To characterize the MSR induced by Tets in vivo, we administered Dox at 500 mg/kg/day (mpkd) in the drinking water to 9-week-old germ-free C57BL/6J mice for 16 days (5, 7), hence eliminating the potential confounding impacts of Dox on the microbiome. Body weight at the time of the sacrifice was not different between the control and Dox-treated animals (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/JCI151540DS1), indicating the absence of obvious adverse effects. As reported in livers of mice maintained under conventional conditions (9), oxidative phosphorylation (OXPHOS) complex activity as well as ATP levels were also reduced by Dox in kidneys of germ-free mice (Figure 1A). Dox elicited organ-specific transcriptional responses, with the expression of quantitatively more and qualitatively different genes being affected in the kidney compared with the liver (Supplemental Figure 1, B and C, and Supplemental Table 1). In Dox-treated kidneys, gene set enrichment analysis (GSEA) (20) revealed the induction of the ATF4-mediated integrated stress response (ISR) (Figure 1B), a common hallmark of the mammalian MSR (21). The MSR features the induction of mitochondrial chaperones and proteases, such as HSPA9 and LONP1, as well as of enzymes mediating adaptation to nutrient deprivation, such as asparagine synthetase (ASNS), which were increased at both the transcript and protein levels (Figure 1, C and D, and Supplemental Figure 1D). In line with the activation of the ATF4/ISR pathway, the kidney displayed increased eIF2α phosphorylation (Figure 1E and Supplemental Figure 1E), which slows down cytosolic cap-dependent translation as a compensation for energy deprivation caused by mitochondrial stress and favors the translation of ATF4 transcripts by cap-independent mechanisms (22). Kidneys of the Dox-treated germ-free mice thus displayed the typical attributes of the ATF4/ISR pathway, a hallmark of the mammalian response to mitochondrial stress.

Doxycycline induces the ATF4 response and the type I IFN response. (A) Biochemical measurement of oxidative phosphorylation (OXPHOS) complexes (CI–CV), citrate synthase (CS), and ATP levels in the kidney of germ-free C57BL/6J male mice raised and maintained in a germ-free environment and that were drinking regular water or water supplemented with doxycycline (Dox) at 500 mg/kg/day (mpkd) for 16 days (n = 4–5). (B and C) Enrichment score plot for the gene set “Reactome Activation of genes by ATF4” (B) and heatmap representing the transcript levels of ATF4/5 targets (C) from kidney transcriptomics data of control versus Dox-treated germ-free mice. (D) Western blot analysis of selected ATF4 targets in the kidneys of germ-free mice (corresponding loading control below, HSP90). (E) Immunoblots of phosphorylated EIF2α (p-EIF2α) and total EIF2α in kidneys of Dox-treated germ-free mice. (F and G) Enrichment score plot for the GO term “Response to type I interferon” (F) and heatmap representing the transcript levels of some IFN-stimulated genes (ISGs) (G) from livers of germ-free mice treated with Dox. (H) Immunoblots of phosphorylated TBK1 (p-TBK1), TBK1, and the ISG proteins CGAS and CXCL10 (corresponding loading control below, vinculin and GAPDH, respectively). (I) Transcript levels of selected ISGs of bone marrow–derived macrophages (BMDMs) (day 6 of differentiation, derived from C57BL/6J mice) treated with Dox at 30 μg/mL for 9 hours (n = 6). (J) Immunoblots of phosphorylated TBK1 (p-TBK1), TBK1, and vinculin as control in BMDMs treated with Dox at 30 μg/mL for 3 hours. (K) Amplification of different mtDNA regions by qPCR in the cytosolic fraction of BMDMs with Dox at 30 μg/mL for 1 hour (n = 10). (L) Levels of IFN-β in the culture medium of BMDMs treated with Dox (30 μg/mL for 14 hours) and/or 2′,3′-dideoxycytidine (ddC, at 100 μM for 72 hours) (n = 8). Statistical analysis: Wilcoxon’s test P values corrected for multiple comparisons with Hommel’s method (A, I, and K) or by 1-way ANOVA followed by Tukey’s post hoc correction (L). *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001. NS, P > 0.05. Error bars represent ±SEM.

