MAP3K8 Regulates Cox-2–Mediated Prostaglandin E₂ Production in the Lung and Suppresses Pulmonary Inflammation and Fibrosis

Idiopathic pulmonary fibrosis (IPF) is an interstitial lung disease with a dismal prognosis, characterized by fibroblast foci and exuberant deposition of extracellular matrix components, leading to the distortion of lung architecture and the deterioration of respiratory functions. The prevailing working hypothesis suggests that the mechanisms driving IPF reflect abnormal, deregulated wound healing in response to persistent, environmentally imposed epithelial damage in genetically predisposed individuals of an older age (1, 2). Although the only effective treatment remains lung transplantation, pirfenidone (3) and nintedanib (4) were found to delay disease’s progression and constitute the current standard of care. Nintedanib is a small molecule that inhibits receptor tyrosine-kinases (RTKs; such as FGFR, VEGFR, and PDGFR), membrane receptors that activate multiple, and frequently overlapping, cellular signaling pathways, including MAPK pathways (4, 5). Although much less is known on the mode of action of pirfenidone, it is generally considered as an antifibrotic and anti-inflammatory agent, regulating the expression and activity of TGF-β and TNF-α that both, variably, rely on MAPK signaling to exert their effector functions (3, 5).

MAPKs, such as ERKs, JNKs, and p38 MAPKs, are protein Ser/Thr kinases that phosphorylate many intracellular targets including cytoskeletal elements, membrane transporters, as well as other kinases and transcription factors, resulting in the regulation of a wide array of cellular functions, including stress response, cell growth, differentiation, proliferation, and apoptosis (6). Aberrant MAPK signaling has been linked to the pathogenesis of many inflammatory, metabolic, or malignant diseases, highlighting the importance of dissecting the regulatory mechanisms governing their activation (6). MAPKs are activated from MAPK kinases (MAP2Ks), in turn activated by MAP2K kinases (MAP3Ks), which provide stimulus- and cell-specific signaling contexts for cellular responses to extracellular stimuli such as peptide growth factors, cytokines, hormones, as well as to endoplasmic reticulum and oxidative stress (6), all inherently linked to IPF pathogenesis.

MAPK kinase kinase 8 (MAP3K8; tumor progression locus 2 [TPL2]; cancer Osaka thyroid [COT] oncogene) is an Ser/Thr MAP3K, at the crossroad of many different signaling pathways. TPL2 was first identified by virtue of its oncogenic transforming activity in cells, and is now incriminated for the regulation of multiple pathogenic pathways in the context of inflammation and cancer (7). MAP3K8 has been reported to promote pleiotropic and even opposing effects, depending on the cell type and the tissue microenvironment. Although MAP3K8 is generally considered as a proinflammatory molecule and an oncogene, many reports have indicated an opposite role in limiting inflammatory and fibrotic responses and suppressing carcinogenesis (7). In the lung, in which Tpl2/Map3k8 is highly expressed, Map3k8 ubiquitous deficiency was shown to exacerbate eosinophilic inflammation when challenged with OVA (8) and to promote urethane-induced lung carcinogenesis (9). Because pulmonary fibrosis is a major risk factor for the development of lung cancer (10), and because different Map3k8-regulated functions are relative with many cellular mechanisms governing the pathogenesis of pulmonary fibrosis, in this report, we examined a possible role of Map3k8 in regulating pathogenetic responses in pulmonary fibrosis.

All mice were bred in the C57Bl6/J background for over 10 generations under specific pathogen-free conditions at the local animal facilities. Mice were housed at 55 ± 5% humidity, 20–22°C under a 12-h light-dark cycle; food and water were given ad libitum. All experimentation, conforming to the Animal Research: Reporting In Vivo Experiments (ARRIVE) guidelines, was approved by the internal Institutional Review Board (no. 95), as well as by the Veterinary Service and Fishery Department of the local governmental prefecture (no. 1121). The generation and genotyping instructions of LysM-Cre (11), Map3k8^f/f, and Map3k8^−/− mice (12) have been described previously. All randomly assigned experimental groups consisted of littermate mice. The health status of the mice was monitored once per day; no unexpected deaths were observed; all measures were taken to minimize animal suffering and distress.

Pulmonary fibrosis was induced by a single intratracheal instillation (with a MicroSprayer aerosolizer) of bleomycin (BLM) (3.2 U/Kg; Nippon Kayaku) into anesthetized (i.p.; xylazine, ketamine, and atropine, 10, 100, and 0.05 mg/kg, respectively) 6- to 8-wk-old male and female mice as previously reported (13–15). It should be noted that beyond dosing and route of administration, the timing, severity, and resolution of BLM-induced fibrosis depends on a number of parameters, such as the specific genetic background (i.e., C57bl6 J or N, genetic drift, vendor) and local health status. The dose used in this study was selected upon prior extensive local testing to minimize lethality (for ethical and practical reasons) while inducing a solid fibrotic profile, analyzed with a plethora of readout assays as previously reported (13) and as described below.

