Salmonella Induced IL-23 and IL-1β Allow for IL-12 Production by Monocytes and Mφ1 through Induction of IFN-γ in CD56+ NK/NK-Like T Cells

Background The type-1 cytokine pathway plays a pivotal role in immunity against intracellular bacterial pathogens such as Salmonellae and Mycobacteria . Bacterial stimulation of pattern recognition receptors on monocytes, macrophages and dendritic cells initiates this pathway, and results in the production of cytokines that activate lymphocytes to produce interferon (IFN)-  . Interleukin (IL)-12 and IL-23 are thought to be the key cytokines required for initiating a type-1 cytokine immune response to Mycobacteria and S almonellae . The relative contribution of IL-23 and IL-12 to this process is uncertain. Methodology/Principal Findings We show that various TLR agonists induce the production of IL-23 but not IL-12 in freshly isolated human monocytes and cultured human macrophages. In addition, type 1 pro-inflammatory macrophages (M  1) differentiated in the presence of GM-CSF and infected with live Salmonella produce IL-23, IL-1  and IL-18, but not IL-12. Supernatants of Salmonella infected M  1 contained more IL-18 and IL-1  as compared with supernatants of M  1 stimulated with isolated TLR agonists, and induced IFN-  production in human CD56 + cells in an IL-23 and IL-1β-dependent but IL-12-independent manner. In addition, IL-23 together with IL-18 or IL-1  led to the production of GM-CSF in CD56 + cells. Both IFN-  and GM-CSF enhanced IL-23 production by monocytes in response to TLR agonists, as well as induced IL-12 production. Conclusions/Significance The findings implicate a positive feedback loop in which IL-23 can enhance its release via induction of IFN-  and GM-CSF. The IL-23 induced cytokines allow for the subsequent production of IL-12 InvivoGen) or 1  g/ml CL087 (tlrl-c87, InvivoGen) in a final volume of 200  l. Supernatants were taken and IL-23 (eBioscience), IL-1  (Biosource) IL-18 (MBL) and IL-12p70 (Sanquin) concentrations were determined by ELISA. For experiments with M  1, cells were harvested with trypsin-EDTA, washed with PBS and seeded at 3.3·10 5 per well in 24-well or 1·10 6 in 12-well culture plates (Corning Life Sciences) and allowed to adhere. Subsequently M  1 were stimulated with TLR agonists, infected with group B Salmonella at a 10:1 multiplicity of infection as described below, or left unstimulated. Twenty-four hours after stimulation, supernatants were collected and cytokine production was determined by ELISA. To test the effect of GM-CSF and IFN-  pre-stimulation, 1·10 5 CD14 + monocytes were cultured for 16 hours with 50-5000 pg/ml GM-CSF, 50-5000 pg/ml IFN-  (Biosource) or left unstimulated. Subsequently cells were stimulated with 100 ng/nl LPS plus 1  g/ml CL075 for 24 hours, or left unstimulated.  1 in the induction of IFN-  production, we used a specific antibody our the supernatants of infected also enhance production. However, we did not determine the role of TNF in this respect. (4) IFN-  and both enhanced production in monocytes


Introduction
Immunity against intracellular bacterial pathogens such as Salmonellae and Mycobacteria depends on the type-1 cytokine pathway (1). This pathway is initiated by bacterial stimulation of pattern recognition receptors on monocytes and macrophages, resulting in the production of cytokines that activate lymphocytes and induce IFN- production. The IFN- in turn activates monocytes and macrophages, to enhance bactericidal effector mechanisms and to further pro-inflammatory cytokine production. Thus, the type-1 cytokine pathway critically depends on the cross-talk between monocytes/macrophages and lymphocytes. Abdi et al. state that IL-12p70 cannot be the primary trigger that initiates Th1 T-cell responses, as it is not produced in response to bacterial stimulation when costimulation in the form of activated T cells or IFN- are absent (2). In addition, in whole blood assays with blood obtained from patients with complete IFN-R deficiency, no IL-12p70 production can be detected in response to M. bovis BCG infection in vitro (3). How the type-1 pathway is initiated, therefore, has remained uncertain.
Interleukin-23 (IL-23) is a cytokine which is produced early in the immune response (4).
Monocytes as well as type 1 macrophages (M1) produce IL-23 in response to the binding of pathogens and pathogen-associated molecular patterns (PAMPs), such as lipopolysaccharide (LPS), to Toll-like receptors (TLRs) (5,6). In contrast to IL-23, for the production of IL-12 in response to PAMPs an additional stimulus such as IFN- is required (5). IL-23 is known to induce IFN- production in naïve T cells, in memory T cells (7) and in NK-like T cells (8), thereby potentially providing the necessary, additional stimulus to induce IL-12 production. Next, IL-12 and IL-18 enhance IFN- production in NK, NK-like T cells and Th1 cells by binding to their respective receptors (9). Though IFN- is not required for the induction of IL-23, the precise role of IFN- in the regulation of IL-23 is not well established.
Granulocyte-Macrophage Colony Stimulating Factor (GM-CSF) is a cytokine produced by NK cells and T cells in response to a variety of stimuli, for instance IL-15 and IL-18 (10). GM-CSF activates monocytes and enhances their bactericidal activity (11)(12)(13). Moreover, monocytes prestimulated with GM-CSF secrete increased quantities of tumor necrosis factor (TNF) and IL-1 when stimulated with LPS (14,15). Furthermore, GM-CSF induces differentiation of human monocytes into M1, a macrophage type that is capable of producing IL-23 (5).
In this study we addressed the roles of the cytokines IL-23, IL-1 , IL-18, GM-CSF and IFN- in the crosstalk between NK/NK-like T cells and monocytes/ macrophages in the early activation of the type-1 cytokine pathway. To this end we determined which TLR agonists induce IL-23 production in human monocytes and macrophages and assessed the roles of GM-CSF and IFN- in the regulation of IL-23 production by human monocytes in response to TLR agonists.  /NK-like T cells, and tested the capacities of IL-23, IL-1 and IL-18 to induce GM-CSF and IFN- in human NK and NK-like T cells.

