Transcriptome Assessment Reveals a Dominant Role for TLR4 in the Activation of Human Monocytes by the Alarmin MRP8

Innate immune mechanisms have a key role in inflammatory diseases such as infections, cardiovascular diseases, inflammatory bowel diseases, and autoimmune diseases. During inflammatory activation of the innate immune system by pathogens, the release of cytokines and chemokines triggers the immune response, often leading to chronic conditions. During the past decade it has become increasingly evident that the immune response is also triggered by endogenous ligands known as damage-associated molecular patterns (DAMPs) or alarmins (1). DAMPs are intracellular molecules primarily involved in cell homeostasis, but they can also act as extracellular danger signals when released by damaged or activated cells. DAMPs are recognized by specific members of the TLR family as well as the multiligand receptor for advanced glycation end products (RAGE) (2, 3). Alarmins such as high-mobility group protein B1 and heat shock proteins orchestrate key events in the inflammatory response not only in infectious diseases, but also, more importantly, in sterile inflammation (4–6).

We have previously characterized the phagocyte-specific myeloid-related proteins (MRPs) MRP8 (S100A8) and MRP14 (S100A9) as DAMPs (7). Both proteins belong to the calcium-binding S100 protein family and are known to form homodimeric as well as heterodimeric MRP8/MRP14 complexes. MRP8 homodimers activate cells of the innate immune system and induce proinflammatory cytokine and adhesion molecule expression in phagocytes and endothelial cells, resulting in leukocyte recruitment and activation (8–10). Studies involving mice deficient for MRP14 have uncovered a pivotal role for MRP8/MRP14 in the promotion of inflammation in many clinically relevant conditions (7, 9, 11–15).

However, the identity of the receptor responsible for MRP8-mediated effects on inflammation is still a matter of debate. RAGE has been described as the receptor for several S100 proteins, including MRP8 and MRP14 (16, 17). In vitro studies have demonstrated that MRP8/MRP14 binds to immobilized RAGE (18, 19), and that RAGE but not TLR4 associates with MRP8/MRP14 in colon tumor cells (20). However, studies using TLR4 or RAGE knockout mouse strains indicate that TLR4 is the main receptor that functions during activation of murine phagocytes (9). Similarly, conflicting data also exist in humans. MRP8/MRP14 colocalizes with RAGE in LNCaP cells (21), and a functional role for MRP8/MRP14–RAGE interaction has been described in inflammation-associated cancer (22). MRP8/MRP14 also promotes tumor cell growth and mediates endotoxin-induced cardiomyocyte dysfunction in a RAGE-dependent manner (23, 24). In contrast to this, MRP8/MRP14 also induces RAGE-independent cell death in several tumor cell lines (25) and displays a TLR4-dependent catabolic effect on human chondrocytes (26).

Recent systematic comparisons of the inflammatory processes of mice and humans have suggested that inflammatory responses in human diseases correlate poorly with the corresponding mouse models. These data clearly underline the need for translational medical research to generate reliable conclusions on human inflammatory diseases (27). However, experimental systems involving primary human phagocytes are limited, as human knockout models are not available and efficiencies of small interfering RNA studies are limited by low transfection and high cell damage/mortality rates. We therefore took a systems biology approach to defining the response of human monocytes to MRP8 and subsequent downstream signaling pathways. We have confirmed the previously elucidated proinflammatory functions of murine as well as human MRP8 at the genome-wide level, and have also uncovered a novel antiapoptotic role for MRP8 in monocytes. Additional bioinformatic analyses clearly demonstrate a dominant role for human TLR4 during MRP8-mediated phagocyte activation. Taken together, our data have identified the TLR4/MRP8 axis as a potent and novel target in humans for future investigations into anti-inflammatory approaches.

Approval was obtained from the Ethics Committee of the Medical Faculty of Muenster for these studies. Human monocytes were isolated from blood samples after leukapheresis according to Roth et al. (28). Purity was always >90% as ascertained by flow cytometry. Cells were cultivated overnight in Teflon bags in McCoy’s 5a medium (Biochrom, Berlin, Germany) supplemented with 15% FCS (Life Technologies, Eggenstein, Germany) prior to use for stimulation or in functional assays.

