Local TNFR1 Signaling Licenses Murine Neutrophils for Increased TLR-Dependent Cytokine and Eicosanoid Production

Innate immune responses can be initiated by a variety of signals broadly classified as either microbial or endogenous, the latter of which are the main drivers of sterile inflammation (1, 2). In responses induced by either class of ligand, recruitment of neutrophils is typically one of the earliest events. Neutrophils represent the most abundant leukocyte in the circulation and are critical for the elimination of microbial pathogens (3). However, neutrophil activation can also lead to tissue damage, such as joint and cartilage erosion during rheumatoid arthritis (4). Neutrophil recruitment can be mediated by multiple factors either produced by host cells, such as chemokines (e.g., CXCL1) and lipid mediators, or released by bacteria and mitochondria, such as formylated peptides. Engagement of specific G-protein coupled receptors expressed by neutrophils (CXCR2, BLT1, and members of the formylated peptide receptor [FPR] family) leads to their extravasation and migration to the site of injury or infection (5). Neutrophils also express many innate immune receptors, including TLR and C-type lectin receptors (CLR) (6), allowing them to recognize a large number of molecular patterns associated with both tissue damage and microbes locally at sites of inflammation.

Engagement of TLRs, CLRs, or FPRs generally leads to neutrophil activation, potentially inducing degranulation, production of reactive oxygen species (ROS) and lipid mediators, and, in some cases, release of the cell’s DNA to form neutrophil extracellular traps. Neutrophils can also produce cytokines and chemokines, such as TNF-α (7) and IL-10 (8), although the latter seems to be specific for mouse neutrophils (as discussed in Ref. 9). However, cytokine production by neutrophils has been largely understudied, in part because neutrophils are generally poor protein producers (10) and because of the species differences mentioned above. Although neutrophils are generally thought to contribute only modestly to cytokine production when compared with macrophages and monocytes, this issue has remained a subject of debate (11, 12).

There is now accumulating evidence that the functional properties of neutrophils, once thought to be fairly stereotyped, can be altered by the local microenvironment during chronic inflammation or within tumors (13). However, it remains unclear how local cues can influence neutrophils during the onset of inflammation. We reasoned that different inflammatory milieus may modify the capacity of neutrophils to produce anti-microbial compounds and pro- and anti-inflammatory cytokines. Such a mechanism could maximize elimination of microbes and limit tissue damage during sterile inflammation.

To address these questions, we compared the functional properties of murine neutrophils recruited to different types of peritoneal inflammation. We found that neutrophils recruited by microbial ligands displayed increased cytokine production compared with naive neutrophils or neutrophils recruited during sterile inflammation. This difference was mediated by a paracrine signal that we identified as TNF-α. Altogether, our results provide evidence for early phenotypic changes in neutrophils and open new avenues to study the regulation of inflammatory responses through the perspective of the most abundant cells present at the onset of inflammation.

C57BL/6 wild type (WT), C57BL/6.SJL (CD45.1) and TNRF single- or double-deficient mice (Tnfrsf1a^−/−, Tnfrsf1b^−/− Tnfrsf1a^−/− × Tnfrsf1b^−/−) were obtained from Jackson Laboratories. Fcer1g^−/− deficient mice (14) were a kind gift from Jessica Hammerman. All mice were bred in our colony at the University of California, Berkeley, and used at 6–12 wk of age. All experiments were approved by the local Animal Care and Use Committee (Animal Use Protocol R298-0816BR).

Bone marrow chimeras were generated by irradiating mice in an X-ray irradiator (900 rads split across two doses, 5 h apart). The irradiated mice were reconstituted by injection of 5 × 10⁶ donor bone marrow cells i.v. immediately after the second irradiation.

Peritonitis was induced by intraperitoneal injection of 100 μl of a 10% w/v sonicated solution of uric acid (15) (Sigma), 500 μg of zymosan or 100 μg trehalose di-behenate (TDB) (both from InvivoGen) in 200 μl PBS. After 16 or 42 h, peritoneal cells were collected by lavage with 5 ml PBS 2% FCS.

Bone marrow was harvested by flushing the femur and tibia of mice with complete RPMI 1640. Cells were passed through a 70 μm strainer and RBC were lysed with ACK (Life Technologies).

