Elevated Mitochondrial Reactive Oxygen Species Generation Affects the Immune Response via Hypoxia-Inducible Factor-1α in Long-Lived Mclk1^+/− Mouse Mutants

Hypoxia-inducible factor (HIF)-1α is an important regulator of the immune response. Several mechanisms have been proposed for how hypoxia, in the immune system and elsewhere, increases expression of the transcription factor HIF-1α. For example, oxygen levels and the depleting effect of mitochondrial oxygen consumption have been proposed (1). However, much evidence implicates another signal coming from the mitochondria, most likely a type of reactive oxygen species (ROS) (2–6). Local hypoxia develops in inflamed and diseased tissues, including tumors, atherosclerotic plaques, rheumatoid arthritic joints, myocardial infarcts, wounds, and at sites of bacterial infection (7). Macrophages and other immune effector cells play important roles at these sites where they respond to the local hypoxia by upregulating HIF-1α; however, the upregulation of HIF-1α can also be triggered by stimuli other than hypoxia, such as cytokines (in T cells) (8) and LPS (in macrophages) (9). HIF-1, in turn (10), stimulates the expression of genes that are important for the effector function of macrophages (11, 12). Recently, it was found that the transcriptional regulator NF-κB is one of the main links between innate immunity and the hypoxic response through its transcriptional control of Hif1a expression (13, 14).

One aspect of the effects of increased HIF-1 is to favor the activation of macrophages along the classic path of activation, which is proinflammatory, as opposed to the alternative pathway of activation, which is less proinflammatory and favors the resolution of inflammation and fibrosis (15). Which of the two possible routes of activation is followed depends on a complex interplay of interactions involving transcription factors, cellular contacts, and diffusible effector (15, 16). Interestingly, it was recently found that the choice between the pathways and the differentiation of the two types of macrophages involves the cellular mechanisms of energy generation. Indeed, in addition to the effect of mitochondrial ROS and HIF-1 expression on classic macrophage activation, it was found that the effect of IL-4 on the inhibition of classic activation and the secretion of inflammatory cytokines, such as TNF-α and IL-6, is crucially dependent on the presence of intact mitochondrial oxidative phosphorylation (17).

Mclk1 (a.k.a. Coq7) encodes a protein that functions as an hydroxylase in the biosynthesis of ubiquinone (18). Yet MCLK1 likely has at least one other function, because Mclk1^+/− heterozygous mutants that carry only one functional copy of the gene display numerous phenotypes in the absence of any detectable change in ubiquinone levels (19). Young Mclk1^+/− mutants show reduced mitochondrial oxygen consumption, reduced electron transport, reduced mitochondrial ATP synthesis, reduced mitochondrial and overall ATP levels, and reduced whole-animal oxygen consumption. These mutants also show increased mitochondrial oxidative stress, by several measures (19). Moreover, despite their reduced mitochondrial function and increased mitochondrial oxidative stress, the physiological condition in these animals leads to a slowing of the development of oxidative biomarkers or aging, such as plasma levels of isoprostanes and 8-hydroxy-deoxyguanosine, as well as to a decrease in the age-associated loss of mitochondrial function and to an increase in lifespan (20, 21).

We sought to take advantage of Mclk1^+/− mice as an in vivo model of increased mitochondrial oxidative stress to further test the evidence that relates mitochondrial function to the control of HIF-1α expression. Furthermore, we attempted to reconcile the positive features of the phenotype of Mclk1^+/− mutants with our observation of defective mitochondria in these animals. We focused on the immune system because of the known links between inflammation and oxidative stress and because immune function is capable of having a global impact on health and survival. In this study, we show that there is increased expression of HIF-1α in Mclk1^+/− mice as well as in the RAW264.7 mouse macrophage cell line after Mclk1 knockdown. In the mice, this was accompanied by an enhanced inflammatory response, including enhanced classic activation of macrophages and higher sensitivity to LPS challenge. Interestingly, macrophages from the mutant mice also have reduced alternative activation, with reduced sensitivity to IL-4 and a low level of arginase expression, which might participate in the bias toward an enhanced inflammatory response. We used the macrophage cell line to demonstrate that altered mitochondrial ROS metabolism resulting from reduced Mclk1 expression is responsible for increased HIF-1α expression and HIF-1α–dependent TNF-α secretion. In light of our findings, we discuss the mechanisms by which altered immune function might positively affect lifespan.

