RAW 264.7 macrophages were stimulated with KDO (100 ng/mL), IFN-β (300 pM), and/or 8-Br (100 µM). These ligands were applied individually or in combinations containing KDO for 60 min and 120 min. Total RNA was extracted using TriPure (Roche, Indianapolis, IN, USA) following the manufacturer's protocol. Each experiment was conducted in duplicate.

Mouse oligonucleotide arrays were fabricated with 15,631 oligomers that were 65 bp or 70 bp long. The oligomers were purchased from Operon (Huntsville, AL, USA) and Sigma-Genosys (St. Louis, MO, USA) and were inkjet-printed onto glass slides by Agilent Technologies (Santa Clara, CA, USA). The list of genes is available through the Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo) under accession number GPL254.

Hybridization and the resulting analyses were performed as previously described [11-13]. Cy5-labeled cRNA (from ligand-treated cells) and Cy3-labeled cRNA (from time-matched controls) were hybridized in the array. Dye-swap labeling was performed for each pair of samples. The arrays were scanned with an Agilent Scanner G2505A (Agilent Technologies), and image files were extracted with background subtraction and dye normalization using Agilent G2566AA Extraction Software version A.6.1.1 (Agilent Technologies). The microarray data used in this study were deposited into GEO (http://www.ncbi.nlm.nih.gov/geo) under the series accession number GSE11449 and platform ID GPL254.

Selected genes were subjected to real-time PCR analysis in duplicate using the 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) following the protocol provided by the manufacturer. The measurement was normalized to a β actin RNA control. The primers of genes are as follows: TNF, F 5'-TCA GCG AGG ACA GCA AGG-3', R 5'-GTG AGT GAA AGG GAC AGA ACC-3'; β-actin, F 5'-CTT TGC AGC TCC TTC GTT GC-3', R 5'-ACG ATG GAG GGG AAT ACAGC-3'. The remaining primer sequences are available upon request.

The Bio-Plex cytokine assay (Bio-Rad Laboratory, Hercules, CA, USA) was used for quantification of cytokines in tissue culture supernatants. In the kit, an antibody directed against each cytokine is covalently coupled to a different color-coded polystyrene bead. The conjugated beads are reacted with a sample containing a known (standard) or unknown amount of cytokines. After unbound cytokines are removed, biotinylated detection antibodies are added to the reaction. The complexes are detected by streptavidin-phycoerythrin, which has fluorescence characteristics. Quantification is carried out by a specialized microtiter plate reader.

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis was performed to resolve a mixture of proteins by size. The resolved proteins were transferred from the gel to a nitrocellulose membrane (Invitrogen, Carlsbad, CA, USA) via electroelution. The blot was processed for detection of specific phosphoproteins with a mixture of antibodies. All of the incubations and washes were conducted at room temperature with gentle shaking using an oscillating platform shaker. A Molecular Dynamics Storm 860 gel and a blot analysis system (Amersham Pharmacia Biotech, Piscataway, NJ, USA) were used.

Features that were differentially expressed in response to ligand stimulation were identified with Linear Models for Microarray Data (LIMMA, http://bioinf.wehi.edu.au/limma). The Benjamini-Hochberg method was used to control the false discovery rate [14]. Four independent experiments were used to identify differentially expressed genes, and p < 0.05 by ANOVA was deemed statistically significant. The average log₂-(treated/control) and the average difference (avg.DIF) were hierarchically clustered to visualize the patterns of gene expression changes and the nonadditivity, respectively [15]. Only features with significant expression changes and significant nonadditive responses at one or more time points were included, respectively. Clustering was done one way across the features with experimental conditions aligned in ligand, followed by time course orders. Euclidean correlation coefficient and complete linkage were used as similarity metrics [16]. The hierarchical clustering program that was used was implemented in the Multiple Experiment Viewer (MeV, www.tigr.org/software/tm4/mev.html).