For neutrophil purification, total bone marrow cells were underlayed with a solution of 68% isotonic Percoll (i.e., 90% Percoll + 10% 10× PBS) and 79% isotonic Percoll, then spun for 30 min at 750 g without brake. Neutrophils were collected at the 68/79 interface. Neutrophil purity was >90% as tested by flow cytometry after Ly6G-staining.

Peritoneal cells, total bone marrow cells, or purified neutrophils were plated in complete RPMI 1640.

For ex vivo experiments, cells were directly restimulated with TLR ligands (Pam3CSK4, 500 ng/ml, LPS, 500 ng/ml, ultrapure flagellin, 1 μg/ml, R848, 1 μM, and CpG-B ODN 1668, 1 μM; all from InvivoGen; Sa23, 4 μM; IDT DNA). After 30 min, brefeldin (GolgiPlug; BD Biosciences) was added to cells before incubation for another 5 h 30 min.

For in vitro experiments, cells were first stimulated as indicated (zymosan, 5 μg/ml, TDB, 1 μg/ml; both from InvivoGen, recombinant TNF-α aa84-235 at the indicated concentrations, from R&D, or supernatants) for 12–16 h, before restimulation as described above.

At the end of the restimulation, cells were washed once in PBS, and stained with anti-Ly6G, Ly6C, CD11b, and CD62L Abs in the presence of FcBlock. Cells were then fixed and permeabilized (Cytofix/perm kit; BD Biosciences) then stained for TNF-α.

For signaling assays, cells were immediately fixed with 1.6% paraformaldehyde at room temperature after restimulation, then washed and resuspended in ice-cold methanol. After 3 h of incubation at −20°C, cells were stained with FcBlock, anti-p-Erk, p-p38, and IκBα for 1 h at room temperature.

For measurements of de novo protein synthesis, cells were incubated for 1 h with 50 μM O-propargyl-puromycin (Jena Bioscience) after overnight stimulation with TNF-α as described above. When indicated, cycloheximide (10 μg/ml; Sigma) was added 30 min before incubation with OP-puromycin. At the end of the incubation, cells were fixed and permeabilized as described above, then stained with Click-IT reaction kit and Alexa488-azide (Invitrogen) according to the manufacturer’s instructions. After two washes, cells were stained with Abs as described above.

A complete list of Abs, fluorophores, clones, and providers is available in Supplemental Table I. All cells were analyzed on an LSR II or LSR Fortessa (BD Biosciences), sorting was performed on a MoFlo (Beckman Coulter) where indicated, and data were analyzed with FlowJo (v9.9.3; TreeStar). Absolute cell counts were obtained by adding 10 μl of CountBright counting beads (Invitrogen) to each sample.

To determine bacterial killing, 10⁵ purified neutrophils were incubated with 10³ serum-opsonized bacteria (Escherichia coli, K12 strain) for 1 h at 37°C in RPMI 1640 (complete w/o Pen/Strep) (16). Cultures were then plated on agar plates and the killing efficiency was calculated as the ratio of CFU to the CFU obtained in the absence of neutrophils.

ROS production was measured by culturing 10⁵ purified neutrophils in HBSS in the presence of 50 μM luminol and 4 U/ml HRP (Sigma). Luminescence was measured on a L384 luminometer every 30 s, for 10 min before and 20 min after the addition of 1 μM fMLF (Sigma).

Bone marrow cells or RAW.264 macrophages were stimulated in the presence of curdlan (100 μg/ml; from InvivoGen) for 12–16 h. Supernatants were harvested after centrifugation for 10 min at 4000 × g, and further filtered through a 0.22 μm syringe filter.

Supernatants were either used directly or concentrated and exchanged into 20 mM Tris (pH 8) using 30 K ultrafiltration membranes (Millipore). Samples were then loaded on a HiTrap Q HP column (GE) using an NGC chromatography system (Bio-Rad) and eluted over 40 ml of a 0–400 mM NaCl gradient, with one fraction collected each milliliter.

Applied Biomics performed 2D gel electrophoresis of the concentrated samples or fractions, spot picking, and mass spectrometry identification of candidates.

Measurements of cytokine levels in supernatants were performed using a custom-ordered Bio-Plex assay (Bio-Rad) according to the manufacturer’s instructions.

