Interplay between NOD1 and TLR4 Receptors in Macrophages: Nonsynergistic Activation of Signaling Pathways Results in Synergistic Induction of Proinflammatory Gene Expression

Innate immune system detects microbial invasion by pattern-recognition receptors (PRR), which sense conservative microbial structures, the well-known pathogen-associated molecular patterns (PAMP). During the past 20 years, mechanisms of action of individual PRRs have been studied in detail. However, microbes usually carry several PAMPs, which activate different PRRs. The parameters of innate immune response during infections are largely determined by PRR interactions (1).

A common type of PRR interaction is synergy, whereby the response to combined stimulation of two PRRs is greater than the sum of responses to stimulation of each individual PRR. In particular, a well-known synergistic interaction is the one between TLRs, which recognize diverse bacterial and viral PAMPs (2), and two members of NOD-like receptors family, NOD1 and NOD2, both of which recognize muropeptide fragments of bacterial peptidoglycan (3). Together, these two groups of receptors play the main role in bacterial pathogen recognition. Synergistic induction of cytokine production by combinations of NOD1/NOD2 and TLR agonists has been demonstrated in human and murine cell cultures in vitro (4–11) and in experimental animals in vivo (12–14). Hyperproduction of cytokines caused by synergistic PRR interactions can potentially lead to the life-threatening systemic inflammatory response underlying sepsis and septic shock (13). In contrast, controlled synergistic interactions of NODs and TLRs can be used therapeutically to enhance the host’s resistance against pathogens. For instance, a combination of a NOD1 and a TLR5 agonist confers an 80% resistance against a lethal dose of Salmonella typhimurium, whereas each individual component is nonprotective (14). When used as a vaccine adjuvant, a combination of a NOD2 and a TLR4 agonist demonstrates stronger activity compared with each individual agonist (15). Together, these data highlight the importance of NOD–TLR synergy. However, mechanisms of synergistic NOD–TLR interactions are largely unknown (16).

A key role in NOD1/2 and TLR4 signaling belongs to the NF-κB family of transcription factors, which includes five proteins (p105/p50, p65/RelA, c-Rel, RelB, and p100/p52) (17, 18). p50, p65, c-Rel, and RelB participate in the so-called canonical pathway of NF-κB activation (17). In resting cells, these proteins form homo- or heterodimers that are bound to inhibitory IκB proteins and sequestered in the cytoplasm. Signaling pathways downstream of NOD1/2 or TLRs activate IκB kinase (IKK), which phosphorylates IκB proteins, targeting them for proteasomal degradation (19–22). When liberated from IκB, NF-κB dimers are transported to the nucleus, where they regulate the expression of 500+ innate immune response genes. The effect on mRNA transcription depends on the composition of dimers; homo- or heterodimers containing p65 and c-Rel promote transcription, whereas p50:p50 homodimers inhibit it (17, 23). In the noncanonical NF-κB activation pathway, the key role is played by p100/p52. This signaling pathway is involved in NOD- and TLR-dependent signaling at relatively late time points (24).

Because NOD- and TLR-dependent signaling pathways converge on IKK, it was attempted to explain NOD–TLR synergy by synergistic IKK activation. According to one study, combined stimulation of NOD2 and TLR2 in monocytoid THP-1 cells results in a stronger and more prolonged IKK activation and, consequently, in a faster and more complete degradation of IκBα protein (20). However, it is not obvious that the convergence of NOD- and TLR-dependent pathways on IKK should lead to its synergistic activation. More likely, each pathway would activate a certain part of the total IKK pool, leading to the summation of effects but not synergy. Indeed, according to another work, combined 20-min stimulation of THP-1 cells by NOD2 and TLR4 agonists results in summation of phosphorylated IKKα/β and IκBα levels, but no synergy is observed at this stage of activation (15).

In this study, we investigated several consecutive stages of activation of human macrophages treated by a NOD1 agonist (N-acetyl-d-muramyl-l-alanyl-d-isoglutamyl-meso-diaminopimelic acid [M-triDAP]), a TLR4 agonist (LPS), and their combination in vitro. We show that up to the arrival of the activation signal in the nucleus, no synergy occurs between the two receptors. However, at the output from the nucleus, synergistic induction of a subset of NOD1- and TLR4-inducible genes is observed.

Recombinant human GM-CSF and IL-1R antagonist were purchased from Miltenyi Biotec (Bergisch Gladbach, Germany) and Life Technologies (Paisley, UK), respectively. Goat polyclonal neutralizing Abs against human TNF, IL-1α, and IL-1β as well as normal goat IgG were from R&D Systems (Minneapolis, MN), and polyclonal neutralizing sheep antisera against IFN-α and IFN-β were from Life Technologies. M-triDAP and lauroyl-d-isoglutamyl-meso-diaminopimelic acid (C12-iE-DAP) were from InvivoGen (San Diego, CA), LPS Escherichia coli O111:B4 was from Merck Millipore (Billerica, MA), SB203580 and polymyxin B were from Sigma (St. Louis, MO), ML130 (Noditinib-1), PF-184, amlexanox, and VX-745 were from Tocris Bioscience (Bristol, UK).