The Mclk1 knockout mutants were described previously (18, 20) and were maintained in the heterozygous state. The C57BL/6J strain was derived from the SV129S6 strain by crossing ≥10 times to C57BL/6J animals obtained from The Jackson Laboratory (Bar Harbor, ME). All procedures were approved by McGill University’s Animal Care and Ethics committees. RAW264.7 cells were purchased from the American Type Culture Collection and maintained in complete medium, containing DMEM (Invitrogen, Carlsbad, CA), in the presence of 10% FBS and penicillin (50 U/ml)-streptomycin (50 μg/ml).

TNF-α and IL-6 ELISA kits and Luminex multiple cytokine assay kits were purchased from R&D Systems (Minneapolis, MN) and BioSource International (Camarillo, CA), respectively. TRIzol reagent was purchased from Invitrogen; the RNeasy Kit, Omniscript RT Kit, and QuantiTect SYBR Green PCR Kit were purchased from Qiagen (Valencia, CA). Liquid alanine aminotransferase (ALT) Reagent Set was purchased from POINTE Scientific (Canton, MI). MitoSOX Red was from Invitrogen/Molecular Probes (Eugene, OR). LPS (L6529), fluorescent latex beads (L1278), and M-CSF (M9170) were purchased from Sigma-Aldrich (St. Louis, MO). Cytotox-ONE kit was from Promega (Madison, WI). Lymphocyte-M (CL5031) was purchased from Cedarlane Laboratories (Hornby, ON, Canada). Nuclear/Cytosol Fractionation Kit and NF-κB RelA polyclonal Ab were purchased from BioVision (Mountain View, CA). Anti-TATA box binding protein (TBP), anti–HIF-1α, and anti-rabbit–HRP Abs were purchased from Santa Cruz Biotechnology (Santa Cruz, CA), Cayman Chemical (Ann Arbor, MI), and Sigma-Aldrich, respectively. The QuantiChrom Arginase Assay Kit was purchased from BioAssay Systems (Hayward, CA). RNAiMAX reagent and stealth RNAi for RNA knockdown experiments were purchased from Invitrogen. The mitochondria-specific ROS scavenger peptide SS31 (H-D-Arg-Dmt-Lys-Phe-NH2) and the control peptide SS20 (H-Phe-D-Arg-Phe-Lys-NH2), which lacks antioxidant properties, were synthesized in the laboratory of Dr. Peter W. Schiller, Clinical Research Institute of Montreal, using a published protocol (22).

Both techniques were performed by standard protocols. For PCR, all measurements were performed in triplicate, and β-actin was used to normalize the data. The primer sequences used for quantitative PCR assays were as follows. Tnfa: 5′-GAACTGGCAGAAGAGGCACT-3′ and 5′-GGTCTGGGCCATAGAACTGA-3′; Il6: 5′-TTCCATCCAGTTGCCTTCTT-3′ and 5′-CAGAATTGCCATTGCACAAC-3′; Arg1: 5′-CGCCTTTCTCAAAAGGACAG-3′ and 5′-ACAGACCGTGGGTTCTTCAC-3′; Hif1a: 5′-GGAGCCTGATGCTCTCACTC-3′ and 5′-TTTGGAGTTTCCGATGAAGG-3′; Pgk1: 5′-CAAGGCTTTGGAGAGTCCAG-3′ and 5′-TGTGCCAATCTCCATGTTGT-3′; Pgc1b: 5′-CCTTCCCAGAACTGGATGAA-3′ and 5′-TCTGGAACTGAGGCTGGTCT-3′; and Actb: 5′-TGCGTGACATCAAAGAGAAG-3′ and 5′-GATGCCACAGGATTCCATAC-3′.

For Western blotting, the density of each band was analyzed using Scion Image software (Scion, Frederick, MD), and the density of TBP was used to normalize the data. All primer sequences can be obtained upon request.