Gene accession numbers, fold-changes of each treatment, and p-value were loaded into the IPA5.0 software (http://www.ingenuity.com). The genes were grouped with their molecular functions and depicted as a network picture, with their direct and indirect relationships represented with the gene ontology by the canonical pathway analysis. All these annotations used Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway maps, and the most significant genes were summarized in the network and as a graph.

We first characterized gene expression profiles after treating RAW 264.7 macrophages with the individual ligands in a kinetic study (60 and 120 min). We found that each of the three ligands, when applied singly, exhibited maximum transcriptional activation after 120 min of stimulation (as analyzed by Limma) (Fig. 1A). We further explored the expression data using k-means/medians clustering (KMC) in MeV, in which we divided the gene expression profiles into clusters in order to identify groups of genes that might participate in similar pathways. The clusters that exhibited changes greater than 2-fold are shown in Fig. 1B. Upon single-ligand treatment, KDO and IFN-β mediated robust transcriptional upregulation, whereas 8-Br stimulation resulted in both up- and downregulation of gene expression. The top-most functions of the genes by all ligand treatment and major networks are given in Fig. 2.

After classifying the genes by KMC, we subcategorized them according to their functions. KDO-regulated genes included transcription factors and genes involved in inflammation, immune responses, signal transduction, cell cycle regulation, cell proliferation, differentiation, metabolism, transport, cell adhesion, and apoptosis (Supplementary Table 1). As expected, KDO-treated cells exhibited a gene regulation pattern similar to that in LPS-treated cells, which included upregulation of Tnf, Il-1α, Il-1β, Cxcl4, and c-Myc. Furthermore, we also observed that KDO regulated several immediate early response genes, such as Egr1, Egr3, Gadd45α, Dusp1, and Dusp2, all of which are reportedly upregulated by LPS [11]. Our results indicated that KDO elicits transcriptional responses that are similar in gene identity and potency to LPS responses.

During the KDO-mediated transcriptional response in RAW 264.7 cells, many of the genes induced at 120 min were candidate feedback repressors of signal transduction [17]. Such repressors were specific to nuclear factor-κB (NF-κB) and Il-1 and include nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor epsilon (Nfκbie, also known as IκBε), nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor zeta (Nfκbiζ, also known as Iκbζ), and interleukin 1 receptor antagonist (Il-1ra). In addition, KDO upregulated Tnfsf9, Tnfaip3, Traf1, and Fas6, indicating that KDO has the potential to elicit inflammation. Its upregulation of Irf1 suggested that KDO also activated interferon-stimulated response elements through the MyD88-independent pathway.

Other KDO-induced genes included IFN-β, CD40, Socs5, Cish, Fosl1, Rgs16, CCR9, Ifit3, and cardiotropin 1 (Ctf1). Among these genes, Socs5 and Cish are feedback inhibitors of the JAK/STAT pathway; both were upregulated after 120 min of stimulation. KDO also promoted Il-10 expression after 120 min of stimulation; this gene contains 2 Creb binding sites in its promoter; so, KDO may have upregulated this gene through enhanced phosphoinositide 3-kinase (PI3K)-v-akt murine thymoma viral oncogene (AKT)-glycogen syntase kinase (GSK) pathway signaling and Creb phosphorylation.

Binding of IFN-β to the Type I-IFN receptor activates the JAK/STAT signaling pathway. Phosphorylated Stat proteins regulate gene transcription by binding to conserved sequences within the promoters of IFN-induced genes. IFN-stimulated Stat activation mediates the expression of a second family of transcription factors, the Irf proteins [18]. Irf proteins also mediate transcriptional responses to IFN stimulation. In our study, IFN-β stimulation was associated with Stat2 and Irf1 production, and recent findings indicate that Irf proteins play critical roles in interferon regulation [19].

IFN-β induced a number of genes involved in inflammatory/immune responses, transcription, signaling, and apoptosis. The most highly responsive genes induced by IFN-β were interferon-inducible GTPase 1 (Iigp1), Ifit1, myxovirus (influenza virus) resistance 2 (Mx2), Ifit2, Cxcl9, Cxcl10, Cxcl11, ISG15 ubiquitin-like modifier (ISG15), Socs1, CD40, and TLR3. Secondary responsive genes included Gbp1, Gbp2, Gbp5, Il-13, and Fas. The fold-changes in the expression of these genes are provided in Supplementary Table 2.

