Interferon regulatory factor 7 impairs cellular metabolism in aging adipose-derived stromal cells

Competition for food and response to infections are the major challenges for survival and evolution (Darwin, 1859; Fumagalli et al., 2011). To face these challenges, organisms have evolved complex strategies of metabolic adaptation and immunity. Their dysregulation (López-Otín et al., 2016; Franceschi et al., 2018b) is the major hallmark of the passing of time over the course of adult life and drastically increases vulnerability to diseases by midlife (midlife switch) (Berry et al., 2017; D'Antona et al., 2010; Manaye et al., 2013; Mori et al., 2012, 2014; Rogers et al., 2012; Schaum et al., 2020; Shoji et al., 2016; Timmons et al., 2019; Wu et al., 2014). Nevertheless, the mechanistic connection between metabolic and immune dysfunction with age is still largely unexplored. Some metabolic factors have been proposed that might sustain chronic inflammation with age (excess of nutrients, mitochondrial debris, misfolded and/or misplaced proteins), but very little is known about whether and how immune effectors might impact on metabolic homeostasis (Franceschi et al., 2018b).

Beside systemic immunity, all nucleated cells are endowed with a cell-autonomous, anciently conserved mechanism of self-defense to counteract microbial infection (MacMicking, 2012; Randow et al., 2013). This works as the first barrier against intruders and may eventually prime a systemic response. Type I interferons (IFN I) are the major effectors of cell-autonomous immunity (MacMicking, 2012). The nine members of the interferon regulatory factor (IRF) family of transcription factors (named IRF1 to IRF9) coordinate IFN I activation and function (Honda and Taniguchi, 2006). Along with IRF3, IRF7 is a master activator of IFN I (Honda et al., 2005) and interacts with other transcription factors (STATs, ATF-2, NFkB, CREB, AP-1) to modulate grading and diversification of IFN I-mediated responses (Levy et al., 2002; Wathelet et al., 1998). The IFN I network reprograms the expression of thousands of interferon-stimulated genes (ISGs) (Schneider et al., 2014; Schoggins et al., 2011), the impact of which on the biology of non-immune cells is largely unexplored (Levy et al., 2002; Wathelet et al., 1998).

We show herein that IRF7 expression increases with chronological age in adipose-derived stromal cells, driving impairment of mitochondrial function, amino acid biogenesis and branched-chain amino acid (BCAA) degradation. Our results identify IRF7 as major causal link between immune and metabolic dysfunctions that occur with age and draw a new paradigm of aging as the ‘sterile’ (i.e. in the absence of viral infection) activation of a cell-autonomous pathway of self-defense that impairs metabolic homeostasis and functions in a cell-autonomous fashion.

Previous studies have shown that most age-related metabolic and neurodegenerative dysfunctions in rodents occur within the first half of life (up to 12 months of age) (Berry et al., 2017; D'Antona et al., 2010; Manaye et al., 2013; Mori et al., 2012, 2014; Rogers et al., 2012; Schaum et al., 2020; Shoji et al., 2016; Timmons et al., 2019). Adipose tissue is the central hub of metabolic (Rosen and Spiegelman, 2006) and immune regulation (Wu et al., 2007), and it represents an early sensor of the passing of time, as changes in its transcriptome precede those of other organs during adult life (Schaum et al., 2020). In a search for cell-intrinsic determinants of homeostatic alterations at middle age, we performed whole-genome transcriptional profiling of adipose-derived mesenchymal stromal cells (MSCs) explanted from 1-month-old (1mo; n=3) and 12-month-old (12mo; n=3) mice. To identify functional categories of genes that changed in a coordinated manner, we applied a gene set enrichment analysis (GSEA) approach (Subramanian et al., 2005) using a pre-compiled collection of genes belonging to Reactome functional annotation sets (Fabregat et al., 2016; Milacic et al., 2012; Vastrik et al., 2007) (Fig. 1A). In GSEA, genes in the dataset are first ranked based on their differential expression between the two experimental groups, and tests are then conducted on the ranked dataset to determine whether the gene sets of interest are randomly distributed or overrepresented among the genes upregulated in one group over the other. Cytokine and IFN signaling pathways were the most enriched functional categories in 12-month-old MSCs, while genes associated with respiratory electron transport had the highest enrichment scores (ES) in 1-month-old MSCs (Fig. 1B). GSEA using mitochondrial-related gene ontology (GO) terms (Ashburner et al., 2000; The Gene Ontology Consortium, 2017) showed a significant enrichment of a large repertoire of mitochondrial genes encoding for structural and functional components, including mitochondrial matrix, ribonucleoproteins, aminoacyl-tRNA synthetases and regulators of mitochondrial biogenesis, in addition to components of the respiratory chain (Fig. 1C, left). Quantitative PCR (qPCR) analysis confirmed the overexpression of multiple oxidative phosphorylation (OXPHOS) genes in a larger cohort of freshly isolated MSCs (Fig. 1C, right). Consistent with increased mitochondrial gene expression, 1-month-old MSCs showed a higher number of mitochondria. In addition, the mitochondria of 1-month-old MSCs displayed a larger size and more structured cristae than those of 12-month-old MSCs (Fig. 1D). Analysis of mitochondrial respiration and function indicated quantitative and qualitative differences between 1-month-old and 12-month-old MSCs (Fig. 1E). Oxidative function was reduced overall in 12-month-old MSCs compared to that of 1-month-old MSCs (two-way ANOVA; P=0.03). Among the different respiratory states analyzed, oxygen consumption in the uncoupled maximal capacity of mitochondria (ET) state was 34% lower in 12-month-old MCSs compared to that in 1-month-old cells (P=0.01; Fig. 1E, panels I and II). Beside these quantitative data, qualitative analysis also revealed further differences in mitochondrial function between the two groups of MSCs. Indeed, data describing respiratory state expressed relative to the maximal ET capacity (flux control ratios, FCRs) indicated that, although no difference was found in the baseline unstimulated ROUTINE state (Fig. 1E, panel III), the fraction of oxygen consumption recorded in the unstimulated dissipative LEAK state in 12-month-old MSCs was higher compared to that of 1-month-old MSCs (56% higher; P=0.03; Fig. 1E, panel IV). Consistently, the net ROUTINE capacity (R-L), determined by physiological substrate uptake and cellular energy requirements, was higher in 1-month-old cells (69% higher; P=0.02). Furthermore, the ET coupling efficiency, accounting for matching between oxygen consumption and ATP production, was lower in the 12-month-old MSCs compared to that in the 1-month-old MSCs (4.2% lower; P=0.03; Fig. 1E, panel V). These results indicate that aging affects mitochondrial respiratory efficiency and leads to increased proton dissipation at the mitochondrial inner membrane.