LPS was injected i.p. into 10–12-wk-old C57BL/6J male mice at doses of 0.01 or 1 mg/kg. Mice were anesthetized and sacrificed 2 h after injection. Blood samples obtained by cardiac puncture were preserved in Li-heparin–containing tubes and centrifuged at 8000 rpm at 4°C for 10 min. The supernatant was aliquoted at 50 μl/tube and stored at −80°C until use. Tissue samples were snap-frozen immediately in liquid nitrogen and kept at −80°C until use.

Ten- to 12-wk-old male C56BL/6J mice were sacrificed, and the peritoneal cavity was washed with 8 ml PBS. PBS recollected after washing was centrifuged at 1100 rpm for 5 min, and the pellet was resuspended in RPMI 1640 supplemented with 10% FBS, 1% penicillin/streptomycin (complete medium). Macrophages were counted and plated at a density of 4 × 10⁵ cells/ml (for H₂O₂ and cytokine assays) or 8 × 10⁵ cells/ml for cytotoxicity assays. After incubation at 37°C in 5% CO₂ for 1 h, the cells were washed twice with PBS, and the adherent cells were kept in culture in prewarmed complete medium overnight before use. For hypoxia experiments, the cells were incubated in 5% O₂ in a hypoxia chamber overnight before extracting RNA or protein. Bone marrow-derived macrophages (BMDMs) were also prepared from 5–7-mo-old male C56BL/6J mice, as described previously (23). Briefly, the bone marrow cells were harvested from mouse femur and tibia and purified using Lymphocyte-M (Cedarlane Laboratories). After overnight culture, the cells were grown for 3 d in the presence of 10 ng/ml M-CSF (Sigma-Aldrich). The secreted TNF-α and IL-6 were measured by ELISA after stimulation with LPS (2.5 ng/ml)/IFN-γ (1 U/ml) in the presence or absence of IL-4 (5 ng/ml) for 24 h.

The concentrations of TNF-α and IL-6 in male mouse plasma and macrophage medium were measured using specific ELISA kits. Plasma IL-1β, GM-CSF, INF-γ, IL-2, -4, -5, -6, -10, and -12, and TNF-α were measured using the Luminex multiple cytokine assay according to the manufacturer’s instruction. All measurements were performed in duplicate and a few extreme values were excluded. Plasma samples were prepared as described above. Liquid ALT Reagent Set (POINTE Scientific) was used to determine the ALT activity, according to the manufacturer’s instructions.

Peritoneal macrophages were cultured in complete medium for 24 h. After washing twice with PBS, the H₂O₂ concentration was measured using an Amplex Red Hydrogen Peroxide Assay Kit, according to the manufacturer’s instruction. All measurements were performed in duplicate. For the phagocytosis assay, peritoneal macrophages were incubated in the presence of latex beads with orange fluorescence for 30 min at a concentration of 1:1000. After fixation, the fluorescence was checked by FACS. For the bacteria-induced cytotoxicity assay, exponential growing phase Escherichia coli (DH10B) were cocultured with macrophages at two concentrations (yielding two multiplicities of infection [MOI]). A Cytotox-ONE kit was used to quantify the bacteria-induced cell death. The release of lactate dehydrogenase of bacteria-treated cells was normalized by untreated controls. All measurements were performed in duplicate, according to the manufacturer’s instructions. For the short-peptide treatment, the freshly prepared peritoneal macrophages were cultured with 5 μM SS20 or SS31 for 24 h. The TNF-α concentration of the medium was measured by ELISA.

RAW264.7 cells were treated at a concentration of 10⁵ cells/ml in serum-free medium using RNAiMAX reagent, according to the manufacturer’s instructions. Forty-eight hours after treatment, RNA was extracted using TRIzol reagent, and nuclear protein was extracted using a Nuclear/Cytosol Fractionation Kit. To measure the mitochondrial superoxide level, the cells were incubated with 5 μM MitoSOX Red for 15 min after 48 h Mclk1 small interfering RNA (siRNA) treatment, then the cells were washed and subjected to FACS analysis. For the LPS challenge experiment, cells were treated with siRNAs for 48 h and then incubated with 50 ng/ml LPS for an additional 90 min. For short-peptide treatment experiments, cells were treated with siRNAs against Mclk1 for 24 h before adding the short peptides to the medium to a final concentration 5 μM, and cultured for another 24 h before analysis. The sequences of all RNAs can be obtained upon request.