Bone marrow transplantation was performed as previously described (16); Map3k8^−/− and wild-type (wt) mice were irradiated once with 1000 rad (J. L. Shepherd Irradiator). Irradiated mice were injected i.v. through the tail vein with 5 × 10⁶ bone marrow cells from male wt or Map3k8^−/− mice. All the irradiated nontransplanted control mice died within 14 d from the time of irradiation, whereas all implanted mice survived, confirming efficient transplantation and reconstitution of the hemopoietic system.

The 16,16-dimethyl PGE₂ (dmPGE₂; Cayman Chemical, Ann Arbor, MI) was dissolved in methyl acetate, which was then evaporated under a nitrogen stream; dmPGE₂ was immediately dissolved in nitrogen-purged ethanol and kept as a stock solution at a concentration of 0.5 mg/ml. Immediately before in vivo administration, the stock dmPGE₂ solution was diluted with saline and kept on ice. dmPGE₂ was administered starting 1 d before BLM via two daily i.p. injections at a concentration of 10 μg/kg.

Respiratory mechanics, recently shown to be an accurate surrogate marker of disease development (13), were analyzed in tracheostomized mice with the FlexiVent ventilator system (SCIREQ) following manufacturer instructions, as previously reported (13).

Bronchoalveolar lavage fluid (BALFs) were obtained by lavaging the lungs of tracheostomized mice with 1 ml of 0.9% sterile sodium chloride (three times). After centrifugation at 100 × g for 10 min (4°C), the first BALF supernatant was stored at −80°C for protein and collagen content determination, and cell pellets from total BALF volume were counted with a hematocytometer after being stained with 0.4% trypan blue solution or used for FACS analysis.

Total protein levels were assessed with the Bradford assay according to manufacturer’s instructions (Bio-Rad Laboratories, Hercules, CA). Absorbance values were converted in mg/ml using a BSA standard curve (BSA 0–2 mg/ml). Total soluble collagen was quantified using the Sirius Red assay protocol; briefly, 50 μl of BALF samples, diluted in 0.5 M acetic acid, were incubated for 30 min with Sirius Red at room temperature (direct red 80; 120 μg/ml in 0.5 M acetic acid). After centrifugation (12,000 × g for 10 min), the absorbance of supernatant was read at 540 nm, and values were converted in micrograms per milliliter according to a standard curve with collagen type I from rat tail (0–500 μg/ml).

For FACS analysis, BALF cells were centrifuged at 1200 rpm at 4^οC and incubated with Fc Receptor Binding Inhibitor C16/32 (eBioscience) for 15 min on ice. Cells were then incubated for 30 min in FACS buffer containing manufacturers’ suggested dilutions of fluorescently labeled mAbs. The following Abs were used for analysis: anti-CD11b–PE, anti-F4/80–PE, anti-CD4–AF700, anti-CD8–allophycocyanin, anti-B220–PerCP–Cy5.5, and anti-Gr1–FITC (all BioLegend, San Diego, CA). Live hematopoietic cells (HCs) were first gated empirically by forward scatter versus side scatter characteristics (SSC). Cells in the hematopoietic gate were then interrogated for surface immunophenotypic markers such as CD8, CD4 (T cells), or B220 (B cells). Neutrophils were recognized as non-autofluorescent highly granular (SSC^hi) cells and, within this gate, were defined as cells F4/80⁺-Gr1^high (R1) or by gating for CD11b⁺ (CD11b versus SSC area [SSC-A]) cells and these were further gated for Gr1^High (neutrophils Gr1 versus SSC-A). CD11b/Gr1^high cells had multilobed nuclei typical of granulocytes, whereas CD11b/Gr1^mid and CD11b/Gr1^low cells had ovoid nuclei typical of monocytes/macrophages. Macrophages were identified as large autofluorescent cells F4/80⁺Gr1^mid/low (R2) or by gating for CD11b⁺ (CD11b versus SSC-A) cells and these were further gated for Gr1^low (macrophages-monocytes) (Gr1 versus SSC-A). All acquisition was performed using a BD FACSymphony flow cytometer. Data analysis was performed with the FlowJo software.

The right lung tissues were fixed in 10% v/v neutral buffered formalin and embedded in paraffin. Four-micrometer lung sections were prepared and stained with H&E with standard protocols. For Sirius Red–Fast Green staining, tissue sections were deparaffinized and stained successively with Fast Green 0.04% and Sirius Red 0.1%/Fast Green 0.04% dissolved in picric acid. Finally, sections were mounted with distyrene plasticizer xylene. Lung tissue imaging was performed using a Nikon Eclipse E800 microscope (Nikon, Shinagawa-ku, Japan) attached to a Q Imaging EXI Aqua Digital Camera, using the Q-Capture Pro 7 software.