The genes associated with the mitochondrial-related GO terms mentioned above (GO_MITOCHONDRIAL MATRIX, GO_MITOCHONDRIAL_PART, GO_RESPIRATORY-CHAIN; Fig. 1C, left) comprised 971 unique transcripts (Fig. 1F). Interestingly, among them, 691 (71%) were identified as IFN targets by the INTERFEROME database, a web-based repository of experimentally verified ISGs (Rusinova et al., 2013). This subset of mitochondrial genes was strongly enriched in 1-month-old MSCs relative to 12-month-old MSCs (Fig. 1G). Strikingly, the leading-edge subset – that is, the genes that contributed the most to the ES – comprised 221 genes mainly related to amino acid biosynthesis and BCAAs (Fig. 1H). This suggests that the age-induced changes in IFN signature might be the driving force of a multifaceted transcriptional reprogramming that impairs mitochondrial function and alters amino acid metabolism.

In line with our transcriptomic data, a comparison of the untargeted metabolomic profiles of 1-month-old and 12-month-old MSCs (Fig. 2) revealed a striking depletion of amino acids at middle age (Fig. 2A). Most of the key enzymes involved in amino acid biosynthesis were upregulated at the transcriptional level in 1-month-old MSCs compared with their expression in 12-month-old MSCs, especially those related to serine and glycine biosynthesis (Fig. 2B). The list of upregulated transcripts includes asparagine synthetase (Asns), an established readout of the activation of Atf4, which is the common transcriptional effector of downstream stress response pathways and the master regulator of amino acid biosynthesis-related genes (Harding et al., 2003).

Genes implicated in amino acid biosynthesis were identified as ISGs, according to the INTERFEROME database (Fig. S1A). Specifically, the promoters of these genes displayed motifs indicating regulation by various transcription factors activated by IFN type I and II regulators (Fig. S1B).

Having identified the transcriptional repression of genes implicated in amino acid biosynthesis, we also assessed the expression of genes implicated in amino acid degradation pathways. Most of the genes assessed had very low levels of expression, although they displayed a slight upregulation with age (Fig. 2C). In sharp contrast, branched-chain amino acid transaminase genes (Bcat1 and Bcat2) displayed a higher expression and were downregulated with age. These Bcat genes encode two enzymes that catalyze the first step of the degradation of BCAAs (valine, leucine and isoleucine; Fig. 2D), yielding glutamate as a by-product. Hence, BCAT proteins are simultaneously implicated in amino acid degradation (of BCAAs) and synthesis (of glutamate). According to the INTERFEROME database, the transcriptional changes of the amino acid-related genes with age were consistent with the effects of IFN treatment (Fig. S1C).

To more robustly assess these transcriptional changes in amino acid metabolic pathways, we conducted a qPCR analysis of a greater number of biological replicates of differently aged MSCs (1-month-old MSC cells, n=8; 12-month-old MSC cells, n=6) (Fig. 2E). Although only Bcat1 and Psat1 showed changes in expression that reached statistical significance, we confirmed a coordinated downregulation with age of BCAA degradation-related and amino acid biosynthesis-related genes and the upregulation of amino acid degradation-related genes. However, we were not able to detect Ido1 in our qPCR analysis.