Western band density was determined using Scion Image software (Scion). Data and graphs were processed using GraphPad Prism 5 software. Paired or unpaired t tests were used. The Welch correction was applied when the variances were different. Bar graphs are expressed as mean ± SE.

We measured the level of HIF-1α in liver nuclei of Mclk1^+/− mutants and control littermates and found that it was significantly elevated (Fig. 1A, 1B). The same samples were used to determine the level of RelA, a component of the transcription factor NF-κB, because of its close functional relationship with HIF-1α and its importance in innate immune function. The expression level of RelA was variable in both genotypes, but, on average, it was not significantly elevated in the mutants (Fig. 1A, 1B). However, there was a strong correlation between the levels of HIF-1α and the RelA expression in each sample from Mclk1^+/− mutants but not in those from the Mclk1^+/+ controls, suggesting that RelA was elevated in at least some mutant animals (Fig. 1C).

RelA and HIF-1α can be induced by LPS (9, 10). To further study whether the altered expression of RelA and HIF-1α in Mclk1^+/− mutants was likely to affect their immune function, we measured the levels of the two proteins in the livers of animals treated with a low dose of LPS (0.01 mg/kg) (Fig. 1A, 1D). The expression of HIF-1α was increased in comparison with untreated animals in both genotypes (compare Fig. 1B and 1D), but it remained substantially higher in the mutants compared with the controls (Fig. 1D). Furthermore, the expression of RelA after treatment became significantly higher, on average, in the mutants (Fig. 1D).

To evaluate whether a specific cell type, such as Kupffer cells, is responsible for the elevation of HIF-1α, we carried out immunohistochemical staining against HIF-1α in liver paraffin sections after LPS treatment. The staining appeared homogenous, suggesting that HIF-1α levels were not increased only in a particular cell type (data not shown). Because HIF-1 is a transcription factor, we also tested by RT-PCR the expression of Pgk1 (phosphoglycerate kinase 1), a HIF-1 target, in the liver of LPS-treated animals (0.01 mg/kg). As expected, Pgk1 expression was significantly elevated in the mutants (Supplemental Fig. 1A).

Macrophages are one of the principal cell types that form the innate immune system and shape its responsiveness. We studied peritoneal macrophages, which were harvested and cultured overnight before analysis. Because of the limited amount of protein that could be extracted from the macrophages of a single animal, macrophages from three mice of each genotype were pooled. Three groups of three mice were used. Under these conditions, the level of RelA could be readily detected and was found to be significantly increased in the mutants (Fig. 2A, 2B). However, HIF-1α was barely detectable (Fig. 2A), and no difference could be scored (Fig. 2B). Nonetheless, to obtain information on the expression of HIF-1α and to overcome the limit of detection by Western blotting, we scored Hif1a mRNA levels by RT-PCR and observed a significant increase in the transcription of Hif1a in the mutants (Fig. 2C).

The expression of HIF-1α and RelA is likely lower in culture than in vivo because of the hyperoxia that cells experience when cultured in atmospheric oxygen (21%), although this condition is called normoxia by convention. In fact, the oxygen tension in healthy tissues is believed to be ∼2.5–9%. Therefore, we wondered whether the increases we observed in Mclk1^+/− mutants could be due to their different sensitivity to oxygen. To mimic in vivo oxygen concentration, we cultured the macrophages in a hypoxia chamber with 5% oxygen overnight. The levels of both proteins in both genotypes increased in response to this relative hypoxia, and the levels of both proteins were higher in the mutants (Fig. 2A, 2D). In fact, the increased level of HIF-1α in Mclk1^+/− mutants could now be scored. Thus, HIF-1α and RelA are expressed at higher levels in Mclk1^+/− mutant livers treated with LPS and in mutant macrophages under relative hypoxia. The increased expression of these two transcription factors that control inflammation suggest that Mclk1^+/− mutants might exhibit an enhanced inflammatory response.