The study was approved by the local ethical committee of the Institute of Immunology (protocol no. 1/18). To generate human macrophages, blood monocytes from anonymous donors were cultured for 6 d in RPMI-1640 (Life Technologies) supplemented with 2-mM l-glutamine (Life Technologies), 10% FCS, and 40 ng/ml GM-CSF.

To assess cytokine production, cells were replated in 96-well flat-bottom plates at 20,000 cells/well and stimulated by different concentrations of M-triDAP and/or LPS in duplicates. For combined stimulation, agonists were added simultaneously. Supernatants were collected at the indicated time points and kept at −70°C until analysis. For RT-PCR, Western blotting, and measurements of glucose consumption and lactate release, macrophages were stimulated in 24-well plates by M-triDAP (10 μg/ml), LPS (10 ng/ml), their combination, or left untreated, and samples were collected at the indicated time points.

Inhibitors were added 15 min before or 60 min after the addition of the agonists. Neutralizing Abs and antisera against TNF (final concentration, 5 μg/ml), IL-1α (5 μg/ml), IL-1β (5 μg/ml), IFN-α (5000 NU/ml), IFN-β (1360 NU/ml), normal goat IgG (5 μg/ml), or IL-1R agonist (500 ng/ml) were added 15 min prior to the addition of agonists.

Experiments with supernatant transfer were done as follows. Macrophages were seeded in six-well plates at 10⁶ cells/well and stimulated with 10 μg/ml M-triDAP, 10 ng/ml LPS, their combination, or left untreated for 1 h. Media were then removed, and cell monolayers were washed twice with 5 ml warm RPMI and recultured for another 3 h with 2 ml complete culture medium without agonists. These supernatants were collected, incubated without or with polymyxin B (25 μg/ml final, 37°C, 10 min), and added neat to unstimulated autologous macrophage monolayers, without or with M-triDAP (10 μg/ml) or LPS (10 ng/ml). After 4 h, these macrophages were collected and used for RT-PCR measurements of cytokine mRNA.

Levels of TNF and IL-6 in cell culture supernatants were quantified by sandwich ELISA using kits from Thermo Fisher Scientific/eBioscience (Waltham, MA).

Glucose consumption and lactate release per cell per 24 h were determined, respectively, by SYNCHRON CX5 PRO Biochemical Analyzer (Beckman Coulter, Brea, CA) and by a colorimetric assay (BioVision, Milpitas, CA), as described (25).

Real-time measurements of extracellular medium acidification rate (ECAR) and oxygen consumption rate were performed using Seahorse XFe96 Analyzer (Agilent Technologies, Santa Clara, CA), as described (25). Briefly, macrophages were seeded in Seahorse XF96 plates (Agilent Technologies) at 16,000 cells per 80 μl complete medium without GM-CSF and allowed to rest overnight. Then, medium was exchanged to the Seahorse assay medium (XF base medium [Agilent Technologies] supplemented by 2-mM l-glutamine, 11-mM d-glucose [Sigma], and 10% FCS). After equilibration of pH in an atmospheric air incubator (37°C), plates were transferred to the analyzer. ECAR and oxygen consumption rate were measured every 9 min (duration of measurement, 3 min). After 3 basal measurements, medium, M-triDAP (final concentration, 10 μg/ml), LPS (final concentration, 10 ng/ml), or their combination was injected, and another 22 measurements were done. Areas under time-response curves (AUC) after the addition of stimuli were calculated, taking the two measurements preceding agonist injection as baseline.

RNA extraction, reverse transcription, and PCR were performed as described (25). Genes studied and primers used are listed in Supplemental Table I. Relative expression was calculated using the 2^−ΔΔCt method, taking unstimulated cells at the earliest time point in the given experiment as the reference and GAPDH as the housekeeping gene.

Macrophages were stimulated in 24-well plates by M-triDAP (10 μg/ml), LPS (10 ng/ml), their combination, or left untreated. Cell specimens were collected every 15 min. Briefly, macrophage monolayers were rinsed with ice-cold PBS and treated with a hypotonic buffer (10 mM HEPES, 1.5 mM MgCl₂, and 10 mM KCl [pH 7.9]) supplemented with cOmplete Mini Protease Inhibitor Mixture (Roche) for 12 min on ice. Nonidet P-40 was then added to the final concentration of 0.6% for 30 s, whereafter cell lysates were vigorously pipetted. Nuclei were sedimented by centrifugation (2000 g, 5 min), and cytosolic fraction was removed. Nuclei were washed once with the above-mentioned hypotonic buffer (2000 g, 5 min) and lysed in Western blot sample buffer (1.6% SDS, 5% glycerol, 0.1 M DTT, 0.02% bromophenol blue, and 0.05 M Tris [pH 6.8]). Cytosolic fraction was supplemented with the same buffer. All samples were heated for 5 min at 90°C and then frozen at −70°C until Western blot analysis. To verify purity of nuclear and cytosolic fractions, both were assayed for α-tubulin and histone H3. The absence of α-tubulin in the nuclear fraction and of histone H3 in the cytosolic fraction was confirmed in all samples.

To assay for phosphorylated proteins, stimulated and unstimulated cells were washed with ice-cold PBS and lysed in an ice-cold buffer containing 150 mM NaCl, 50 mM Tris (pH 8.0), 1% Triton X-100, and a mixture of protease and phosphatase inhibitors (MS-SAFE, Sigma).