Total RNA was extracted from the left lung lobe from each mouse using the Tri Reagent (Molecular Research Center), followed by DNase treatment (RQ1 RNAse-free DNase; Promega, Madison, WI) according to the respective manufacturers’ instructions. cDNA was synthesized from 2 μg of total RNA in a 20-μl reaction using Moloney murine leukemia virus reverse transcriptase (Promega). Quantitative RT-PCR (Q-RT-PCR) was performed using SoFAst EvaGreen Supermix in a Bio-Rad CFX96 Touch Real-Time PCR Detection System (Bio-Rad Laboratories). Values were normalized to β2-microglobulin (B2M). Primers used, as well as the product size (bp) were as follows: col1a1 (forward: 5′-CTA CTA CCG GGC CGA TGA TG-3′; reverse: 5′-CGA TCC AGT ACT CTC CGC TC-3′; 188 bp), map3k8 (forward: 5′-TTA GCC CAA GAC ATG AAG AC-3′; reverse: 5′-ACT CAG CAA TGT TCT CAT GC-3′; 117 bp), and b2m (forward: 5′-TTC TGG TGC TTG TCT CAC TGA-3′; reverse: 5′-CAG TAT GTT CGG CTT CCC ATTC-3′; 104 bp). The annealing temperature for all primers was 58°C. Values were calculated according to the 2^{ΔΔcycle threshold} method.

As the estimated half-life of PGE₂ in vivo is less than 15 s (17), we quantified PGE₂ through its metabolites, as recently reported (18). Briefly, 300 μl of BALFs were deproteinized with acetone. After vigorous mixing for 4 min and centrifugation at 2000 × g for 10 min at 4°C, samples were transferred to a clean 15-ml glass vial, mixed with 800 μl of hexane by vigorous mixing for 30 s and centrifuged for 10 min at 2000 × g at 4°C. The lower phase was acidified to pH 3.5 with formic acid and then mixed with chloroform. After three vigorous mixing for 30 s and centrifugation for 10 min at 2000 × g at 4°C, the lower chloroform phase was kept for 15 min at −80°C to separate any residual from the upper phase. Samples extracted in chloroform were evaporated to dryness using a speed-vac concentrator (model no. SC110-120; Savant Systems, San Diego, CA) and redissolved in 100 μl of methanol/100 mM ammonium acetate pH 8.5 (9:1). The dmPGE₂ (Cayman Chemicals) was used as an internal control in each sample from the beginning of the above extraction procedure at a final concentration of 2.5 ng/ml. Samples were analyzed by direct infusion in an LTQ Orbitrap XL mass spectrometer (Thermo Fisher Scientific, Waltham, MA) with electrospray ionization in negative mode. Capillary temperature was 275°C, spray voltage was 3.5 kV, sheath gas was set at 40 U, and sweep gas was at 8 U. The resolution was 100 K, providing high accuracy for prostanoid measurements. Precursor ion masses were used for prostanoid profiling, whereas lipid identity was confirmed by precursor ion fragmentation using collision induced dissociation. Precursor m/z 297.1530 was used for 13,14-dihydro-15-keto-tetranor-PGE₂ and tetranorPGE₁, stable metabolites of PGE₂. Precursor m/z Precursor m/z 325.2021 was used for metabolites 2,3-dinor-11b-PGF2α, 2,3-dinor-PGE₁, dinor-PGF2α, and 2,3-dimensionalinor-8-iso-PGF2α.

The raw microarray datasets GSE32537 and GSE47460 (that contains data from two different platforms, GPL6480 and GPL14550, which were processed separately) were downloaded from Gene Expression Omnibus (GEO), and intensity values were background corrected with subtraction or normexp methods implemented into limma (19) and oligonucleotide (20) Bioconductor packages, respectively, and robust multi-array average (rma) normalized. Afterwards, sample outliers were removed based on a principal component analysis (PCA) plot created with arrayQualityMetrics Bioconductor package (21). Control probes, as well as probes mapping to more than one HUGO gene symbol were removed from further analysis. Intensity values were then summarized at the gene level using a weighted average value of all probes/transcript clusters representing each gene (weights sum up to the unit). Subsequent differential expression analysis was performed using the moderated t test statistics algorithm provided by limma R package.

scRNAseq GSE122960 dataset was processed with Seurat R package version 3.1.2 (22). Adaptation of the original reported code provided in this study was used to reanalyze the GSE122960 hierarchical feature-barcode matrices. Because of the Seurat update since the original GSE122960 publication (23), samples integration was performed with a two-step application of the standard Seurat v3 integration pipeline: integration of the donor and IPF samples separately prior to donor-IPF samples final integration. All analyses were performed on R version 3.6.2, and the datasets were retrieved from the GEO repository.

Statistical significance was assessed with the Prism (GraphPad) software, as detailed at each figure legend.