We tested whether mutant macrophages were activated, with the typical increased antibacterial and inflammatory effector functions, as a result of the increased expression of HIF-1 and NF-κB (10, 12, 24, 25). We used peritoneal macrophages cultured overnight to measure Tnfa mRNA expression by quantitative RT-PCR (Fig. 3A), as well as TNF-α and H₂O₂ secretion (Fig. 3B, 3C). All three measures were significantly elevated in the Mclk1^+/− cells. We also tested bacteria-induced cell death, because increased resistance is expected from activated macrophages (26, 27). In fact, Mclk1^+/− cells were more resistant at two different MOI (Fig. 3D). Lastly, we tested the phagocytic capacity of macrophages by scoring latex bead uptake (28), which was again increased in Mclk1^+/− cells (Fig. 3E). This increase in classic macrophage activation is consistent with the increased level of HIF-1α expression.

IL-4 favors alternative macrophage activation and inhibits classic activation (15, 29). It was reported that the inhibitory effect of IL-4 on classic macrophage activation of BMDMs is dependent on intact mitochondrial function (17). Therefore, we wondered whether the Mclk1^+/− mutant phenotype, which, in addition to increasing mitochondrial ROS, reduces mitochondrial oxidative phosphorylation (19), would affect the sensitivity of macrophages to IL-4. BMDMs from each genotype were prepared from the femur and tibia and grown for 3 d in the presence of 10 ng/ml M-CSF. After stimulation with LPS (2.5 ng/ml) and IFN-γ (1 U/ml), IL-4 treatment (5 ng/ml) of wild-type cells reduced the secretion of TNF-α and IL-6 to 55% and 41%, respectively, of that of untreated cells (i.e., 45% and 59% inhibition; Fig. 3F). In contrast, IL-4 treatment of Mclk1^+/− mutant cells only reduced the secretion of TNF-α and IL-6 to 74% and 60%, respectively, of that of untreated cells (i.e., 26% and 40% inhibition; Fig. 3F). These findings suggest that the activated, inflammatory phenotype that we detected in Mclk1^+/− mutant macrophages is the result of the effect of high mitochondrial ROS on HIF-1α expression and a reduced ability to differentiate along the alternative path of activation as the result of reduced oxidative phosphorylation.

It was reported that PPARγ coactivator 1β (PGC-1β) primes macrophages for alternative activation and strongly inhibits proinflammatory cytokine production (17). Therefore, we used quantitative RT-PCR to investigate whether the expression of PPARγ coactivator 1β was altered in the liver of LPS-treated mutant animals (0.01 mg/kg) and in mutant macrophages; no significant effect was observed (Supplemental Fig. 1B, 1C).

The increase in classic activation and the resistance to the effects of IL-4 in the macrophages of Mclk1^+/− mice predicts a decrease in arginase activity and Arg1 gene expression, which is necessary for collagen synthesis during fibrosis, and is typical of alternative activation (15). As expected, we found a significant decrease in arginase activity (Fig. 3G) and Arg1 gene expression (Fig. 3H) in Mclk1^+/− macrophages.

The elevated levels of expression of HIF-1α and RelA in liver and peritoneal macrophages, as well as the partial resistance of macrophages to the inhibition of inflammatory cytokine secretion by IL-4 in vitro, predict an altered inflammatory response in Mclk1^+/− mutants. We first used quantitative RT-PCR to measure the expression of the inflammatory cytokine genes Tnfa and Il6 in the liver of naive animals (Fig. 4A, 4B). The expression levels of both cytokines were increased in Mclk1^+/− mutants, but only the increase observed for Tnfa reached statistical significance. Similarly, the plasma levels of both cytokines, measured by ELISA, seemed to be elevated (Fig. 4C, 4D), but the differences did not reach significance; this is likely due to the very low levels found in unstimulated animals and to the large animal-to-animal variability that we always observe in the mutants but not in the wild-type (see below).

Because the expression of HIF-1α was higher in Mclk1^+/− mice after challenge with 0.01 mg/kg LPS, we treated animals in the same way before testing the circulating levels of four inflammatory cytokines (TNF-α, IL-1β, IL-6, and GM-CSF) and six Th1/Th2 cytokines (IFN-γ, IL-2, -4, -5, -10, and -12), using the Luminex multiple cytokine assay (BioSource). As shown in Fig. 4E and Supplemental Table I, significantly more TNF-α, IL-6 and -12, and GM-CSF were found in Mclk1^+/− mice, on average. The elevations of plasma TNF-α and IL-6 in Mclk1^+/− mice were confirmed by ELISA (Supplemental Fig. 2). The level of IL-1β was also elevated (Fig. 4E, Supplemental Table I), but the difference did not reach statistical significance.