Western blotting was performed as described (25). Primary rabbit Abs against human phospho-Akt (pT308 and pS473), phospho-p70-S6K (pT389), phospho-p38 (pT180/pY182), phospho-MNK1 (pT197/pT202), phospho-eIF4E (pS209), NF-κB p65/RelA, c-Rel, p105/p50, p100/p52, IκBα, A20, aconitate decarboxylase 1 (ACOD1)/IRG1, and histone H3 as well as primary mouse Abs against phospho-ERK1/2 (pT202/pY204) were from Cell Signaling Technology (Danvers, MA); anti-human IκBε rabbit mAb was from Abcam (Cambridge, UK); and anti-human α-tubulin mAb (clone DM1A) was from Novus Biologicals (Centennial, CO). Secondary Abs labeled with HRP were from Jackson ImmunoResearch Laboratories (West Grove, PA). The staining was developed using Clarity Western ECL Substrate (Bio-Rad Laboratories, Hercules, CA) and detected by an AI600 imager (Amersham Biosciences, Amersham, UK). Densitometry was performed using ImageJ freeware (http://imagej.nih.gov/ij). A linear relationship between the amount of protein and output signal (integrated gray intensities of bands) was confirmed in additional experiments by assessing serially diluted samples (data not shown).

Experimental protocol was as described (26), with modifications. Briefly, macrophages were seeded into eight-well poly-d-lysine–coated chamber slides (SPL Life Sciences, Pocheon-si, South Korea) at 5 × 10⁴ cells/well and allowed to adhere overnight. The next day, cells were stimulated with M-triDAP (10 μg/ml), LPS (10 ng/ml), their combination, or left untreated. At desired time points, cells were fixed with 4% paraformaldehyde, permeabilized with 95% ice-cold ethanol, and stained with rabbit anti-human p65 Abs (Cell Signaling Technology, catalog no. 8242). Staining was developed using donkey anti-rabbit IgG NorthernLights NL493-conjugated Ab (R&D Systems). Slides were mounted with ProLong Gold Antifade Medium containing DAPI (Thermo Fisher Scientific) and examined using Leica DM LB fluorescent microscope (Leica Microsystems, Wetzlar, Germany). Images were analyzed using ImageJ. Nuclear regions of interest were set according to DAPI staining, and cytoplasmic regions of interest were set according to cytoplasmic p65 staining. At least 100 cells per condition were examined. For each cell, mean intensities of p65 staining in the cytoplasm and in the nucleus were measured, and nucleus/cytoplasm ratios (NCR) were calculated. Nuclear NF-κB–positive cells were defined as those where NCR exceeded mean NCR + 2 SD of unstimulated cells.

To assess the effect of NOD1 and TLR4 agonist combination on parameters of cell activation, we calculated indices of synergy (IS) using the following formula: IS = Response_M-triDAP+LPS/(Response_M-triDAP + Response_LPS), where response is the difference between stimulated and basal values of the parameter.

Agonist interaction was considered 1) synergistic if IS was significantly higher than 1 (p < 0.05 in Student t test), 2) additive if IS was not significantly different from 1, or 3) infra-additive if response to the agonist combination was significantly lower than the sum of responses to individual agonists (IS < 1, p < 0.05) but not lower than the response to each individual agonist. Statistical analysis was performed using GraphPad Instat 3.06 и GraphPad Prism 6 (GraphPad Software, San Diego, CA).

Statistical analysis was done using paired Student t test unless otherwise indicated. Correlations were analyzed using Pearson test.

First, we determined levels of TNF and IL-6 in macrophage supernatants at 3 and 24 h of stimulation with different concentrations M-triDAP and LPS added separately or in combination (simultaneously). At 3 h, no synergistic interactions were observed across the entire ranges of agonist concentrations (Fig. 1A, 1B, 1E, 1F). Interestingly, M-triDAP at even the lowest concentration tested (0.1 μg/ml) somewhat inhibited the response to 1–10 ng/ml LPS (Fig. 1A, 1E). At 24 h, however, strong synergistic effects of agonists were observed (Fig. 1C, 1D, 1G, 1H). Thus, the synergistic response to NOD1+TLR4 agonist combination develops relatively late upon activation. In subsequent experiments, we used M-triDAP at 10 μg/ml and LPS at 10 ng/ml, i.e., concentrations at which the two agonists produce a robust synergistic effect at 24 h. Using these agonist concentrations, we performed an additional four experiments to confirm a time-dependent increase of IS for TNF levels in supernatants from 1.09 ± 0.21 at 60 min (p = 0.2 for difference from 1) to 2.04 ± 1.05 at 240 min (p = 0.047).