To further clarify these findings, we interrogated a recently published resource of single-cell transcriptomic data from 3-, 18- and 24-month-old mice (Consortium, 2020) (Fig. 2F). Specifically, we assessed the expression of genes related to mitochondria structure and function (the 971 genes indicated in Fig. 1F), amino acid biosynthesis (those listed in Fig. 2B), amino acid degradation (those listed in Fig. 2C) and BCAA degradation (those assigned to KEGG pathway mmu00280), in different deposits of adipose tissue (brown adipose tissue, BAT; gonadal adipose tissue, GAT; mesenteric adipose tissue, MAT; and subcutaneous adipose tissue, SCAT) either as an average of all cell types or as individual values of single resident MSCs, both in male and female animals. At least in males, we confirmed a strong anti-correlative relationship between age and expression of mitochondria-, BCAA degradation- and amino acid biosynthesis-related genes. Surprisingly, in sharp contrast with our model, amino acid degradation-related genes also displayed a negative correlation with age, although to a lesser extent. Of note, in line with data from our model, amino acid degradation-related genes were found to be expressed at very low level. Expression of Ido1, in particular, was undetectable. Taken together, these findings suggest that, although the effect of age on expression of amino acid degradation-related genes might be variable and context dependent, the downregulation of mitochondria-, amino acid biosynthesis- and BCAA degradation-related genes is inherently associated with aging in males.

In a search for molecular determinants underlying the transcriptional and metabolic dysregulation described above, we interrogated our microarray dataset (1-month-old versus 12-month-old MSCs, n=3 each) for the enrichment of gene sets comprising the targets of 615 different transcription factors (GSEA MSigDB collection c3.tft). The most highly enriched gene sets were related to the targets of the IRF family of transcription factors (Fig. 3A). Among the nine known members of the IRF family, the only ones readily upregulated in 12-month-old MSCs compared to their expression in 1-month-old MSCs were IRF7 and, to a lesser extent, IRF9 (Fig. 3B). This is in line with the STRING database (Szklarczyk et al., 2017), which reports IRF7 and IRF9 as displaying the highest frequency of co-expression among the IRFs (Fig. 3C). We confirmed a remarkable upregulation of IRF7 mRNA and protein (Fig. 3D) in 12-month-old MSCs compared to the expression in 1-month-old MSCs. Previously identified IRF7-specific targets (Puthia et al., 2016) were also significantly enriched in older MSCs (Fig. 3E).

The increase of IRF7 expression with age was not restricted to MSCs, since we found IRF7 protein to be upregulated in skeletal muscle (both gastrocnemius and quadriceps), adipose tissue and lungs (but not the spinal cord) of 8-month-old mice (n=3) compared to 1-month-old mice (n=3) (Fig. 3F, upper panels). qPCR analysis confirmed the downregulation of mitochondria-, amino acid biosynthesis- and BCAA degradation-related genes in 8-month-old adipose tissue and lung, whereas amino acid degradation-related genes, such as Got1 and Ido1, were upregulated or unchanged (Fig. 3F, lower panels).

To further support our findings, we assessed how the transcriptional change of Irf7 with age might correlate with expression of metabolic genes using a recently published dataset that comprises RNA-seq profiling of multiple organs (gastrocnemius, kidney, liver and hippocampus) from differently aged (6, 9, 12, 18, 21, 24 and 27 months) rats (Shavlakadze et al., 2019) (Fig. 3G). Once again, the analysis confirmed the upregulation of IRF7 expression with age, and its inverse correlation with expression of mitochondria- and BCAA-related genes in different organs (gastrocnemius, kidney and liver, but not hippocampus), whereas how age correlates with expression of amino acid biosynthesis- and degradation-associated genes remains more controversial.

In addition, we surveyed the expression of Irf7 transcripts across additional publicly available datasets from the Gene Expression Ominibus (GEO) database (Fig. S2) and confirmed upregulation of Irf7 expression at either middle or old age in most tissues and organs (skeletal muscle, adipocytes, heart, kidney, brain and lung) across multiple species. We then interrogated the same GEO datasets for the enrichment of gene sets related to mitochondrial biogenesis and function, translation and amino acid metabolism (Fig. 3H). In line with the data from MSCs, younger samples across the GEO datasets displayed a strong enrichment of genes related to mitochondrial components, mitochondrial translation machinery and, to a lesser extent, aminoacyl-tRNA biosynthesis. Conversely, older samples displayed an enrichment of genes related to amino acid deprivation. Collectively, these data suggest that IRF7 expression increases with age in multiple tissues across different species and correlates with the downregulation of mitochondria- and BCAA degradation-related genes.