Cytokine induction and measurement
To determine cytokine production by monocytes, CD14 + beads isolated cells were seeded in a 96-

IFN- and GM-CSF production
To determine GM-CSF and IFN- production in NK and NK-like T cells, overnight-rested CD56

Results
Monocytes and M1 produce IL-23, but not IL-12, in response to various TLR agonists.
LPS is known to induce IL-23, but not IL-12, production in M1 (5). To determine whether various TLR agonists are capable of inducing IL-23 or IL-12 production in monocytes and M1, we first tested these capacities in freshly isolated human monocytes. Unstimulated monocytes did not produce IL-23, whereas Zymosan A (agonist for TLR2/6) induced minimal IL-23 production, LPS (TLR4) and CL075 (TLR8/7) each induced small amounts of IL-23, while LPS and CL075 synergized in the induction of IL-23 production in monocytes (Fig. 1A). IL-23 production was also observed in response to LPS plus flagellin (TLR5) but the amount released was similar to that induced by LPS alone. Of note, the amounts of IL-23 produced by CD14 + monocytes varied somewhat between different donors, the relative amount released upon the various TLR stimulations was however similar (data not shown). In none of the supernatants IL-12p70 was detected, given a detection limit of the IL-12p70 ELISA of 3 pg/ml (data not shown).
Next, we compared the amount of IL-23 and IL-12 produced by monocytes to those by cultured to the TLR agonists. Again, no IL-12p70 production was detected in response to any of the stimuli (data not shown). To verify that the Mφ1 used in these experiments are capable of producing IL-12p70 they were stimulated with LPS in combination with IFN- Large amounts of IL-12p70 were detected in the supernatants (data not shown).

Infection of M1 with live Salmonella induces IL-23, but not IL-12, production.
To determine whether not only PAMPs such as the TLR agonists, but also live pathogens induce IL-23 rather than IL-12 production in M1, we infected cultured M1 with live Salmonella and assessed cytokine release in the culture supernatants. M1 exposed to and containing ingested Salmonellae produced IL-23 whereas M1 left unstimulated did not (Fig. 1C). Of note, Salmonella infected M1 produced markedly more IL-23 as compared with M1 stimulated with large amounts of LPS ( Fig 1C). Similar to stimulation with TLR agonists, we did not detect any IL-12p70 in the supernatants of infected M1 (data not shown). Moreover, in a related project, micro-array analysis of M1 infected with live Salmonella for 1, 2, 4, 8 and 24 hours revealed that infection did not induce IL12A (IL-12p35) transcription, whereas IL12B (IL-12p40) and IL23A (IL-23p19) transcription were both upregulated in response to Salmonella (16).

TLR stimuli and infection with live Salmonella induce IL-18 and IL-1 production.
We have shown previously that for IL-23 to induce IFN- production in NK-like T cells, an additional stimulus such as IL-18 is required (8). Similar to IL-18, IL-1 can costimulate for IFN- production (17,18). To determine whether the production of IL-23 occurs in concert with that of IL-18 or IL-1, production in response to LPS and infection varied markedly between donors; M1 of all donors produced both IL-18 and IL-1 in response to these stimulations ( Fig. 2A and B). Salmonellainfected M1 produced markedly more IL-18 and IL-1 as compared with cells stimulated with LPS ( Fig. 2A and B).

