IFNγ and IFNγ mimetics prevent IFN-I-mediated TB susceptibility by regulating iron metabolism and lipid peroxidation

Reviewer #1 (Public review):

Summary:

This study examines how type I IFN and IFN-γ exert opposing effects on macrophage responses relevant to TB. Using bone marrow-derived macrophages from genetically susceptible B6.Sst1S mice, the authors describe a persistent pathological activation state induced by TNF and characterized by sustained type I IFN signalling, oxidative stress and lipid peroxidation. They show that IFN-γ priming limits several features of this state and propose altered iron metabolism as one mechanism underlying this protective effect. They then use a computational cell-state approach to identify pharmacological interventions that may mimic aspects of IFN-γ activity. In particular, CDK4/6 inhibition with trilaciclib and activation of retinoic acid signalling with ATRA appear to act through complementary mechanisms and, when combined at low concentrations, improve control of intracellular M. tuberculosis.

Strengths:

A major strength of the study is the combination of several complementary approaches, including genetic susceptibility, cytokine signalling, oxidative stress, iron and lipid metabolism, transcriptomics, computational modelling and pharmacological perturbation. Together, these experiments build a coherent model of macrophage dysfunction.

The evidence that type I IFN signalling contributes to maintenance of the pathological state is particularly convincing within the TNF stimulation model. Blocking the type I IFN receptor after the phenotype has developed restores responsiveness to IFN-γ and prevents further accumulation of lipid-peroxidation products. The authors also provide evidence that persistence does not simply reflect continued TNF signalling, since blockade of the TNF receptor after 24 h does not abolish the elevated lipid-peroxidation phenotype. Another strength is that the computational analysis generates experimentally testable predictions, and two mechanistically distinct interventions identified by this approach are subsequently validated in macrophages.

Weaknesses:

There are, however, several limitations that affect the strength and scope of the conclusions.

(1) First, the use of the terms "persistent" and especially "self-sustaining" would be better supported by a more complete time-course analysis.

(2) Second, the proposed central role of ferritin-mediated iron sequestration in the protective effect of IFN-γ is not yet demonstrated directly. The data clearly link IFN-γ treatment to ferritin induction and reduced labile iron, but the causal contribution of ferritin itself remains to be established.

(3) Third, an important limitation is the connection between the mechanistic model developed with TNF stimulation and actual M. tuberculosis infection. Most of the mechanistic analysis, including type I IFN super-induction, lipid peroxidation, ferritin induction, labile iron and HIF1α regulation, is performed in TNF-stimulated macrophages. The infection experiments show that IFN-γ improves bacterial control and that low-dose trilaciclib plus ATRA reduces intracellular bacterial burden, but they do not establish that M. tuberculosis infection induces the same pathological circuit, or that these interventions improve bacterial control by acting through that circuit. The study therefore defines a convincing TNF-driven macrophage phenotype with relevance to bacterial control, but the broader conclusion that this mechanism underlies IFN-dependent susceptibility to TB remains only partially supported.

(4) Finally, the therapeutic implications go beyond the experimental evidence currently presented, since all of the pharmacological experiments are performed in cultured macrophages and there is no in vivo validation.

Conclusion:

Overall, this study proposes an interesting framework for understanding how inflammatory activation may become maladaptive in susceptible macrophages and how IFN-γ may combine antimicrobial activation with protection from oxidative damage. The convergence between IFN-γ, iron metabolism, lipid peroxidation and the pharmacological perturbations identified computationally is a clear strength. However, the causal role of ferritin, the operation of the proposed circuit during M. tuberculosis infection, and the in vivo relevance of the pharmacological strategy remain to be established. These limitations leave the mechanistic and translational evidence incomplete, while the study itself remains potentially important.

