Circadian clock regulates the host response to Salmonella

To determine whether the circadian clock regulates the host response to infection with S. Typhimurium, we infected wild-type (WT) mice by oral gavage either at 10:00 AM (day, early rest phase; zeitgeber time 4, ZT4) or at 10:00 PM (night, early active phase; ZT16). At 48, 60, 72, and 78 h postinfection (p.i.), mice were killed, and tissue samples collected for bacteriology, histopathology, and gene expression analyses (Fig. 1). We observed that S. Typhimurium colonization significantly changed with time of infection, especially at later time points. Notably, S. Typhimurium numbers were significantly increased in the colon content of mice infected at 10:00 AM in comparison with 10:00 PM at both 72 and 78 h p.i. (Fig. 1A); colonization of Peyer’s patches and spleen was also significantly higher in mice infected during the day (Fig. S1A). Next, we determined whether the degree of the host response to infection changed with the time of inoculation. Histopathology showed that ceca from mice infected during the day were on average more inflamed than those from mice infected during the night, at 48, 72, and 78 h p.i. (Fig. 1 B and C; Figs. S1B and S2). In contrast, at 60 h p.i., a mild increase in inflammation was observed in mice infected at night, suggesting that the time of infection was not the only variable influencing the inflammatory response. Reduced signs of cecal inflammation were characterized by low-grade submucosal edema and neutrophil influx (Fig. 1 B and C and Figs. S1B and S2). Major differences were also found in the levels of cryptitis and surface erosions (Fig. S1B), which constitute other classical signs of inflammation (10–12). To ascertain whether these differences were accompanied by changes in gene expression, we analyzed expression of proinflammatory cytokine Tnfα, neutrophil chemoattractant chemokine (C-X-C motif) ligand 1 (Cxcl-1), and antimicrobial peptide lipocalin-2 (Lcn-2) in the cecum of infected and uninfected mice and found significant differences during the course of infection (Fig. 1D and Table S1). Consistent with our previous findings (11–13), mice infected during the day showed increased expression of these genes compared with uninfected controls, and the expression was dependent both on time of infection and time of death (Fig. 1D). Genes in mice infected in the morning (ZT4) or at night (ZT16) showed maximal expression when killed during the day (48 and 72 h for mice infected at ZT4; 60 h for mice infected at ZT16). This finding was particularly notable for Tnfα because its expression was increased at both 48 and 72 h in mice infected at ZT4 (ZT4 death) relative to those infected at ZT16 (ZT16 death), and the opposite trend was observed at 60 h (ZT16 death for mice infected at ZT4; ZT4 for mice infected at ZT16). In line with this observation is the absence of significant changes in expression in mice killed 78 h after infection, corresponding to ZT10 for mice infected at ZT4 and ZT22 for mice infected at ZT16 (Fig. 1D). Additionally, the overall expression levels of the three proinflammatory genes analyzed were reduced in mice infected at ZT16 (yellow and gray graphs of Fig. 1D), consistently with the differences observed in colonization level and pathology score.

To establish whether circadian transcription follows a normal profile during infection we analyzed the expression of the clock gene period 2 (Per2). Per2 oscillation in uninfected mice demonstrated proper synchronization (Fig. 1D). Notably, Per2 expression was progressively down-regulated after infection (Fig. 1D).

These results suggest that a circadian mechanism may regulate components of the innate immune response to acute Salmonella infection, also revealing a profound repression of circadian components during infection.

Our finding that the inflammatory response to S. Typhimurium infection is time-of-day–dependent in vivo prompted us to determine whether we could reproduce similar results in vitro in a less complex system. As detailed earlier, macrophages are involved in the first response to S. Typhimurium infection. Moreover, macrophages have an efficient clock machinery, and LPS-activated pathways in these cells are under tight circadian regulation at multiple levels (3, 4). However, bone marrow-derived macrophages (BMDMs) do not synchronously oscillate after 1 wk of differentiation in vitro (Fig. S3A), and LPS stimulation at different circadian times leads to similar levels of Il-6 expression in these asynchronous cultures (Fig. S3B). We therefore tested whether BMDMs could be synchronized by common methods such as dexamethasone or high-serum treatments (Fig. S3 C and D). Oscillation of circadian genes Per2 (Fig. S3 C and D), cryptochrome 1 (Cry1) and brain and muscle ARNT-like protein 1 (Bmal1) (Fig. S3C) confirmed entrainment of these cells. To better evaluate the contribution of the circadian system to the expression of proinflammatory genes, we added LPS (1 μg/mL) to synchronized macrophages at different times of their circadian cycle and followed the expression profile of Il-6 (Fig. S3D). As expected, we obtained different curves of expression depending on the time of LPS administration, with minimal induction for administration at T18 or T30 and major induction at T12 or T24, where T0 is the time when synchronization began. Notably, expression levels appear to be dependent both on the time of treatment and on the time of collection, as suggested by our in vivo data.