Neutrophils were lysed in RNAzol (MRC). RNA was purified according to the manufacturer’s instructions and concentrated using a column (RNA Clean and Concentrator; Zymo Research). cDNA was prepared with iScript cDNA synthesis kit (Bio-Rad), and quantitative PCR was performed with SYBR Green Master mix (Bio-Rad) on a StepOnePlus thermocycler (Applied Biosystems). Primers were obtained from PrimerBank (17) and synthesized by IDT (see Supplemental Table II).

Neutrophils were harvested and purified from bone marrow as described above, and plated at 5 × 10⁶ cells/ml in complete RPMI 1640 with 10 ng/ml TNF-α. After 12 h, cells were restimulated for 30 min with Pam3CSK4 or fMLF. Eicosanoids in neutrophils and secreted into the culture medium were quantified via liquid chromatography–tandem mass spectrometry (LC-MS) as described before (18–20). In brief, cold methanol was added to cultured neutrophils in media to stabilize lipid mediators. Deuterated internal standards (PGE2-d4, LTB₄-d4, 15-HETE-d8, LXA₄-d5, DHA-d5, AA-d8) were added to samples to calculate extraction recovery. The LC-MS/MS system consisted of an Agilent 1200 Series HPLC, Kinetex C18 minibore column (Phenomenex, Torrance, CA), and AB Sciex QTRAP 3200 mass spectrometer. Analysis was carried out in negative ion mode, and lipid mediators quantitated using a scheduled multiple reaction monitoring mode using specific and established transition ions.

Statistical significance was assessed using either a two-tailed t test for unpaired data points (comparison of two groups), or in the case of multiple comparisons, an ANOVA followed by Tukey’s post test. All tests were performed using Prism (v6; GraphPad).

To investigate how different inflammatory conditions may influence neutrophil functions, we harvested neutrophils from the peritoneal cavity of mice injected with uric acid or zymosan to mimic sterile or microbial inflammation, respectively. As expected, neutrophils recruited to the peritoneal cavity by either stimuli displayed signs of priming, as evidenced by increased ROS production (Fig. 1A, top panel). We also observed no difference in bacterial killing when neutrophils elicited with uric acid or zymosan were coincubated with serum-opsonized bacteria (Fig. 1A, bottom panel). Next, we assessed neutrophil cytokine production by intracellular TNF-α staining following restimulation with the synthetic TLR2 ligand Pam3CSK4. Surprisingly, only neutrophils harvested from mice injected with zymosan produced significant amounts of TNF-α, whereas neutrophils harvested from mice injected with uric acid only produced low amounts of this cytokine (Fig. 1B). We also examined TNF-α production by inflammatory monocytes restimulated in the same well. Interestingly, we did not observe any changes in levels of TNF-α produced by inflammatory monocytes, but neutrophils elicited in the presence of zymosan produced similar levels of TNF-α as inflammatory monocytes (Fig. 1B). The observed increase in neutrophil cytokine production could be cell-intrinsic or mediated by differences occurring during the restimulation, such as different cell populations present in the peritoneal lavage. To distinguish between these possibilities, we injected congenically distinct mice with uric acid or zymosan and then mixed and restimulated the harvested cells with TLR2 ligands (Fig. 1C). In these conditions, neutrophils harvested from mice injected with zymosan clearly produced higher levels of TNF-α than cells from mice injected with uric acid, demonstrating that the increase in cytokine production is intrinsic to neutrophils. Given that neutrophils recruited to the peritoneum in either condition were capable of efficient ROS production and bacterial killing, a phenotypical change referred to as priming, we decided to refer to the increase in cytokine production observed only in zymosan elicited neutrophils as licensing, to clearly distinguish between the two phenomena.

As neutrophils are among the first immune cells to infiltrate the peritoneum during local inflammation, we hypothesized that they directly sense signals delivered by zymosan in the peritoneum, resulting in an increase in their ability to produce cytokines. To test our hypothesis, we used bone marrow as a source of naive neutrophils and pretreated bone marrow cells with zymosan before restimulation with Pam3CSK4. Similar to our in vivo experiments, we observed a striking increase in the TNF-α production by neutrophils pretreated with zymosan before stimulation with Pam3CSK4 (Fig. 1D–E).