Human embryonic kidney (HEK)293 cells stably transfected with human TLR4-CD14-MD2 or TLR2 genes were purchased from InvivoGen (San Diego, CA) together with control cells transfected with empty vectors. HEK293 cells stably transfected with human RAGE were provided by M. E. Bianchi (Division of Genetics and Cell Biology and Division of Regenerative Medicine, Stem Cells and Gene Therapy, San Raffaele University and Scientific Institute, Milan, Italy) and used as described for TLR4-expressing HEK293 cells (9). Cells were grown in DMEM (Biochrom) supplemented with 4.5 g/l glucose and 10% FCS at 37°C in 5% CO₂.

Human monocytes (1.6 × 10⁶/ml) or HEK293 cells (5 × 10⁵/ml) were stimulated with 5 μg/ml human MRP8, 1 ng/ml LPS, 10 ng/ml TNF-α, or 1 ng/ml IL-1β for 4 h.

HEK293 cells stably expressing TLR9 were transiently transfected with an NF-κB–driven luciferase reporter and wild-type (WT) RAGE or mCherry (control) using GeneJuice (Novagen) as transfection reagent. Cells were incubated at 37°C for 24 h to allow protein expression. Cells were subsequently stimulated with a dose gradient of CpG2006 (Metabion) from 63 to 250 nM either alone or together with 5 μg/ml MRP8. As controls, cells were also stimulated with MRP8 alone (5 μg/ml), LPS (200 ng/ml), or human TNF-α (1 ng/ml). Luciferase assay readout was done by incubating cell supernatants with coelentrazine (Promega) and subsequent measurement of luminescence with a SpectraMax i3 plate reader (Molecular Devices).

ER-Hoxb8 macrophage progenitors derived from WT, TLR4^−/−, RAGE^−/−, and MyD88^−/− C57BL/6 mice were generated as previously described (29). Cells were maintained in RPMI 1640 medium (Biochrom) supplemented with 10% FCS, 1 μM estradiol (E-2758; Sigma-Aldrich, Steinheim, Germany), and 20 ng/ml GM-CSF or 1% culture supernatant from a GM-CSF–producing cell line (B16 melanoma cells). Macrophage differentiation was induced by removal of estradiol. ER-Hoxb8 macrophages were used on day 5 of differentiation and were stimulated (2.5 × 10⁶) with 5 μg/ml murine MRP8 for 4 h.

LPS from Escherichia coli 055:B5, polymyxin B, and anti–β-actin (AC15) Ab were obtained from Sigma-Aldrich. Annexin V–allophycocyanin (31490016X2) and recombinant murine GM-CSF were purchased from ImmunoTools (Friesoythe, Germany) and staurosporine from Alexis Biochemicals (San Diego, CA). Anti–Bcl-x_L (610212) Ab was obtained from BD Biosciences (San Diego, CA). Abs against cleaved caspase-3 (9661), cellular inhibitor of apoptosis 1 (cIAP-1; 4952), IκBα (4812), phospho-p38 (9211), total p38 (9212), phospho-MEK (9121), phospho-ERK1/2 (9101), and total ERK (9102) were purchased from Cell Signaling Technology (Danvers, MA).

MRP8 was expressed in E. coli BL21(DE3) cells and purified as previously described (9, 30). Briefly, pET11/20 expression vector containing human or murine MRP8 cDNA was used to transform E. coli BL21(DE3) bacteria. Bacteria were grown at 37°C in 2× YT for 24 h. Afterward, bacteria were harvested, lysed, and the inclusion bodies were prepared. The inclusion bodies pellet was dissolved in 8 M urea buffer, and to establish proper refolding samples were adjusted to pH 2.0–2.5 first by adding hydrochloric acid. After 60 min incubation at room temperature, samples were stepwise dialyzed to get adapted to pH 7.4 for refolding in the presence of 2 mM DTT. After centrifugation (10 min, 60,000 × g, 4°C) to pellet-aggregated material, samples were further dialyzed and applied to anion exchange column and gel filtration chromatography. The MRP8 protein concentrations were determined by UV absorption at 280 nm using a specific absorption coefficient of 1.02 (human MRP8) or 0.40 (murine MRP8) ml/mg⁻¹/cm⁻¹, respectively. Possible endotoxin contaminations were determined by a Limulus amebocyte lysate assay (BioWhitaker, Walkersville, MD) and could either not be detected or were <1 pg LPS/μg MRP8 in the different batches.