We next ascertained whether disruption of the circadian clock could affect the function of some central component of the LPS response. Circadian gene expression in mice carrying a mutation of the master circadian regulator Clock [Clock mutant or clock/clock mice (19)] is both phase-shifted by 8 h as well as reduced compared with WT littermates (20). Clock mutant mice also fail to rhythmically express a number of immunoregulatory genes in the liver (21). Therefore, we followed the timing of expression of different cytokines after LPS stimulation of BMDMs isolated from either WT or Clock mutant mice (Fig. 2). Remarkably, we observed an overall reduction in the fold induction of the proinflammatory genes Il-6, Il-1β, Tnfα, Cxcl-1, Ifn-β, and Chemokine (C-C motif) ligand 2 (Ccl2) in response to LPS stimulation in BMDMs isolated from Clock mutant mice compared with WT mice (Fig. 2A); BMDMs isolated from Clock mutant mice also exhibited reduced secretion of IL-6 and TNF-α after 24 h of LPS stimulation (Fig. 2B). A comparable reduction in the levels of IL-6 in the supernatant of BMDMs isolated from Clock mutant mice, compared with WT mice, was obtained after infection with S. Typhimurium (Fig. 2C). Because the induction of IL-6 is largely dependent on LPS stimulation of the TLR4 pathway, we also infected BMDMs with an isogenic S. Typhimurium strain carrying a mutation in the lipid A acylation pathway (msbB mutant), which impairs signaling through TLR4 (11). As predicted, secretion of IL-6 was reduced in BMDMs infected with the msbB mutant compared with infection with WT S. Typhimurium. In contrast, both S. Typhimurium WT and the msbB mutant elicited similar low levels of secretion of IL-6 (Fig. 2C) in BMDMs isolated from Clock mutant mice, indicating that TLR4 signaling in response to S. Typhimurium was not induced in the absence of a functional clock. Next we sought to investigate whether the secretion of the proinflammatory cytokine IL-1β was also impaired in Clock mutant mice. IL-1β is directly induced by TNF-α, causing its expression to peak later than Tnfα, and mature IL-1β secretion requires both TLR4 and NLR signaling (22). Upon S. Typhimurium infection of BMDMs, LPS activation of TLR4 induces pro–IL-1β expression, which is followed by NLR-mediated signaling through ice protease-activating factor/NLR family CARD domain-containing protein 4 (Ipaf/Nlrc4) in response to Salmonella T3SS-1 secretion of flagellin into the cytosol (22). As expected from these earlier studies, S. Typhimurium strains carrying either a mutation in msbB (unable to signal through TLR4), a mutation in both flagellin genes (fliC fljB mutant, unable to signal through Ipaf/Nlrc4), or a mutation in T3SS-1 (invA mutant, unable to signal through Ipaf/Nlrc4) induced secretion of lower levels of IL-1β compared with S. Typhimurium WT (Fig. 2D) in BMDMs from WT mice. In contrast, infection of BMDMs from Clock mutant mice with S. Typhimurium WT resulted in a marked reduction of IL-1β secretion. Moreover, all S. Typhimurium mutants elicited similar low levels of IL-1β in Clock mutant BMDMs. Collectively, we observed that secretion of proinflammatory cytokines in Clock mutant BMDMs was particularly low and comparable to levels elicited by S. Typhimurium mutants designed to evade activation of particular components of the inflammatory response. Thus, an intact circadian machinery appears to be required for the induction of proinflammatory cytokines in vitro in response to Salmonella infection.

Because the circadian clock is necessary for a robust proinflammatory response in isolated macrophages, we explored its involvement in the host response to Salmonella in vivo. We infected Clock mutant mice at 10:00 AM (ZT4) or 10:00 PM (ZT16) (Fig. 3A). Differently from WT mice, similar numbers of S. Typhimurium were recovered 72 h p.i. (Fig. 3A). Similar results were also obtained when Clock-deficient mice (Clock^−/−) and their WT littermates were infected, thus confirming that a functional CLOCK protein is necessary for the observed effect. (Fig. S4A). Furthermore, a significant difference was found in the colonization of Peyer’s patches between WT mice infected at day and Clock mutant mice, even if these lymphatic structures were significantly enlarged in Clock mutant mice during infection (Fig. S4B). Moreover, the histopathology revealed higher inflammation in Clock mutant mice infected at night compared with mice infected during the day (Fig. 3 B and C; Fig. S4 C and D). Altogether, these results suggest that a circadian mechanism regulates components of the host response upon Salmonella infection and that clock disruption largely affects this response.