To examine a possible role of Map3k8 in the regulation of pathogenic signal transduction pathways during fibrogenesis, we first sought to quantify its mRNA expression levels during the development of BLM-induced pulmonary inflammation and fibrosis. To this end, BLM (3.2 U/Kg) was intratracheally administered to 6- to 8-wk-old C57Bl6/J mice, which were then sacrificed 7, 14, and 21 d post-BLM administration, roughly corresponding to the (postacute) inflammatory, fibrotic, and resolution phases of disease development. As expected, surviving (Fig. 1A) mice that received BLM lost weight (Fig. 1B), accompanied by a gradual increase in pulmonary edema and inflammation, as indicated from the total protein concentration and the infiltrating cell numbers of the BALF (Fig. 1C, 1D, respectively). Col1a1 mRNA expression, as determined with Q-RT-PCR in lung tissue from the same mice, was also found gradually increasing post-BLM (Fig. 1E), as also reflected at the soluble collagen levels in the BALF (Fig. 1F), determined with the Sirius Red assay. Accordingly, histopathological analysis of the lungs of mice post-BLM administration indicated the increasing presence of peribronchiolar and parenchymal fibrotic regions (Fig. 1G), resulting in impaired respiratory mechanics, such as resistance, tissue elasticity, and static compliance, as measured with FlexiVent (Fig. 1H–J). All disease signs peaked 14 d post-BLM, subsiding at 21 d (Fig. 1C–G). Q-RT-PCR mRNA analysis (in the same samples as in Fig. 1E) indicated a gradual decrease in pulmonary Map3k8 expression in lung tissue in inverse correlation with disease development (Fig. 1K).

To translate the findings into the human disease, we performed in silico reanalysis of the three largest publicly available expression profiling datasets (GSE32537 and GSE47460) at GEO interrogating differential expression in the lung tissue of 119, 122, and 38 IPF patients in comparison with 50, 91, and 17 controls, respectively (24, 25). Raw data were background corrected and rma normalized, and outliers were removed based on PCA plots (Fig. 2A, 2B, 2D–G). Differential expression analysis indicated a modest but highly statistical significant decrease of MAP3K8 mRNA levels in IPF (log₂ FC/FDR-corrected p values for the different datasets GSE32537, GSE47460-GPL6480, and GSE47460-GPL14550 are respectively −0.459/4.65 × 10⁻¹⁰, −0.658/1.06 × 10⁻⁴, and −0.668/1.46 × 10⁻¹³; Fig. 2), in agreement with the modeled disease (Fig. 1). Thus, the development of pulmonary fibrosis, in both humans and mice, is accompanied with a transcriptional (or posttranscriptional) downregulation of MAP3K8 mRNA levels in the lung.

As MAP3K8 is widely expressed in different pulmonary cell types exerting pleiotropic effects (7), we next visualized MAP3K8 mRNA expression in single pulmonary cells in the normal or fibrotic human lung, via the web-tool available at nupulmonary.org/resources and reanalysis of the relative single-cell RNA-sequencing (scRNAseq) dataset GSE122960 (23). As evident in Supplemental Fig. 1A, MAP3K8 is expressed at both nonhematopoietic (nHCs) and HCs, and most notably in monocytes/macrophages. More specifically, reduced MAP3K8 expression was detected in the SPP1⁺ macrophage subcluster 1, which most likely correspond to monocyte-derived macrophages (p = 6.11 × 10⁻²⁴; Supplemental Fig. 1B). In addition, the same data also suggest MAP3K8 downregulation in the (aSMA⁺) myofibroblast cell cluster (p = 0.0002; Supplemental Fig. 1C), further corroborating a transcriptional downregulation of MAP3K8 expression in IPF.

To validate and dissect a possible Map3k8 role in pulmonary fibrosis, BLM was administered to Map3k8 ubiquitous knockout mice (Map3k8^−/−) (12) and wt littermate controls. Map3k8^−/− mice present with no apparent pulmonary phenotype or impaired respiratory functions upon healthy conditions (Fig. 3; Sal control groups). BLM administration to Map3k8^−/− mice resulted in increased lethality (Fig. 3A) and in greater weight loss in the surviving mice (Fig. 3B), suggesting increased systemic disease burden in comparison with littermate wt mice. Pulmonary edema and inflammation were also found significantly increased in surviving Map3k8^−/− mice (Fig. 3C, 3D, respectively), accompanied by increased collagen expression (Fig. 3E, 3F). Accordingly, tissue fibrosis was notably expanded (Fig. 3G), resulting in further deterioration of respiratory functions upon BLM administration (Fig. 3H–J).

As Map3k8 is expressed from both nHCs and HCs (Supplemental Fig. 1), we next interrogated the relative contribution of Map3k8 expression from nHCs and HCs in disease protection. To that end, wt and Map3k8^−/− mice were irradiated to abolish HCs, followed by injections of bone marrow cells isolated from wt or Map3k8^−/− mice to reconstitute the immune system (Fig. 4A). The generated chimeric mice bearing HCs with the genetic background of the donor (Fig. 4B), were then administered BLM and disease severity was assessed 14 d post-BLM, at the peak of the disease. Map3k8 deficiency in HCs resulted in a bigger effect in the systemic manifestations of the disease, as reflected in lethality (Fig. 4C) and weight loss (Fig. 4D). Map3k8 deficiency in either nHCs or HCs in surviving mice increased pulmonary edema (Fig. 4E) and collagen expression (Fig. 4G, 4H) in comparison with wt/wt controls, indicating that Map3k8 expression from both nHCs and HCs exert some protective role in disease pathogenesis. Noteworthy, the accumulation of inflammatory cells in the BALF was more pronounced when Map3k8 expression was missing from the HCs (Fig. 4F). Accordingly, and in agreement with the systemic manifestations (Fig. 4C, 4D), Map3k8 deficiency in HCs was sufficient to exacerbate collagen deposition and to distort lung architecture (Fig. 4I), suggesting a main protective role in pulmonary fibrosis for Map3k8 expression in HCs.

