Introduction A common feature of many disease processes is pathological cell death leading to tissue dysfunction. Two processes that are becoming increasingly recognized as inducers of pathological cell death are ER and oxidative stress (Chandra et al., 2000; Kim et al., 2008). Many of the diseases that feature ER and oxidative stress are associated with aging and/or obesity, including diabetes, atherosclerosis, renal disease, and neurodegenerative disease, and are becoming epidemic among humans in modern day society (Martínez, 2006; Kregel and Zhang, 2007; Kim et al., 2008). Nonetheless, progress in translating this understanding into useful therapeutic strategies has been disappointing. As one example, there is good evidence that the clinical progression of advanced atherosclerosis involves both ER and oxidative stress (Glass and Witztum, 2001; Tabas, 2010). However, ER stress involves many signaling pathways with differing pathological and physiological functions (Malhotra and Kaufman, 2007b; Ron and Walter, 2007), which makes targeting this process for atherosclerosis complex and challenging. Likewise, there are many sources of cellular oxidative stress in atherosclerotic lesional cells, and human trials using vitamin E as an antioxidant have failed to suppress the incidence of cardiovascular disease (Kris-Etherton et al., 2004; Williams and Fisher, 2005). Similar issues exist with other diseases driven by ER stress–induced apoptosis and oxidative stress. An important approach to this problem is to gain a more in-depth understanding about molecular and cellular mechanisms and links between ER stress and oxidative stress, particularly in the process of cell death and in settings relevant to disease processes (Malhotra and Kaufman, 2007a). In this study, we investigate these mechanisms using cultured macrophages as an in vitro model and the kidney in ER-stressed mice as an in vivo model. The macrophage model is relevant to advanced atherosclerosis (Tabas, 2010), and the kidney model is relevant to several renal diseases (Kitamura, 2008). With regard to macrophages, ER stress, often in combination with other proapoptotic hits, leads to apoptosis in part through the ER stress effector CCAAT/enhancer binding protein homologous protein (CHOP; Feng et al., 2003a; Thorp et al., 2009; Tabas, 2010). An important mechanism linking CHOP with apoptosis involves induction of the ER oxidase ERO1α, which causes activation of the ER calcium channel inositol 1,4,5-triphosphate (IP3) receptor (IP3R) and release of ER calcium (Li et al., 2009). The released calcium activates the enzyme calcium/calmodulin-dependent protein kinase II (CaMKII), which triggers both death receptor and mitochondrial pathways of apoptosis (Timmins et al., 2009). Oxidative stress can also be a trigger for apoptosis in macrophages, although the detailed molecular mechanisms are not well understood (Forman and Torres, 2002). There is evidence that ER stress–induced macrophage apoptosis occurs in advanced atherosclerosis and contributes to plaque necrosis, a critical process in atherothrombotic vascular disease (Feng et al., 2003b; Thorp et al., 2009; Tabas, 2010). Oxidative stress has also been implicated in atherosclerosis, but its importance in advanced lesional macrophage death and plaque necrosis is not known (Glass and Witztum, 2001; Bonomini et al., 2008). With regard to the renal model, several different inducers of renal dysfunction result in ER stress–induced apoptosis of renal tubular epithelial cells, podocytes, and other cells in the kidney (Kitamura, 2008). Renal disease can also be associated with oxidative stress (Locatelli et al., 2003), but, as mentioned above, its possible links to ER stress are not known. Using these two models, we show here that the aforementioned branch of ER stress signaling involving calcium and CaMKII induces NADPH oxidase (NOX) and oxidative stress, which are necessary for ER stress–induced apoptosis. We also provide evidence that CHOP is amplified by NOX/oxidative stress–mediated activation of a double-stranded RNA–dependent protein kinase (PKR)–CHOP pathway and that this amplification pathway is important for the final apoptotic response. Results ER stress stimulates oxidative stress in macrophages, which is dependent on CHOP, proapoptotic calcium signaling, and NOX Using two atherosclerosis-relevant inducers of ER stress, cholesterol loading and 7-ketocholesterol (Tabas, 2002; Myoishi et al., 2007), we determined whether oxidative stress was part of the previously elucidated CHOP–calcium–CaMKII pathway of apoptosis (Li et al., 2009; Timmins et al., 2009). Dihydrodichlorofluorescein (DCF) staining was used as a measure of intracellular peroxide accumulation (Wolber et al., 1987). We found that both inducers caused a substantial increase in the percentage of DCF-positive cells in wild-type (WT) macrophages, and the increase was much less in Chop−/− macrophages (Fig. 1 A). We next examined the calcium arm of the pathway using the intracellular calcium chelator BAPTA-AM. The increase in DCF-positive cells was decreased substantially by the chelator, which paralleled its inhibitory effect on annexin V staining as a measure of apoptosis (Fig. 1 B). DCF positivity and annexin V staining were also blocked by the CaMKII inhibitor KN93 but not by the structurally related inactive homologue KN92 (Fig. 1 C). We also used RNAi to silence CaMKII-γ, which is the isoform expressed in macrophages (Timmins et al., 2009). Two separate Camk2g siRNAs, which decrease Camk2g by ∼60 and 75% (Timmins et al., 2009), gave similar results as KN93 (Fig. 1 D). Thus, oxidative stress, like apoptosis, is stimulated by ER stress in macrophages and is dependent on CHOP, intracellular calcium, and CaMKII. Figure 1. ER-stressed macrophages undergo oxidative stress, which is dependent on CHOP, calcium, and CaMKII. (A) Peritoneal macrophages from Chop+/+ (WT) and Chop−/− mice were incubated for 15 h without sterol (Con), under cholesterol-loading conditions (CHOL), or with 7-ketocholesterol (7KC). Intracellular peroxide accumulation was then assayed by DCF fluorescence. Three fields for each sample were quantified and expressed as a percentage of DCF-positive cells. (B) Macrophages were pretreated for 1 h with 5 µM BAPTA-AM or with equivalent volumes of vehicle (Veh). The cells were then incubated for 15 h under control or cholesterol-loading conditions and also with BAPTA-AM or vehicle control as indicated. Intracellular peroxide accumulation and apoptosis were assayed by DCF fluorescence (green) and Alexa Fluor 594–conjugated annexin V (red), respectively. Bar, 20 µm. (C) Macrophages were pretreated for 1 h in the absence or presence of 5 µM of the CaMKII inhibitor KN93 or the inactive analogue KN92, followed by incubation for 14 h without sterol under cholesterol-loading conditions or with 7-ketocholesterol. DCF fluorescence and annexin V staining were then assayed. (D) Macrophages were transfected with two different Camk2g siRNA constructs. After 72 h, the cells were incubated for 30 h under control or cholesterol-loading conditions and then assayed for DCF fluorescence and annexin V staining. Bars with the same symbols are not significantly different from each other, whereas bars with different symbols are significantly different from each other. n = 3 for each experimental group. scrRNA, scrambled RNA. Data are presented as means ± SEM. One well-documented source of oxidative stress in macrophages is NOX (Iles and Forman, 2002). To test the relevance of NOX in ER-stressed macrophages, we first determined whether expression of the major NOX isoform in macrophages, NOX2, was affected by ER stress. We found that both cholesterol loading and 7-ketocholesterol induced Nox2 mRNA (Fig. 2 A, black bars). To test the causative role of NOX2 in oxidative stress and annexin V staining, we used siRNA against Nox2, which was effective in decreasing the mRNA to non–ER stress levels in the cholesterol- and 7-ketocholesterol–treated cells (Fig. 2 A, gray bars). This level of Nox2 suppression significantly decreased both DCF staining and annexin V staining in ER-stressed macrophages (Fig. 2 B). As an independent approach, we compared macrophages isolated from WT versus Nox2−/− mice and found similar effects on oxidative stress and apoptosis, using either annexin staining or TUNEL for the apoptosis assay (Fig. 2, C and D). These data indicate that NOX plays an important role in the oxidative stress and apoptotic responses of macrophages subjected to ER stress. Figure 2. Oxidative stress and apoptosis in ER-stressed macrophages are dependent on NOX2. (A and B) Macrophages were transfected with Nox2 siRNA and, after 72 h, were incubated without sterol (Con), under cholesterol-loading conditions (CHOL), or with 7-ketocholesterol (7KC). The cells were assayed for Nox2 mRNA by RT-QPCR after 8 h (A) and DCF fluorescence and annexin V staining after 30 h (B). (C) Macrophages from WT or Nox2−/− mice were incubated for 14 h without sterol, under cholesterol-loading conditions, or with 7-ketocholesterol and then assayed for DCF and annexin V staining. (D) Macrophages from WT or Nox2−/− mice were incubated under cholesterol-loading conditions for 24 h or with 7-ketocholesterol for 20 h and