The 8-Br-regulated genes were involved in transcription, signaling, inflammation, apoptosis, metabolism, proteolysis, transport, and cell cycle regulation (Supplementary Table 3). Treating cells with 8-Br upregulated several immediate early transcriptional targets, including nuclear receptor subfamily 4 group A member 2 (Nr4a2), Gadd45 beta, Dusp1, Pde4b, Crem, Cebpb, Vegf, Bag3, and v-maf musculoaponeurotic fibrosarcoma oncogene family protein F (Maff).

Surprisingly, we observed that 8-Br treatment downregulated transcription of Rgs16, Tnf, Cxcl4, Ptger4, Tnfsf9, c-Myc, and Map3k8 by more than 2-fold (Fig. 3A). 8-Br can influence the mitogen-activated protein kianse (MAPK) pathway by downregulating Mapk3k8. Map3k8 activates the kinases p38 and JNK and the transcription factors NF-κB and NF-AT, although its specialized role is to activate the extracellular signal-regulated protein kinase (ERK) pathway downstream of most TLR signals [20-23]. We observed 8-Br-mediated downregulation of 2 proto-oncogenes, c-Myc and Map3k8, and a cell cycle gene, Gadd45α (Supplementary Table 3), which emphasizes that 8-Br cAMP may also have important inhibitory roles in cancer cell proliferation [24, 25].

We next hypothesized that the co-downregulated genes may have common transcription factor binding sites; therefore, we examined their promoter regions using TOUCAN (http://homes.esat.kuleuven.be/~saerts/software/toucan.php) and analyzed the genes downregulated by 8-Br using MotifMogul (http://xerad.systemsbiology.net/MotifMogulServer). We found AP1 and/or NF-κB subunit binding sites in the upstream regions of all of the downregulated genes, which adds to evidence from previous studies [26, 27]. Among these genes were several classical targets of NF-κB (for example, Tnf) and AP-1 (for example, c-Myc and Gadd45α); these genes are highlighted with arrows in Fig. 3A and shown in greater detail in Fig. 3B. Since protein kinase A (PKA) activation by cAMP suppresses the MAPK cascade through phosphorylation of Raf [28], we can not exclude the possibility that elevated cAMP may, in part, reduce the transcription of the targeted genes by modulating protein binding to AP-1 sites, by inhibiting the ERK pathway through Raf, and/or by specifically affecting NF-κB subunits. Importantly, Il-1β was up-regulated quickly, whereas Il-1α and Il-10 were not affected prior to 120 min of stimulation. This differential mRNA regulation of genes that are all involved in inflammation suggested that cAMP utilized more than one distinct mechanism to alter the expression of specific sets of genes.

To understand how the presence of one or more ligands modified the transcriptional responses to the other ligands, we evaluated features that exhibited non-additive responses during the double-ligand studies and grouped them into three categories: significant responses 1) to both ligands; 2) to either ligand A or ligand B; 3) to neither ligand (Table 1). A total of 433 features showed non-additive responses above threshold levels in at least one of the double-ligand studies.

In all of the double-ligand studies involving KDO, we found differences in gene transcription that were not present when the ligands were applied individually. Applying KDO together with IFN-β augmented IFN-β-mediated transcriptional responses. In contrast, when KDO was combined with 8-Br, each ligand abrogated the single-ligand response of the other.

KDO enhanced IFN-β-dependent signaling pathways in a time-dependent manner. Since TLR4 activation can also produce IFN-β, the KDO-mediated enhancement suggested cooperation between exogenously applied and endogenously produced IFN-β (Supplementary Table 4). Several IFN-β and KDO primary responsive genes were upregulated, including genes that activate the immune system. In some cases, the IFN-β gene signature dominated over the KDO gene signature. For example, IFN-β inhibited two KDO primary responsive genes, Il-1ra and Nfκbζ, but KDO did not change the expression of any of the interferon-responsive primary genes. The expression levels of select genes were validated by using QRT-PCR (Fig. 4).