The above data points to IRF7 as the causal connection between increased interferon signaling and dysregulated cellular metabolism over adult life. To definitively test this hypothesis, we knocked-down IRF7 in 12-month-old MSCs (shIRF7; Fig. 4A) and executed a new whole-genome microarray profiling. Genes that had significantly (t-test; P<0.05) downregulated expression in shIRF7 MSCs were enriched in functional categories related to the antiviral immune response, thus clearly validating the loss of IRF7 function (Fig. 4B; Fig. S3). On the other hand, genes significantly upregulated in 12-month-old shIRF7 MSCs were predominantly enriched in mitochondrial- and electron transport-related genes (Fig. 4B; Fig. S3). To identify the functional categories of genes that were altered in aging and reversed by IRF7 knockdown, we performed GSEA with the Reactome collection of gene sets and plotted the ES values of 12-month-old control (Ctrl) MSCs versus shIRF7 MSCs against the respective ES value of 1-month-old versus 12-month-old MSCs. Strikingly, interferon signaling pathway- and oxidative phosphorylation-related gene sets hit the top up and top down positions, respectively (Fig. 4C). In addition, genes related to ‘mitochondrial-matrix’, ‘respiratory chain’ and ‘mitochondrial-part’ GO categories were significantly enriched in 12-month-old shIRF7 cells compared to 12-month-old control cells (Fig. 4D). In accordance with the transcriptional data, 12-month-old shIRF7 cells showed a significantly greater number of mitochondria displaying more structured cristae compared to controls (Fig. 4E).

To determine whether, besides affecting the expression of mitochondrial gene sets and mitochondrial morphology, IRF7 depletion is also relevant for mitochondrial function, we compared, using high-resolution respirometry (HRR), the rates of oxygen consumption of intact 12-month-old shIRF7 cells and 12-month-old control cells, in which metabolic responses are essentially based on endogenous substrates (Fig. 5A–C). A coupling control protocol was used to investigate the different respiratory states (ROUTINE, LEAK and ET). The ET state induced by stepwise titration of the uncoupler (CCCP) was significantly increased by 25% (P=0.02) in 12-month-old shIRF7 cells compared to that in controls (Fig. 5A). Data normalized relative to the ET (FCRs) showed that 12-month-old shIRF7 cells had reduced oxygen consumption in the basal ROUTINE (P=0.004) state as well in the dissipative LEAK state (∼50% reduction; P=0.02) (Fig. 5B). Lower oxygen consumption by 12-month-old shIRF7 cells in the ROUTINE and LEAK states was accompanied by an increase of the ET coupling efficiency, expressed as 1−(LEAK/ET) (32% increase; P=0.038; Fig. 5C). These results indicate that IRF7 knockdown improves mitochondrial respiratory efficiency, probably due to a greater integrity of mitochondrial inner membranes and, thus, a reduced proton leak.

To dissect the impact of IRF7 knockdown on mitochondrial fitness in greater depth, we permeabilized 12-month-old shIRF7 and control cells to allow the entry of specific substrates and compared their mitochondrial respiration using HRR assays in the ROUTINE, LEAK, OXPHOS and ET respiratory states. To investigate the effects of IRF7 depletion on fatty acid- and NADH-dependent pathways we sequentially added octanoylcarnitine fatty acid (FA) and malate (ETF-linked); malate and glutamate (complex I-linked); and succinate substrates (complex II-linked). These conditions, following ADP stimulation, allow determination of the relative contribution to OXPHOS of β-oxidation, complex-I and complex-II, respectively. Qualitative analysis of FCRs allowed us to evaluate the relative contribution of the various complexes to mitochondrial respiration and highlighted significative qualitative differences between 12-month-old shIRF7 cells and control cells. Fatty acid-dependent pathways in the OXPHOS state (P_OM) were activated at significantly lower levels in 12-month-old shIRF7 cells compared to activation in 12-month-old control cells (19% lower; P=0.02; Fig. 5D). In addition, investigating the effect of addition of glutamate (G) in the OXPHOS respiratory state, using the control factor P_OMG−(P_OM/P_OMG), we observed a remarkable stimulation in 12-month-old shIRF7 cells compared to 12-month-old ctrl cells (80% increase; P=0.02; Fig. 5E), suggesting that IRF7 knockdown renders mitochondria more responsive to glutamate stimulation.