To better characterize the response of macrophages to combined agonist treatment, we studied mRNA expression of several cytokines, chemokines, and MX1 as a type I IFN response marker. All of these genes were induced by M-triDAP and by LPS, except for MX1, which was induced only by LPS (Fig. 2A). Upon combined stimulation, synergistic enhancement of TNF, IL6, IL1B, IL12B, and IL23A mRNA expression was observed but only at relatively late time points (4 and 10 h); at 1 h, the effect of agonist combination was nonsynergistic (Fig. 2A). Thus, the synergistic enhancement of TNF and IL-6 secretion shown above is due to synergistic enhancement of cytokine mRNA expression, which develops between 1 and 4 h of combined agonist treatment. Because cytokine secretion lags behind mRNA expression, synergy at the level of mRNA was observed earlier than at the level of secreted cytokines. Of note, CCL4 and MX1 mRNA did not show a significant synergistic induction at any time point studied (Fig. 2A).

To more precisely define the time frame of synergy development, we analyzed kinetics of TNF and IL6 mRNA expression during the first 4 h of stimulation. In the presence of M-triDAP or LPS, levels of TNF mRNA reached a maximum at 120 ± 30 min or 100 ± 17 min, respectively, and then began to decrease (Fig. 2B, 2F). Accumulation of TNF mRNA in M-triDAP–stimulated cells lagged behind that in LPS-stimulated cells, probably because of differences in subcellular localization of the respective receptors. During combined stimulation, levels of TNF mRNA initially followed the same track as in LPS-stimulated cells but from the 120th minute, became significantly higher than those in LPS-stimulated cells and peaked at 150 ± 30 min (i.e., significantly later than upon LPS stimulation) (Fig. 2B, 2F). A significant synergistic effect (IS > 1, p < 0,05) was registered from the 150th minute onward (Fig. 2C). Similar trends were observed for IL6 mRNA expression (Fig. 2D, 2E), except that this mRNA was induced with a slower kinetics than TNF mRNA (Fig. 2F). Thus, events leading to synergistic enhancement of TNF and IL6 mRNA expression take place between the 90th and 150th minutes of combined stimulation.

NF-κB proteins are important regulators of proinflammatory cytokine gene expression (27, 28). NF-κB signaling is not a one-time event but an oscillatory process with alternating phases of NF-κB transport to and from the nucleus. These NF-κB oscillations are most typical for TNF signaling but can also be observed upon TLR stimulation (29–32). We therefore determined levels of p65, c-Rel, and p50 in macrophage nuclei by Western blotting every 15 min during the initial 210 min of stimulation (i.e., the time frame of synergy development). Usually, we observed a single wave of nuclear NF-κB translocation (Fig. 3A–I) and sometimes possibly the beginning of a second wave continuing beyond the time frame of the experiment (e.g., Fig. 3E). As expected, nuclear levels of NF-κB peaked generally 40–50 min earlier than TNF mRNA, and nuclear accumulation of NF-κB in M-triDAP–stimulated cells lagged behind that in LPS-stimulated cells (Table I), similarly to the kinetics of TNF and IL6 mRNA.

The combined NOD1+TLR4 stimulation did not have a synergistic effect on the nuclear translocation of p65, c-Rel, and p50 (Fig. 3A–I). In the case of c-Rel, signals from NOD1 and TLR4 interacted additively (IS ≈ 1) and in the case of p65 and p50, infra-additively (IS < 1). p50 showed the lowest IS values (close to 0.5 most of the time; Fig. 3I). To obtain an integral assessment of nuclear NF-κB accumulation, we calculated AUCs from the 60th until the 180th minutes of stimulation (AUC_60–180) (i.e., the period when synergistic enhancement of TNF mRNA expression develops). As shown in Table I, combined treatment yielded a significantly greater accumulation of p65 than LPS treatment and a greater accumulation of c-Rel than M-triDAP treatment. Upon combined treatment, AUC_60–180 of c-Rel responded additively (IS ≈ 1), whereas AUC_60–180 of p65 and p50 responded infra-additively (IS < 1), which is in line with IS values at individual time points.

We did not detect p100/p52 in macrophage nuclei during the first 210 min of stimulation (data not shown), which speaks against the role of noncanonical NF-κB activation pathway in the development of NOD1–TLR4 synergy.

Increased levels of nuclear p65 and c-Rel in M-triDAP+LPS–stimulated cells could be due to increased nuclear translocation of p65 in individual cells or due to increased proportions of responding cells. Because the latter parameter cannot be determined by Western blotting, we assessed nuclear translocation of p65 in single cells by immunofluorescence (Supplemental Fig. 1A). Kinetics of p65 NCR obtained by immunofluorescence was roughly similar to that of nuclear p65 measured by Western blotting. M-triDAP+LPS combination produced significantly higher NCRs than either agonist alone, although the effect of agonist combination was infra-additive (IS = 0.5–1; Supplemental Fig. 1B, 1C). At the top of the response (60–90 min), nearly 100% of M-triDAP–, LPS-, or M-triDAP+LPS–stimulated macrophages were nuclear p65 positive, indicating that all cells responded even to a single agonist (Supplemental Fig. 1D). At later time points, however, higher percentages of nuclear p65–positive cells were observed among macrophages stimulated by M-triDAP+LPS than by individual agonists (Supplemental Fig. 1D). Additionally, we assessed NCRs selectively in nuclear p65–positive cells and again found this parameter to be significantly higher in macrophages stimulated by agonist combination than by either agonist alone (Supplemental Fig. 1E). Thus, both immunofluorescence and Western blotting showed that macrophages stimulated by M-triDAP+LPS attain higher levels of nuclear p65 than cells treated by individual agonists; however, the effect of agonist combination is nonsynergistic.