We next performed experiments with purified bone marrow neutrophils and observed similar results, suggesting that neutrophils can acquire the capacity to produce cytokines directly, in the absence of accessory cells (Fig. 1F) Of note, as Percoll purified neutrophils are 90–95% pure, we also sorted neutrophils by FACS to 97–99% purity and restimulated them as above (Fig. 2). Although we did observe slightly lower levels of cytokine production, perhaps due to a decrease in viability after sorting, licensing was similar in total bone marrow (input) and highly purified cells. Using purified cells also allowed us to directly measure cytokines secreted by neutrophils in the supernatant. Strikingly, induction of the anti-inflammatory cytokine IL-10 and several chemokines (CCL3-4, CXCL1-2) produced by neutrophils followed a pattern similar to TNF-α (Fig. 1G). Moreover, other CLR ligands, such as curdlan or the synthetic Mincle/MCL ligand TDB, also strongly increased secretion of cytokines in response to TLR2 ligands (Fig. 1G, top left).

Altogether, these experiments suggest that the ability of neutrophils to produce cytokines is modulated during certain types of inflammation. In particular, the low cytokine production exhibited by naive neutrophils can be greatly increased by microbial signals. We propose to call this process neutrophil licensing to distinguish it from neutrophil priming, the rapid differentiation process that neutrophils undergo when recruited by either microbial or sterile inflammatory signals.

Next, we sought to determine the nature of the signal that licenses neutrophils for cytokine production. Our experiments with purified neutrophils argued for direct sensing of ligands by neutrophils (Fig. 1F), but these results did not rule out the possible contribution of other cells in vivo. The synthetic CLR ligand TDB is recognized by Mincle and MCL, and both receptors depend on the adaptor protein FcεRIγ for signaling as well as cell surface expression (21). In agreement with these previous studies, we observed TDB-induced neutrophil licensing in WT but not Fcer1g^−/− mice (Fig. 3A). To determine if hematopoietic cells are responsible for ligand sensing and neutrophil licensing in vivo, we generated bone-marrow chimeras with WT mice and mice deficient for Fcer1g. In chimeras generated with Fcer1g^−/− bone marrow, TDB injection in the peritoneum failed to induce an increase in the TNF-α production by the recruited neutrophils (Fig. 3B, 3C, left panels). The requirement for Fcer1g was restricted to hematopoietic cells, consistent with a model in which direct sensing of the ligand by neutrophils is sufficient for licensing. However, in mixed bone marrow chimeras receiving both WT and Fcer1g^−/− cells, injection of TDB led to a similar increase in neutrophil cytokine production in both WT and Fcer1g^−/− cells (Fig. 3B, 3C, right panels), even though the latter population cannot directly recognize TDB. This result raised the possibility that neutrophil licensing can also occur via paracrine signaling. Indeed, Fcer1g^−/− neutrophils could be licensed when cocultured with WT cells in vitro (Fig. 3D). Interestingly, at low ratios of WT/Fcer1g^−/− cells, we noticed a reduction in the TDB-induced increase in TNF-α production in both WT and Fcer1g^−/− neutrophils (Fig. 3D). This result suggests that even though WT cells can directly sense TDB, the paracrine signal is necessary for efficient licensing, and that this signal becomes limiting with low numbers of WT cells.

Paracrine activation can be mediated either by soluble factors or by cell to cell interactions. If a soluble signal is sufficient for neutrophil licensing, we reasoned that supernatant from activated cells should be able to license naive neutrophils even in the absence of the original microbial ligand. To test this hypothesis, we took advantage of the fact that the curdlan, another Dectin-1 agonist, is composed of large particles that can easily be removed from a solution by filtration. Indeed, although curdlan can directly induce neutrophil licensing, passage through a 0.22 μm filter abrogated this effect (Fig. 4A, top panels). However, when we treated bone marrow cells with curdlan, collected and filtered the supernatant of those cells, and transferred the supernatant to naive cells, we observed robust neutrophil licensing, demonstrating the presence of a soluble licensing factor (Fig. 4A, bottom panels). Next, we tested whether other cell lines might be able to produce a soluble factor with the same properties to generate enough material for purification and identification. The macrophage cell line RAW 264.7 also produced significant levels of a licensing factor after treatment with curdlan (Fig. 4B). Treatment with proteinase K completely abrogated the activity of the supernatant, strongly suggesting that the soluble factor is a protein (Fig. 4B).

