|Home | About | Journals | Submit | Contact Us | Français|
In advanced atherosclerosis, macrophage apoptosis coupled with defective phagocytic clearance of the apoptotic cells (efferocytosis) promotes plaque necrosis, which precipitates acute atherothrombotic cardiovascular events. Oxidative and endoplasmic reticulum (ER) stress in macrophages are important causes of advanced lesional macrophage apoptosis. We now show that pro-apoptotic oxidative/ER stress inducers trigger another stress reaction in macrophages, autophagy. Inhibition of autophagy by silencing ATG5 or other autophagy mediators enhances apoptosis and NADPH oxidase-mediated oxidative stress, while at the same time rendering the apoptotic cells less well recognized by efferocytes. Most importantly, macrophage ATG5 deficiency in fat-fed Ldlr−/− mice increases apoptosis and oxidative stress in advanced lesional macrophages, promotes plaque necrosis, and worsens lesional efferocytosis. These findings reveal a protective process in oxidatively stressed macrophages relevant to plaque necrosis, suggesting a mechanism-based strategy to therapeutically suppress atherosclerosis progression and its clinical sequelae.
In advanced atherosclerosis, death of lesional macrophages, coupled with defective phagocytic clearance of the dead cells (“efferocytosis”), promotes the formation of plaque necrosis (Tabas, 2010). Plaque necrosis is a key distinguishing feature of and likely causative process in the small percentage of atherosclerotic lesions that cause acute atherothrombotic vascular disease, such as myocardial infarction, sudden cardiac death, and stroke (Virmani et al., 2002). In vivo data suggest that endoplasmic reticulum (ER) and oxidative stress play important roles in advanced lesional macrophage death (Lusis, 2000; Moore and Tabas, 2011). When macrophages are exposed to atherosclerosis-relevant factors that trigger these stress reactions, such as oxysterols, oxidized phospholipids (oxPLs), or unesterified “free” cholesterol (FC), ER stress-induced apoptotic pathways are activated (Lusis, 2000; Moore and Tabas, 2011). Moreover, the NOX2 subunit of NADPH oxidase is induced, leading to assembly of active NADPH oxidase complex on lysosomes and pro-apoptotic oxidative stress (Li et al., 2010; Seimon et al., 2010). In response to cell death, efferocytosis is normally rapid and efficient and thereby prevents post-apoptotic necrosis and inflammation (Henson et al., 2001), but for reasons that are not yet known, efferocytosis is defective in advanced atherosclerosis (Schrijvers et al., 2005; Tabas, 2010).
In an attempt to increase our understanding in these areas, we decided to explore the process of autophagy, which can affect both apoptosis and efferocytosis in other settings (Eisenberg-Lerner et al., 2009; Qu et al., 2007). In this regard, we set out to explore how autophagy may effect these processes in advanced atherosclerosis by genetically preventing the autophagic response in macrophages exposed to oxidative/ER stressors in vitro and in advanced atherosclerotic lesions in vivo. We show that blocking autophagy renders macrophages more susceptible to cell death, worsens the recognition and clearance of the dead cells by efferocytes, and promotes plaque necrosis in a mouse model of advanced atherosclerosis.
Primary murine macrophages from mice transgenic for a GFP-tagged version of the autophagy effector LC3-II were treated for 6 h with 7-ketocholesterol (7KC), an atherosclerotic lesional oxysterol that promotes oxidative stress, ER stress, and apoptosis in macrophages (Lizard et al., 1998; Myoishi et al., 2007; Li et al., 2010). The 7KC-treated macrophages showed a punctate pattern of fluroscence that is indicative of autophagy induction (Mizushima et al., 2010) (Figure 1A). At the ultrastructural level (Figure 1B), 7KC-treated macrophages had numerous double-membrane structures that were characteristic of autophagosomes. We also explored macrophages incubated with KOdiA-PC, an oxPL typical of those in oxidized LDL, in combination with low-dose thapsigargin, which induces ER stress by inhibiting calcium re-entry into the ER (Seimon et al., 2010). These macrophages also showed the presence of double membrane structures. In contrast, there were very few double membranes in untreated macrophages or in 7KC-treated macrophages from Atg5fl/flLysmcre+/− mice, which lack the key autophagy protein ATG5 (Hara et al., 2006). To further document autophagy, we assayed the expression of the LC3-II, which is converted from LC3-I during the early stages of autophagy (Mizushima et al., 2010) (Figure 1C, lanes 1–3, 7–9). When macrophages were incubated with 7KC for 4 h, the expression level of LC3-II was higher than baseline, whereas only the precursor LC3-I was detectable in ATG5-deficient macrophages.