The amino acids depleted in 12-month-old cells were all partially restored upon IRF7 knockdown, with the exception of the BCAAs leucine and isoleucine (Fig. 6A). However, we did not find coordinated changes in the transcription of genes involved in either amino acid biosynthesis or degradation (Fig. 6B). This drives us to speculate that the rescue of the amino acid pool after IRF7 knockdown relies on alternative pathways. The fact that, unlike other amino acids, leucine and isoleucine were not restored suggests that IRF7 knockdown allows the BCAA degradation pathway to feed into the TCA cycle (anaplerosis) to sustain an accelerated flux of intermediates (oxalacetate) for amino acid biosynthesis (Fig. 6C). In line with this model, genes involved in BCAA degradation were drastically enriched in 12-month-old shIRF7 cells compared to control cells (Fig. 6D). Remarkably, we found a significant upregulation of the gene encoding PP2Cm (Ppm1K), the mitochondrially localized enzyme that activates the branched-chain α-keto acid dehydrogenase (BCKDH) complex, which is the rate-limiting enzyme of the BCAA degradation pathway, and loss of which has been reported to impair mitochondrial function, increase oxidative stress, and promote abnormal cardiac and neural development (Lu et al., 2007, 2009). We also found increased transcript levels of krueppel-like factor 15 (Klf15) (Fig. 6D), which has been previously shown to be a master inducer of BCAA degradation genes and crucial for prevention of mitochondrial dysfunction and oxidative stress in heart (Gray et al., 2007; Sun et al., 2016a,b). Also, IRF7 knockdown completely cleared the age-induced methylmaleic acid (citraconic acid), which is a by-product and established diagnostic biomarker of deranged isoleucine degradation (Duran et al., 1978) (Fig. 6E). BCAA degradation may feed the TCA cycle by providing succinyl-CoA, which is converted to succinate by succinate CoA-ligase GDP-forming beta subunit (SuclG2). Accordingly, succinate was more abundant in 12-month-old shIRF7 cells compared to levels in control cells, and SuclG2 expression was significantly upregulated (P=0.03; Fig. 6E). Oxalacetate did not accumulate in 12-month-old shIRF7 cells compared with levels in control cells, despite expression of the enzyme that generates it from malate, malate dehydrogenase 1 (encoded by Mdh1), being markedly upregulated (P=0.07) and expression of the enzyme that converts oxalacetate to citrate (citrate synthase, encoded by Cs) being downregulated in the 12-month-old shIRF7 cells (Fig. 6E). This suggests that the rate of oxalacetate generation equals the rate of its cataplerotic diversion toward extra-TCA cycle pathways. Accordingly, expression of the cytoplasmic and mitochondrial isoforms of the enzyme that converts oxalacetate to aspartate (glutamic oxalacetic transaminase-1 and -2, encoded by Got1 and Got2, respectively) were slightly but coordinately upregulated in shIRF7 cells. Finally, the vast majority of the enzymes involved in the TCA cycle were upregulated in 12-month-old shIRF7 cells (paired t-test; P=0.06), which supports an increased TCA cycle rate (Fig. 6C).

Confirming that IRF7 knockdown can drive the increase in TCA cycle rate, INTERFEROME analysis revealed that all the TCA cycle genes that were upregulated in 12-month-old shIRF7 cells are ISGs and, as such, might be regulated upon IFN treatment (Fig. S4A,B). The same applied for genes involved in BCAA degradation (Fig. S4C,D).

Collectively, our data indicate that aging induces the sterile activation of IRF7-mediated metabolic programming, impairing mitochondrial and amino acid biogenesis (Fig. S5). Since viruses rely on the bioenergetic and biosynthetic machinery of the host for their replication (Walsh and Mohr, 2011), it can be postulated that IFN signaling counteracts viral spread by shutting down ‘core’ metabolic pathways of the host and that an overwhelming response may lead to severe impairment of cell and organ functions.

The detrimental effects of aging manifest early in the course of adult life (Berry et al., 2017; D'Antona et al., 2010; López-Otín et al., 2013; Mori et al., 2012, 2014; Rogers et al., 2012; Shoji et al., 2016; Timmons et al., 2019). The molecular determinants of aging have recently been proposed to follow a non-linear progression rate, with a boost at middle age (midlife switch) (Timmons et al., 2019) that drastically impairs body fitness and increases vulnerability to metabolic, cardiovascular, locomotory and neurodegenerative disorders, among many others. As such, according to the ‘geroscience hypothesis’, the discovery of novel therapeutic strategies capable of preventing the general process of aging is a more straightforward challenge than fighting each specific age-related disorder individually (Franceschi et al., 2018a).

We show herein that the increased expression of IRF7 during aging leads to the sterile activation of cell-autonomous pathways of self-defense, which impairs crucial metabolic functions, such as mitochondrial and amino acid biogenesis and BCAA degradation. Our results contribute to draw a novel paradigm in cell biology, setting a causal link between cell-autonomous pathways of self-defense and mitochondrial-centered metabolic pathways, such as respiration, amino acid biosynthesis and BCAA degradation.

IFNs regulate the transcription of thousands of genes (ISGs) (MacMicking, 2012; Randow et al., 2013). Although the anti-viral effect of some of these genes is partially clear, the effect of their global repatterning on host cell biology is virtually unexplored. Nevertheless, the IFN-mediated response must be finely tuned, as its dysregulated activation may become detrimental (Baruch et al., 2014; Channappanavar et al., 2016; Davidson et al., 2014; Major et al., 2020; Teijaro et al., 2013). Our results drive us to speculate that one of the major aims of the IFN-mediated response is to shut down the bioenergetic and biosynthetic machinery of the host, which would be otherwise usurped by the virus for its own replication (Walsh and Mohr, 2011). The cost:benefit ratio for the host is never obvious, but the opportunity for the cell to embark on such a potentially harmful program might be justified when a real microbial attack occurs. What is less clear is why and how age might activate this program in the absence of viral infection.