To identify potential candidates, we used two parallel approaches. First, we performed 2D-gel electrophoresis on supernatants from untreated RAW 264.7 cells and cells treated with curdlan, to identify proteins secreted specifically in the presence of curdlan (Fig. 4C). We also performed anion exchange chromatography on the supernatant and tracked the licensing activity across 40 fractions. Most of the licensing activity eluted in a single peak around 250 mM NaCl (Fig. 4D). We then included that fraction in a 2D-gel electrophoresis assay together with the original active and control supernatants, to detect proteins that are specifically enriched after purification by anion exchange chromatography (Fig. 4E). We then selected spots corresponding to proteins induced by curdlan and enriched in the fraction eluted at 250 mM NaCl for identification by mass spectrometry (Fig. 4E).

TNF-α itself was present in the candidate proteins identified by purification, so we tested whether recombinant TNF-α can directly induce neutrophil licensing. Bone marrow neutrophils incubated with recombinant TNF-α before restimulation produced significantly higher levels of TNF-α after restimulation with Pam3CSK4 (Fig. 4F, top panels). This effect was dose dependent and robust at concentrations above 10 ng/ml (Fig. 4F, bottom panels). Importantly, treatment with TNF-α alone was not sufficient for cytokine production by neutrophils, but required subsequent activation by a TLR2 ligand (Fig. 4F).

TNF-α can signal through two related receptors, TNFR1 and TNFR2 (encoded by Tnfrsf1a and Tnfrsf1b, respectively). Deletion of Tnfrsf1a completely abrogated the licensing induced by TNF, whereas deletion of Tnfrsf1b had little or no effect, demonstrating the critical role of TNFR1 in the licensing of neutrophils (Fig. 4G). Of note, deletion of Tnfrsf1a had a significant but partial effect on the cytokine production of neutrophils treated with zymosan (Fig. 4G, middle column), suggesting that some microbial ligands may, directly or through another soluble factor, increase cytokine production to some extent, but that the bulk of this increase is mediated by paracrine TNFR1 signaling.

Altogether, our results suggest that TNF-α, through a TNFR1-dependent signaling pathway, mediates an increase in the ability of neutrophils to produce TNF-α itself in response to microbial ligands.

To address definitively whether TNF-α can directly license neutrophils, we examined cytokine production of WT and TNFR1-deficient cells in mixed cultures stimulated with CLR ligands and Pam3CSK4. Under these conditions, we did not observe any rescue of TNF-α production by Tnfrsf1a^−/− cells when cultured with WT cells (Fig. 5A). These results contrast with what we observed in mixed cultures of WT and Fcer1g^−/− cells (Fig. 3D) and confirm that TNF-α directly modifies neutrophil cytokine production, rather than causing the induction of another downstream paracrine signal.

Next, we tested whether the effect of TNF-α-TNFR1 signaling on TNF-α production could be extended to other cytokines produced by neutrophils. Prestimulation of neutrophils with TNF-α was sufficient to increase the production of CXCL1-2, IL-10, and to some extent CCL3-4 in purified neutrophils treated with Pam3CSK4 (Fig. 5B). Again, TNF-α stimulation itself did not induce any cytokine production on its own but enhanced TLR-induced production.

A primary function of neutrophils is to generate eicosanoids such as leukotrienes and PGs, which initiate and amplify acute inflammation, adaptive immune responses, vascular leakage, and pain (22). Formation of eicosanoids by neutrophils is tightly regulated and usually triggered by G-protein–coupled receptors such as FPR and C5AR. To assess whether treatment with TNF-α could also modulate eicosanoid production, we measured lipid mediator release in neutrophils treated with Pam3CSK4 for 30 min, after a 12 h incubation with or without TNF-α. As expected, TNF-α or TLR stimulation alone did not induce significant production of leukotriene B₄ (LTB₄) and PGE₂; however, we observed a synergistic effect of these two stimuli that induced significant levels of LTB₄ and PGE₂ release (Fig. 5C). As a control, we also treated neutrophils with fMLF, and, as expected, observed robust lipid mediator production independently of TNF. These results suggest that TNF treatment directly modulates the outcome of TLR signaling in neutrophils, namely licensing TLRs to trigger the release of LTB₄ and PGE₂, which are potent mediators of neutrophil and effector T cell recruitment.