Cytokine assays, conducted following a combination of LPS and IFN-β treatment, also demonstrated non-additive enhancement of Il-6, Il-10, RANTES, and TNF protein levels (Fig. 5A). This increase could be due to synergistic production of transcription factors in response to these ligands, apart from their classical transcription factors. It is likely that the rapid, greater-than-2-fold increase in Socs1 and Cish production may have contributed to the attenuated expression of the KDO secondary responsive genes [29, 30]. Il-10 is an anti-inflammatory cytokine regulated through the PI3K-AKT-GSK-CREB pathway [31, 32], and we noted its expression after 120 min of stimulation, suggesting that the PI3K pathway may have been activated after 60 min of stimulation.

KDO- and IFN-β-treated cells exhibited upregulation of many interferon-regulatory factors (IRFs), including Irf1, Irf4, and Irf8. Irf4 inhibits the expression of MyD88-dependent IRF genes by competing with IRF5 for binding to the same region of MyD88 [33]. Combined KDO and IFN treatment also upregulated genes with IRF3 binding sites in their promoters; i.e., Ifit1, Ifit2, Ifit3, Isg15, and Gbp1. Our gene profiling results suggested that IRF3 is sufficient to induce antiviral responses at the initial activation time point, consistent with an earlier report that demonstrated a role for IRF3-dependent genes in establishing antiviral responses in Jurkat cells [34].

Interestingly, two of the genes that exhibited non-additive responses are produced only by IFN-γ GAS elements: Irf8, previously known as interferon consensus sequence-binding protein 1 (ICSBP), and transporter associated with antigen processing-1 (Tap1). Upregulation of GAS-driven genes requires the Stat1 homodimer exclusively but not NF-κB; they are both produced by IFN-γ. Interestingly, in our study, we did not observe elevated expression of IFN-γ or STAT1 mRNA, but we still observed the production of GAS-driven genes.

Two phosphorylation sites, Tyr-701 and Ser-727, differentially regulate Stat1, and phosphorylation of both sites is required for maximal transcriptional activity [35, 36]. A previous study demonstrated that Stat1 is phosphorylated on both Tyr-701 and Ser-727 in LPS-treated cells after 2 h [37]. We surmise that this enhanced phosphorylation of STAT1 may be mediated by KDO and IFN-β. However, when we conducted STAT1 phosphoprotein measurements with KDO and IFN-β in combination, we observed a 3-fold upregulation in phosphorylated STAT1 after 30 min (Fig. 5B). This increase in phosphorylated STAT1 during double-ligand stimulation was lower than the increase in phosphorylation of STAT1 induced by IFN-β alone (10-fold). Our previous studies with LPS- and IFN-γ-treated RAW 264.7 cells also showed non-additive increases in Irf8 and Tap1 gene expression [11]. This elevated GAS gene expression definitively indicated that KDO and IFN-β activate GAS promoters, either directly or indirectly.