We conclude that upon simultaneous triggering of NOD1 and TLR4, nuclear NF-κB translocation is regulated nonsynergistically.

Transcription factors are known to function in temporal waves, when pre-existing (“primary”) transcription factors activated directly downstream of signaling receptors trigger the expression of secondary transcription factors, which may interact with primary transcription factors resulting in amplification of the response. We first analyzed mRNA expression of the five NF-κB genes (NFKB1 [p105/p50], NFKB2 [p100/p52], NFKB3 [p65/RelA], REL [c-Rel], and RELB). All of them were upregulated by 4 h of stimulation (Supplemental Fig. 2A). Of note, combined NOD1+TLR4 stimulation produced an additive but not synergistic induction of these genes’ expression (Supplemental Fig. 2A).

IRF5 and C/EBPδ are transcription factors that co-operate with NF-κB at inducing expression of proinflammatory genes (33, 34). In particular, IRF5 facilitates activation-induced looping of the TNF gene, a mechanism proposed to explain sustained production of TNF by LPS-activated dendritic cells (35). IRF5 protein was present in macrophage nuclei, but its levels in the nucleus as well as IRF5 mRNA expression remained unaltered upon stimulation (Supplemental Fig. 2A). C/EBPδ is a “second-wave” transcription factor induced by LPS treatment (33, 36). CEBPD mRNA was inconsistently induced by M-triDAP or LPS alone, but their combination produced a synergistic induction of this gene (Supplemental Fig. 2A). Increased expression of NF-κB and C/EBPδ induced by agonist combination may contribute to sustained NF-κB signaling and enhanced cytokine production at relatively late stages of macrophage activation.

Proteasomal degradation of IκB is a prerequisite for nuclear NF-κB translocation. IκB proteins possess certain selectivity toward specific NF-κB proteins; in particular, IκBα binds the most common p65:p50 dimers, whereas IκBε prefers p65:p65 and p65:c-Rel dimers (17). IκBα is rapidly resynthesized because of increased transcription of NFKBIA gene, which is itself NF-κB inducible (37). The newly synthesized IκBα is transported to the nucleus, where it binds NF-κB proteins facilitating their dissociation from promoters and export from the nucleus (23, 38). IκBα therefore facilitates termination of NF-κB signaling. Another negative regulator of the NF-κB pathway is A20 deubiquitinase (encoded by TNFAIP3 gene, which is also NF-κB inducible) (39). A20 removes activating K63-linked polyubiquitin chains from different signaling proteins including IKKγ and helps terminate NOD- and TLR-dependent NF-κB activation (40–42).

In LPS-treated macrophages, levels of IκBα in the cytoplasm reached a minimum (<5% of baseline) after 25 ± 9 min of stimulation (Fig. 3J, 3K, Table I). During M-triDAP treatment, IκBα degradation was less complete, and a minimum of cytoplasmic IκBα was delayed till 45 ± 26 min. Upon combined treatment, a minimum of IκBα in the cytoplasm was achieved as early as upon LPS treatment (Fig. 3J, 3K, Table I). As expected, the phase of low cytoplasmic IκBα coincided with the phase of rapid NF-κB accumulation in the nucleus. IκBα degradation was succeeded by IκBα resynthesis, with end point levels of cytoplasmic IκBα being 1.5–2.5 times higher than basal levels (M-triDAP < LPS ≈ M-triDAP+LPS) (Fig. 3J, 3K). Resynthesis was due to rapid upregulation of NFKBIA mRNA expression (Fig. 3L). Of note, combined NOD1+TLR4 stimulation had an additive but not synergistic effect on NFKBIA mRNA expression (Fig. 3L).

Nuclear levels of IκBα rose from the 90th minute of agonist treatment (Fig. 3M, 3N). During this period, IS for nuclear IκBα dwelled at around 0.5 (Fig. 3O), indicating no summation of responses to M-triDAP and LPS. AUC_60–180 for nuclear IκBα in M-triDAP+LPS–stimulated cells was lower than in LPS-stimulated cells, with IS ≤ 0.5 (Table I), pointing at a mutually inhibitory effect of M-triDAP and LPS on nuclear translocation of IκBα. From these data, it appears that a relative deficiency of IκBα may exist in the nuclei of macrophages when the synergistic effect develops.

Degradation of cytoplasmic IκBε was also observed with all types of stimulation (Supplemental Fig. 2B). The M-triDAP+LPS combination did not synergistically enhance IκBε degradation in the cytoplasm (Supplemental Fig. 2C). TNFAIP3 (A20) mRNA and protein expression was also upregulated in M-triDAP– and LPS-stimulated macrophages, but similarly to NFKBIA induction, no synergy between M-triDAP and LPS was observed (Supplemental Fig. 2A, 2D, 2E). Altogether, negative modules of NF-κB signaling are activated nonsynergistically upon simultaneous NOD1 and TLR4 triggering.