Reviewer #2 (Public review):

Summary:

The authors have carried out extensive transcriptomic, phenotypic and modelling-based analyses to provide novel insights into the interaction of the type I and II interferon programs in the determination of macrophage activation status and resistance to Mtb infection and infection-mediated damage. They demonstrate how an antagonistic effect between the two programs goes beyond classical downstream immune signalling pathways to lipid peroxidation maintained in a sustained autocrine manner and generation of a persistent pathological activation state (pPAS). Based on these analyses, the authors propose a therapeutic strategy of boosting specific pathways that increase oxidative stress resilience to reduce inflammatory pathology without suppressing host defenses for bacterial control and resisting pPAS. The conceptual framework may prove applicable to interferonopathies and to other bacterial and viral infections, though this remains to be tested.

Strengths:

(1) The study uses macrophages from a disease-relevant genetic murine model in which the sst1 locus drives the formation of necrotic pulmonary granulomas resembling human TB lesions- pathology not seen in standard C57BL/6 mice. This provides a genetically defined comparison between susceptible and resistant macrophages on an otherwise identical background, allowing the authors to attribute differences in activation state to a single locus rather than to strain-level variation.

(2) The experimental design isolates the phenomenon of interest: the TNF withdrawal and restimulation scheme allows the authors to establish that the aberrant activation state persists after removal of the initiating stimulus, rather than simply reflecting ongoing stimulation. The timed IFNAR blockade at 2, 12 and 24 h similarly separates initiation of the state from its maintenance.

(3) Lipid peroxidation is assessed through two orthogonal readouts: 4-HNE immunostaining for accumulated adducts and linoleamide alkyne click chemistry for ongoing synthesis. These, coupled with ROS and labile iron pool measurements, isotype antibody controls, parallel B6 and B6.Sst1S comparisons, and an anti-TNFR control, help in establishing that the phenotype is independent of continued TNF signalling. The convergence of these independent measures gives confidence in the peroxidation phenotype itself.

(4) The cSTAR analysis is applied here using regression rather than classification, generating a continuous DPD_TB score that correlates with measured Mtb fold change and thus provides a quantitative transcriptomic metric of macrophage priming state. Critically, the pathway predictions arising from this analysis (CDK4/6 inhibition and RAR activation) were tested and confirmed experimentally. The inferred network topology further predicted synergy between these two interventions, which bore out experimentally as an approximately ten-fold reduction in the effective dose of each agent in controlling Mtb during infection.

Weaknesses:

(1) Figure 2C is difficult to interpret as presented. The row labels ("No TNF"/"TNF") use different terminology from the corresponding conditions in panel A ("TNF withdrawal"/"TNF restimulated"), and "TNF" appears on both axes referring to different phases of the experiment; no timepoint is given on the panel itself, unlike neighbouring panels. Harmonising the labels with panel A and stating the harvest timepoint would help the reader. More substantively, the remaining lipid peroxidation readouts in this figure (panels D-G) are all at early timepoints of TNF stimulation (2-24 h) rather than during withdrawal, which limits what they can say about sustained autocrine signalling. Extending these assays to the later timepoints used in Figure 1 would considerably strengthen the claim that IFN-I maintains, rather than only initiates, the pathological state. The same applies to Figure 2G, where the contribution of itaconate to 4-HNE accumulation over time is not yet resolved.

(2) Several of the pathways implicated here are reported to behave differently between murine and human macrophages, and between macrophage subsets (alveolar versus monocyte-derived macrophages), during Mtb infection. This does not diminish the findings in this model, but it does bear on how broadly they can be generalised.

a) Type I interferon responses differ by species and by macrophage subset across multiple reports. Since the proposed model depends on autocrine IFN-I signalling reaching a threshold sufficient to sustain the pathological state, these differences in IFN-I output are worth keeping in mind when interpreting the wider significance of the findings.

b) Similarly, itaconate production in murine BMDMs is 20-fold higher than in LPS-activated human monocyte-derived macrophages and 50-fold higher than in LPS-activated alveolar macrophage-like cells, and Mtb infection of these human macrophages very weakly induces ACOD1 with almost no detectable itaconate (PMID 41797714). This is relevant to the Acod1/4-OI arm of the mechanism.