Supplementary Table 5 shows the genes regulated upon double-ligand treatment with KDO and 8-Br. Importantly, combined KDO and 8-Br application increased the transcription of Tnf but did not affect Ccl4, Map3k8, or c-Myc (Fig. 3A). Since 8-Br downregulated these genes during the single-ligand application studies, these results suggested that the KDO-induced pathway attenuated these 8-Br-induced changes. We classified the direct targets of CREB (major TF induced by cAMP analogs) and NF-κB (direct target of KDO) to see which of the major transcription factors are affected (Table 2). For CREB, we searched in the website developed by Marc Montminy, which takes advantage of CHIP-on-chip data (http://natural.salk.edu/CREB). For potent NF-κB sites, we looked in the website http://www.bu.edu/nf-kb. Surprisingly, both of these transcription factors are not affected by double ligand treatment of KDO and 8-Br, reinsisting that KDO directly inhibits 8-Br-induced changes. CREB can be induced to a certain extent through the ERK pathway by KDO; however, such a claim remains to be proven in the future studies. This approach further raises a question on by which mechanism, if not CREB and NF-κB, KDO attenuates 8-Br induced changes, which also remains to be tested. One possible explanation for this phenomenon is that KDO stimulation rapidly downregulated the 8-Br-induced increase in intracellular cAMP through the action of phosphodiesterases (PDEs) that transform cAMP into inactive AMP [38]. In support of this idea, LPS stimulation of mouse peripheral leukocytes induced Pde4b mRNA accumulation and increased PDE4 activity [39]. This increase in PDE4 activity leads to a decrease in cAMP, thus removing the cAMP constraint and allowing for full induction of Tnf mRNA. PKA phosphorylation inhibits Raf1 directly [40, 41], and this inhibition occurs downstream of Ras [42, 43] to inhibit Ras-dependent signaling to ERK [44]. Hence, Raf can be the crosstalk point between the KDO and 8-Br signal transduction pathways.

Supplementary Table 6 shows the transcriptional responses in RAW 264.7 cells upon triple-ligand stimulation. We observed the following effects: 1) upregulation of the primary and secondary KDO- and IFN-β-responsive genes; 2) production of various inflammatory cytokines and their receptors (Il-1a, Il-1b, Il-6, Il-20, Il-17rb, Il-15, Il-18, Il-13, Il-4ra, Il-10ra, and Tnf) and the anti-inflammatory cytokine Il-10; 3) induction of apoptosis-inducing transcription factors (Fosl-1, Atf3, Egr1, Egr2, Fos-b, Jun dimerization protein [Jdp], and Irf14); and 4) up-regulation of multiple G-protein-coupled receptors, like leukotriene B4 receptor1 (Ltb4r1), complement component 5a receptor 1 (C5ar1), chemokine (c-c) receptor 1 (Ccr1), Gpr132, Gpr65, and calcitonin receptor-like (Calcr1). In these triple-ligand studies, the transcription factors exerting the greatest regulatory control over gene expression were NF-κB, Ap1, Nfat, Stat2, and Stat3. Our data demonstrated that concomitant stimulation with all three ligands induced KDO and IFN-β primary and secondary responsive genes, but it repressed 8-Br-induced transcriptional changes.

We observed an interesting transcriptional change in gene clusters (Fig. 6). NF-κBζ is weakly expressed in cells but induced to a marked extent upon LPS stimulation and regulates a set of gene expression, including Il-6 [45]. However, a recent study demonstrated that Il-6 promoter can also be activated by unknown molecules acting far from the transcriptional active site by cis-regulatory molecules [46]. We observed the same phenomenon when we stimulated RAW 264.7 cells with KDO. NF-κBζ was upregulated by KDO at 60 min, and Il-6 was not induced by KDO till 120 min, which is in line with a previous observation [45]. This observation is very novel to be explained. Triple-ligand studies showed upregulation of Il-6 as early as 60 min (about 6-fold). This indicates that though KDO-induced pathways need de novo synthesis of NF-κBζ for Il-6 production, it does not require de novo synthesis of NF-κBζ for the induction of Il-6 when combined with IFN-β and 8-Br. We then looked for the possible ligand that could have mediated this and found that KDO and 8-Br never produced Il-6 on single-ligand stimulation. In that case, the only remaining candidate is IFN-β. IFN-β upregulated Il-6 in our experiment at 120 min (TLR3 can produce Il-6 due to the late phase of NF-κB activation). IFN-β also increased mRNAs of Atf3, Ccl4, and Map3k8 at 120 min. It is of interest to know what pathway regulates these genes in future studies, since most of them were reported to be regulated by a MyD88-dependent pathway during TLR signaling. Either the presence of a putative STAT1 site is sufficient to induce the MyD88-dependent genes or IFN-β might have crosstalk with the MAPK pathway.