We observed a great spread of values for cytokine levels in the mutants, with many animals having cytokine levels that were not different from those of wild-type animals. However, the levels of cytokines in the mutants were correlated with each other. Fig. 4F and 4G illustrate the correlations between the levels of TNF-α and IL-6 and TNF-α and IL-12, respectively. Animals that have elevated levels of TNF-α also have elevated levels of IL-6 and -12. This indicates that some mutant animals had higher levels of several cytokines, but other animals did not have elevated cytokines. In conclusion, the inflammatory response in Mclk1^+/− mice is variable but is hypersensitive to low levels of LPS in many individual animals, and cytokine levels also seem to be spontaneously elevated in some animals.

The inflammatory response can result in chronic and acute tissue damage. Plasma ALT activity, a measure of tissue damage, was used to estimate the extent of such damage in naive animals as well as in animals 2 or 24 h after treatment with a high dose of LPS (1 mg/kg) (Fig. 5A). No significant differences in ALT levels between the genotypes were detected. We also evaluated the effect of 1 mg/kg LPS on cytokine levels 2 h after treatment. As expected, the levels of several cytokines were significantly higher than after treatment with 0.01 mg/kg LPS, but there was no significant difference between the genotypes with this harsher treatment (Supplemental Table I). The inflammatory reaction triggered by treatment with a high level of LPS represents an important stress, and animals treated in this way can lose a substantial fraction of their body weight (30). We determined the effect of 1 mg/kg of LPS on body weight 24 h after the treatment (Fig. 5B) and observed that Mclk1^+/− mutants lost a significantly smaller fraction of their body weight as a consequence of their response. Thus, the hair-trigger inflammatory response that we observed in Mclk1^+/− mutants is not obviously deleterious at low or high levels of LPS stimulation.

To verify the causal links suggested by the analysis of the phenotype of Mclk1^+/− mice, we used the RAW264.7 mouse macrophage-like cell line, which is a model for the study of macrophages, HIF-1α, and mitochondria (10, 31–35). The levels of nuclear HIF-1α were significantly increased upon siRNA knockdown of Mclk1 (Figs. 6A, 6B, 7B). Mclk1 knockdown also significantly elevated TNF-α levels in the medium of the cells after 48 h without LPS treatment or 1.5 h after stimulation with 50 ng/ml LPS (Figs. 6C, 6D, 7C). These increases in TNF-α were entirely abolished by concomitant knockdown of Hif1a (Fig. 6C, 6D). The levels of Hif1a mRNA were also monitored in these experiments on TNF-α secretion (Fig. 6E, 6F). The effect of Mclk1 knockdown on HIF-1α seems to be posttranscriptional, because Hif1a mRNA levels were not altered by Mclk1 knockdown, but the effects on TNF-α were clearly Hif1a expression dependent.

We first established that Mclk1 knockdown in RAW264.7 cells increased mitochondrial ROS levels as expected from the phenotype of Mclk1^+/− mice. We used MitoSOX Red fluorescence to monitor mitochondrial superoxide levels. Superoxide is the ROS produced initially as a result of mitochondrial electron transport, and its production is increased as the result of inefficient or abnormal function of the mitochondrial respiratory chain. Mitochondrial superoxide is detoxified by superoxide dismutase (SOD) 2, the mitochondrial Cu/Zn SOD. Therefore, we used siRNA knockdown of Sod2 as a control, because this would be expected to increase mitochondrial superoxide levels (Fig. 7A). Mclk1 and Sod2 knockdown led to a clear increase in MitoSOX Red fluorescence. Although superoxide is the ROS initially produced in the mitochondria, it is rapidly detoxified by SOD2 into peroxide (H₂O₂), which, in contrast to superoxide, is able to cross membranes. Most antioxidants are not specific for the mitochondria or for a single ROS species. Thus, they exercise powerful effects on many cellular functions that can be affected by ROS or the redox state and, therefore, could affect HIF-1α and TNF-α levels in a variety of indirect ways. Thus, to test whether the altered ROS metabolism induced by Mclk1 knockdown is responsible for the increase in HIF-1α and TNF-α levels, we used a mitochondrially targeted H₂O₂-specific antioxidant peptide (Szeto-Schiller peptide SS31) and its inactive homolog (SS20) as a control (36, 37). SS31 treatment, but not SS20 treatment, fully abolished the effect of Mclk1 on HIF-1α and TNF-α (Fig. 7B, 7C). Comparable effects were also obtained in peritoneal macrophages (Fig. 7D). These findings fully support the notion that a ROS-based signal from mitochondria, likely H₂O₂ itself, can regulate HIF-1α levels and that the effects of reduced MCLK1 levels in mice and cultured cells is mediated by altered mitochondrial ROS metabolism.