In the liver, eIF2α phosphorylation and the ATF4/ISR program were not induced (Supplemental Figure 1, F and G, and Supplemental Table 2). The liver transcriptome, however, indicated that Dox induced the type I IFN response (Figure 1, F and G), which was confirmed at the protein level by the increased expression of 2 IFN-stimulated genes (ISGs), cyclic AMP-GMP synthetase (CGAS) and C-X-C motif chemokine ligand 10 (CXCL10), and the increased phosphorylation of TANK-binding kinase 1 (TBK1) (Figure 1H and Supplemental Figure 1H). The type I IFN response is an innate immune pathway that, upon sensing viral DNAs, activates cGAS/STING/TBK1 signaling, culminating in the secretion of the type I IFNs, IFN-α and IFN-β, and the induction of the expression of ISGs (23). Of note, the type I IFN response was also induced in kidneys, but to a lesser extent (Supplemental Figure 1, I and J, and Supplemental Table 2).

Similarly, in mouse bone marrow–derived macrophages (BMDMs), a highly relevant model to investigate innate immune signaling in vitro, Dox induced the expression of ISGs and triggered the phosphorylation of TBK1 (Figure 1, I and J, and Supplemental Figure 1K). Mitochondrial DNA (mtDNA) effusing from the mitochondria into the cytosol was previously shown to elicit the type I IFN response in the context of mitonuclear genomic instability caused by the loss of function of transcription factor A mitochondrial (TFAM) (12). We also detected increased levels of mtDNA in the cytosol of Dox-treated BMDMs (Figure 1K), which underpins the activation of antiviral signaling resulting in the secretion of IFN-β from these BMDMs (Figure 1L). Finally, Dox-induced secretion of IFN-β was abrogated by the nucleoside analogue, 2′,3′-dideoxycytidine (ddC) (Figure 1L), which gradually leads to depletion of mtDNA (24), demonstrating that cytosolic release of mtDNA contributes to the activation of antiviral signaling.

We then set out to identify Tets with lower antimicrobial activities that could be more easily developed for clinical use. To this end, we screened in C. elegans a library of 52 position-modified Tet derivatives (Supplemental Table 3) that are clinically used, are synthetic intermediates, or derivatives specifically synthesized to probe initial structure-activity relationships among Tets that elicit the MSR (Figure 2, A–C), most of them having very limited antibacterial activity (Table 1). We screened the compounds for induction of the UPRmt using a C. elegans hsp-6:gfp reporter strain with Dox administration and cco-1 RNAi feeding as positive controls, as previously described (25) (Figure 2C). Out of the 52 Tet derivatives tested (Supplemental Table 3), representing clinically relevant and C2–C10 position–modified compounds (26–32) (Figure 2, A and B), we identified 9-TB and anhydrotetracycline 2 (ATc) as the strongest activators of the UPRmt (Figure 2C). We then compared detailed dose responses of Dox, 9-TB, and ATc to induce a GFP signal in the C. elegans hsp-6:gfp reporter strain in an automated microfluidic device (33). 9-TB and ATc were again in this system more efficacious at lower doses to induce the UPRmt relative to Dox (Figure 2D and Supplemental Figure 2A), with 9-TB surpassing the robust UPRmt activation caused by cco-1 RNAi feeding (Figure 2, C and D).