Because the deletion of Map3k8 in HCs had a higher effect in the number of inflammatory cells in the BALFs of BLM-challenged mice (Fig. 4F), we next performed basic (nonexhaustive) FACS analysis of immune cells in the BALFs (Fig. 5A). Map3k8 deletion had minor effects in neutrophilic infiltration (Fig. 5B); accordingly, Map3k8^−/− mice had minor and nonconsistent effects in modeled acute lung disorders, such as LPS-induced lung injury and ventilation-induced lung injury (data not shown). However, Map3k8 deletion in either of the cellular reservoirs promoted the accumulation of macrophages, which was found to be more pronounced upon deletion in HCs (Fig. 5B). Moreover, Map3k8 deletion in HCs, but not in nHCs, further promoted lymphocyte accumulation (Fig. 5C).

Given the suggested role(s) of Map3k8 in macrophage responses, its reduced expression in fibrotic lungs (Figs. 1, 2) and IPF macrophages (Supplemental Fig. 1), and the finding that Map3k8 deficiency in HCs exacerbates BLM-induced pulmonary inflammation and fibrosis (Figs. 4, 5), we next genetically deleted Map3k8 specifically in macrophages (and granulocytes) by mating the conditional (floxed; ^f/f) knockout mouse for Map3k8 (Map3k8^f/f) (12) with a transgenic mouse strain expressing the Cre recombinase under the control of the LysM promoter (TgLysM-Cre) (11). TgLysM-Cre has been reported with an 80–95% deletion efficiency in macrophages (11, 26), whereas LysM-mediated Cre expression, per se, has been previously reported not to have any effects on lung development and architecture or BLM-induced pulmonary fibrosis (26). BLM was then administered to LysMMap3k8^−/− mice and littermate controls and disease severity was assessed with the standardized readout assays at the peak of the disease. Disease development in LysMMap3k8^−/− mice was found exacerbated in all readout assays, including respiratory functions (Fig. 6). Furthermore, FACS analysis of inflammatory cells in the BALF in the absence of Map3k8 expression from macrophages (Fig. 7) phenocopied the HC deletion (Fig. 5), indicating a major protective role for macrophage (LysM⁺) Map3k8 expression in the disease pathogenesis as well as confirming the seminal contribution of macrophages in pulmonary fibrosis.

Among the different cellular pathways affected by Map3k8 that could possibly play a protective role in pulmonary fibrosis, Map3k8 has been reported to regulate cyclooxygenase-2 (Cox-2)–mediated PGE₂ expression from arachidonic acid (AA) (Fig. 8A) (12, 18). PGE₂ is a bioactive eicosanoid that is considered as a proinflammatory mediator; on the contrary, in the lung, PGE₂ has an established role in limiting fibrotic processes (27); however, its local regulation remains relatively unexplored and its cellular origin uncertain.

To examine if Cox-2 mRNA levels are affected by the genetic deletion of Map3k8, we performed Q-RT-PCR analysis in lung tissue samples from genetically modified mice at the peak of the disease post-BLM administration. Ubiquitous genetic deletion of Map3k8 (Fig. 8B) was found to downregulate Cox-2 mRNA levels (Fig. 8C). A similar expression profile was also observed upon the genetic deletion of Map3k8 in nHCs and HCs in chimeric mice (Supplemental Fig. 2A, 2B) as well as in macrophages (Supplemental Fig. 2C, 2D), thus confirming 1) Map3k8 expression from both nHCs and HCs, 2) efficient Map3k8 targeting, as well as 3) the regulatory role of Map3k8 in Cox-2 expression. Accordingly, AA metabolism was found to be disturbed in Map3k8^−/− mice upon BLM-induced pulmonary fibrosis that presented with increased levels of AA (Fig. 8D) and reduced levels of PGE₂ in their BALFs (Fig. 5E–G), as deduced, given its extreme instability from its metabolites, according to an established and recently published procedure (18). Similar results were obtained in chimeric mice (Supplemental Fig. 3), thus establishing Map3k8 as a major regulator of PGE₂ production in the lungs and suggesting HCs as the major PGE₂ producers in the lung.

