We first investigated the effect of TLR5 on TLR4 signaling in vivo, by exposing Tlr5-deficient mice or wildtype controls to LPS via intraperitoneal administration. As expected, this led to substantial lung inflammation in wildtype mice, which was ameliorated in the absence of TLR5 (Figure 1A). TLR5 deficiency also ameliorated cellular influx and lung injury as evidenced by lung lavage protein (Figure 1B). Furthermore, expression of inflammatory cytokines in the lung was significantly ameliorated in Tlr5-deficient mice (Figure 1C,D). This was largely mirrored in a reduction of inflammatory gene expression in the liver (Figure 1—figure supplement 1A). To ensure that our results were not affected by obesity-, microbiome- or breeding-related immune perturbations, we performed experiments with mice that were either purchased from a commercial vendor (using C57BL/6 as controls) or bred in our NIEHS colony (using wild-type littermates as controls) and treated some mice with neomycin to reduce bacterial burden in the intestinal tract. Our results did not vary regardless of mouse provenance or antibiotic treatment (Figure 1—figure supplement 1B).

To determine if the TLR5 effect on TLR4 signaling has broader biological and clinical relevance in the lung, we explored TLR5-mediated effects on sterile lung injury after exposure to the ambient pollutant, O₃. O₃ exposure is associated with increased morbidity and mortality in human patients with cardiopulmonary disease (Katsouyanni et al., 1995; Levy et al., 2005); furthermore, it is now understood that TLR4 mediates the development of inflammation and airway hyperresponsiveness (AHR) after O₃ exposure (Garantziotis et al., 2010). We used an O₃ dose that is equivalent to human exposure during a high-O₃ day (Slade et al., 1997; Hatch et al., 1994). Tlr5-deficient mice had ameliorated airway cytokine expression and almost abolished AHR after O₃ exposure (Figure 2A,B). Because hyaluronan is the endogenous danger-associated molecular pattern that activates TLR4 and mediates the response after O₃ exposure (Garantziotis et al., 2010; Garantziotis et al., 2009), we then investigated the effect of TLR5 on hyaluronan signaling. Tlr5-deficient mice had substantially reduced inflammatory gene induction and significantly diminished AHR after instilled ultrapure (pharmaceutical grade) hyaluronan exposure (Figure 3A,B). We then compared the effects of Tlr4-, Tlr5- and Myd88 deficiency on the hyaluronan response in vitro. We incubated tracheal rings of these strains with short fragments of hyaluronan, which induces hyperresponsiveness to methacholine-induced constriction through a TLR4-MyD88 pathway (Garantziotis et al., 2010; Li et al., 2011) We found that deficiency in either of the 3 molecules abolished the response to hyaluronan (Figure 3C) suggesting that TLR4 and TLR5 are non-redundant and necessary for the hyaluronan-MyD88 signaling response.

We then performed a more global analysis of the TLR5 effects on TLR4 signaling. We analyzed gene expression patterns using the NanoString platform (www.nanostring.com) and utilizing the Mouse Innate Immunity Panel Codeset (Ns_Mm_Myeloid_v2.0) and focused specifically on 242 genes that were more than 2-fold upregulated after LPS exposure (Supplementary files 1a and 1b and Figure 1—figure supplement 2) and sorted them according to magnitude of TLR5 effect. We showed that the presence or absence of functional TLR5 is associated with differential regulation of immune genes in this panel. Interestingly, there was a linear correlation between the magnitude of the TLR5 effect and the proportion of genes that are either published or predicted to be downstream of the NFκB pathway (Figure 1—figure supplement 2C): among the genes that were 70–90% upregulated in Tlr5-sufficient mice compared to Tlr5-deficient mice, almost 90% were in the NFκB pathway, while this proportion fell to 50% among the genes that were not different between genotypes. (R² = 0.89, p=0.0013). This may suggest that TLR5 preferentially impacts gene expression downstream of NFκB activation. Alternatively, because these data are from whole lung tissue, this could reflect decreased recruitment of inflammatory cells, although we are not aware that inflammatory cells preferentially express NFκB pathway genes. In aggregate, the presented results support that TLR5 promotes TLR4 signaling in several models of TLR4 activation through pathogen- or danger-associated molecular patterns (PAMPs or DAMPs, that is LPS or hyaluronan respectively) and promotes TLR4-mediated inflammation and airway hyperresponsiveness in vivo.