By NOD1–TLR4 cross-talk, we assume scenarios in which NOD1 stimulation upregulates the expression of proteins specifically involved in TLR4 signaling, and vice versa. These proteins are, most obviously, receptors themselves and their proximal adapters. We found that TLR4, NLRC1 (NOD1), TIRAP, and TICAM1 (TRIF) mRNA expression was not affected by any type of stimulation, MYD88 expression was modestly induced by M-triDAP+LPS, whereas TICAM2 (TRAM) and RIPK2 (RIP2) were induced by M-triDAP, LPS, and especially by their combination, although nonsynergistically (Fig. 4). TRAM functions together with TRIF in the TRIF-dependent signaling pathway downstream of TLR4; therefore, upregulation of TRAM in the absence of upregulation of TRIF might not impact the strength of the TLR4 signal. In contrast, upregulation of RIP2 as the sole NOD1 adapter may impact NOD1 signaling. RIP2 has been reported to be degraded after NOD1 or NOD2 triggering, a mechanism limiting the activation signal from these receptors (43). Transcriptional upregulation of RIP2 upon combined stimulation might counterbalance this degradation, resulting in augmented and/or prolonged NOD1 signaling.

It is known that some aspects of innate immune cell activation require a continuous presence of PRR agonists, whereas others do not (33). To test whether continued NOD1 stimulation is required for NOD1–TLR4 synergy, we inhibited RIP2 and NOD1 using SB203580 and ML130/noditinib-1, respectively (44, 45). SB203580 is a dual RIP2/p38 inhibitor that inhibits RIP2 with a lower IC₅₀ than it does p38 (44). At 5 μM, SB203580 interfered with macrophage response to M-triDAP but not to LPS, indicating specific inhibition of RIP2 but not p38 (Fig. 5A). When applied 60 min after the addition of agonists, SB203580 strongly inhibited M-triDAP– and M-triDAP+LPS–induced TNF mRNA expression after 4 h of stimulation and strongly reduced the synergistic effect (Fig. 5A). ML130, a specific NOD1 inhibitor, had a similar effect (Fig. 5A). We noticed that ML130 did not completely block M-triDAP–induced responses and did not completely abolish synergy, which could be due to the minor activity of M-triDAP toward NOD2 receptor (46), which also signals through RIP2 (47). Therefore, we repeated these experiments using C12-iE-DAP, a more specific NOD1 agonist (48). Similarly to M-triDAP, C12-iE-DAP strongly synergized with LPS, and this synergy was totally cancelled by ML130 added either before or 1 h after agonists (Fig. 5B). Thus, continuous NOD1–RIP2 signaling between 1 and 4 h of stimulation is required to produce the NOD1–TLR4 synergistic effect.

We also tested whether synergy could be abolished by inhibition of selected pathways downstream of NOD1/RIP2 and TLR4. An IKKβ inhibitor (PF-184) did not affect TNF mRNA expression at 4 h but inhibited it at 1 h of agonist treatment (Fig. 5C). VX-745, a specific p38 inhibitor, lowered M-triDAP–, LPS-, and combination-induced TNF mRNA expression after 4 h of stimulation (Fig. 5D). However, whereas both PF-184 and VX-745 dampened responses to M-triDAP and LPS, responses to M-triDAP+LPS remained synergistic in the presence of these inhibitors (Fig. 5C, D), indicating that NF-κB and p38 pathways are required for TNF mRNA expression but are not sufficient to produce synergy. Amlexanox, an inhibitor of TBK1/IKKε kinases that acts in the TRIF–TBK1–IRF3 pathway downstream of TLR4 (49), did not affect TNF expression induced either by separate agonists or by their combination (Fig. 5C).

Cytokines acting auto- or paracrinely, most commonly TNF, IL-1, and type I IFNs, have been reported to mediate some of the effects of PRR agonists on innate immune cells (50–52). However, neutralization of secreted TNF, IL-1α/β, IFN-α, and IFN-β had no effect on TNF, IL-6, and IL-1β production and/or expression by macrophages activated/induced either by single agonists or by their combination (Supplemental Fig. 3A, 3B). Additionally, we collected supernatants from M-triDAP– and LPS-stimulated macrophages (see Materials and Methods for details) and added them together with LPS or M-triDAP to resting macrophages. Supernatants from M-triDAP–treated cells did not enhance the macrophage response to LPS (Supplemental Fig. 3C). Interestingly, supernatants from LPS-treated macrophages enhanced response to M-triDAP, but this effect disappeared when residual LPS in these supernatants was neutralized by polymyxin B (Supplemental Fig. 3C). Altogether, these data indicate that soluble mediators secreted by macrophages are unlikely to contribute significantly to NOD1–TLR4 synergy.

We also studied other aspects of macrophage activation by the combination of M-triDAP and LPS. Upregulation of glycolysis appears to be an integral part of macrophage response to inflammatory stimuli, including M-triDAP and LPS (25). Enhanced glycolysis supplies energy and carbon backbone for the synthesis of proteins (including cytokines and antimicrobial peptides), phospholipid membranes, etc. (53). Macrophages upregulate ECAR, a key indicator of glycolysis, within 1 h after the addition of TLR or NOD1/2 agonists (25, 54, 55). Interestingly, this rapid boost of ECAR does not depend on de novo gene expression (55) and is presumably mediated by Akt-dependent translocation of hexokinase-II to the outer mitochondrial membrane (56). In this study, we confirmed that both M-triDAP and LPS boost ECAR in a dose-dependent fashion (Fig. 6A, 6B). Upon combined stimulation, synergy was seen only with the combination of the lowest agonist concentrations (Fig. 6C). In all other conditions, additive or infra-additive effects on ECAR were observed. In line with this, M-triDAP or LPS alone enhanced glucose consumption and lactate release, two main biochemical parameters of glycolysis; however, no synergy was observed upon combined stimulation (Fig. 6D, 6E). Thus, glycolysis is not boosted synergistically upon simultaneous NOD1 and TLR4 triggering.