Human TNF-α, IL-1β, and IL-6 were analyzed in culture supernatants from human monocytes (1.6 × 10⁶ cells/ml) using commercial ELISA kits (BD Biosciences). Mouse TNF-α and IL-1β were analyzed in culture supernatants of Hoxb8 cells (2.5 × 10⁶ cells/ml) using commercial ELISA kits (mouse TNF-α from BD Biosciences; mouse IL-1β from eBioscience, San Diego, CA).

Total RNA was isolated using the NucleoSpin RNA II kit (Macherey-Nagel, Düren, Germany), and hybridization to Affymetrix (Santa Clara, CA) Human Genome 133 Plus 2.0 Gene Chip arrays was performed as previously described (31). Data were imported into the Partek Genomics Suite v6.6 (Partek, St. Louis, MO) using robust multiarray average prior to batch correction. Robust multiarray average is an algorithm used for background correction, log₂ transformation, and quantile normalization of Affymetrix expression data. A one-way ANOVA test was performed to calculate the 1000 most variable and differentially expressed genes across the different stimulation groups. Differentially expressed genes were defined by a fold change >2.5 or <−2.5 and a false discovery rate (FDR)–corrected p value of <0.05. To visualize the structure within the data, we performed hierarchical clustering (HC) and principal component analysis (PCA) on the 1000 most variable genes, with default settings in the Partek Genomics Suite, on FDR-corrected p values according to the expression values of the samples across the conditions. In addition, we performed co-regulation analysis (CRA) based on Pearson correlation coefficients using BioLayout Express 3D (32), as well as Pearson correlation coefficient matrix (PCCM) analysis, which was plotted as a heat map (33). To identify the differences and similarities between MRP8- and LPS-stimulated conditions, differentially expressed genes were visualized with BioVenn (34) and SigmaPlot version 10.0 (Systat Software, San Jose, CA) as a ratio-ranked (log₂) plot. To determine enhanced or inhibited biological functions of MRP8-stimulated monocytes, we used the 200 most strongly upregulated and downregulated genes to generate and visualize networks based on gene ontology (GO) enrichment analysis (GOEA) using BiNGO, EnrichmentMap, and Word Clouding in Cytoscape (33). Data are provided in Gene Expression Omnibus (GSE56681; http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE56681).

Total RNA was isolated using a NucleoSpin RNA II kit (Macherey-Nagel). One microgram of total RNA was reverse-transcribed to cDNA using RevertAid H Minus transcriptase (Fermentas, St. Leon-Roth, Germany). Primers used are listed in Supplemental Table I.

Quantitative real-time PCR (qRT-PCR) was performed using SYBR Green (PeqLab, Erlangen, Germany), and data were acquired with the CFX 384 system and CFX Manager software version 3.0 (Bio-Rad, Munich, Germany) as previously described (35). Each assay was set up in duplicate and the relative expression was calculated using 2^ΔCT, and the housekeeping genes ribosomal protein L (RPL) and GAPDH were used as controls.

Monocytes were treated as described previously (36). Monocytes prestimulated with MRP8 were treated with 400 nM staurosporine for an additional 6 h or were left untreated. The percentage of apoptotic and necrotic cells was determined by staining with allophycocyanin-conjugated annexin V and propidium iodide and analyzed by flow cytometry using a FACSCalibur flow cytometer (BD Biosciences) and FlowJo software version 7.6.5 (Tree Star, Ashland, OR).

Cells were lysed in high-salt buffer containing a protease inhibitor mixture or a nuclear extraction kit (Epigentek, Farmingdale, NY) was used for cell fractionation, and equal amounts of protein were separated on SDS-polyacrylamide gels and transferred to nitrocellulose membranes. Membranes were probed with primary Abs overnight at 4°C. Primary Abs were detected with HRP-conjugated secondary Abs and developed with ECL.

All values given throughout are expressed as mean ± SEM or mean ± SD from at least three independent experiments. Statistical analyses were performed using t tests and Mann–Whitney U tests. A p value <0.05 was judged to be significant.