Selecting Tet derivatives that induce UPRmt in C. elegans. (A) Structural locants of the Tet scaffold and UPRmt-active and -inactive compounds by chemically modified positions (based on activity of the hsp-6:gfp reporter, C). (B) Chemical structures of the Tet derivatives shown in A and C. (C) Representative images of the induction of the UPRmt in the C. elegans hsp-6:gfp reporter strain (25) exposed to the indicated Tet derivatives at 68 μM (except for 9-TB, which is at 17 μM) since the parental L4 stage. Dox and OXPHOS loss of function through feeding cco-1 RNAi serve as positive controls. The pictures show the progeny at day 2–3 of adulthood (similar exposure time for all images; GFP fluorescence in top raw, differential interference contrast [DIC] in bottom raw). (D) Dose-response for the UPRmt activation (hsp-6:gfp reporter strain) upon exposure to different concentrations of Dox, 9-TB, ATc, or treatment with cco-1 RNAi using an automated microfluidic device (33) (n = 14–16). Statistical analysis was performed by 1-way ANOVA followed by Bonferroni’s post hoc correction. ***P ≤ 0.001. Error bars represent ±standard deviation (±SD).

Minimum inhibitory concentrations (μg/mL) of bacterial growth for indicated Tet derivatives and bacterial strains

In addition, other derivatives modified along the upper periphery spanning positions C2, C4, C5, C6, C13, and aromatic positions C7–C9 also activated the UPRmt, such as compounds 3–5, although their effect was not as pronounced as that of 9-TB and ATc (Figure 2, A–C). In contrast, clinically used minocycline 7, Nuzyra 14 (see Supplemental Table 3), Tygacil 23, or derivatives based on the minocycline scaffold did not activate the UPRmt (16 compounds), while C5–C9 derivatives of sancycline only mildly activated the UPRmt (see Supplemental Table 3). Additionally, compounds modified at the lower periphery, spanning positions C10, C11, C12-C1, and the A-ring C2 carboxamide did not induce the activity of the GFP reporter, showing the importance of this integrated phenolic keto-enol system in maintaining UPRmt activity (34, 35).

We then characterized the pharmacology of Dox, 9-TB, and ATc in the human embryonic kidney (HEK293T) cell line. 9-TB and ATc also generated a more robust MSR response than Dox, as reflected by their impact (up to almost 2-fold stronger) on the mitonuclear protein imbalance (Figure 3A), an imbalanced ratio between mitochondrial and nuclear encoded OXPHOS subunits, underpinning the induction of the MSR (5). Furthermore, 9-TB and ATc reduced the basal oxygen consumption rate (OCR) in a dose-dependent and more pronounced fashion than Dox (Figure 3B). The induction of transcripts for the mammalian MSR signature genes was likewise more prominent with 9-TB and ATc (Figure 3C). In mouse BMDMs, lower doses of 9-TB (1.88 μg/mL) and ATc (3.75 and 7.5 μg/mL) were also superior to Dox (at 7.5 and 15 μg/mL) in inducing the ISG and MSR genes (Figure 3D and Supplemental Figure 3A) and the secretion of IFN-β (Figure 3E). Knocking out ATF4 in mouse embryonic fibroblasts (MEFs) showed that Tets induced the MSR and most ISG genes in an ATF4-dependent manner (Supplemental Figure 3B). Taken together, these studies in C. elegans, mouse BMDMs, human HEK293T cells, and MEFs ascertained the identification of non-antimicrobial Tets with higher potency to trigger the MSR and type I IFN response, relative to our benchmark antibacterial Tet, Dox.

Tet derivatives induce the MSR and type I IFN signalling in mammalian cells. (A–C) Tet derivatives induce a mitochondrial/nuclear protein imbalance and the MSR in HEK293T cells (human) treated for 24 hours at the indicated concentrations. (A) Immunoblots of HEK293T cells for the OXPHOS subunits ATP5A (encoded in nuclear DNA) and MTCO1 (encoded in mtDNA) with tubulin serving as a control. Quantification of the relative MTCO1/ATP5A ratio is shown on the right. (B) Oxygen consumption rate of HEK293T cells exposed to different concentrations of Dox, 9-TB, or ATc (n = 8). (C) Transcript levels of the indicated MSR genes measured by RT-qPCR (n = 4). (D and E) Tet derivatives induce transcript levels of the indicated ISGs (D) and stimulate IFN-β secretion (E) after 24 hours of treatment at the indicated concentrations in mouse BMDMs (day 6 differentiation) (n = 4). Statistical analysis was performed by 1-way ANOVA (B, D, and E) or 2-way ANOVA (C) followed by Tukey’s post hoc test . *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001. Error bars represent ±SEM.