then assayed for apoptosis by TUNEL staining. Bars with the same symbols are not significantly different from each other, whereas bars with different symbols are significantly different from each other. scrRNA, scrambled RNA. n = 3 for each experimental group. Data are presented as means ± SEM. The role of both CaMKII and NOX in ER stress–induced oxidative stress and apoptosis raised the question as to whether the two enzymes were in the same or different pathways. Consistent with their being in the same pathway, inhibition of CaMKII activity with KN93, or genetic targeting of the kinase through the use of Camk2g−/− macrophages, blocked Nox2 mRNA induction by cholesterol loading or 7-ketocholesterol (Fig. 3 A). We showed previously that CaMKII activation requires an ERO1α–IP3R1–calcium release pathway (Li et al., 2009), which predicts that IP3R1 and ERO1α would play a role in Nox2 induction. For this purpose, we transfected macrophages with an Ero1a siRNA that decreases ERO1α expression by ∼50% or an Ip3r1 siRNA that decreases IP3R1 expression by ∼60% compared with scrambled RNA (Li et al., 2009) and found that each was able to suppress Nox2 mRNA induction (Fig. 3, B and C). Figure 3. Nox2 induction by ER stress is dependent on CaMKII, ERO1α, IP3R1, and JNK. (A, left) Macrophages were pretreated for 1 h in the absence or presence of 5 µM of the CaMKII inhibitor KN93, the inactive analogue KN92, or vehicle control (Veh), followed by incubation for 8 h without sterol (Con) or under cholesterol-loading conditions (CHOL). Nox2 mRNA was then measured by RT-QPCR. (A, right) Peritoneal macrophages from WT or Camk2g−/− mice were incubated for 8 h without sterol, under cholesterol-loading conditions, or with 7-ketocholesterol (7KC) and then assayed for Nox2 mRNA. (B) Macrophages were transfected with scrambled RNA (scrRNA) or Ero1a siRNA, which, after 72 h, led to an ∼50% decrease in ERO1α expression as assessed by immunoblotting (Li et al., 2009). The cells were then incubated an additional 8 h without sterol or with 7-ketocholesterol and then assayed for Nox2 mRNA. (C) Macrophages were transfected with scrambled RNA or Ip3r1 siRNA. After 72 h, IP3R1 expression was decreased by 60% as assessed by immunoblotting (Li et al., 2009). The cells were then incubated without sterol or under cholesterol-loading conditions for an additional 8 or 30 h and then assayed for Nox2 mRNA or DCF fluorescence, respectively. (D, left) Macrophages were pretreated for 1 h with 10 µM of the JNK inhibitor SP600125 or vehicle control and then incubated for 8 h without sterol or under cholesterol-loading conditions, also in the absence or presence of SP600125. (D, right) WT or Jnk2 −/− macrophages were incubated under control conditions or for 4 or 8 h under cholesterol-loading conditions. Nox2 mRNA was then assayed. (E) Macrophages were pretreated for 1 h with the JNK inhibitor SP600125 (SP) or vehicle control and then incubated for 15 h without sterol, under cholesterol-loading conditions, or with 7-ketocholesterol, also in the absence or presence of SP600125. DCF fluorescence and annexin V staining were then assayed. Bars with the same symbols are not significantly different from each other, whereas bars with different symbols are significantly different from each other. n = 3 for each experimental group. Data are presented as means ± SEM. Our previous work showed that a key downstream effector of CaMKII-induced apoptosis is the mitogen-activated protein kinase JNK (Timmins et al., 2009). We found here that the JNK inhibitor SP600125 blocked Nox2 mRNA induction, DCF positivity, and annexin V staining in ER-stressed macrophages (Fig. 3, D and E). The Nox2 mRNA data were confirmed using macrophages from Jnk2−/− mice (Fig. 3 D, right graph). In summary, the data in Figs. 1–3 suggest that NOX plays a critical yet previously unknown role in the ER stress–CHOP–calcium–CaMKII–JNK pathway of apoptosis. Evidence for CHOP amplification through NOX/oxidative stress–mediated activation of PKR The simplest way to integrate our previous findings (Li et al., 2009; Timmins et al., 2009) with the new data described thus far would be a linear pathway in which CHOP and CaMKII are upstream of NOX as follows: ER stress→CHOP→IP3R activation→ER calcium release→CaMKII activation→NOX activation→oxidative stress. However, the possibility of a bidirectional pathway was suggested by surprising data showing that absence of NOX2 actually suppressed both CaMKII activation/phosphorylation and CHOP induction in cholesterol-loaded or 7-ketocholesterol–treated macrophages (Fig. 4 A). Note that the transcriptional inducer of CHOP in