Because previous microarray studies revealed that ∼10% of genes follow circadian oscillation in almost all tissues (7), we set out to gain further insight into the circadian clock regulation of the host response to infection by analyzing global changes in gene expression by microarray analysis. To determine whether circadian regulation occurs at the transcriptional level in response to infection in vivo, we analyzed the differential gene expression profile by microarray analysis of the cecum in WT and Clock mutant mice 72 h p.i. with S. Typhimurium or mock control at different circadian times (day, ZT4; night, ZT16). Our analysis identified four main clusters of transcripts with different patterns of expression, which we numbered clusters 1–4 (Fig. 4A and Dataset S1). The gene expression profile in cluster 1 was characterized by the induction of genes linked to inflammation and immunity in response to infection, including genes involved in immune cell activation and antimicrobial responses. Genes in cluster 2, related to glycolipid biosynthesis and protein transport, were up-regulated during infection. In contrast, genes involved in lipid and steroid metabolism (cluster 3) and dendritic cell maturation (cluster 4) were mostly down-regulated in response to infection.

Notably, several genes in all four clusters were regulated in a time-specific and clock-dependent manner during infection (Fig. 4A, Datasets S2, S3, and S4). Moreover, a detailed pathway analysis identified several transcripts with a different day–night expression in WT mice (Datasets S1 and S2). Specifically, we observed the most relevant changes of gene expression for transcripts involved in the antimicrobial response (Fig. S5A), acute inflammation (Fig. S5B), leukocyte chemotaxis, and antigen presentation and processing (Fig. 4A and Dataset S1). Consistent with our observations, the analysis of genes in cluster 1 revealed a general low level of induction of the inflammatory response in Clock mutant mice, a response that was slightly higher at night (Fig. 4A, Fig. S5 A and B, and Dataset S3). These observations were further validated through analysis of a subset of cluster 1 genes by quantitative real-time PCR (Fig. 4 B and C and Table S1). In particular, many of the proinflammatory (Tnfα, Il-17, Cxcl-1) and antimicrobial [Lcn2, regenerating islet-derived 3 γ and β (Reg3γ and β)] genes analyzed were up-regulated at night upon infection, but to lower levels than during the day. In contrast, levels of expression in Clock mutant mice were mostly similar between day and night, with only a couple genes [Tnfα and S100 calcium binding protein A8 (S100a8)] that were expressed at higher levels at night than during the day.

Because expression of different cytokines follows a circadian rhythm (2–4), we investigated whether intrinsic temporal changes may influence gene expression in the intestine of both WT and Clock mutant mice (Fig. S6A and Dataset S5). As expected, a higher proportion of genes oscillated in the cecum from WT uninfected mice compared with Clock mutant mice (Fig. S6 A and B), including circadian genes, metabolic genes, and genes involved in the immune response (Fig. S6 A and C and Dataset S5). Because we observed that some transcripts (Tnfα, Cxcl-1, and Lcn2) did not show differences between day and night, whereas other genes (Il-17, Reg3β, and Reg3γ) were strongly up-regulated at night (Fig. 4C), we sought to determine the absolute expression levels of each gene (Fig. S7). This analysis showed that the expression level of the majority of genes analyzed was lower in Clock mutant mice in both basal condition and after infection (Fig. S7). Together, these results suggest that circadian regulation influences cytokine expression at different levels, by modifying both basal and induced transcription of pro- and anti-inflammatory genes during acute infection.

Little is known about the metabolic and physiological changes that occur during the circadian cycle in infected mice. Our data suggest that the rhythmicity of circadian and metabolic genes is suppressed or attenuated during infection (Fig.4A and Dataset S1). Furthermore, we noticed that genes peaking at different circadian times also undergo opposite transcriptional regulation after infection (Fig. 4A). For instance, cluster 2 genes massively oscillate in basal conditions and are strongly induced after infection in WT mice; however, their cycle is inverted and their induction is completely abolished in Clock mutant mice. Similarly, clusters 3 and 4 contain genes that are up-regulated at basal levels at day or night, respectively, but are repressed following infection (Fig. 4A). Within these categories are the clock genes (Dataset S1 and Fig. S8). Consistent with previous studies (23), circadian components [Bmal1, also called Aryl hydrocarbon receptor nuclear traslocator-like (Arntl), period proteins, cryptochromes, neuronal PAS domain protein 2 (Npas2), nuclear receptor subfamily 1, group D, member 1 (Nr1d1)] oscillated in uninfected mice and showed either a marked reduction of the oscillation or a phase-shifted rhythm in Clock mutant mice (Fig. S8 A and B). Also, genes of the main metabolic pathways were found in these clusters, as predicted by previous studies (24). Moreover, the time-dependent changes observed in the transcription of genes involved in dendritic cell development and leukocyte function, as well as other immune genes included in cluster 4 (Fig. 4A and Dataset S1), suggest a possible circadian control of the development of innate and adaptive immunity following infection.