We performed a global gene expression analysis to define the cellular response of human monocytes to MRP8 treatment. Monocytes from individual donors were stimulated with MRP8 or left untreated to assess changes in gene expression patterns. Using hierarchical clustering of the most variable genes (p < 0.05) within the dataset, we demonstrate that MRP8 induces a strong transcriptional response (Fig. 1A). This was further corroborated by PCA showing that MRP8-stimulated monocytes were distinct from unstimulated cells (Fig. 1B). We analyzed 22,277 transcripts using Affymetrix U133 arrays and identified 1,512 differentially expressed genes (ANOVA test; fold change >2.5 or <−2.5, FDR-corrected p < 0.05), 516 of which were upregulated and 996 downregulated by MRP8 (Supplemental Table II). Differential regulation of 13 selected genes identified by microarray analysis was validated by quantitative real-time PCR (qRT-PCR; Supplemental Table III). Additionally, we confirmed our findings at the protein level for three secreted cytokines (IL-6, TNF-α, and IL-1β) by ELISA of cell culture supernatants (Fig. 1C). Notably, RNA expression of both potential MRP8 receptors, TLR4 and RAGE, showed no alterations in MRP8-treated monocytes.

To elucidate the biological processes induced by MRP8, we performed GOEA based on the 200 most differentially expressed genes followed by network visualization of enriched GO terms using BiNGO and EnrichmentMap (Fig. 1D). As expected, the overwhelming majority of genes were related to immune cell activation, including categories such as NF-κB signaling, chemotaxis, cell migration, and inflammatory response, as well as leukocyte activation and signal transduction. However, we also identified novel aspects of MRP8 biology, such as the induction of genes involved in apoptotic processes in leukocytes (Fig. 1D). In fact, when deconvoluting the genes from the GO terms, it was found that a large number of antiapoptotic genes were induced in response to MRP8 stimulation (Supplemental Table III). To address whether the MRP8-induced effects on cell survival identified in silico were indeed functionally relevant, we confirmed the upregulation of antiapoptotic genes by qRT-PCR (Supplemental Table III) and demonstrate that expression of antiapoptotic proteins cIAP-1 and Bcl-x_L were significantly upregulated in monocytes after treatment with MRP8 compared with unstimulated cells. We also observed decreased activation of caspase-3 during staurosporine-induced apoptosis after MRP8 stimulation (Fig. 1E). We quantified apoptosis rates by staining for annexin V and in addition necrosis with propidium iodide (Fig. 1F). MRP8-stimulated monocytes displayed higher spontaneous survival rates compared with untreated cells, and they were also significantly protected from staurosporine-induced apoptosis as indicated by GOEA. These antiapoptotic effects are known for TLR agonists as LPS, and the MRP8 stimulation induces similar effects (Fig. 1E, 1F).

To rule out endotoxin contamination in protein preparations, we strictly screened all protein batches for potential contamination by LPS (9). Our recombinant-prepared protein was shown to be entirely free of any contaminating E. coli proteins (Fig. 2A). Endotoxin was not detectable in our protein preparations by the Limulus amebocyte assay (detection limit 1 pg/μg protein). This indicates that the maximal possible endotoxin contamination of MRP8 protein preparations is <5 pg LPS in the experiments, which alone did not induce any cytokine induction (data not shown). Furthermore, MRP8-induced effects could not be inhibited by addition of polymyxin B (50 μg/ml) at concentrations that efficiently blocked the action of 1 ng/ml LPS. Finally, heat-inactivated MRP8 failed to stimulate monocytes, whereas LPS was still fully active at these conditions (Fig. 2B).