The final stage of autophagy, autophagosome-lysosome fusion, leads to lysosomal degradation of autophagosomal content and proteins, including LC3-II itself. This process of “flux” through lysosomes can be documented by showing a higher level of LC3-II in the presence vs. the absence of the lysosomal inhibitor bafilomycin A1 (Mizushima et al., 2010). We found that 7KC treatment of WT macrophages, but not ATG5-deficient macrophages, increased the difference between the minus- and plus-bafilomycin LC3-II signal, indicative of autophagic flux through lysosomes (Figure 1C, lanes 4–6, 10–12). Similar data were obtained in macrophages incubated with KOdiA-PC/thapsigargin and in macrophages loaded with FC (data not shown). These combined data provide strong evidence that 7KC and other inducers of macrophage apoptosis relevant to atherosclerosis induce autophagy and promote autophagic flux through lysosomes.
To test the role of autophagy in apoptosis, we treated wild-type (WT) and Atg5fl/flLysmcre+/− mice with a number of different inducers and found markedly higher cell death in the autophagy-defective macrophages (Figure 2A–B and S1A). Note that Figure 2A includes macrophages treated with thapsigargin plus apolipoprotein(a), an atherogenic protein in humans that induces apoptosis in ER-stressed macrophages through its bound oxPLs (Clarke et al., 2009; Seimon et al., 2010). Apoptosis was also enhanced in a more acute model of autophagy inhibition in which ATG5 or the autophagy effector Beclin1 was silenced using siRNA (Figure 2C).
As shown in Fig. 2B, thapsigargin alone at the dose and time used here showed only a very slight, non-significant trend toward increased apoptosis over baseline, and apoptosis was not significantly increased in ATG5-deficient macrophages. However, tunicamycin, which induces ER stress by inhibiting protein glycosylation, did induce apoptosis, and this was further enhanced by ATG5 deficiency (Figure 2D). In this context, we explored a mouse model in which i.v. tunicamycin induces apoptosis of peritoneal macrophages in vivo (Li et al., 2009) and found that apoptosis was ~2-fold higher in Atg5fl/flLysmcre+/− vs. control Lysmcre+/− mice (Figure 2E).
To test the importance of autophagic flux through lysosomes in the anti-apoptotic effect of autophagy, WT macrophages were treated with 7KC either in the absence or presence of bafilomycin A1 or the combination of pepstatin and E64D, which is another means to inhibit autophagolysosomal degradation. Apoptosis was substantially increased in the presence of the lysosomal inhibitors (Figure 2F), suggesting that the protective effect of autophagy requires the full process of autophagic flux through lysosomes.
Many macrophages in atherosclerotic lesions are “foam cells” filled with cholesteryl ester (CE). Although CE loading per se does not cause apoptosis (Feng et al., 2003a), we asked whether it would affect the increase in KOdiA-PC/thapsigargin-induced apoptosis caused by autophagy inhibition (Figure S1B). The data show that in foam cells, as in non-foam cells, ATG5 deficiency enhanced KOdiA-PC/thapsigargin-induced apoptosis. Interestingly, ATG5 deficiency also significantly increased apoptosis in non-treated foam cells, suggesting that CE loading, while not inducing apoptosis by itself, renders the cells more sensitive to the pro-apoptotic effect of autophagy inhibition.