To extend our comprehension of the transcriptional pathways participating in the circadian activation of the host defense against infection, we used a computational modeling approach to predict the transcriptional regulatory networks involved in the control of genes in each of the four clusters identified in our genomic profiling analysis (the complete representation is available at www.ics.uci.edu/~baldig/CLOCK/salmonella/). In cluster 1, we identified main synergistic nodes connecting transcription driven by BMAL1:CLOCK (ARNTL:CLOCK) to critical inflammatory pathways (Fig. 4D and Fig. S9A). The gene node graphics show transcription factors with significant changes in expression between different conditions (blue, WT infected vs. uninfected day; green, WT infected vs. uninfected night; brown, Clock mutant infected vs. uninfected day; orange, Clock mutant infected vs. uninfected night; P < 0.05; Fig. 4D and Fig. S9 A and B). In agreement with previous reports, NF-κB and hypoxia inducible factor 1, alpha subunit (HIF-1α) are the transcription factors with the largest numbers of connections (25, 26). Both NF-κB and HIF-1α share many target genes with the transcription factor BMAL1 (ARNTL), which cooperates with CLOCK in regulating circadian transcription. Notably, HIF-1α appeared to mediate significant changes in all four conditions that we analyzed. Nodes were much larger in WT mice compared with Clock mutant mice, thus indicating that many pathways regulated by HIF-1α also require a functional clock system. This observation is of particular interest because cross-talk between hypoxic and circadian pathways has been proposed, and Hif-1α is thought to be a clock-controlled gene (27).

We next sought to determine whether circadian changes influence Salmonella growth, which is known to be enhanced by intestinal inflammation, in part because Salmonella is resistant to some antimicrobial proteins secreted in the inflamed gut (11). As shown in Fig. 1D and Fig. 4B and Figs. S5A and S7, the gene Lcn2, which encodes for lipocalin-2, an antimicrobial peptide that sequesters the essential nutrient iron and thereby reduces the growth of susceptible bacteria (28), has a different day–night expression following infection. S. Typhimurium is resistant to LCN2 because it encodes for the synthesis and uptake of a high-affinity iron chelator, termed salmochelin (29). High expression of LCN2 in the inflamed gut enhances Salmonella colonization over competing microbes, whereas a LCN2–sensitive mutant in the salmochelin receptor IroN has a colonization defect and is outgrown by Salmonella WT in mice expressing Lcn2, but not in mice carrying a Lcn2 mutation (11). Thus, we hypothesized that WT S. Typhimurium would have a greater colonization advantage in competition with the iroN mutant during the day, when Lcn2 is highly induced, whereas we would recover both strains at similar levels at night, when Lcn2 is expressed at lower levels. Infection of WT mice at ZT4 or ZT16 with a 1:1 mixture of WT S. Typhimurium and the iroN mutant confirmed this prediction. Salmonella WT outcompeted the iroN mutant in the colon content collected 72 h p.i. from mice infected during the day (Fig. 4E). In contrast, S. Typhimurium WT had no significant advantage over the iroN mutant at night (Fig. 4E), when Lcn2 was induced at lower levels. These results further suggest that circadian expression of Lcn2, and possibly other antimicrobial peptides, influences Salmonella colonization and its competitive advantage over susceptible microbes.

Mice kept in 12:12 light:dark conditions were pretreated with streptomycin (0.1 mL of a 200 mg/mL solution in sterile water) intragastrically, as described (10–12) 24 h before inoculation with WT S. Typhimurium [100 µL containing ∼1 × 10⁹ colony-forming units (cfu)] or with a 1:1 mixture of S. Typhimurium WT and iroN mutant strain, at two different circadian times (10:00 AM, ZT4; 10:00 PM, ZT16). Mice were euthanized at the time point indicated (48–78 h p.i.). For enumerating bacterial cfu, homogenates of colon content, Peyer’s patches, mesenteric lymph nodes, and spleen were plated on agar plates containing the appropriate antibiotics [LB plus carbenicillin (Carb), LB plus naladixic acid (Nal)]. Competitive index was calculated by dividing the output ratio (cfu of WT/cfu of iroN mutant) by the input ratio (cfu of WT/cfu of iroN mutant). Mice used as control were housed in the same conditions as infected mice, pretreated with streptomycin, and left uninfected. Additional information can be found in SI Materials and Methods.