Several studies documented that ROS originating from the mitochondria in response to hypoxia, likely H₂O₂, are involved in the stabilization of HIF-1α (2–4). We showed in this study that Mclk1^+/− mouse mutants, which sustained elevated mitochondrial oxidative stress (19), exhibited elevated levels of HIF-1α: 1) HIF-1α is elevated in the livers of mutant mice, even in the absence of any inducing treatment; 2) treatment of the animals with a very low dose of LPS triggered an increase in HIF-1α in wild-type and mutant mice but to a much greater extent in the mutants; and 3) the effect of mild hypoxia on HIF-1α was much more pronounced in peritoneal macrophages from Mclk1^+/− mice than from sibling controls. Furthermore, RAW264.7 macrophage-like cells in which Mclk1 had been knocked down by siRNA treatment displayed elevated HIF-1α levels that required elevated mitochondrial H₂O₂. To our knowledge, these findings provide the first evidence for the action of mitochondrial ROS on HIF-1α expression and function in an intact animal and provide a model to study these mechanisms further. Although we investigated the consequences of a reduction in MCLK1 levels on HIF-1α, the level of mitochondrial changes observed in Mclk1^+/− could readily be produced by other conditions, such as toxins, anticancer drugs, hypoxia, hyperthermia, or deliberate pharmacological intervention.

A number of observations with Mclk1^+/− mutants suggest a mechanism for the amplification of the effect on HIF-1 of a potentially small initial increase in mitochondrial ROS generation. These observations include increased classic macrophage activation; partial resistance to alternative macrophage activation; increased expression of inflammatory cytokines; increased expression of the NF-κB subunit RelA, including in macrophages; and increased Hif1a gene expression in macrophages. The mechanism of amplification might include the following steps: HIF-1α activates macrophages, and more generally favors the inflammatory response (11), which leads to the secretion of cytokines, such as TNF-α, IL-1β, and others, which, in turn, activates the expression of NF-κB (38), which is a transcriptional activator of Hif1a (13). We did not observe a transcriptional effect in RAW264.7 cells, which is consistent with the notion that the effect of Hif1a transcription is a secondary one via cytokine expression in the animals. The reduction of mitochondrial oxidative phosphorylation that is observed in Mclk1^+/− mutants (19) might further contribute to the amplification of the effect of mitochondrial ROS on HIF-1 by partially inhibiting the effect of IL-4 in triggering alternative over classic activation of macrophages, because the net effect of this inhibition should be an increase in classic activation. To fully establish a causal relationship between the increased levels of HIF-1α and the enhanced inflammatory response that we observed in Mclk1^+/− mutants, we plan to study the genetic interactions between Mclk1 and Hif1a in double-mutant combinations.

The inflammatory response is a powerful weapon against infection. As such, it can also have deleterious effects, such as the temporary incapacitation of the organism, as well as inducing tissue damage. Mclk1^+/− mutants show a tendency for elevated markers of inflammation in the absence of experimental stimulation. Possibly, this is due to a stronger than normal reaction to the numerous minor infections that mammals sustain and combat successfully on a continuous basis. Furthermore, at a very low dose of LPS (0.01 mg/kg), Mclk1^+/− mutants display a much greater increase in inflammatory cytokine expression in the liver and plasma than that seen in controls. However, we did not observe that this increased reactivity is obviously deleterious to the organism. For example, we did not observe any difference in plasma ALT levels between the genotypes, even after a high dose of LPS (Fig. 5A). In fact, the weight loss that is a typical deleterious consequence of inflammation, which is brought about by LPS treatment, was significantly less marked in the mutant mice (Fig. 5B).