The fact that macrophages exposed to the pro-apoptotic inducers used here eventually die suggests that either the prolonged effect of pro-apoptotic processes overwhelms the protective effect of autophagy or that death occurs because the autophagic response eventually decreases. Indeed, we found that LC3-II flux through lysosomes was lower in macrophages exposed to 7C for 16–18 h vs. 6 h (Figure S1C), and this was not associated with a general defect in lysosomal function (Figure S1D). However, when the late decrease in flux was prevented by treatment with rapamycin at the 12-h timepoint or by adenoviral-mediated transduction with ATG7 (Pattison et al., 2011) (Figure S1E), 7KC-induced apoptosis was not decreased (Figure S1F). These data suggest that the late decrease in autophagolysosomal flux is an effect rather than a trigger of apoptosis and that the protective effect of autophagy is eventually overwhelmed by ongoing pro-apoptotic processes.
NADPH oxidase-mediated oxidative stress is a major mechanism of apoptosis in macrophages exposed to the inducers used in this study (Li et al., 2010; Seimon et al., 2010). To evaluate its role in autophagy-inhibited macrophages, we first compared WT and ATG5-deficient macrophages for DCF staining, which fluoresceces in the presence of peroxide in a manner parallel to NADPH oxidase activation in ER-stressed macrophages (Li et al., 2010). In all models tested, the percentage of DCF-positive cells was substantially higher in the autophagy-defective group (Figure 2G). Similar data were obtained using a FACS assay for cells stained with CellRox™, which is non-fluorescent in the reduced state but exhibits excitation/emission maxima at 640/665 nm upon oxidation (Figure S2A).
We next used a genetic approach to test the role of NADPH oxidase in the enhancement of apoptosis by autophagy inhibition. The experiment is based on the idea that there are two “components” of apoptosis in autophagy-inhibited macrophages: the “basal” level that occurs in autophagy-competent cells, which is suppressed by NOX2 deficiency (Li et al., 2010), and the additional increment above this level that occurs in autophagy-inhibited cells. If this incremental component is also NOX2-dependent, NOX2 deficiency should not only decrease the high level of apoptosis seen in autophagy-inhibited cells to a very low level, but it should also eliminate the difference in apoptosis between autophagy-competent and autophagy-inhibited cells. We therefore compared KOdiA-PC/thapsigargin-induced apoptosis in Nox2+/+ vs. Nox2−/− macrophages treated with scrRNA or Atg5 siRNA to inhibit autophagy (Figure 2H). The data show that NOX2 deficiency markedly inhibited apoptosis in ATG5-silenced macrophages, and the low level of residual apoptosis was similar to that in autophagy-competent (scrRNA) Nox2−/− macrophages. Similar results were obtained using 7KC as the apoptosis inducer (data not shown). In theory, NOX2 deficiency could protect macrophages by increasing autophagy, but this was not the case (Figure S2B). These data are consistent with the hypothesis that autophagy inhibition enhances apoptosis, at least in part, by enhancing NADPH oxidase-induced oxidative stress.
To further explore this hypothesis, we considered two recent findings: (a) NADPH oxidase-induced oxidative stress and apoptosis in macrophages are associated with the assembly of NADPH oxidase complex on lysosomes (Seimon et al., 2010); and (b) autophagy in macrophages can participate in the clearance of phagocytosed bacteria in NADPH oxidase-containing phagolysosomes (Huang et al., 2009), and so NADPH oxidase complex itself may be disrupted in the process as part of homeostatic mechanism to protect the cells against excessive oxidative stress. If so, NADPH oxidase-activated macrophages, such as those occurring in the models explored here, would depend on autophagy to limit the oxidative damage. To test this idea, we used the fluorescence microscopy assay established in the aforementioned study to quantify the association of the NADPH oxidase subunit p47phox with lysosomes (Seimon et al., 2010), with the prediction that autophagy inhibition would disallow the disruption of NADPH oxidase complex and thus increase p47phox-lysosome association. Consistent with our previous data (Seimon et al., 2010), incubation of WT macrophages with KOdiA-PC/thapsigargin, but not either factor alone, increased p47phox-lysosome association (Figure S2C, black bars). As predicted by the hypothesis, p47phox-association was increased by ~55% in ATG5-deficient macrophages (Figure S2C, grey bars), which was not simply due to an increase in Nox2 induction (Figure S2D). Importantly, inhibition of autophagolysosomal flux with bafilomycin in WT macrophages was able to mimic the effect of ATG5 deficiency, and p47phox-lysosome association was not further increased in this setting by ATG5 deficiency. Similar results were obtained with 7KC (data not shown). In summary, the combined data in this section and in previous studies (Li et al., 2010; Seimon et al., 2010; Huang et al., 2009) support the idea that increased apoptosis in autophagy-inhibited cells involves increased NADPH oxidase-induced oxidative stress, possibly through a process that increases lysosome-associated p47phox.