To examine if the decreased PGE₂ levels upon Map3k8 deletion contribute to the observed exacerbated fibrotic phenotype, dmPGE₂, a stable analogue of endogenous PGE₂, was administered (i.p.; 10 μg/kg; twice daily) to Map3k8^−/− and wt littermate mice undergoing BLM-induced pulmonary fibrosis development. dmPGE₂ administration restored the exacerbated fibrotic responses in Map3k8^−/− mice, as indicated from all readout assays (Fig. 9), as well as decreased the severity of the BLM-induced disease in wt mice (Fig. 9). Noteworthy, PGE₂ administration also attenuated inflammatory influx (Fig. 10), integral to fibrosis development in this animal BLM model.

MAP3K8 mRNA was found downregulated in both IPF patients and the corresponding animal model, whereas public scRNAseq data mining suggested that MAP3K8 is expressed in both nHCs and HCs in the human lung. Genetic deletion of Map3k8 in mice exacerbated BLM-induced pulmonary fibrosis, whereas bone marrow transfer experiments indicated that although Map3k8 expression in both nHCs and HCs exert some protective role in disease pathogenesis, Map3k8 in HCs has a more dominant role. Moreover, macrophage-specific deletion of Map3k8 was shown to be sufficient to exacerbate disease severity, thus confirming a major role for macrophages in fibrotic responses in the lung and a role for Map3k8 in the homeostasis of their effector functions. Map3k8 deficiency was further shown to decrease Cox-2 mRNA expression, followed by a decrease in PGE₂ expression in the lung, whereas exogenous administration of (dm)PGE₂ reversed the exacerbated fibrotic response of Map3k8 ^−/− mice and decreased disease severity in wt mice.

The expression and/or activation of MAP3K8 has been shown to be modulated by multiple inflammatory mediators through their cognate receptors, such as TLRs, CD40, TNFR1, IL-1R, as well as different GPCRs. At steady state, MAP3K8 is inactive through the association with NF-κB p105 and A20-binding inhibitor of NF-κB 2 (ABIN2). Activation of IκB kinase (IKKβ) leads to the phosphorylation of p105 and its subsequent ubiquitination and proteasomal degradation, thus releasing MAP3K8, which is then phosphorylated, in turn activating its downstream targets and thus participating in the regulation of inflammatory responses (7). However, and given the central role of Map3k8 in regulating inflammatory and homeostatic responses, additional levels of regulation of the Map3k8 levels and effector functions have been reported at the genetic and epigenetic level. MAP3K8 mRNA expression has been found decreased in intestinal myofibroblasts isolated from the inflamed ileum of inflammatory bowel disease patients (12). Decreased MAP3K8 mRNA expression was also detected in the lung tissue of lung cancer patients correlating with poor survival, suggested to be imposed through miRNA-370 which targets MAP3K8 transcripts for degradation (9). As shown in this study, decreased MAP3K8 mRNA levels were detected in the lungs of mice upon BLM-induced pulmonary fibrosis (Fig. 1) and of IPF patients (Fig. 2), indicating transcriptional or posttranscriptional downregulation of Map3k8 expression upon fibrinogenesis in the lung, notwithstanding additional means of regulation via protein-protein interactions and/or phosphorylation cascades.

Irrespectively of the regulatory mode of Map3k8 expression, low Map3k8 levels have been associated with reduced Cox-2 and PGE₂ levels in different cell types and pathophysiological situations, including inflamed adipocytes (28), intestinal myofibroblasts (12), and activated macrophages (29); however, no such correlation has been reported in the context of pulmonary fibrosis. As shown in this study, reduced Cox-2 and PGE₂ levels always accompanied Map3k8 genetic deficiency and the associated exacerbated fibrotic phenotype (Figs. 8, Supplemental Fig. 2), which could be reverted by the administration of exogenous PGE₂ (Figs. 9, 10). PGE₂ administration also prevented disease development in wt mice, as previously reported (30, 31), whereas no gross pathologic effects were observed (data not shown) (30, 31).

Therefore, Map3k8 plays a central role in the regulation of local PGE₂ production in the lung. However, additional means of regulation of PGE₂ production have been proposed, via soluble mediators and epigenetic changes (reviewed in Ref. 27). Moreover, little is known on the relative contribution of different cell types in the overall PGE₂ production, although the current dogma, based mostly on in vitro experiments, suggests alveolar epithelial cells and fibroblasts as the main producers in the lung (27). However, as shown in this study in vivo (Fig. 4), both nHCs and HCs produce PGE₂, whereas the overall contribution of HCs (likely macrophages) in pulmonary levels was found higher in these experimental settings.