Which of the receptors present on human monocytes is responsible for MRP8-mediated effects during inflammation is still a matter for debate. To define the relevant MRP8 receptor in human monocytes, we compared the genome-wide transcriptomes of MRP8-stimulated monocytes with human monocytes activated by LPS, TNF-α, or IL-1β. Following previously developed approaches (33), we performed several bioinformatic analyses, including CRA (Fig. 3A), PCA (Fig. 3B), HC (Fig. 3C), and PCCM analyses (Fig. 3D), to determine the relationship of MRP8-induced transcriptional reprogramming with the other stimuli. CRA revealed a clear group-based network structure with control samples separated from stimulated samples. Moreover, the MRP8-stimulated samples were most closely related to LPS stimulation, whereas TNF- and IL-1β–stimulated samples were placed in a subcluster (Fig. 3A). These findings were corroborated by PCA, again showing that MRP8 stimulation is most closely related to LPS stimulation (Fig. 3B). Using the most variable genes within the dataset, we found using HC that most gene clusters were similarly regulated by MRP8 and LPS (Fig. 3C). Finally, PCCM analysis also confirmed that MRP8 stimulation was most closely related to LPS stimulation of monocytes (Fig. 3D). We further visualized the groups of genes differentially expressed after MRP8 or LPS stimulation using a Venn diagram (Fig. 3E) and a ratio-ranked (log₂) plot (Fig. 3F). Most genes differentially regulated by LPS were also regulated by MRP8 (purple area), although MRP8 induced an even stronger response (red area; Fig. 3E). Using the ratio-ranked (log₂) plot, we confirmed that the large number of genes significantly induced or repressed by MRP8 stimulation that were not significantly altered in response to LPS (red dots) did, however, show the same regulation trends. This provides further support for the notion that MRP8 and LPS trigger a very similar transcriptional reprogramming of monocytes. Among the genes most significantly induced were IL-6, CCL20, and IL-1α, all of which are well known to be induced by LPS via TLR4 signaling. The similarity of gene expression patterns, together with the induction of common cytokines in MPR8- and LPS-stimulated human monocytes, has provided us with evidence that MRP8 is a specific ligand of human TLR4.

To further confirm the role of TLR4 in MRP8-mediated stimulation of human monocytes, we analyzed pathway members downstream of TLR4 signaling. Several members of the MAPK family are activated by phosphorylation during TLR4 signaling. Accordingly, we detected activation of MEK, ERK1/2, and p38 in MRP8-stimulated monocytes (Fig. 4A). We also observed an induction of IκBα degradation (Fig. 4A), resulting in NF-κB activation and translocation (Fig. 4B). Furthermore, qRT-PCR demonstrated that inflammatory cytokines such as TNF-α, IL-6, and IL-8, which are known to be induced via MyD88- and IκBα/NF-κB–dependent pathways, are upregulated after MRP8 treatment. This is in addition to cytokines such as CD80, IFN-γ–inducible protein 10 (IP-10), and monokine induced by IFN-γ (MIG), which are involved in the MyD88-independent, Toll/IL-1R domain–containing adapter inducing IFN-β/IFN regulatory factor 3–dependent TLR4 signaling pathway (Fig. 4C). Overall, this clearly demonstrates that both known TLR4 signaling pathways are activated by MRP8 in a similar manner to LPS.

Because we could not detect substantial amounts of RAGE expression on the human monocytes by flow cytometry and Western blot analyses (data not shown), we used HEK293 cells stably transfected with TLR4/CD14/MD2 or RAGE as well as TLR2 as controls for specificity to verify that TLR4 is the primary receptor for human MRP8. HEK293^{TLR4/CD14/MD2} cells showed a strong induction of proinflammatory cytokines, including TNF-α, IL-8, and MCP-1, on stimulation with MRP8 and LPS as positive controls as demonstrated by qRT-PCR, which is in accordance with results derived from human monocytes. In contrast, HEK293^RAGE, HEK293^TLR2, and HEK293^MOCK cells showed no response, neither to MRP8 nor to LPS stimulation. HEK293^TLR2 cells responded well to lipoprotein stimulation, providing a positive control for TLR2 activation (Fig. 5A, 5B). As a positive control for RAGE-dependent signaling we used the HEK293^RAGE/TLR9 model recently described (37). Although RAGE nicely increased CpG2006-induced TLR9 response, MRP8 neither antagonized nor amplified this effect. Of note, MRP8 alone did not activate RAGE in this system (Fig. 5C).

To prove that signaling pathways are identical in the human and murine systems, we stimulated estrogen-regulated Hoxb8 progenitor cell–derived macrophages from WT, TLR4^−/−, RAGE^−/−, and MyD88^−/− mice on day 5 of macrophage differentiation. RAGE^−/− cells responded to MRP8 stimulation in a manner similar to WT controls, as demonstrated by induction of TNF-α, IL-1β, and IL-6 shown by qRT-PCR at the RNA level (Fig. 5D) and TNF-α and IL-1β at the protein level (Fig. 5E). In contrast, removal of TLR4 or the downstream signaling adaptor molecule MyD88 abolished activation by MRP8.