In preparation for the atherosclerosis project using Atg5fl/flLysmcre+/−Ldlr−/− mice, we first showed that the aortic root lesions of Western diet (WD)-fed GFP-LC3-Ldlr−/− mice contain cells in macrophage-rich regions that express ATG5 and have macrophages that display a punctate pattern of GFP-LC3 fluorescence typical of autophagy (Figure 3A). We then looked for another autophagy protein, p62/SQSTM1, which is known to accumulate when autophagy flux through lysosomes is defective, i.e., through decreased p62 lysosomal degradation (Bjorkoy et al., 2009) (Figure 3B). p62 immunofluorescence was present in ~10–15% of intimal cells in macrophage-rich areas of lesions of 8-wk WD-fed Ldlr−/− mice, and this value increased as lesions progressed in 12-wk and 16-wk WD-fed mice. Interestingly, p62 staining was observed in a few of the acellular necrotic areas in 16-wk lesions, which likely reflects the residual debris from macrophages that had accumulated p62 before they died (Ball et al., 1995; Feng et al., 2003b). Note that the mRNA for Sqstm1 did not increase between 8 and 16 weeks of WD, indicating that the increase in p62 was not due to transcriptional induction (Figure 3B, right graph). These data further substantiate the presence of autophagy in atherosclerotic lesions and are consistent with the concept that autophagic flux through lysosomes decreases as plaques progress.
To test the functional role of macrophage autophagy in advanced atherosclerosis, we studied 12-wk and 16-wk WD-fed Lysmcre+/−Ldlr−/− (control) and Atg5fl/flLysmcre+/−Ldlr−/− mice. Despite no difference in body weight, plasma lipids and lipoproteins, and plasma insulin between the two groups of mice (Figure S3A–B), lesion and necrotic area in the aortic root were higher in Atg5fl/flLysmcre+/−Ldlr−/− vs. control mice at both time points (Figure 3C–D and S3C). The increase in lesion size appeared to be due to the increase in necrotic areas, because the numbers of lesional macrophages and smooth muscle cells were not increased in the experimental cohort (data not shown). This finding is consistent with the mechanistic hypothesis described herein as opposed to an alternative possibility in which a different process, such as macrophage influx or SMC proliferation, might be increased by autophagy inhibition.
Consistent with the mechanistic data in Figures 2–3, quantitative analysis revealed that macrophage-rich regions of the Atg5fl/flLysmcre+/−Ldlr−/− plaques had more cells positive for TUNEL, activated caspase-3, DHE, and p47 than similar regions in control Lysmcre+/−Ldlr−/− plaques (Figure 4A–B and S4A–B). The increase in apoptosis in the experimental group was not associated with an increase in lesion Chop expression (Figure S4C), which is consistent with the macrophage data in Fig. 1C.
The increase in plaque necrosis in the ATG5-deficient mice raised the possibility that the macrophages in these mice had a defect in efferocytosis (Henson et al., 2001; Tabas, 2010). We therefore conducted an in-situ efferocytosis assay, which quantifies the ratio of TUNEL-labeled apoptotic cells in lesions that are associated with F4/80-labeled macrophages vs. unassociated (“free”) (Thorp et al., 2008). The data show a marked decrease in the associated:free ratio in the lesions of the mutant mice, indicative of defective efferocytosis (Figure 4C). To examine this point in a more controlled in vitro setting, we first compared the ability of ATF5-deficient vs. control macrophages to engulf apoptotic cells and found equal competency (data not shown). We therefore tested whether apoptotic cells that die with defective autophagy are poorly engulfed by efferocytes. To this end, WT or ATG5-deficient macrophages, or WT macrophages treated with bafilomycin A1, were rendered apoptotic by incubation with 7KC and then labeled with Alexa-595-annexin V (red). Equal numbers of labeled apoptotic cells were added to Calcein-AM (green)-labeled WT macrophages as a source of efferocytes. The data show that there was less efferocytosis of apoptotic cells that were either ATG5-deficient or had inhibited autophagolysosomal flux (bafilomycin) (Figure 4D). The defect appeared to be in apoptotic cell-efferocyte binding, because the inhibitory effect of apoptotic cell ATG5 deficiency on efferocytosis was recapitulated in a 4°C binding assay (Figure 4E). We also found that when the late decrease in autophagic flux in WT cells undergoing apoptosis was bolstered by rapamycin or adeno-Atg7 (see Figs. S1C), their efferocytosis was increased (Figure S4D). Thus, autophagic flux through lysosomes in macrophages undergoing apoptosis enhances their recognition by phagocytes, and this process can be enhanced further by bolstering autophagy.