c) In a cross-species comparison of Mtb-infected macrophages, cholesterol homeostasis genes (including HMGCS1, IDI1, LSS) were significantly upregulated in human alveolar macrophages but downregulated in subcutaneous BCG-exposed murine alveolar macrophages (PMID 41208107)- the opposite direction to the lipid biosynthesis suppression treated here as a defining pPAS feature. The same group reports that murine AMs lack c-Maf and IL-10 whereas murine BMDMs express both (PMID 40073087), indicating that the autocrine anti-inflammatory brake on IFN-I responses is itself subset-dependent.

d) Finally, the cSTAR network predictions were inferred from human THP-1 perturbation data, but tested only in murine BMDMs. Establishing how this circuit operates in human macrophages, and during Mtb infection rather than TNF stimulation alone, would be a valuable extension of the work.

(3) The causal relationships linking IFN-I, lipid peroxidation, ROS and loss of IFNγ responsiveness could be drawn together more clearly. These elements are each established, but the connections between them are not always demonstrated directly. IFN-I appears to promote peroxidation through Acod1/itaconate and suppression of lipid biosynthesis rather than through iron, since neither IFNβ nor IFNAR blockade alters the labile iron pool. This would suggest two separable inputs to 4-HNE rather than a single pathway. This raises a further question about the persistent state itself: the labile iron pool rise appears to be TNF-driven, yet all labile iron measurements are made at 24 h in the continued presence of TNF and none under the withdrawal condition, so it is unclear whether elevated catalytic iron is sustained once the initiating stimulus is removed. Similarly, while IFNAR blockade reduces peroxidation, the reciprocal arm is not tested. An antioxidant or iron chelator could be used to ask whether peroxidation in turn drives Ifnb1 super-induction. In the absence of this information, the proposed feedback loop remains partly inferred. It would considerably strengthen the manuscript if the authors could clarify, through additional experiments or in the text, how the labile iron pool and lipid peroxidation relate to one another and what sustains each of them after TNF withdrawal.

(4) Reading across the manuscript, the labile iron pool emerges as the variable most consistently associated with the phenotype. Every protective intervention tested converges on it. By contrast, the alternative candidate mechanisms do not track with outcome. Lipid biosynthesis genes are suppressed by IFNγ yet induced by both trilaciclib and ATRA, all three of which are protective. GPX4 is unchanged under IFNγ and trilaciclib. The Acod1/itaconate axis cannot account for it since IFNγ priming blocks 4-HNE accumulation induced by exogenous itaconate. Yet, a direct causal role for iron is never tested. Additionally, IFNβ induces 4-HNE with the labile iron pool entirely unchanged, indicating at least one route to lipid peroxidation that bypasses catalytic iron. Focusing on iron handling would make the manuscript's message more coherent and its therapeutic argument more compelling.

Reviewer #3 (Public review):

Summary:

Araveti et al. demonstrate that Type I Interferon (IFN-I) and Interferon-gamma (IFN-γ) play opposing roles in regulating lipid peroxidation and host resistance during Mycobacterium tuberculosis (Mtb) infection. While both the two pathways drive inflammation, they differ fundamentally in cell protection. IFN-I signaling triggers a destructive, self-amplifying cycle catalysing the generation of reactive oxygen species (ROS) and lipid peroxidation, which ultimately impairs the host's ability to clear Mtb. Conversely, the authors demonstrate that IFN-γ couples antimicrobial activation with cytoprotection. It primes macrophages to fight the mycobacteria while simultaneously shielding them from oxidative stress. It achieves this by sequestering iron, which successfully prevents ROS from converting into damaging lipid peroxidation products.

Strengths:

Ultimately, this study highlights a critical biological distinction: IFN-γ safely balances inflammatory activation with cellular defense, whereas IFN-I promotes uncontrolled pathological damage. This divergent coupling of inflammation and cytoprotection carries major consequences for disease progression and host survival.

Weaknesses:

This study demonstrates all the findings in specific mouse strains. How these translate in human macrophages is not well characterised, thus raising the issue of its overall impact in tuberculosis disease.