Map3k8^−/− mice have been previously reported to be protected from LPS/d-gal–induced endotoxic shock (32), attributed to defective TNF and chemokine receptors expression from macrophages (32, 33). Map3k8^−/− mice were also found protected from experimental encephalomyelitis, attributed to defective IL-17 signaling (34), as well as from Con A (ConA)–induced, T cell–dependent and TNF-mediated liver inflammation and injury, attributed to defective NKT effector functions (35). However, and on the contrary, Map3k8 ubiquitous genetic deficiency has been reported to elevate inflammation and to exacerbate fibrosis in the small intestine, liver, and lung following Schistosoma mansoni infection (36), as well as to promote intestinal inflammation and tumorigenesis (12, 37). In agreement, in this report, genetic deletion of Map3k8 was shown to exacerbate BLM-induced pulmonary fibrosis (Fig. 3). A similar protective role was previously shown for urethane-induced lung carcinogenesis (9), thus extending the mechanistic similarities between IPF and lung cancer (10) and suggesting a major homeostatic role for Map3k8 in lung pathophysiology and fibrotic responses that may underlie carcinogenesis.

Map3k8 is expressed from both nHCs and HCs in the lung (Supplemental Fig. 1) (7), and genetic deficiency in either cellular reservoir was shown to exacerbate the modeled disease (Fig. 4), suggesting that MAP3K8-regulated homeostatic mechanisms are active in both compartments. Although MAP3K8 is thought to be expressed mainly in HCs, it can also be detected in other cells (7), including adipocytes (28) and fibroblasts (Supplemental Fig. 1) (12). Decreased MAP3K8 expression was visualized in IPF (aSMA/ACTA2⁺) pulmonary myofibroblasts (Supplemental Fig. 1B, 1C), whose accumulation is a major hallmark of pulmonary fibrosis (1, 2). Decreased MAP3K8 expression was previously shown for inflammatory bowel disease intestinal myofibroblasts (12), corelating with reduced COX-2 expression and PGE₂ production, suggesting similar effects in IPF myofibroblasts. Accordingly, lung fibroblasts isolated from IPF patients have been suggested to express reduced levels of COX-2 and PGE₂ (38, 39). Moreover, alveolar epithelial cells have been proposed to produce PGE₂, via Cox-2, which target adjacent fibroblasts (40). In turn, PGE₂ has been shown to decrease fibroblast differentiation and collagen production (reviewed in Ref. 27) and to promote their apoptosis (41), likely through the EP2 receptor and disruption of calcium signaling (42). Therefore, the observed exacerbated fibrotic response of chimeric mice lacking Map3k8 in nHCs cells (Fig. 5) can be partly attributed to the attenuation of the antifibrotic effects of PGE₂ on lung fibroblasts.

However, the bone marrow transfer experiments indicated, for the first time (to our knowledge) on a quantitative basis in vivo with tandem mass spectrometry, that HCs are the major source of Map3k8-regulated, Cox-2–mediated, PGE₂ production in the lung in modeled pulmonary fibrosis (Figs. 8, Supplemental Fig. 3). Moreover, Map3k8 deletion in HCs, phenocopied in the macrophage-specific deletion, was sufficient for disease exacerbation (Figs. 4, 6), corroborating the important role of Map3k8 in the regulation of macrophage responses, as well as the role of pulmonary macrophages, per se, in fibrosis development in the lung. Macrophages are well recognized as essential players in IPF pathogenesis, as they comprise the major immune cell type populating the lungs of IPF patients, whereas depletion of circulating monocytes, via genetic or pharmacologic means, attenuates fibrosis severity in animal models (43–47). Beyond their well-established roles in apoptotic cell clearance and the production of profibrotic mediators such as TGFb and IL-13, macrophages secrete numerous cytokines and chemokines thus modulating the immune response, as well as matrix metalloproteinases responsible for extracellular matrix remodeling and resolution. However, their mode of action remains controversial because they are highly heterogenous and exhibit remarkable plasticity (48, 49).

Pulmonary macrophages can be grouped into two broad subsets based on their anatomic location: alveolar macrophages, which line the surface of alveoli, and interstitial macrophages (IMs) localized between the alveolar epithelium and the vascular endothelium. Furthermore, mouse alveolar macrophages include tissue-resident cells, which are long-lived, self-renewing cells that arise from fetal progenitors (50, 51), as well as monocyte-derived alveolar macrophages (Mo-AMs) (52, 53). Mo-AMs originate postnatally from circulating monocytes, are recruited via a CCL2/CCR2 axis, exhibit an proinflammatory expression profile, and have been suggested to be the main macrophage subgroup driving pulmonary fibrosis in mice (54, 55). IMs also originate from monocytes, and some from the yolk sac are reprogrammed epigenetically by the local microenvironment, although it has been suggested that they may serve as an obligatory intermediate between blood monocytes and alveolar macrophages (44, 56). Three distinct populations of IMs have been suggested (IM1-3) based on surface markers exhibiting differential turnover rates (57). The advent of scRNAseq has revealed even greater macrophage diversity, and several novel macrophage subsets have been suggested, defined by the expression of different chemokines and cell-surface markers such as M-CSF/M-CSFR, SPP1/MERTK, CX3CR1, Ly6C^hi, CD171, Fra-2, CD300c2, and Lrp5. However, a consensus has not been reached because data analysis of scRNAseq data and cluster identification have not yet developed standard operational procedures and commonly used cell-specific markers. Given the differential roles that have been suggested for the different macrophage subsets and to possibly identify in which macrophage subtype Map3k8 is found downregulated, we performed reanalysis of the publicly available scRNAseq dataset GSE122960 (23) (Supplemental Fig. 1). MAP3K8 was found downregulated in a profibrotic macrophage subpopulation characterized by elevated expression of SPP1, likely corresponding to the previously reported macrophage subset SPP^hi in humans (58), as well as to the Mo-AMs in mice (54), reported to drive the pathogenesis of the disease (54, 58). However, de novo scRNAseq of both lung and BALF cells in the absence of Map3k8 will be further required to fully appreciate the role of Map3k8 in macrophages subsets and to obtain additional mechanistic insights.