mtDNA instability–driven innate immunity can potentiate resistance to viruses (12) and mediates the antiviral immune response against the IFV (36). We thus asked whether the Tet-induced MSR enables mice to survive infection by IFV. We hence subjected 8-week-old female BALB/cN mice to either mock (1 group, n = 10) or intranasal inoculation with 175 PFU of the IFV H1N1 PR8 strain (3 groups). The 3 infected groups (n = 10 each) were given vehicle, Dox (at 40 mpkd), or 9-TB (at 1 mpkd) by daily intraperitoneal injection, from preinoculation day –3 (Supplemental Figure 4A). Dox and 9-TB improved the survival to the infectious challenge, with 50% of the mice treated with Dox or 9-TB recovering (Figure 4A). The improved health of the Tet-treated cohorts was further supported by the recovery of body weight loss, and their improved clinical score (Figure 4B and Supplemental Figure 4C). In contrast, the IFV infection was lethal to all control mice by day 11 after inoculation. Strikingly, on day 7 after infection no significant difference in viral titer in the lung tissue was observed (Figure 4C). Similarly, when mice were infected with a much higher viral load (1000 PFU; Supplemental Figure 3B), Dox and 9-TB delayed mortality and the decline in health (Figure 4, E and F, and Supplemental Figure 4E), again in the absence of an impact on the viral titer in the lungs on day 5 after infection (Figure 4G). The Tet-induced MSR did not cause obvious adverse effects (Supplemental Figure 4D), yet decreased the levels of interleukin 6 (IL-6) in both 175 and 1000 PFU experiments (Figure 4, D and H), and of some other markers of tissue stress and damage (Supplemental Figure 4F) (37–39). These results demonstrate that the Tet-induced MSR increases the survival of mice to a lethal IFV infection by improving tolerance, rather than by reducing viral load, which is reflective of resistance to the virus.

Tets mediate disease tolerance to IFV in mice. (A–D) Eight-week-old BALB/cN mice were injected with Dox (40 mpkd) or 9-TB (1 mpkd) and intranasally infected with 175 PFU of IFV H1N1 PR8, as described in Supplemental Figure 4A. Survival (A) and clinical score (B) were followed for 16 days after infection (n = 10). On day 7 after infection, viral titers in lung lysates (C, n = 5) and IL-6 levels in plasma (D, n = 6) were measured (n = 5). (E–H) Eight-week-old BALB/cN mice were injected with Dox (50 mpkd) or 9-TB (12.5 mpkd) and intranasally infected with 1000 PFU of IFV H1N1 PR8, as shown in Supplemental Figure 4B. Survival (E) and clinical score (F) were followed over 10 days after infection (n = 10). On day 5 after infection, viral titers in lung lysates (G) and IL-6 levels in plasma (H) were measured (n = 5). Dashed horizontal lines in C and G indicate the lower limit of detection (LLD). Statistical analysis was performed by 1-way ANOVA followed by Tukey’s post hoc test. For survival curves in A and E, statistical analysis was performed by log-rank (Mantel-Cox) test. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001. NS, P > 0.05. Error bars represent ±SEM.

To assess the impact of Tets on the microbiome we transiently individually caged animals and longitudinally collected feces before (day –4, before IFV inoculation), 3 days after (day 0, just before IFV inoculation), and 6 days after (day 3, after IFV inoculation) the start of daily administration of 9-TB or Dox. We then extracted DNA from feces and performed whole-metagenome sequencing. While the composition and diversity of the bacterial communities showed no differences between groups before treatment, the gut bacterial community of mice treated with Dox showed a significant difference in composition compared with both untreated mice and mice treated with 9-TB after 3 and 6 days of Tet treatment (respectively day 0 and day 3 after inoculation), as assessed by permutational multivariate analysis of variance (perMANOVA) and visualized by nonmetric multidimensional scaling (NMDS) (Figure 5A and Supplemental Table 8). This was reflected by a lower bacterial species diversity in Dox-treated mice in terms of both Shannon diversity index (SDI) and richness (Figure 5B). In contrast, no differences were observed between 9-TB–treated mice and untreated mice at any time point, suggesting that that the administered dose of 9-TB does not affect the mice gut microbiota in vivo (Figure 5, A and B, and Supplemental Table 8).

