (n = 5 biological replicates).

Data information: A representative blot is shown. In C and D, data are cumulative results of 2 and 3 independent experiments, respectively. In Western blots, the protein expression fold change was calculated relative to β-Actin and depicted above the blots and a representative blot was shown. Data are mean ± SEM. Unpaired t-test with Welch’s correction.

Source data are available online for this figure.

We next analyzed the corresponding protein expression changes for the FMRP-regulated mRNA targets from the same experiment in Fig 4A and B. As expected, ABCA1, ABCG1, MerTK, and LRP1 protein expression levels were induced in Fmr1−/− BMDMs (Fig 4C and Appendix Fig S3C). To further assess the impact of IRE1-mediated FMRP phosphorylation on the translation of FMRP’s targets, we overexpressed the FMRP phosphorylation-deficient mutant (STSA) or WT-FMRP in Fmr1−/− MEFs and treated with PA to induce ER stress. ABCA1, LRP1, and MerTK expression levels were reduced in the Fmr1−/− MEF expressing WT-FMRP, but not in Fmr1−/− MEFs expressing STSA-FMRP (Fig 4D and Appendix Fig S3D). Taken together, our data demonstrate that IRE1-mediated FMRP phosphorylation tunes cholesterol transporters and efferocytosis regulators expression in macrophages.

FMRP knock-down and IRE1 kinase inhibition alleviates atherosclerosis

Our findings demonstrate that both IRE1 kinase inhibition and FMRP deficiency result in increased RCT, reduced foam cell formation, and enhanced efferocytosis in vivo, suggesting that FMRP deficiency in mice leads to protection from atherosclerosis. We tested this notion using Fmr1−/− and Fmr1+/+ mice in which hyperlipidemia was induced by combining AAV_PCSK9 injection with a WD as described above (Fig 5A). Although there was a very slight decrease in body weight ratio; the plasma glucose, total plasma cholesterol (TPC), lipoprotein levels, and the number of circulating, major type of immune cells were indistinguishable between the Fmr1−/− and Fmr1+/+ genotypes (Appendix Fig S4A–F). Yet, FMRP deficiency resulted in a significant reduction in atherosclerotic lesions in en face aorta preparations (Fig 5B). FMRP deficiency did not alter aortic root lesion area despite a significantly decreased foam cell area (as assessed by Oil Red O staining) (Fig 5C and D). We next asked what may be contributing to this phenotype. The necrotic core area in the lesions from Fmr1−/− mice was significantly less than in Fmr1+/+ lesions (Fig 5E), indicating improved AC clearance by Fmr1−/− macrophages in plaques. We also observed a significant reduction in lesion macrophages (as assessed by the anti-monocyte macrophage 2 (MOMA-2)-stained area) and the number of apoptotic cells (TUNEL-stained) per macrophage area (Appendix Fig S4G and H). We further investigated whether apoptosis is altered in Fmr1−/− and AMG-18-treated macrophages. There was no significant change between the groups (Appendix Fig S4I), supporting the notion that the primary consequence of inhibiting IRE1-FMRP signaling is efficient clearance of apoptotic cells through increasing efferocytosis capacity. Additionally, we detected no change in smooth muscle area (stained with smooth muscle actin (SMA)), but there was a significant increase in the collagen content of Fmr1−/− lesions when compared to Fmr1+/+ lesions, suggesting this could be the reason why aortic root lesion area is not significantly altered (Appendix Fig S4J and K).

Figure 5. FMRP-deficiency alleviates atherosclerosis

Atherosclerosis experiment design in Fmr1+/+ and Fmr1−/− mice that were injected with AAV_PCSK9 and fed WD (16 weeks). Lesion area calculated from en face aorta, stained with ORO (n = 12–13 mice per group; Scale bar = 5 mm). Total plaque area was calculated from hematoxylin & eosin (H&E)-stained aortic root sections (n = 8 mice per group; Scale bar = 300 µm). Foam cell area was calculated from ORO-stained aortic root sections (n = 8 mice per group; Scale bar = 300 µm). Necrotic area was calculated from H&E-stained aortic root sections (n = 8 mice per group; Scale bar = 100 µm). Atherosclerosis experiment design in myFmr1+/+ and myFmr1−/− mice that were injected with AAV_PCSK9 and fed WD (16 weeks). Lesion area calculated from en face aorta, stained with ORO (n = 9 mice per group; Scale bar = 5 mm). Total plaque area was calculated from H&E-stained aortic root sections (n = 9–6 mice per group; Scale bar = 300 µm). Foam cell area was calculated from ORO-stained aortic root sections (n = 9–6 mice per group; Scale bar = 300 µm). Necrotic area was calculated from H&E-stained aortic root sections (n = 9–6 mice per group; Scale bar = 100 µm).

Data information: Data are mean ± SEM; Mann Whitney U test.

Source data are available online for this figure.

To approach the role of macrophage FMRP in atherosclerosis, we generated a myeloid Fmr1-deficient mouse model (myFmr1−/−) and induced hyperlipidemia (Fig 5F). As expected from the systemic deletion data shown in Fig 5A, the body weight, plasma glucose and TPC were indistinguishable between the myFmr1−/− and myFmr1+/+ genotypes (Appendix Fig S4L–N). in general, myeloid-specific FMRP deficiency paralleled the results obtained for Fmr1−/− mice: it resulted in a significant reduction in atherosclerotic lesions in en face aorta preparations (Fig 5G). Myeloid-specific FMRP deficiency did not alter aortic root lesion area but significantly reduced foam cell area (Fig 5H and I). The necrotic core area in the lesions from myFmr1−/− mice was also significantly less than in myFmr1+/+ lesions (Fig 5J), indicating improved AC clearance by Fmr1−/− macrophages in plaques and supporting our observations that FMRP deficiency increases macrophage efferocytosis in vitro and in vivo (Fig 3). Thus, comparing the data obtained with the Fmr1−/− mice with those obtained from the myFmr1−/− mice indicates that the atheroprotective effects observed in Fmr1−/− mice are mostly, if not exclusively, due to FMRP’s role in myeloid cells such as macrophages.

To test the notion that IRE1 functions upstream of FMRP, we next investigated the impact of IRE1 kinase inhibition on atherosclerosis. To this end, we fed Apoe−/− mice with WD for 12 weeks and injected them with AMG-18 or vehicle once daily for the last 4 weeks of WD (Fig 6A). We observed no significant differences in body weight, plasma glucose or TPC, and the number of circulating immune cells between the groups (Appendix Fig S5A–D). Consistent with an earlier publication that determined the effective and non-toxic dose for AMG-18 in mice, the inhibitor engaged its molecular target effectively in the treatment group (as assessed by reduced IRE1 autophosphorylation) (Appendix Fig S5E). IRE1 kinase inhibition led to a decrease in atherosclerotic lesions in en face aorta preparations (Fig 6B). As shown above for the Fmr1−/− mice, the aortic root lesion area was not different between the groups (Fig 6C), but foam cell area as well as necrotic core area were significantly decreased (Fig 6D and E).