In addition, we assessed the activation of signaling pathways controlling different aspects of cell metabolism, such as the Akt–mTORC1–p70 pathway, which mediates upregulation of protein synthesis and glycolysis (56, 57), and the p38/ERK1/2–MNK1/2–eIF4E pathway, which controls mRNA stability and protein translation (58, 59). Levels of phosphorylated Akt, p70, p38, ERK, MNK, and eIF4E were assessed by Western blotting 1 and 4 h after the addition of M-triDAP, LPS, or their combination. Upon combined stimulation, levels of the phospho-proteins examined were either unchanged in relation to baseline, or additive/infra-additive effects were observed (Fig. 6F). Synergy was not observed for any phosphoprotein studied.

We also studied expression of ACOD1, an important metabolic regulator strongly induced by LPS (60, 61). This gene was only weakly induced by M-triDAP, in agreement with our previous work (25); however, the M-triDAP+LPS combination had a strong synergistic effect on ACOD1 mRNA and protein expression (Fig. 6G, 6H).

Over the course of this study, we have analyzed the expression of 25 genes. Nineteen genes were inducible by NOD1 and/or TLR4 agonists, but only a subset of them showed a uniform synergistic induction upon combined NOD1+TLR4 stimulation (IS > 1, p < 0.05). We noticed that the synergistically induced genes (group 1) had a lower basal expression and a higher induction after 4-h stimulation with M-triDAP+LPS as compared with nonsynergistically induced genes (group 2) (Fig. 7A). IS for individual genes at 4 h positively correlated with their inducibility at 4 h and inversely with basal expression (Fig. 7B, 6C).

In this study, we have studied several layers of macrophage activation by simultaneously applied NOD1 and TLR4 receptor agonists. Our results are summarized in Fig. 8. First, we show that up to the entry of NOD1- and TLR4-dependent activation signals into the nucleus, no synergy occurs between these signals. Signaling events downstream of NOD1 and TLR4 that precede or do not require mRNA transcription, such as nuclear translocation of NF-κB, activation of Akt- and MAPK-dependent signaling pathways, and upregulation of glycolysis, can be summed up but not synergistically enhanced. Inhibition of NF-κB or p38-dependent pathways dampens responses to M-triDAP and LPS but does not cancel synergy. It appears that each receptor acts independently to activate a portion of the total pool of signaling molecules, which are largely shared by the two receptors. Some of the signaling modules shared by NODs and TLRs may be fully activated even by a single agonist; in that case, even summation of signals would not be observed (as may be the case with p50 nuclear translocation). Our data are not consistent with an earlier model whereby NOD–TLR synergy occurs at the level of IKK activation and IκBα degradation (20). Although we did not directly measure IKK activation, its consequence (i.e., nuclear translocation of NF-κB) was regulated nonsynergistically.

Second, although the NOD1+TLR4 agonist combination generally induces a higher gene expression than either agonist alone, only a subset of the M-triDAP– or LPS-inducible genes shows a synergistic induction (Fig. 7A). These synergistically responding genes tend to have a lower basal expression and a higher inducibility as compared with those responding nonsynergistically. The synergistically regulated subset of genes includes those coding for proinflammatory cytokines, such as TNF, IL1B, IL6, IL12B, and IL23A (i.e., most common read-outs used to define synergy). These genes belong to different groups with respect to timing and regulation of expression. For example, TNF is a primary response gene and its expression in LPS-stimulated murine macrophages is independent of de novo protein synthesis, whereas IL6, IL12B, and IL23A are secondary response genes that require de novo protein synthesis for full-scale expression (62, 63). At the same time, different NF-κB–regulated genes can be upregulated either synergistically or nonsynergistically (e.g., TNF and NFKBIA, respectively).

Third, the synergistic enhancement of proinflammatory cytokine expression develops within a certain time frame (i.e., after the 90th minute of combined agonist treatment) and is fully apparent at 4 h (at mRNA level) or ≥4 h (at the protein level). Kinetic data on IκB degradation, NF-κB nuclear translocation, and TNF mRNA expression in combination-treated cells show that very early stages of cell activation (up to ∼60 min) are dominated by the LPS response. For instance, LPS stimulation alone causes degradation of nearly all IκBα within 25 min, and simultaneous application of M-triDAP adds little more to this response (Fig. 3J, K). Similarly, levels of TNF mRNA in LPS-stimulated and in M-triDAP+LPS–stimulated macrophages follow the same track up to the 90th minute (Fig. 2B). In macrophages stimulated by LPS or M-triDAP alone, TNF mRNA reaches a peak at around 90–120 min and then decreases, whereas in M-triDAP+LPS–stimulated cells, it keeps rising, apparently because of continued TNF transcription. After the 150th minute, TNF mRNA in combination-stimulated macrophages remains at a plateau, presumably because of sustained transcription and/or increased mRNA stability. Experiments with SB203580 and ML130 showed that a continued “firing” through NOD1/RIP2 at this stage is required for the development of synergy (Fig. 5A, B).