The data in this study focus on functional relationships between autophagy and two critical cell biological processes in atherosclerotic plaque necrosis, M apoptosis and defective efferocytosis. We found that when autophagy was inhibited in vitro and in vivo, apoptosis was increased and recognition of the autophagy-inhibited dead cells by efferocytic phagocytes was decreased. We infer from these data that autophagy can induce two independent protective actions, which in normal physiology may function to delay cell death in case repair is possible but also ensure that the cells are efficiently cleared by neighboring phagocytes if repair is not possible.
The findings herein raise a number of interesting mechanistic questions for future study. First, the type of autophagy that is involved in protection from apoptosis and oxidative stress and in mediating efficient efferocytosis needs to be defined. The finding of double membranes in a our cell culture model indicates that autophagy is occurring, but it is theoretically possible that an additional single-membrane process involving LC3-II, ATG5, and beclin1, such as LC3-associated phagocytosis (LAP) (Martinez et al., 2011; Florey et al., 2011), is also operational. Moreover, whether the effects on apoptosis, oxidative stress, and efferocytosis involve classic macroautophagy or, for example, chaperone-mediated autophagy or mitophagy, remains to be established. While our data implicate the role of NADPH oxidase, they do not role out other sources of oxidative stress, such as mitochondria, which could be affected by mitophagy (Scherz-Shouval and Elazar, 2011). Second, how the factors studied here, and those that actually function in atherosclerotic lesions, induce autophagy is not known. While ER stress has been reported to induce autophagy (Kroemer et al., 2010), we found that silencing key molecules in the IRE1 and PERK branches of the UPR did not block autophagy initiation or flux through lysosomes (data not shown). Third, the mechanism involved in defective efferocytosis of autophagy-inhibited apoptotic macrophages requires further study. The data here indicate that autophagosomal flux through lysosomes in cells undergoing apoptosis somehow increases their recognition by phagocytes, perhaps by increasing the expression of a cell-surface recognition molecule. However, we found no evidence of a defect in PS externalization or presentation of three other efferocytosis ligands, calreticulin, ERp57, and pentraxin 3, in autophagy-inhibited macrophages (data not shown).
Future studies will likely reveal additional roles of autophagy in atherosclerosis that are cell type- and lesion stage-specific. For example, previous work has shown that intimal smooth muscle cells can acquire morphological features suggestive of autophagy (De Meyer and Martinet, 2009), although the functional significance of this finding is unknown. Links between autophagy and inflammation have been implicated in other disease models (Levine et al., 2011), which could be important in atherosclerosis if such processes occurred in lesional cells. A recent study revealed that autophagy can play a role in the hydrolysis of stored cholesteryl ester droplets in macrophages, thus facilitating cholesterol efflux (Ouimet et al., 2011), but the relevance of this finding to atherosclerosis remains to be determined. Knowledge gained from autophagy studies in atherosclerosis may be useful for devising mechanism-based therapeutic strategies. Although increasing autophagy above the induced level with rapamycin or ATG7 transduction did not protect macrophage in vitro, autophagy appears to be defective in advanced atherosclerosis, and so restoring autophagy to the normally induced level may be protective. Moreover, boosting autophagy did improve efferocytosis in vitro, which may translate into a beneficial effect in advanced atherosclerosis (Tabas, 2010). Such strategies would be unique, because autophagy may be the macrophage’s natural way to combat two complementary processes in plaque necrosis, macrophage apoptosis and defective efferocytosis.