Nevertheless, a generalized shutdown of crucial metabolic functions may help to explain many aspects of the pathophysiology of aging-related disorders. Mitochondrial and amino acid biogenesis are common outcomes of the ancient evolutionarily conserved programs referred to as the integrated stress response (ISR) (Harding et al., 2003; Pakos-Zebrucka et al., 2016). These pathways act constitutively as quality control mechanisms and are massively invoked to face multiple acute or chronic stress conditions, such as nutrient deprivation, an oxidative environment, proteotoxic stress and overloading of the endoplasmic reticulum (Harding et al., 2003; Quirós et al., 2017). The moderate stimulation of these pathways is the basis of the beneficial effects of mild stress interventions, such as dietary restriction, in preventing aging and even prolonging life span (Zid et al., 2009; Mazière et al., 1999). Mitochondria are both targets and master regulators of these stress response programs. Beyond their bioenergetic role, mitochondria have evolved as signaling organelles that instruct nuclear transcription, cytosolic protein translation and metabolic rewiring to coordinate cellular adaptation to environmental changes (Cardamone et al., 2018; Chandel, 2015; Quirós et al., 2017; Suhm et al., 2018; Wrobel et al., 2015). Hence, maintenance of proper mitochondrial activity is a priority challenge for cells and organs, and one that should be accomplished in favor of stress-induced adaptive rearrangements (Harding et al., 2003; Pakos-Zebrucka et al., 2016). The observation that mitochondria of 12-month-old shIRF7 cells showed a reduced OXPHOS capacity sustained by fatty acids suggests that these cells may accumulate lower levels of reactive oxidant species (ROS), which are generally accumulated during β-oxidation (Mazière et al., 1999). This condition could explain the improved quality of mitochondrial membranes, consistent both with the enhanced coupling efficiency and improved morphology of the mitochondrial membranes. We cannot say at this stage whether the increase in mitochondrial number and function might be the direct effect of IRF7 knockdown or just a consequence of a chain of events that result in a youthful phenotype. What is clear is that changes in IRF7 and IFN signaling with age might impact on cellular metabolism that involves mitochondria-centered pathways such as cellular respiration, amino acid biosynthesis and BCAA degradation.

BCAA degradation has been extensively implicated in aging and aging-related diseases. Specifically, an increased amount of circulating BCAAs and related metabolites is strongly associated with obesity, type 2 diabetes and heart disease (Felig et al., 1969; Newgard et al., 2009; Huffman et al., 2009; Li et al., 2017; Lynch and Adams, 2014; Turer et al., 2009). Conversely, increased expression of genes related to BCAA degradation correlates with insulin sensitivity and boosts fatty acid oxidation (Lerin et al., 2016; Papathanassiu et al., 2017). BCAA supplementation supports cardiac and skeletal muscle mitochondrial biogenesis, prevents oxidative damage and promotes exercise capacity in middle-aged animals (D'Antona et al., 2010).

Previous studies have explored the reciprocal causal connections between mitochondria and inflammation, and their possible role in aging. Recent studies have shown that mitochondrial disruption may trigger inflammatory reactions (Garaude et al., 2016; Jin et al., 2014; Zhang et al., 2010), whereas pro-inflammatory mediators may contribute, in turn, to alteration of mitochondrial activity (López-Armada et al., 2006). However, mechanisms modeled in these studies implicate the systemic recruitment of immune cells (neutrophils and macrophages) (Franceschi et al., 2017; Jin et al., 2014), circulating pro-inflammatory cytokines (TNF, IL6 and IL1) (López-Armada et al., 2006; van Horssen et al., 2017), cell debris (Zhang et al., 2010) or DNA-damaging chemical mediators (such as reactive oxygen species and reactive nitrogen species) (Garaude et al., 2016; Kim et al., 2010) and have ultimately contributed to establishing the prevailing notion that aging-related deterioration of tissue function and mitochondrial integrity relies on a complex bi-directional interaction between systemic immunity and tissue-resident cells. We provide herein an additional, cell-intrinsic perspective that envisions the age-related mitochondrial dysfunction of non-immune cells as the consequence of ‘cell-autonomous’ alterations of immunity-related signaling pathways. IRF7 functions as the molecular trigger of these pathways during aging. Whether the increased expression of IRF7 with age is primed by extrinsic or intrinsic factors is not clear. Nevertheless, its impact on the transcription of immune and mitochondrial genes is cell autonomous (as it persists in cultured cells over multiple passages) and reversible, thus ruling out the involvement of the genetic damage eventually induced by ROS and NOS, which has previously been implicated in connecting inflammatory responses with mitochondrial integrity. The role of IRF7 as a regulator of immune cell function has been extensively investigated by thousands of studies. In sharp contrast, its expression and function in non-immune cells remains unexplored. Our data demonstrate that the role of IRF7 extends beyond the protection against external intruders and that its gain of function may indeed represent an intrinsic paradigm for the degeneration that occurs during adulthood. Our results suggest that IRF7 is an ideal target for the development of therapies that are potentially effective against multiple and intersecting aspects of the aging phenotype.