The question is whether nonsynergistic interaction of NOD1- and TLR4-dependent signals at the level of nuclear NF-κB translocation can result in synergy at the level of cytokine mRNA transcription. As can be concluded from Fig. 3, higher concentrations of p65 and especially of c-Rel are maintained in the nuclei of macrophages stimulated by M-triDAP+LPS as compared with cells treated with individual agonists. Theoretically, this may result in a more stable occupation of promoters by p65 and c-Rel, leading to a higher expression of respective genes. However, it is not obvious that this expression would be synergistic. Notably, some NF-κB–regulated genes such as NFKBIA and TNFAIP3 do not show synergistic induction upon combined treatment.

Contributing to the development of synergy could be enhanced expression of several NF-κB proteins (Supplemental Fig. 2A) and relative deficiency of negative regulation of NF-κB signaling. For instance, although nuclear levels of c-Rel and p65 in M-triDAP+LPS–stimulated cells respond additively or infra-additively (that is, they are higher than in cells stimulated with M-triDAP or LPS alone), nuclear levels of IκBα upon combined stimulation are even lower than upon LPS stimulation (IS ≤ 0.5, Table I). Thus, at the time when TNF or IL6 mRNA expression is synergistically boosted (90–150 min of stimulation), relative deficiency of IκBα may exist in the nucleus, facilitating more stable interaction of NF-κB proteins with cytokine gene promoters and enhanced transcription. The reason for the deficiency of nuclear IκBα in M-triDAP+LPS–stimulated cells is unclear; no deficit of IκBα in the cytoplasm is observed at these same time points (Fig. 3J, 6K).

However, it is more likely that additional transcription factors cooperating with NF-κB are needed for the synergistic induction of a subset of NF-κB–regulated genes. These may be secondary transcription factors that have low initial expression levels but are rapidly upregulated upon combined NOD1+TLR4 stimulation. The time between the addition of agonists and development of the synergistic effect (90 min and later) should be sufficient for de novo mRNA expression and translation of these putative transcription factor(s). As a result, nuclear concentrations of the secondary transcription factors might exceed a certain threshold, resulting in a qualitative change of transcription (a switch-on effect). These transcription factors remain to be identified. A candidate could be C/EBPδ, which is synergistically induced at the transcriptional level by the M-triDAP+LPS combination (Supplemental Fig. 2A) and has been shown to control the duration of LPS-induced cytokine mRNA transcription (33).

An additional mechanism studied in this article is the cross-talk between NOD1- and TLR4-dependent signaling pathways. The principles of inter-PRR cross-talk were first described by Bagchi et al. who studied TLR–TLR interactions (64). They showed that a synergy between pairs of TLRs develops when the two receptors use different proximal adaptors (essentially, MyD88 or TRIF) but share distal parts of signaling pathways. A similar principle may apply to NOD1 and TLR4, which use different adaptors (RIP2 in case of NOD1; MyD88, TIRAP, TRIF, and TRAM in case of TLR4). Specifically, TLR4-dependent signals may enhance signaling through the NOD1 pathway by upregulating RIPK2 mRNA expression and preventing loss of RIP2 protein, which is otherwise observed after NOD1 stimulation (43). Supporting this idea is cancellation of the NOD1–TLR4 synergy by SB203580. Potentially, triggering of NOD1 and TLR4 may upregulate the expression of a broader set of genes and proteins that reciprocally augment signaling through TLR4- and NOD1-dependent pathways, respectively.

Innate immune responses are supported by metabolic reprogramming (i.e., extensive rewiring of innate immune cell metabolism). In this study, we show that those aspects of metabolic reprogramming that do not require de novo gene expression, such as the rapid boost of glycolysis (55), are regulated nonsynergistically upon combined NOD1 and TLR4 stimulation (Fig. 6). By contrast, processes that depend on de novo gene expression, such as ACOD1 expression, can be regulated synergistically. ACOD1 enzyme uses aconitate, a Krebs cycle intermediate, to generate itaconate, which is a powerful antimicrobial and immunomodulatory compound (60, 61). Therefore, combined activation of NODs and TLRs may result in synergistic enhancement of bactericidal activity of macrophages, as recently demonstrated by Zhou et al. (11).

In sum, we show the complexity of the response of human macrophages to the combination of NOD1 and TLR4 agonists. Up to the stage of target gene transcription, signals from NOD1 and TLR4 interact nonsynergistically. Furthermore, only a subset of NOD1 and TLR4 target genes is subject to synergistic induction. Activation of the NF-κB pathway alone is not sufficient to explain the synergistic effects. It appears that NOD1–TLR4 synergy, and perhaps synergy between other pairs of NODs and TLRs, is the result of many overlapping processes that include amplification of positive regulatory events and insufficiency of negative regulatory mechanisms.

The authors have no financial conflicts of interest.