9-TB does not impact gut microbiome and shows encouraging effects when therapeutically administered. (A) Comparison of bacterial community composition by nonmetric multidimensional scaling (NMDS) based on the Bray-Curtis dissimilarity. (B) Comparison of bacterial species diversity in terms of Shannon diversity index (SDI) and richness. The lower and upper hinges are the first and third quartiles. The middle line is the median. The upper and lower whiskers respectively represent the highest and lowest values that are within 1.5× IQR from the hinge, where IQR is the interquartile range (i.e., distance between the first and third quartile). Data points beyond whiskers are considered outliers. Statistical significance assessed by Kruskal-Wallis test and post hoc Wilcoxon’s test with P values adjusted for multiple comparison using the Holm-Bonferroni method. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. NS, P > 0.05. (C and D) Eight-week-old BALB/cN mice (n = 12) were infected intranasally with 760 PFU of IFV H1N1 PR8 and injected with 9-TB (0.05, 0.025 mpkd), as described in Supplemental Figure 5A. Survival (C) and clinical score (D) were followed for 14 days after infection (n = 12). Statistical analysis was performed by log-rank (Mantel-Cox) test (C) or 1-way ANOVA followed by Tukey’s post hoc test (D). Error bars represent ±SEM.

To further investigate the clinical relevance of the Tet derivatives we focused on 9-TB and administered 9-TB in a therapeutic mode starting on day 1 after inoculation with 760 PFU of the IFV H1N1 PR8 strain (Supplemental Figure 5A). Of note, the different death kinetics and survival proportion with regard to the high viral load (760 PFU) in Figure 5, C and D and Supplemental Figure 5, A–D in comparison with Figure 4, A and E at lower viral load (175 PFU) are due to the fact that both experiments were run with different viral batches. Nevertheless, although not significant, the trend of 20% survival upon administration of 2 very low doses of 9-TB (0.025 and 0.05 mpkd) are highly encouraging (Figure 4, C and D) and suggest that these Tet derivatives can trigger tolerance to IFV in a clinically relevant setting. Future investigations are thus needed to optimize the timing and doses of Tet derivatives and refine their therapeutic potential in viral infections.

To gain insight into the mechanisms underlying the Tet-induced disease tolerance, we analyzed the transcriptome of the lung, as well as that of the liver and kidney (Supplemental Figure 6A), 2 organs often affected by the multiorgan failure syndrome seen after infection by respiratory viruses like IFV or SARS-CoV-2 (40). In each tissue, principal component analysis (PCA) separated noninfected from IFV-infected mice along the first dimension, PC1, whereas 9-TB had a more pronounced and less variable effect, with better clustering and further separation, along the second dimension relative to control and Dox transcriptomes (Figure 6A and Supplemental Figure 6B).

9-TB counteracts the inflammatory and lung-damaging effects of IFV infection. (A) Principal component analysis (PCA) of lung RNA-Seq transcriptomes collected on day 7 after infection of BALB/cN mice with 175 PFU IFV H1N1 PR8 (n = 5–6). (B) Gene set enrichment analysis (GSEA) results for Gene Ontology (GO) gene sets modulated in the comparison between 9-TB–treated versus control IFV-infected mice. A positive normalized enrichment score (NES) corresponds to an overall upregulation, while a negative NES indicates downregulation, of the corresponding gene set. (C) Revigo plot summarizing the main themes in the significantly enriched GO Biological Process (GOBP) sets among genes induced by IFV infection and downregulated by 9-TB (left panel), and genes downregulated by IFV infection and induced by 9-TB (right panel). The size of the bubbles (top right legend) is proportional to the number of annotations for the GO term (i.e., frequency) in the GO annotation database, with more general terms displaying larger bubbles. (D) GSEA results of the RNA-Seq data showing the directionality (increase or decrease) of the modulated lung cell transcript profiles based on common markers shared by both human and mouse lung cell types derived from extant single-cell transcriptomic data (43). The α value (transparency) represents the –log10(adjusted P value) of the enrichment. *Adjusted P < 0.05.