We then investigated the effect of TLR5 in human TLR4 signaling. In humans, a dominant-negative TLR5 single nucleotide polymorphism (SNP) (rs5744168, TLR5^392STOP) (Hawn et al., 2003) is found with a prevalence of 8–10% in Caucasians and 3% in African Americans. We hypothesized that carriers of this SNP may have reduced TLR4-mediated inflammation. We used the Environmental Polymorphisms Registry (Chulada et al., 2011), a NIEHS-supported cohort, to recruit carriers of this allele, as well as ‘wildtype’ controls. Others have reported that whole blood from rs5744168 minor-allele carriers does not differ in the response to LPS compared to ‘wildtype’ (West et al., 2013), and we confirmed this finding (Figure 4—figure supplement 1A). We believe this happens because whole blood consists of different cell types, which differentially express TLR5, thereby confounding the effect on TLR4 signaling. We then investigated the effect of this TLR5 SNP on purified, primary monocyte-derived macrophages. Macrophages from rs5744168 minor-allele carriers had a decreased response to flagellin and ultrapure LPS, but not Pam3CSK4 (Figure 4A and Figure 4—figure supplement 1B), thus confirming that human TLR5 genetic variation specifically determines the response to LPS. This was not due to altered expression of TLR4 or CD14, which was not changed by the rs5744168 genotype (Figure 4—figure supplement 1C).

We then investigated the effect of TLR5 in O₃-induced inflammation in healthy human volunteers. We exposed human volunteers to O₃, and isolated alveolar macrophages through bronchoscopy 24 hr after exposure, which represents the peak of O₃-induced inflammation and symptoms in humans. TLR5 expression was modestly increased in alveolar macrophages of human volunteers after O₃ exposure (Figure 4B, p=0.05 by Wilcoxon pairwise signed rank test). There was no association between TLR5 expression and TLR4 expression after O₃ exposure. We found, that TNFα expression by alveolar macrophages after O₃ exposure was not increased in any of the TLR5-deficient individuals (rs5744168 minor-allele carriers), while it was increased in wildtype-allele carriers (Figure 4C and Figure 4—figure supplement 1D).

To investigate the mechanistic role for TLR5 in the response to LPS-induced TLR4 activation, we then compared primary bone marrow derived macrophages (BMDM) from Tlr5-deficient and -sufficient mice. Tlr5-deficient BMDM had significantly decreased expression of TNFα and IL-6 after ultrapure LPS exposure in vitro, by an average of 30–50% (Figure 5A). To ensure that the altered response was not due to LPS contaminants despite the ultrapure preparation, we also assayed Tlr4-deficient BMDM and saw no response to ultrapure LPS (Figure 5A). The effect of Tlr5-deficiency was also observed when using other sources of Tlr4 activation, such as Monophosphoryl Lipid A (Figure 5—figure supplement 1A). To ascertain that our observation was not due to an off-target effect of genetic Tlr5 ablation on macrophage biology, we used RAW264.7 cells, a murine macrophage cell line, which are naturally deficient in Tlr5 (McDonald et al., 2007). We transfected these cells with a murine Tlr5 construct, or empty vector, and noted that the Tlr5-transfected cells had significantly higher TNFα expression in response to ultrapure LPS (and the TLR5 ligand flagellin, as expected). The TLR5-dependent effect was specific to LPS exposure, as responses to poly(I:C) (TLR3 ligand), Pam3CSK4 (TLR2 ligand), and ODN (TLR9 ligand) were not affected by the presence of TLR5 (Figure 5B).

We then interrogated the effect of TLR5 on TLR4 signaling. TLR4 signals through both MyD88 and TRIF pathways, whereas TLR5 is so far only known to signal through MyD88 (Kawai and Akira, 2010). We found that, in primary murine and human BMDMs, MyD88 co-immunoprecipitated with TLR5 after ultrapure LPS exposure (Figure 6A, Figure 6—figure supplement 1). This association was directly dependent upon TLR4 activation, since it was not observed in Tlr4-deficient primary BMDM (Figure 6B). TLR4, as expected, also immunoprecipitated with MyD88 after LPS exposure (Figure 6B, Figure 6—figure supplement 1). Furthermore, the association of MyD88 with TLR4 and TLR5 was specific, since other TLR like TLR6 and TLR7 did not immunoprecipitate with MyD88 after LPS exposure (Figure 6B and C). MyD88 signaling occurs through the complexing of Myd88 with IRAK-4 which stabilizes formation of the so-called Myddosome (Gay et al., 2011). Tlr5-deficient BMDM had decreased immunoprecipitation of IRAK-4 with MyD88 (Figure 6C). There was also significant reduction in the phosphorylation of IKKα/β, IκB, p65, JNK1/2 and ERK1/2 in Tlr5-deficient BMDMs compared with wildtype cells during early activation (Figure 7A,B, Figure 7—figure supplement 1A), which is MyD88- but not TRIF-dependent (Kawai et al., 2001; Yamamoto et al., 2003). Nuclear IRF3 (a specific readout of TRIF-dependent signaling Honda and Taniguchi, 2006) was not affected in Tlr5-deficient cells (Figure 7—figure supplement 1B). In aggregate, these results support that TLR5 directly interacts with MyD88 after LPS exposure and enhances MyD88-dependent TLR4 signaling by promoting efficient assembly of Myddosome.