the phospho-eIF2α (P-eIFα) pathway, ATF4, was also suppressed in Nox2−/− macrophages, whereas neither phospho–PKR-like ER kinase (PERK) nor Xbp1 splicing was suppressed in Nox2−/− macrophages (Fig. 4, A and B). These data are consistent with the conclusion that NOX2 deficiency was specifically suppressing a non-PERK–activated P-eIF2α pathway rather than global ER stress. We therefore looked at another eIF2α kinase, PKR, and found that phosphorylation of this kinase, which is a measure of its activation, was substantially suppressed in Nox2−/− macrophages (Fig. 4 A). Figure 4. Evidence for CHOP amplification through NOX/oxidative stress–mediated activation of PKR. (A) Macrophages from WT or Nox2−/− mice were incubated under cholesterol-loading conditions (CHOL) or with 7-ketocholesterol (7KC) for the indicated times. Lysates were then immunoblotted for phospho-CaMKII (p-CaMKII), total CaMKII (T-CaMKII), CHOP, phospho-PERK (p-PERK), phospho-PKR (p-PKR), and GAPDH. (B) Macrophages from WT or Nox2−/− mice were incubated under cholesterol-loading conditions for 8 h (ATF4 experiment) or 9 h (XBP1 experiment). Nuclei were isolated and immunoblotted for ATF4 and nucleophosmin (Np) loading control, or RNA was extracted and assayed for spliced and unspliced Xbp1 and Gapdh loading control. (C) Macrophages were transfected with scrambled RNA (scrRNA) or Pkr siRNA. After 72 h, the cells were incubated for the indicated times under cholesterol-loading conditions. Lysates were then immunoblotted for PKR, phospho-eIF2α (P-eIF2α), CHOP, and β-actin loading control. (D) Macrophages were transfected with scrambled RNA or Pkr siRNA. After 72 h, the cells were incubated for an additional 8 h without sterol (Con) or with 7-ketocholesterol and then assayed for Nox2 mRNA. The data are displayed as Nox2 mRNA levels relative to those for the control scrambled RNA group. (E, left) Macrophages from WT, Chop+/− , or Chop−/− mice were incubated for 14 h without sterol, under cholesterol-loading conditions, or with 7-ketocholesterol and then assayed for annexin V staining. (E, right) Macrophages were transfected with scrambled RNA or Pkr siRNA. After 72 h, the cells were incubated for an additional 30 h without sterol, under cholesterol-loading conditions, or with 7-ketocholesterol and then assayed for DCF fluorescence and annexin V staining. In this graph, * indicates P 70% of the intimal cells were F4/80-positive macrophages. Total RNA was isolated using the RNAqueous-Micro kit (Applied Biosystems) and reverse transcribed into cDNA using SuperScript III First-Strand Synthesis SuperMix (Invitrogen). QPCR was performed as described previously (Feng et al., 2003a). Immunohistochemistry and immunofluorescence analyses Mouse kidneys were immersion fixed in 10% neutral-buffered formalin overnight followed by embedding in paraffin. 5-µm paraffin-embedded sections were immunostained using anti-NOX2 antibody (Santa Cruz Biotechnology, Inc.) and the rabbit staining system (ImmunoCruz; Santa Cruz Biotechnology, Inc.). Sections were counterstained with DAPI and mounted with antifade reagent (ProLong Gold). Images were captured using an inverted microscope (Axiovert 200M; Carl Zeiss, Inc.). For immunofluorescence detection of superoxide, the frozen kidney or aorta sections were incubated with 2 µM DHE for 30 min at 37°C and then washed for 5 min with PBS. Sections were counterstained with DAPI and mounted with antifade reagent (ProLong Gold). DHE is a hydrophobic uncharged compound that can cross intracellular membranes and binds to superoxide, which can be monitored as a bright red fluorescence (Laurindo et al., 2008). Renal function tests Serum was collected after centrifuging clotted blood at 3,000 g for 20 min. The creatinine concentration in serum was measured using a commercially available kit (Creatinine Enzymatic Assay kit; Bio-quant, Inc.). The albumin concentration in the urine was measured by a competitive ELISA (Exocell) and normalized to urine creatinine levels, which were measured as using the Creatinine Enzymatic Assay kit. Statistics Data are presented as means ± SEM. For most of the bar graph data, n = 3 for each experimental group. However, for the bar graphs depicting fmax/f0 data, n = ∼30 per group. Analysis of variance followed by Tukey post-test (Prism 4 version 4.03; GraphPad Software, Inc.) was used to determine statistical significance among all groups. Bars with the same symbols are not significantly different from each other, whereas bars with different symbols are significantly different from each other. P-values for statistically different values ranged from P < 0.05 to P < 0.0001.