GSEA showed that 9-TB significantly downregulated multiple inflammatory and immune-related terms in the lungs (Figure 6B and Supplemental Figure 6D), such as “immune response,” “T cell activation,” or “B cell activation.” We then sought to characterize how 9-TB reversed the effect of IFV infection on transcript levels (Supplemental Figure 6A). We thus assessed which Gene Ontology (GO) Biological Processes (GOPB) terms were enriched in the intersection of gene sets changed in opposite directions by infection and 9-TB, respectively (Supplemental Figure 6C). As summarized by Revigo representation (41) (Supplemental Figure 6A), inflammatory, immune, and apoptotic processes were the main enriched terms among genes induced by IFV and downregulated by 9-TB (Figure 6C). Infection by IFV leads to lung epithelial cell dysfunction and downregulation of genes implicated in cilia or tight junctions, which underpin failures of mucociliary clearance and barrier function that contribute to the pathogenesis of ARDS (19). Accordingly, multiple gene sets related to lung development and to lung cell function and structure were decreased by IFV infection and their expression was restored by 9-TB (Figure 6B, Supplemental Figure 6E, and Figure 6C). Altogether, the results show that 9-TB elicits disease tolerance to IFV mainly by counteracting inflammation and the loss of lung epithelial cells and structures, processes that directly determine the severity of infection by respiratory viruses.

To estimate the impact of the infectious challenge or Tet treatment on the lung cell types, we used single-cell RNA sequencing (scRNA-Seq) transcriptomic profiles of mouse (42) and human (43) lung cell populations from 2 independent studies studying IFV infection; one of these studies furthermore established cell markers that overlap between mouse and human lung cell types (43). Using these cell profiles to perform GSEA on our data confirmed that 9-TB reverted the loss of multiple cell types crucial to lung function, such as club, ciliated, and alveolar epithelial cells (Figure 6D), whereas it decreased several classes of immune cells, such as neutrophils, natural killer cells, and monocytes, all contributing to tissue damage upon IFV infection (17). Dox showed similar, but more discrete, tendencies toward changes in cellular patterns (Figure 6D). These observations were confirmed when using a different set of scRNA-Seq profiles of IFV-infected mouse lungs (Supplemental Figure 7A) (42).

9-TB also downregulated multiple immune-related and inflammatory gene sets in liver and kidney (Supplemental Figure 7B). In particular, as shown through Revigo analysis, these immune and inflammatory terms were enriched among the group of liver genes induced by IFV and downregulated by 9-TB (Supplemental Figure 7, C and D). In the 3 organs studied, Dox led to a weaker downregulation of many of these terms, as shown by GSEA (Supplemental Figure 7B), suggesting that it does not lower systemic IFV-driven inflammation as efficiently as 9-TB, which is also consistent with its more moderate impact on IL-6 plasma levels (Figure 4D). Furthermore, Dox induced gene sets involved in cytopathic processes and fibrogenesis in liver and kidney (Supplemental Figure 7B), suggesting an improved safety profile of 9-TB, relative to Dox, at doses showing similar efficacy. Further investigations will be needed to establish whether the relative upregulation of extracellular matrix/collagen gene sets by 9-TB in lungs (i.e., the site of highest tissue damage due to the infectious challenge) corresponds to proper healing and tissue repair mechanisms (Supplemental Figure 7). Taken together, our transcriptomic data highlight that Tet-induced mitohormesis (and in particular 9-TB) elicits disease tolerance to IFV by preventing IFV-associated lung damage and by dampening inflammatory responses not only in lungs, but in liver and kidney as well.