Finally, we assessed the kinetics of the effects of TNF-α and of other ligands. Although TNF-α has been reported to prime ROS production in minutes, we observed that the increase of cytokine production was maximal only after hours of incubation with TNF-α or zymosan (Fig. 5D). In the case of the ligand TDB, increased cytokine production required over 8 h of stimulation, likely reflecting the fact that licensing occurs entirely indirectly in this case and first requires accumulation of sufficient amounts of TNF-α in the culture media. This result suggests that TNF-α has long-term effects on neutrophil phenotype (i.e., licensing), in addition to its short-term effect on ROS production (i.e., priming), the latter of which is mediated by phosphorylation of the NADPH oxidase (23).

We also tested whether the continuous presence of TNF-α was required to maintain licensing. To this end, cells were washed and cultured in fresh media for increasing amounts of time before restimulation. We still observed a strong increase in cytokine production in neutrophils that had been treated with TNF-α and subsequently washed and cultured in fresh media although the licensing effect decreased as the rest duration increased (Fig. 5E). Of note, the prolonged licensing observed in vitro is consistent with what we observed in ex vivo neutrophils, in which increased cytokine production is maintained even after the cells are taken from the inflammatory milieu of the peritoneum (Fig. 1C). These results argue that TNF-α can durably modify neutrophil activity.

Finally, we assessed the role of TNFR1-mediated licensing in a model of zymosan-induced peritonitis in vivo. As described above, injection of zymosan leads to neutrophil accumulation in the peritoneum at 16 h, and to an increase in their capacity to produce TNF-α (Fig. 1B). In TNFR1-deficient mice, we observed a comparable recruitment of neutrophils (Fig. 6A), but, in agreement with our in vitro results, the recruited neutrophils displayed a lower production of TNF-α upon restimulation (Fig. 6B). Interestingly, 42 h after zymosan injection, i.e., during the resolution of inflammation, the number of neutrophils present in WT mice was significantly higher than in Tnfrsf1a^−/− mice (Fig. 6A). Altogether, these experiments argue that neutrophil licensing, likely through increased cytokine or lipid mediator production, contributes to the maintenance of inflammation through increased neutrophil recruitment. Of note, it is also possible that the different inflammatory conditions in WT and Tnfrsf1a^−/− mice alter the half-life of neutrophils, which would also influence their numbers, but we did not observe any positive effect of TNF-α treatment on neutrophil survival in vitro (Fig. 6C), leading us to favor the hypothesis of an increased recruitment.

In the experiments presented so far, we relied on TLR2 ligands to stimulate neutrophils and measure cytokine production. To determine whether increased cytokine production is a general feature of neutrophil licensing, we tested ligands for other TLRs. Although the magnitude of responses to TLR ligands varied drastically, TNF-α pretreatment increased cytokine production for all TLRs tested (Fig. 7A). In particular, TLR4 and TLR13 ligands yielded similar cytokine production patterns as TLR2. Of note, increased cytokine production in response to other TLR ligands was also observed with pretreatments other than TNF-α, such as zymosan and TDB (data not shown).

Engagement of TLRs leads to the activation of IL-1R–associated kinase (IRAK) family members downstream of MyD88 or toll/IL-1R domain-containing adapter inducing IFN-β (TRIF), and ultimately to the activation of the NF-κB and MAPK pathways. Therefore, we measured the degradation of IkBα and phosphorylation of Erk and p38 in neutrophils after TLR2 stimulation to examine whether TNF-α pretreatment affected downstream signaling. We detected robust phosphorylation of Erk following TLR2 stimulation but did not see any effect of pretreatment with TNF-α on this response (Fig. 7B). Similarly, degradation of IkBα in response to TLR2 was not affected by pretreatment. Interestingly, we did not observe phosphorylation of p38 in neutrophils, even though we could detect it reliably in inflammatory monocytes (Fig. 7B, right panel). Altogether, we found no evidence that TNF-α–mediated licensing of neutrophils modifies the signaling directly downstream of TLR activation.