The inguinal fat pads of FVB male mice aged 1 (n=8) and 12 (n=8) months were surgically excised, weighed and processed to isolate the adipose-derived stromal cells, according to standard protocols (Peroni et al., 2008). Briefly, the extracellular matrix was digested with collagenase (1 mg/ml) and centrifuged to obtain a high-density pellet, the stromal vascular fraction (SVF). After centrifugation, the SVF cell number was determined using light microscopy, and the cells were plated at a concentration of 10⁵ cells/cm² in DMEM medium (GIBCO) with high glucose concentration (GLUTAMAX I, GIBCO; 10% FCS, GIBCO; 100 U/ml penicillin; and 100 μg/ml streptomycin). After 2–3 weeks of culture, a homogeneous cell population was obtained. The cells were identified as MSCs on the basis of their immunophenotype. Specifically, the positivity of CD106 (VCAM1), CD73 (NT5E), CD29 (ITGB1), CD44, CD90 (THY1) and the lack of hematopoietic (as detected using CD45, CD14, CD11c, CD123 and CD34 monoclonal antibodies) and endothelial cell markers (as detected using CD31 monoclonal antibody) were assessed by means of cytofluorimetric analysis. Animals were handled in accordance with the regulations of the Italian Ministry of Health and the European Communities Council (86/609/EEC) directives.

Whole-genome microarray analysis of MSCs from differently aged mice was performed using the NimbleGen Gene Expression system. Briefly, total RNA was isolated from samples using the Qiagen RNeasy kit (Qiagen), following the manufacturer's instructions. RNA was used for cDNA synthesis, followed by labeling of the cDNA with Cy3. The labeled cDNA samples were hybridized to the Mus musculus 12×135K Array (Roche NimbleGen), which represents 44,170 mouse genes. The single color NimbleGen arrays were scanned using a GenePix 4400A Microarray Scanner. The data were extracted from scanned images using NimbleScan software, and the Robust Multichip Average (RMA) algorithm was used to generate gene expression values. Hybridization, scanning and normalization of the data were performed as a service by the Functional Genomic Center of the University of Verona (Verona, Italy). Whole-genome microarray of shIRF7 12-month-old MSCs was performed as service by the UTSW Genomics Sequencing & Microarray Core Facility (Dallas, TX, USA) on an Affymetrix platform (Clariom S Mouse Array).

Immunoblots were performed according to standard procedures in RIPA buffer (150 mM NaCl, 10 mM Tris pH 7.5, 1% NP40, 1% deoxycholate and 0.1% SDS) supplemented with phosphatase and protease inhibitors (Sigma-Aldrich). Samples were resolved on Tris-glycine 4–20% gradient SDS–PAGE gels (BIO-RAD), blotted on Protran 0.2 μm membrane (Whatman), incubated with anti-IRF7 (EPR4718; 1:500; ab109255; Abcam), anti-β-actin (Abcam, ab6276; 1:1000) or anti-GAPDH (Covance, #MMS-580S; 1:1000) antibody and developed with ECL (Amersham).

RNA was extracted using the Qiagen RNAeasy Plus mini kit. cDNA was generated by Superscript II (Invitrogen) and used with SYBR Green PCR master mix (Applied Biosystems) for real-time qPCR analysis. Assays were performed using an Applied Biosystems Step-One Real-Time PCR System. Primers used for qPCR amplification are detailed in Table S1.

Samples were fixed using 2% glutaraldehyde in Sorensen buffer pH 7.4 for 2 h, post-fixed in 1% osmium tetroxide in an aqueous solution for 2 h, dehydrated in graded concentrations of acetone, embedded in Epon-Araldite (Società Italiana Chimici) and cut with an Ultracut E Ultra-microtome (Reichert, Wien, Austria). At the end of the dehydration process, samples were positioned in a multi-well grid for electron microscopy and observed using a TEM Morgagni 268D (FEI Philips). Quantification of mitochondrial size was performed using ImageJ software (NIH, Bethesda, MD) on the images taken at the same magnification from 20 randomly selected fields.

IRF7 was silenced in 12-month-old aged MSCs by lentiviral transduction. Four different shIRF7 shRNAs in the pLKO.1 vector were used, named #1 (Dharmacon; RMM3981-201795252), #2 (Dharmacon; RMM3981-201798737), #3 (Dharmacon; RMM3981-201788865) and #4 (Dharmacon; RMM3981-201789942). Recombinant lentiviruses were generated in 293T cells transfected using the second-generation packaging vectors psPAX2 (Addgene #12260) and pMD2.G (Addgene #12259) using Lentifectin transfection reagent (ABM Good). Stable shIRF7 cells were selected with 1 mg/ml puromycin. 12-month-old MSCs transduced with empty (Addgene #10878) or shRNA scramble pLKO.1 (Addgene #1864) vectors were used as controls.

Cells were trypsinized, washed with phosphate-buffered saline (PBS) solution and centrifuged. Supernatants were removed, and cells were resuspended in PBS. Aliquots containing 10⁶ cells were used for the metabolite extraction. Metabolites were extracted using 760 μl of a precooled MeOH:water mixture (1:0.9, v/v). Extracts were vortexed until complete dissolution of the pellet, and 400 μl of ice-cold chloroform was added. Cell extracts were then homogenized with a cell disrupter at 30 Hz for 10 min, then centrifuged at 2200 g for 5 min at 4°C. After layer separation, the polar and the nonpolar fractions were transferred to new precooled tubes. The aliquots were then evaporated to dryness using a nitrogen flow and stored at −80°C until analysis. Hexadecanoic acid as internal standard was added to the dried extract. The samples were first oximated by addition of 30 μl methoxyamine hydrochloride solution in pyridine (20 mg/ml), mixed in a vortex mixer and subsequently shaken for 90 min at 30°C. Afterwards, 30 µl of N,O-Bis(trimethylsilyl)trifluoroacetamide (BSTFA) with 1% trimethylchlorosilane (TMCS) was added, and the derivatization was performed at 70°C for 60 min prior to gas chromatography–time of flight mass spectrometry (GC–TOF/MS) analysis. An alkane standard mixture (C8–C20; Merck KGaA) was added as internal standard.