C57BL/6J, Ldlr −/−, Nox2 −/−, and ob/ob mice were purchases from Jackson Laboratory (Bar Harbor, Maine). The GFP-LC3 transgenic and Atg5fl/fl mice (Mizushima and Kuma, 2008; Hara et al., 2006) were created by Dr. Noboru Mizushima, Tokyo Medical and Dental University, and sent to us from the laboratory of Dr. Beth Levine (University of Texas, Southwestern). The Lysmcre+/− mice (Clausen et al., 1999) were from Dr. Irmgard Föster (Technical University of Munich). The Ldlr−/−, Atg5fl/fl, and Lysmcre mice were bred with each other to generate the mice used for this study. For the atherosclerosis study, 8-wk-old female mice were placed on a Western-type diet (TD88137; Harlan Teklad) for the indicated number of weeks.
Apoptosis was assayed in cultured macrophages by staining with Alexa 488-conjugated Annexin V or using the TUNEL (TdT-mediated dUTP nick-end labeling) protocol, as described previously (Seimon et al., 2010). Apoptotic cells in atherosclerotic lesions were labeled by TUNEL using the in situ cell death detection kit TMR-red (Roche Diagnostics) according to the manufacturer’s protocol. Only TUNEL-positive cells that co-localized with DAPI-stained nuclei were counted as being positive. Lesional apoptosis was also assayed by immunofluorescence using anti-activated caspase-3 primary antibody. TUNEL, annexin V, and activated caspase-3 staining were viewed using an Olympus IX-70 inverted fluorescent microscope. Representative fields (4–6 fields containing ~2000 cells) were photographed using an Olympus DP71 CCD camera. The number of Annexin V-positive cells were counted and expressed as a percent of the total number of cells in at least 4 separate fields from duplicate wells. For TUNEL analysis of lesions, DAPI and TUNEL images were merged using Photoshop analysis software (Adobe Systems). The number of cells positive for TUNEL or activated caspase-3 was quantified as a percentage of total lesional cells in macrophage-rich regions.
For morphometric lesion analysis, sections were stained with Harris’ hematoxylin and eosin. The total lesion area and necrotic area were quantified as previously described (Seimon et al., 2009). For plaque necrosis, boundary lines were drawn around regions that were free of H&E staining, and area measurements were obtained using image analysis software. A 3,000-μm2 threshold was implemented in order to avoid counting regions that likely do not represent substantial areas of necrosis. Using this method, a 97% agreement in the percent necrotic area was calculated between our 2 independent observers. The methods for DHE and p47phox staining appear in Supplemental Experimental Procedures.
Values are given as means ± S.E.M. (n = 4 unless otherwise noted in the figure legends); absent error bars in the bar graphs signify S.E.M. values smaller than the graphic symbols. Comparison of mean values between groups was evaluated using a two-tailed student t-test, Mann-Whitney U test, or ANOVA. P values less than 0.05 was considered significant. Refer to the legend of Figure S3C for additional statistical methods specific for the analysis of the data in that panel.
This work was supported by NIH grants HL106019, HL075662, and HL054591 (I.T.) and by fellowships of the Netherlands Organization of Scientific research (Rubicon) and the international atherosclerosis society (J.C.S.). The authors acknowledge Dr. Noboru Mizushima and his institution, Tokyo Medical and Dental University, for allowing the use of Atg5fl/fl mice and GFP-LC3 transgenic mice for these studies. The authors thank Dr. Edward Thorp and George Kuriakose, Columbia University, for helpful advice on the efferocytosis experiments and for technical support, respectively; Dr. Beth Levine, University of Texas, Southwestern, for her advice and support during the course of this project; Dr. Douglas R. Green, St. Jude Children’s Research Hospital, for helpful discussions; Dr. Marlys Koschinsky, University of Windsor, Ontario, for providing the recombinant apo(a); and Drs. Rajasekhar Ramakrishnan and Steve Holleran for help with statistical analysis.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.