Because signaling is not affected by licensing, we reasoned that increased cytokine production could be due either to epigenetic changes leading to increased mRNA production, or posttranscriptional regulation of cytokines. To discriminate between these possibilities, we measured the levels of mRNA for cytokines affected by licensing. In the case of TNF-α, CCL3-4 and CXCL1-2, stimulation with Pam3CSK4 alone was sufficient to induce a maximal or near maximal upregulation of the respective mRNA. On the other hand, IL-10 mRNA was strongly induced by TNF-α alone, and we observed a synergistic effect with additional TLR stimulation (Fig. 7C). Altogether, these results suggest that cytokine production was largely regulated at the posttranscriptional level. It is noteworthy that IL-10, the main anti-inflammatory cytokine produced by neutrophils, did not fit this pattern and is strongly upregulated at the transcriptional level, suggesting that the kinetics of pro- and anti-inflammatory gene induction in neutrophils may differ.

In the case of proinflammatory cytokines, it remained unclear how increased cytokine production is achieved when mRNA levels are largely unchanged following licensing. Neutrophils have been reported to be poor protein producers, so we tested whether licensing changes this aspect of neutrophil function. To that end, we measured incorporation of an analog of puromycin in nascent polypeptides. First, we confirmed that protein synthesis in neutrophils occurs at much lower levels than in monocytes (Fig. 7D, left panel). Indeed, using cycloheximide as a background control, we only observed a low level of puromycin incorporation in neutrophils, compared with inflammatory monocytes cultured in the same conditions. These results are in agreement with a recent study of protein synthesis across bone marrow cells (10). Treatment with TNF-α did not noticeably increase the incorporation of puromycin in neutrophils (Fig. 7D, right panel), suggesting that changes in global translation levels are unlikely to account for the difference in cytokine production we observed. Thus, our data suggest that the increased protein production we observed in licensed neutrophils may be due to selective translation or import of proinflammatory cytokines, and a selective transcriptional upregulation of the anti-inflammatory cytokine IL-10. Taken together, these experiments argue that although naive neutrophils are competent for TLR signaling and cytokine mRNA induction, efficient secretion of cytokines requires an additional licensing step mediated by TNFR1 signaling.

In the current study, we have focused on the ability of neutrophils to produce cytokines during different inflammatory conditions. We found that different inflammatory milieus enable neutrophils to produce high concentrations of cytokines, chemokines, and lipid mediators in response to TLR ligands, and propose describing this phenomenon as licensing. It is important to note that the priming of neutrophil functions by several factors, including TNF-α and integrin cross-linking, increases ROS production and enhances bacterial killing (23, 24). However, priming does not necessarily increase cytokine production and is mechanistically distinct from licensing, as we discuss below. Indeed, after licensing, we observed a ∼10-fold increase in cytokine production, raising TNF-α production of neutrophils to levels equivalent to that of inflammatory monocytes stimulated similarly. Even more strikingly, licensed neutrophils could produce eicosanoids in response to TLR stimulation, even though this activity is normally gated by G protein-coupled receptors.

We also demonstrated that licensing is mediated by TNF-α in a paracrine manner. Importantly, production of TNF-α is increased in licensed neutrophils, raising the possibility that neutrophil cytokine production is regulated by a feed-forward loop. In agreement with that observation, we observed that neutrophils stimulated ex vivo either all robustly produced TNF-α in the presence of microbial ligands, or all produced low amounts of this cytokine (Fig. 1B). This unimodality is consistent with the hypersensitivity associated with feed-forward loops, and disappeared in the absence of paracrine signaling in TNFR1-deficient cells (Fig. 4G). Our results thus suggest a form of quorum sensing in neutrophils, where recognition of microbial ligands will increase local concentrations of TNF-α and license other neutrophils migrating to this site for increased cytokine production as long as TLR ligands are present.