We then investigated whether TLR5 participates directly in the TLR4 signaling complex, or whether it affects TLR4 signaling indirectly. We first evaluated whether TLR5 affects TLR4 cell surface expression and trafficking using bone marrow-derived macrophages (BMDM) from genetically deficient or wild-type mice. There were no differences in basal levels of cell surface TLR4 or CD14 between Tlr5-deficient and –sufficient BMDM, nor in LPS-induced internalization of TLR4 (Figure 8—figure supplement 1A–D). Because we were unable to find commercially available validated antibodies against TLR4 that could be used in co-immunoprecipitation experiments, we utilized an induced expression system using tagged TLR4 and TLR5 in HEK293 cells and found that TLR4 and TLR5 reciprocally co-immunoprecipitated in transfected HEK293 cells (Figure 8A). We then overexpressed TLR4 and TLR5 in HeLa cells and confirmed their interaction through a Proximity Ligation Assay (Figure 8—figure supplement 1E). Furthermore, we exposed TLR5-hemagglutinin tagged expressing RAW264.7 cells to biotin-tagged LPS, and (after thoroughly washing the cells) could co-precipitate TLR5 with LPS (Figure 8B). These results suggest that TLR5 directly participates in the TLR4 signaling complex after LPS exposure, and, in aggregate, support a functional interaction of TLR5 with TLR4 in the response to environmental injury.

C57Bl/7J mice and B6.129S1-Tlr5^tm1Flv/J (TLR5-deficient) mice were purchased from the Jackson Laboratory (Bar Harbor, ME). The TLR5-deficient allele was generated in the 129S1 genome and subsequently backcrossed to C57BL/6 before being transferred to the Jackson Laboratory by the donating investigator. When possible, wildtype littermate control mice were used in our study in parallel to commercial wildtype C57BL/6J controls. To ensure that our results were not due to locality-influenced microbiome changes, we repeated experiments in mice that were bred at the NIEHS vivarium, as well as mice purchased from the Jackson Laboratory and studied within 1 week of arrival. In some experiments, mice received neomycin water to control gut microbiome. Results were comparable independent of provenance or antibiotic dosing. Mice were given access to water and chow ad libitum and were maintained at a 12 hr dark-light cycle. No differences in body weight were observed at the ages studied (6–12 weeks old). All experiments are approved by the NIEHS Institutional Animal Care and Use Committee.

Mice received ultrapure E. coli O111:B4 LPS (List Biological Labs, Campbell, CA) (50 μl of 1 mg/ml in PBS), or control PBS vehicle only, by oropharyngeal aspiration and were phenotyped 24 hr later. For systemic LPS exposure, mice received 10 mg/kg LPS or PBS by intraperitoneal injection and were phenotyped 24 hr later. In other experiments, mice were exposed to 2 ppm ozone for 3 hr, in a chamber with 20 exchanges/hour, 50–65% relative humidity and a temperature of 20–25° C as previously described (Garantziotis et al., 2009) and were phenotyped 24 hr later. Control mice received filtered air in an identical setup. In other experiments, mice received 50 µl of a 3 mg/ml solution of sonicated, LPS-free, pharmaceutical-grade hyaluronan with an average molecular weight of 100–300 kDa (derived from Healon, Abbott Laboratories, Abbott Park, IL) or PBS vehicle by retropharyngeal aspiration, and were phenotyped 2 hr later.