IL-23 and IL-1 synergize in the induction of IFN- in CD56 + cells.
Infection of M1 with live Salmonella resulted in the production of IL-23, IL-18 and IL-1, IL-23 is reported to induce IFN- production in NK, NK-like T cells and in T cells, in synergy with IL-18 (7,8,19). As mentioned before, IL-1 is known to enhance IL-12 induced IFN- production (17,18).
Therefore, we tested whether perhaps IL-23 in combination with IL-1 could also induce IFN- production in CD56 + (NK and NK-like T) cells. Each cytokine alone was not able to induce IFN- production (Fig. 3). However, when IL-23 and IL-1 were combined, they synergized in inducing IFN- production (Fig. 3 Fig 4). Surprisingly, the cells producing IFN- in response to IL-23 plus T cells only IL-23 plus IL-1β resulted in significant IFN- production (Fig 4).
Since IL-23 in combination with IL-1 is known to promote the production of IL-17 in memory T cells (20), we also analyzed IL-17 production by CD56 + cells in response to IL-23 and IL-1. However, in none of the supernatants IL-17 was detected (data not shown).

Supernatants of Salmonella infected M1 induce IFN- production in primary CD56 + cells.
So far we found that IL-23, IL-18 and IL-1, but not IL-12p70, are present in supernatants of Salmonella infected M1. We showed that IL-23 can induce IFN- production in CD56 + when IL-1 is present. Previously, we have shown that CD56 + cells produce IFN- when stimulated with IL-23 plus IL-18, in the absence of IL-12 (8). Based on these data we expected that supernatants from Salmonella infected M1 would be able to induce IFN- production in CD56 + cells. Supernatants of Salmonella infected M1 were indeed able to induce IFN- production in primary CD56 + cells (Fig.   5A). Neutralization of LPS in these supernatants did not alter their ability to induce IFN- production (data not shown).
Since IL-18 is induced in small amounts only, even after infection by Salmonella, we determined whether IL-18 was a limiting factor in the induction of IFN- in CD56 + cells by supernatants of M1 infected with Salmonella. The addition of recombinant IL-18 to the supernatants increased the IFN- production by these CD56 + cells (Fig. 5A).

IL-23 and IL-1, but not IL-18, in supernatants are critical for the induction of IFN- production in
To verify the role of IL-23 in the induction of IFN- by supernatants of Salmonella infected M1, we neutralized IL-23 using an antibody binding the p40 subunit of IL-23. (Note that although the p40 antibody can also neutralize IL-12, we have shown above that no IL-12 was present in these supernatants). When IL-23 was thus neutralized, IFN- production was effectively abrogated ( supernatant did not affect the production of IFN- significantly (Fig. 5B), suggesting IL-18 is not crucial in this respect. In contrast, neutralization of IL-1 reduced the capacity of the supernatants to induce IFN- production significantly (Fig. 5B).

IL-23, in synergy with IL-18 or IL-1, induces GM-CSF production in CD56 + NK/NK-like T cells.
IL-18 is known to induce GM-CSF production in NK cells in combination with various cytokines (10).
To determine whether IL-18 or IL-1, combined with IL-23 can induce GM-CSF production in human

GM-CSF and IFN- prime monocytes for enhanced IL-23 production.
M1 generated by culturing CD14 + monocytes for six days in the presence of GM-CSF are strong producers of IL-23 compared with freshly isolated CD14 + monocytes ((5) and above). To investigate whether GM-CSF or IFN- (known to enhance the expression of the IL-23 subunit IL-12p40 (21)) could prime monocytes directly for enhanced IL-23 production in response to stimulation with heat killed Salmonella, we pre-stimulated CD14 + cells for 16 hours with GM-CSF or IFN- and assessed IL-23 production. Stimulation of monocytes with GM-CSF or IFN- alone did not induce IL-23 production ( Fig. 7A and 7B). However, sixteen hours of pre-stimulation of monocytes with as little as 50 pg/ml GM-CSF enhanced IL-23 production in response to heat killed Salmonella (Fig. 7A), with a dose dependent priming effect (Fig. 7A). Similar to GM-CSF, pre-stimulation of monocytes with IFN- also enhanced IL-23 production in response to heat killed Salmonella, again in a dose dependent manner (Fig. 7B).  Table 1 Percentages of IFN- producing CD56 + cells after stimulation with cytokines. -IL-23   Donor  Fraction  -IL-1β  IL-18  -IL-1β  IL-