GC–TOF/MS was performed using an Agilent 7890B GC (Agilent Technologies, USA) and Pegasus (BT) TOF/MS system (Leco Corporation, USA) equipped with an Rxi-5ms column (30 m×0.25 mm×0.25 μm; RESTEK, USA), stationary phase 5% diphenyl–95% dimethyl polysiloxane. High-purity helium (99.999%) was used as the carrier gas at a flow rate of 1.20 ml/min⁻¹. Samples were injected in splitless mode at 250°C. The chromatographic conditions were: initial temperature 70°C, 2 min isothermal, 6°C/min up to 160°C, 10°C/min up to 240°C, 20°C/min up to 300°C, 6 min isothermal. MS parameters: electron impact ionization source temperature (EI, 70 eV) was set at 250°C; scan range 40/630 m/z, with an extraction frequency of 30 kHz. The chromatograms were acquired in TIC (total ion current) mode. Mass spectral assignment was performed using the ChromaTOF BT software (Leco Corporation, USA) by matching with NIST MS Search 2.2. Libraries implemented with the MoNa Fiehns Libraries. Statistical and pathway studies were carried out using the MetaboAnalyst 4.0 tool (www.metaboanalyst.ca/).

To analyze mitochondrial respiration in MSCs (including those transfected with either control or shIRF7 vectors) we used an Oxygraph-2K (Oroboros Instruments, Innsbruck). Instrumental and chemical background fluxes were opportunely calibrated as a function of oxygen concentration using DatLab software (Oroboros Instruments, Innsbruck). All the measures were performed comparing 12-month-old control and 12-month-old shIRF7 cells in parallel in the same experiment. Mitochondrial respiration was evaluated both in intact and permeabilized cells in different respiratory states (ROUTINE, LEAK, OXPHOS and ET) (Calabria et al., 2019; Erich, 2020).

Cells resuspended in DMEM (250,000 cells/ml) were analyzed applying a coupling control protocol (Pesta and Gnaiger, 2012) to determine the rate of oxygen consumption in the various respiratory states: ROUTINE (R), LEAK respiration (L), electron transfer capacity (ET) and residual oxygen consumption (ROX) (Pesta and Gnaiger, 2012). After stabilization of the rate of oxygen consumption in the ROUTINE state, respiration was supported with addition of pyruvate and malate (5 mM and 2 mM, respectively; Sigma-Aldrich), and the ATP-synthase inhibitor oligomycin (Omy; 2 µg/ml; Sigma-Aldrich) was added to obtain a measure of LEAK respiration. After this step followed the titration of CCCP (Sigma-Aldrich, #C-2759) to obtain maximum oxygen flux (ET capacity). Finally, rotenone (Rot) and antimycin A (Ama) were added to specifically inhibit complex I and complex III, respectively, in order to obtain the ROX.

The respiration of permeabilized cells was determined using a modified substrate–uncoupler–inhibitor titration (SUIT) protocol (Pesta and Gnaiger, 2012). To investigate the contribution of different mitochondrial complexes (I–IV) to respiratory capacity, cells were permeabilized with the mild detergent digitonin (12 µg/10⁻⁶ cells; Sigma-Aldrich), this procedure allows specific substrates and adenylates to enter the cells and reach mitochondria. The optimal concentration of digitonin was evaluated in preliminary experiments as that suitable to achieve full permeabilization of cells allowing the access of substrates and ADP to mitochondria without compromising mitochondrial function. To evaluate the contribute of fatty acids to mitochondrial activity, octanoylcarnitine and malate (OM; 0.8 mM and 2 mM, respectively) were added in the ROUTINE state (R_OM). Following permeabilization with digitonin, the OXHPOS capacity (P_OM) was induced by adding ADP (5 mM; Sigma-Aldrich), and addition of glutamate (G; 10 mM) was used to determine complex I activity (P_OMG). To also activate complex II in OXPHOS, succinate (S) was added (P_OMGS; 10 mM succinate; Sigma-Aldrich). The presence of all these substrates (OMGS) allows detection of the respiratory activity of linked complex I, ETF and complex II. The contribution of complexes III and IV to respiratory activity is always present, although they were not stimulated by the addition of specific substrates.

To analyze the electron transfer system (ET_OMGS) capacity, stepwise titration with the uncoupler CCCP (0.5 µM steps; Sigma-Aldrich) was performed. Rot and Ama (2 and 2.5 µM, respectively; Sigma-Aldrich) were added to inhibit complex I (ET_S) and III, determining the ROX. Raw data were analyzed using the DatLab 6 Program (Oroboros Instruments).

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.256230