Airway responsiveness to 25–100 mg/ml methacholine (MCh; Sigma) was measured 24 hr following O₃ or 2 hr following hyaluronan challenge. Briefly, mice were anesthetized with urethane (2 g/kg; Sigma), tracheotomized with a tracheal cannula (Harvard Apparatus) with Luer adapter, and mechanically ventilated on a 42°C water-heated pad at a rate of 150 breaths/min, a tidal volume of 10 ml/kg and a positive end-expiratory pressure (PEEP) of 3 cm H₂O with a computer-controlled small animal ventilator (FlexiVent, Scireq, Montreal, Canada). To block spontaneous breathing, mice were given pancuronium bromide i.p. (0.8 mg/kg; Sigma-Aldrich) 5 min prior to assessment of airway responses. To measure airway responsiveness, a single-frequency forced oscillation waveform, followed by a broadband forced oscillation waveform (matched to the animal breathing frequency) were applied using the Flexiware 7.6 software default mouse inhaled dose-response script. The resulting pressure, volume, and flow signals were fit to either the Single Compartment or Constant Phase model of the lung to obtain total respiratory system resistance (R_rs) and elastance (E_rs) or Newtonian resistance (R_n, generally understood as proximal airway resistance), tissue damping (G, generally understood as peripheral tissue resistance), and tissue elastance (H), respectively (Irvin and Bates, 2003). The peak response at each dose was averaged and graphed along with the average baseline measurement for each group.

TLR4-FLAG M2 or MyD88-V5 and TLR5-HA were overexpressed in HeLa cells grown on glass coverslips. 24 hr post-transfection the cells were fixed with 4% paraformaldehyde (PFA) for 10 min at room temperature (RT) and blocked with 10% normal goat serum for 1 hr at RT. The cells were next permeabilized with 0.1% TritonX-100 in goat serum for 15 min at RT and incubated with primary antibodies (dilution 1:1000) against epitope-tags overnight: rabbit anti-FLAG M2 (Cell Signaling), mouse anti-V5 (Invitrogen) and mouse anti-HA (Sigma). Duolink, based on in situ proximity ligation assay (PLA), was performed according to manufacturer instructions (Sigma).

Bone marrow was collected from tibias and femurs of wildtype and Tlr5-deficient mice and cell single preparations were made. Growth medium for maturation of BMDM consisted of DMEM-F12 containing 10 mM L-glutamine, 10% embryonic stem cell qualified fetal bovine serum, 1% antibiotic and antifungal mix and 30 ng/mL murine M-CSF. Cells were cultivated in an incubator at 37°C, 5% CO₂ for up to 7 days with medium change every 48 hr after first medium change 72 hr after platting. Tlr5-deficient BMDM were evaluated for responsiveness to flagellin and were found to be unresponsive (Figure 5—figure supplement 1A).

We used HeLa cells and Raw264.7 cells. Neither cell line is among the list of commonly misidentified cell lines. All cell lines were procured by ATCC and were free of mycoplasma contamination.

In order to access cell surface expression of TLR4, Wildtype, Tlr4-deficient or Tlr5-deficient BMDM were harvested, washed with PBS and were exposed to ultrapure LPS for 0,15,30, 60 or 90 min. Cells were washed with cold PBS and gently lifted from the culture dishes using a cell lifter. Cells numbers were estimated and cells were aliquoted in 1 × 10⁶ cells per tube in the FACS buffer (0.5% BSA, 0.1% NaN3, and 2 mM EDTA in PBS). Cells were blocked for 20 min on ice in a blocking solution (FACS buffer, 10% species specific serum, and 1% FCR block). Cells were stained using APC anti-mouse CD284 (TLR4) Antibody (clone SA15-21), anti-CD14 Antibody (Biolegend) or isotype controls for 30 min on ice. Cells were washed two times with 1 mL FACS buffer after staining, suspended in 500 uL FACS buffer containing 1 mg/mL propidium iodide (to identify dead cells) and analyzed on a BD FACSAria II equipment.

Wildtype or Tlr5-deficient BMDM were harvested, washed once with cold PBS, and lysed for 30 min at 4°C in 1% TritonX-100, 20 mM HEPES (pH 7.4), 150 mM NaCl, 1.5 mM MgCl₂, 2 mM EGTA, protease and phosphatase inhibitors (Roche). Cellular debris was removed by centrifugation at 16,000rcf for 10 min. For immunoblotting, cell extracts were fractionated by SDS-PAGE and transferred to Immobilon-P transfer membranes (Millipore), using either a wet transfer apparatus (Bio-Rad) or with a dry transfer system (iBlot) from Invitrogen. Immunoblot analysis was performed, and the bands were visualized with HRP-coupled goat anti-rabbit, goat anti mouse, or donkey anti-goat Ig as appropriate (Rockland), using the ECL western blotting detection system (GE Healthcare). Protein levels were equilibrated with the Protein Assay Reagent (Bio-Rad).