IL-23 plus IL-1 stimulated CD56 + cells prime monocytes for IL-12p70 production.
To produce IL-12p70 in response to LPS, IFN- is needed as an extra stimulus. CD56 + cells produce IFN- in response to IL-23 plus IL-1 , therefore we wanted to know whether supernatants of IL-23/ IL-1 stimulated CD56 + cells could prime monocytes for IL-12p70 production in response to LPS.
Monocytes were stimulated with LPS, plus or minus supernatants of IL23/IL-1 stimulated CD56 + cells or, as a positive control, IFN- Monocytes stimulated with LPS alone did not produce IL-12p70, whereas cells stimulated with LPS in combination with IFN- or supernatants of IL-23 plus IL-1 stimulated CD56 + cells both produced IL-12p70 (Fig. 8A). To assess the role of IFN- present in the supernatants of IL-23 plus IL-1 stimulated CD56 + cells in the priming effect on IL-12p70 production by monocytes, we used an IFN- neutralizing antibody. Neutralization of IFN- greatly diminished the priming effect of these supernatants on LPS induced IL-12p70 production by monocytes (Fig. 8B).
Of note, IL-23 plus IL-1 did not directly prime monocytes to produce IL-12p70 in response to LPS (data not shown). Data are means ± SD of triplicates in one representative experiment of five. has been reported to enhance IFN- production (18,24). Therefore, in our experiments, TNF in the supernatants of infected M1 may also enhance IFN- production. However, we did not determine the role of TNF in this respect. In this study we report on a positive feedback of IL-23 production, but negative feedback mechanisms may exist as well. For instance, IL-4 and IL-10 are both reported to inhibit the induction of IL-23 and may serve as a negative feedback (28). IL-4 may be important in balancing IL-23 and IL-12 production as IL-4 can enhance IL-12p70 production (29)(30)(31). It would be interesting to explore the effects of these and other potential negative regulators of IL-23.
To reach the conclusion on a positive feedback loop in which IL-23 enhances its own expression via the induction of IFN- and GM-CSF, the following pitfalls of this study need to be considered. Firstly, in supernatants of M1 infected with live Salmonella, we neutralized IL-23 using an antibody which is able to neutralize IL-12p70 as well. Though we did not detect IL-12p70 and IL-12p35 mRNA expression was not induced by infection with Salmonella, one should bear in mind that the detection limit of the IL-12p70 ELISA used was 3 pg/ml and that we can not exclude that the effect observed after neutralizing the IL-23 is due to the neutralization of small, undetectable amounts of IL-12p70. Secondly, the concentrations of the TLR agonists used may exceed physiological relevant conditions. For example, the concentrations of Salmonella LPS added to the cells used in this study are likely to differ from LPS concentration when M1 are being infected with live Salmonella. During sepsis 5-10 pg/ml LPS can be detected in the blood (32). Thirdly, between donors we observed interindividual differences in cytokine production in response to the stimulations used. This may reflect differences in responsiveness to stimulation and differences in capacity to produce cytokines between cells obtained from different donors. We observed for example remarkable interindividual differences in the subset of IFN- producing CD56 + cells in response to IL-23 in combination with IL-1β or IL-18.
In addition to the elucidation of the positive feed-back loop of IL-23 expression we observed a synergy between LPS (TLR4) and flagellin (TLR5) and between LPS and CL075 (TLR8/7) with respect to the induction of IL-23 production in monocytes and M1. During an infection, synergy between TLRs is likely of importance because pathogens express multiple TLR agonists. LPS and flagellin for instance, which in synergy induce large quantities of IL-23 in M1, are both expressed by Salmonella. Synergy between LPS (TLR4) and R848, another TLR8/7 agonist, has been observed in the induction of IL-12 and IL-23 in human DCs (33). Both TLR8 and TLR4 are implicated in the resistance against M. tuberculosis (34,35), suggesting the synergy we observed between the TLR4 and the TLR8/7 agonists may be relevant in mycobacterial infections.
In conclusion, we have shown that various TLR agonists and infection with Salmonella can induce IL-23, IL-18 and IL-1, but not IL-12, production in monocytes and M1. Furthermore, our findings indicate that a positive feedback loop exists in which IL-23 can enhance its own production via the induction of IFN- and GM-CSF, which both prime monocytes for enhanced IL-23 production.
Last, IL-23, in combination with IL-1, could prime monocytes to produce IL-12p70 in response to LPS, via the activation of CD56 + cells, thereby amplifying the type-1 cytokine pathway to IFN- production.
IL-23 plus IL1 induced IFN- production in CD56 + cells, which consist of CD56 + /CD3 -NK cells and CD56 + CD3 + NK-like T cells. In half of the donors IFN- production was observed in both NK cells and NK-like T cells, whereas in the other half IFN- production was observed only in the NK cells. Possible explanations for these interindividual differences may be previous (recent) exposure to different pathogens or genetic differences between donors; these may be elucidated when individual donors are assayed at regular intervals over a longer time period to determine whether or not the individual cytokine production profiles are constant.