Increasing further heterogeneity and complexity, macrophages can also get polarized, depending on local stimuli, toward two highly dynamic and overlapping states of activation, classically activated (by IFN-γ and TNF) M1 macrophages or alternatively activated (by IL-4, IL-13, etc.) M2 macrophages. M1 macrophages, which secrete TNF, IL1b, and IL-6, are considered proinflammatory, whereas M2 macrophages (that can be further subclassified into a–d), which secrete IL-10 and TGF among others, are thought to suppress inflammation while promoting fibroproliferation and uncontrolled repair. Map3k8 deficiency has been suggested to promote macrophage polarization to M2, following S. mansoni infection (36), suggesting yet another profibrotic mechanism that could be regulated by Map3k8 in macrophages. The promotion of macrophage M2 polarization upon reduced levels of Map3k8 in IPF macrophages could influence not only macrophage wound healing effector functions and fibrosis, but also T cell physiology and functions. As shown in this study, Map3k8 genetic deletion in HCs (Fig. 3) or specifically in macrophages (Fig. 4) increased inflammatory cells in BALFs, including T cells. Moreover, Map3k8 ubiquitous genetic deletion has been shown to promote Th2 polarization of the T cell response in the lung (8, 36), whereas IPF is considered as a type II disease and type 2 cytokines, such as IL-4 and IL-13, are elevated in IPF (59). T cell functions can be also modulated by yet another monocytic subtype, namely circulating myeloid-derived suppressor cells. Increased numbers of myeloid-derived suppressor cells have been found in both IPF patients and the BLM model, suggested to modulate IPF progression by orchestrating immunosuppressive and profibrotic networks (60, 61). Immunosuppressive signals have been also suggested to emanate from a sessile sub population of alveolar macrophages via direct communication, through Cx43, with alveolar epithelial cells (62). Moreover, macrophages have been suggested to direct the metabolism and homeostasis of adjacent cells (63) such as fibroblasts whose activation includes glycolytic reprogramming (64).

Notwithstanding the Map3k8 regulation of PGE₂ production, Map3k8 has been previously suggested to modulate, via MAPK pathways, the expression of various cytokines and chemokines (65). More specifically, Map3k8 genetic deletion has been reported to decrease the levels of TNF and IL-1b (32), IL-6, CXCL2 (MIP-2), and CCL2 (MCP-1) on nonmyeloid cells (66), as well as CCR1, CCR2, and CCR5 on macrophages (33). All of these proinflammatory mediators have been found deregulated in pulmonary fibrosis (48, 49, 65), including the BLM-induced disease (data not shown) (15, 26); among them, CCL2 has been shown to be critical for macrophage recruitment (43). mRNA levels of TNF, Il-1b, IL-6, CXCL2, and CCL2 were found variably decreased in the lung tissue in the absence of Map3k8 at the peak of the BLM-induced fibrosis (data not shown). Therefore, and although RT-PCR in whole-lung tissue at the peak of the disease is not the optimal methodology for this purpose, the effect of Map3k8 deletion on the levels of proinflammatory mediators cannot account for the increased fibrosis post-BLM administration in Map3k8 null mice in these settings.

Further studies, employing inducible deletion of Map3k8 and/or the use of other fibrotic animal models that bypass inflammation (i.e., adenoviral delivery of TGFb), will be required to dissect the effects of Map3k8/Cox-2/PGE₂ axis on fibrosis, per se. Moreover, further lipidomic analysis in IPF and its animal model will be necessary to fully appreciate the pleiotropic effects of various lipids (leukotrienes, thromboxanes, and PGs, as well as lysophospholipids) on pulmonary inflammation and fibrosis.

In conclusion, Map3k8 was shown to have a major role in healthy lung homeostasis, regulating, among others, the local production of PGE₂ and its antifibrotic effects. The development of fibrosis, in both humans and mice, entails a transcriptional downregulation of Map3k8 expression in different pulmonary cells, and especially macrophages. Down regulation of Map3k8 levels results, possibly among others, in the abrogation of Cox-2–mediated PGE₂ production and the alleviation of its lung specific antifibrotic effects.

We thank G. Kollias (Biomedical Sciences Research Center Fleming, National and Kapodistrian University of Athens) for the Map3k8^f/f mice and M. Roulis (Yale University) for critical reading of the manuscript.

The authors have no financial conflicts of interest.