For coimmunoprecipitations, cells were harvested, washed once with cold PBS, and lysed in a TritonX-100-containing buffer (0.5% TritonX-100, 20 mM HEPES (pH 7.4), 150 mM NaCl, 1.5 mM MgCl₂, 2 mM EGTA, protease and phosphatase inhibitors (Roche). Cell extracts were incubated with 1 μg of Ab (anti-HA, Sigma) or normal IgG (negative control) for 2 hr, followed by incubation for 12 hr with 30 μl of protein G-Sepharose beads (prewashed and resuspended in lysis buffer at a 1:1 ratio). After incubations, the beads were washed four times with lysis buffer, separated by SDS-PAGE, and analyzed by immunoblotting. For TLR4, TLR5, TLR6 and TLR7 antibodies, blocking agent was 5% BSA and antibody dilution was 1:10,000. For Myd88 antibody, blocking agent was 5% milk and antibody dilution was 1:1000. Antibodies used were TLR4: Cat#482300 (Life Technology). TLR5: Cat#PA1-41139 (Invitrogen), TLR6: Cat#AF1533 (R and D Systems), TLR7: Cat# MAB7156 (R and D Systems), Myd88: Cat#4283 (cell signaling).

ELISA assay was performed using either R and D Duoset assay kits or Luminex multiplex assays according to manufacturer recommendations.

The complete details of the human exposure studies and the subject characteristics were previously published (Frush et al., 2016). After obtaining informed consent through a Duke University Institutional Review Board approved protocol, healthy human subjects were exposed to filtered air and ozone (200 parts per billion) in a crossover challenge designed study. Exposures were for 135 min, during which participants alternated between resting and walking on a treadmill at 2–3 mph to mimic an individual performing mildly strenuous activity under ambient conditions. Ozone was created from a 100% O2 source by cold plasma corona discharge (Ozotech, Yreka CA), and mixed with filtered air before addition to chamber and was continuously monitored. The order of filtered air or ozone exposure was randomized for every participant, with at least a 21 day washout period. Approximately 20 hr after exposure, participants underwent a flexible bronchoscopy with bronchoalveolar lavage. Following bronchial alveolar lavage, human alveolar macrophages were isolated. After red cell lysis and counting, the macrophages were re-suspended in medium (RPMI1640 with 10% heat-inactivated FBS, 100 units/ml penicillin, and 100 μg/ml streptomycin) and plated in a 24 well plate at a density of 200,000 cells per well. The cells were maintained in a CO2 incubator at 37° C for 2 hr. After 2 hr, the medium was replaced to remove non-adherent cells and then 2 hr later, the supernatant was collected and the cells were harvested for RNA. RNA extraction was performed using the Fourth Edition Qiagen Protocol (Qiagen, RNeasy Mini Kit, 4th edition, Valencia, CA), followed by DNase treatment (DNase I, Ambion, Austin, TX) and cDNA synthesis (BioRad). RT-PCR was performed on an ABI SDS 7500 (Applied Biosystems) using SYBR Green Reagent (Clontec Laboratories Inc, Mountain View, CA). TLR5 expression was determined in comparison to the 18 s RNA housekeeping gene and the data reported as fold change over the matched filter air sample for each individual subject. The following primers were used for RT-PCR: 18 s (Fwd: GTAACCCGTTGAACCCCATT, Rev: CCATCCAATCGGTAGTAGCG); TLR4 (Fwd: GGCCATTGCTGCCAACAT, Rev: CAACAATCA CCTTTCGGCTTTT), TLR5 (Fwd: TGTATGCACTGTCACTCTGACTCTGT, Rev: AGCCCCGGAACTTTGTGACT). Human TNF-α was measured from the cell supernatants via ELISA (MAX Standard Set kit, BioLegend) according to manufactures instructions. Readings were taken using BMG LABTECH Omega (Software Version 1.20). In a second study, participants were invited according to their rs5744168 genotype. Peripheral blood was drawn, and monocyte-derived macrophages isolated after 7 days in culture and exposed to ultrapure LPS or flagellin as indicated.

Data are represented as mean ± s.e.m. and were analyzed depending on experimental design by either analysis of variance (one-way or two-way ANOVA) followed by Tukey’s post hoc test or by unpaired t test with Welch's or Holm-Sidak correction as appropriate. Tlr5 gene expression and TNF-α production data from ozone expose human volunteer macrophages is presented as individual values and analyzed by Wilcoxon pairwise signed rank test.

All clinical studies described in this work were approved by the Institutional Review Boards of the NIEHS and Duke University respectively. Written informed consent was received from all participants prior to inclusion in the described studies.