The aging process impairs immune system function (39). Thus, under imperfectly aseptic conditions, part of the mortality of caged mice is likely due to infections by microorganisms to which they have good resistance when young, but which decreases when weakened by senescence. Thus, the increased lifespan of Mclk1^+/− could be the result of greater resistance to infectious microorganisms. Interestingly, in the survival experiment with Mclk1^+/− mutants conducted previously in the background used in this study (C57BL/6J), we observed a surprisingly short lifespan for the control Mclk1^+/+ mice (20) compared with previously reported results for animals in this background (40, 41). This suggests that environmental conditions, such as exposure to microorganisms, were limiting the survival of the wild-type animals in this particular experiment, yet Mclk1^+/− mutants had significantly better survival than their Mclk1^+/+ siblings, possibly because of their hair-trigger immune response.

We hypothesize that another mechanism linked to infection and immunity, beyond actual better immediate survival from infection, could contribute to the greater lifespan of the mutants. The inflammatory reaction, and the subsequent development of the immune response, protects against infection but also injures the tissues of the infected animals. Thus, the immune response needs to be terminated and the tissue repaired. However, these processes are imperfect; therefore, infection can lead to chronic inflammation and fibrosis (42). Animals that sustain repeated serious infections and the subsequent immune response might suffer from a gradual accumulation of fibrotic lesions, as well as suffer more severely from age-dependent diseases, many of which have the characteristic of chronic inflammation. The link between inflammation and the aged phenotype has been studied most extensively in humans (43). By producing a relatively strong reaction at the slightest sign of infection, the hair-trigger inflammatory response of Mclk1^+/− mutants might prevent severe or chronic infection. Such a hair-trigger response might not be favored by natural selection because of various energetic and behavioral costs to the animals, but it might allow Mclk1^+/− mutants, which are fed and sheltered from predators in the animal facility, to sustain fewer severe infections and the subsequent damage inflicted by the microorganisms and the full run of the immune reaction against them. Thus, in the long-term, a hair-trigger response might lead to a slower accumulation of permanent damage or chronic inflammation due to infection. These considerations suggest that the normal immune response contributes to physiological aging, as assessed by biomarkers of oxidative stress, which are also increased by inflammatory states. Therefore, it is of interest that Mclk1^+/− mutants display dramatically improved global biomarkers of aging (21). This suggests the possibility that the accumulation of oxidative damage that is observed with aging is, in part, the consequence of the damage resulting from chronic infection and inflammation (44).

In humans, an association has been observed between lower childhood mortality and an increase in longevity that is typical of cohorts born closer to modern times (45). The low childhood mortality has been interpreted as resulting from a lower rate of infection, possibly due to better sanitation. This, in turn, has been interpreted as pointing to the deleterious consequences of inflammation for longevity, especially in light of the fact that many age-dependent diseases have an inflammatory component. Our findings suggest the possibility of a different interpretation of the same observations. The low rate of childhood infection and mortality might be the result of a stronger or more sensitive immune response in addition to better sanitation. Indeed, the characteristics of the immune response also depend on environmental factors, such as nutrition (46). Thus, a hair-trigger inflammatory response that results in a low rate of severe infection throughout life might ultimately favor longevity for humans as well. It is possible that the inflammatory age-dependent diseases that are the most prevalent diseases in long-lived modern humans are a trade-off that limits further increases in lifespan.

We thank Eve Bigras, Geraldine Gabriel, and Nadia Prud’homme for expert technical assistance and Drs. Ciro A. Piccirillo and Jérôme Lapointe for critically reading the manuscript. We thank Dr. Peter W. Schiller, Clinical Research Institute of Montreal, for drawing our attention to the SS peptides and for providing samples of SS31 and SS20.

Disclosures The authors have no financial conflicts to report.