Consistent with this idea, we also observed licensing of the production of chemokines and lipid mediators involved in neutrophil recruitment (CXCL1-2, LTB₄). Altogether, the results presented in this study point to a model for escalation of inflammation, where the local licensing of neutrophils allows them to recruit and activate additional neutrophils and other immune cells. Importantly, because activation of this positive feedback loop requires both local licensing signals and the persistent presence of ligands for TLRs, the elimination of microbes and decrease in the availability of TLR ligands would lead to a termination of the cytokine production. There is, however, evidence for activation of TLRs by endogenous ligands, which may mediate a continuous neutrophil recruitment and cytokine production in the context of inflammatory diseases such as rheumatoid arthritis (4). It is interesting to note that TNF-α is a critical mediator of this disease, and our results suggest that its role on neutrophils may be involved in the persistence of inflammation. In the context of normal responses, the fact that cytokine production by neutrophils requires licensing over a few hours may also be important to limit inflammation, as our results would predict that neutrophils will produce very few cytokines in the context of a transient presence of TLR ligands, for example after a wound infected by an easily eliminated commensal. Conversely, anti-TNF-α therapies are also associated with increased risks of infection, and our results suggest that defects in neutrophil licensing or accumulation may play a part in these side effects (25).

Somewhat expectedly, licensed neutrophils also produce increased levels of the regulatory cytokine IL-10, which may concomitantly limit inflammation. Interestingly, IL-10 was mostly regulated at the transcriptional level, whereas proinflammatory cytokines were mostly regulated at the posttranscriptional level. This discrepancy may lead to differences in the kinetics of expression of these proteins both in vivo and in vitro, and may underlie the differences observed in previous reports on neutrophil cytokine production (8, 26). In particular, Zhang et al. (8) measured neutrophil cytokine production 24 h poststimulation, and found the response to be dominated by IL-10, suggesting that anti-inflammatory cytokine production is more sustained than proinflammatory cytokine production. It is also important to note that human neutrophils differ from their murine counterparts, in particular with regards to IL-10 production, and it will be interesting to compare the effects of licensing in human cells.

In this study, we demonstrate that TNF-α is a critical regulator of neutrophil protein and eicosanoid synthesis. It is interesting that TNF-α has long been used as a priming agent for neutrophil studies (24). In this setting, TNF-α treatment leads to a rapid but transient increase in the ability of neutrophils to produce ROS through phosphorylation of the NADPH oxidase and other proteins (23). However, we did not observe any difference in ROS production by neutrophils elicited with uric acid or zymosan, even though only the latter are able to produce large amounts of cytokines. This observation is consistent with the idea that priming can be induced by integrin cross-linking (27), such as when cells are migrating to the peritoneum. Altogether, our data suggest that priming of ROS production for bacterial elimination and licensing of cytokine production are distinct, but not mutually exclusive, phenomena. Specifically, we hypothesize that priming will occur rapidly as neutrophils migrate to most inflammatory sites, whereas licensing will be restricted to sites of prolonged inflammation, for example in the presence of a pathogen.

Mechanistically, our results suggest that TNF-α directly increases the ability of neutrophils to produce proinflammatory cytokines, independently of transcriptional regulation. Previous reports have shown that priming of neutrophils, including with TNF-α, can induce a transient transcriptional upregulation of cytokine genes (28), but this effect was only modest at later time points in our study. Although further studies will be necessary to fully understand the mechanisms involved, it is interesting to note that translation in mature neutrophils is largely affected by the loss of nucleoli (29), defects in mRNA splicing (30), and overall lower translational ability (10). Our results would seem to suggest that TNF-α can reverse some of these hallmarks of neutrophil differentiation, and allow them to produce cytokines at levels consistent with those seen in monocytes. However, because we did not observe global changes in translation in a puromycin-based assay, we favor the hypothesis that there is a selective increase in translation or secretion of cytokines. It is also possible that translation elongation is regulated by licensing, as this would not be revealed by puromycin labeling.

In summary, our work shows that neutrophil cytokine production is locally regulated, and that this regulation is unexpectedly mediated by neutrophils themselves. Although the study of cytokine production by neutrophils has generally been limited to cells activated in vitro, our results argue for a reassessment of the role of neutrophils in the early shaping of the inflammatory microenvironment. These studies may pave the way for the development of new therapies in the context of inflammatory diseases.

We thank members of the Barton laboratory, Gabrielle Reiner, Hélène Beuneu, and Béatrice Bréart for helpful comments on the manuscript, Julie Choe and Jeanette Tenthorey for technical help with protein purification, and Hector Nolla, Alma Valeros, and Kartoosh Heydari for assistance with flow cytometry.

The authors have no financial conflicts of interest.