Moreover, blocking the TLR4 receptor with the mAb HTA125 or TLR4 antagonist Rhodobacter sphaeroides LPS in human monocytes confirmed our findings. Both agents efficiently inhibited the MRP8- as well as the LPS-induced secretion of TNF-α (Fig. 5F). These results confirm TLR4 as the dominant receptor of MRP8-mediated activation of monocytes and macrophages.

DAMPs are a heterogeneous group of proteins that are released upon cell stress and tissue damage and activate the innate and adaptive immune systems. Recent evidence has indicated that uncontrolled release of DAMPs is pivotal in many inflammatory and autoimmune diseases (1). However, the underlying molecular mechanisms mediating DAMP-induced inflammation are currently not well defined. The contribution of various DAMP receptors such as TLR4 and RAGE has also been a controversial topic of discussion (17). Especially in the case of MRP8 and MRP14, numerous studies have described various contributions of both receptors to several different aspects of inflammation. Both proteins belong to the DAMP family, and they promote inflammatory processes during infections, allergies, and autoimmune diseases (1). However, the initial events during protein binding to its receptor are not yet clear. Several studies have demonstrated interactions of MRP8/MRP14 with RAGE, focusing mainly on signaling pathways involving activation of MAPK p38 or the transcription factor NF-κB, including cellular mechanisms downstream of the MRP8/MRP14–RAGE interaction. However, no structural data regarding the ligand-receptor recognition sites have been described so far (21, 22, 24, 38). Similar MRP8/MRP14–RAGE interactions were observed in a study done by Boyd et al. (23) on purified cardiomyocytes, during which RAGE was coimmunoprecipitated with MRP8/MRP14. Binding of MRP8 and/or MRP14 to both TLR4 (9, 19) and RAGE (18, 19) has been demonstrated in BIAcore studies, and strong binding constants in the low nanomolar range could be calculated for both receptors. Although MRP8 has been unequivocally demonstrated to be an active component for phagocyte stimulation in vitro (9), the biologically relevant complex forms of MRP8/MRP14 in vivo, however, are still to be defined. Furthermore, it is not clear whether MRP8/MRP14 interacts directly with certain amino acids within the RAGE receptor or to carboxylated glycans covalently attached to RAGE (16, 19). For isolated MRP14, similar findings were found. Although it could be shown that MRP14 binds to both receptors TLR4 as well as RAGE (19), cytokine induction could only be observed for MRP14-TLR4–dependent activation in a cell line culturing model (39). Studies on phagocytes from TLR4 mutant mice clearly indicate that TLR4 is mainly responsible for MRP8-dependent activation in these cells (9). Further evidence of this TLR4 dependency comes from two studies that show that in MRP14-deficient mice, or when TLR4 is blocked with TAK-242 (a small molecule TLR4 inhibitor), the cellular response to MRP8/MRP14 is almost completely abrogated (12, 26). In one of these studies, blocking of RAGE or carboxylated glycans by specific Abs did not change the MRP14-mediated expression of matrix metalloproteinases, type II collagen, or cytokines (26).

There are also no consistent data suggesting that either RAGE or TLR4 is the dominant receptor during MRP8/MRP14-mediated tumor genesis (18, 20, 40). MRP8/MRP14 promotes tumor growth by inducing so-called myeloid-derived suppressor cells, which in turn suppress adaptive immunity and thereby facilitate tumor growth. However, the relevance of either RAGE or TLR4 to any of these studies has not been investigated so far (40, 41).

Whether murine data are always transferable to the human system has been controversial. Recent systematic comparisons of inflammatory processes between mice and humans have suggested that inflammatory responses in human diseases correlate only poorly with the corresponding mouse models (27). To better understand the underlying molecular mechanisms of inflammation in humans, we have in the present study taken a systems biology approach utilizing the transcriptome of phagocytes to generate MRP8-receptor interaction models. We focused on MRP8 in this study because this molecule has been unequivocally identified as a biologically active component of MRP8/MRP14 complexes (9). We have used microarray technology to define the MRP8-induced transcriptome in human monocytes and determined a dominant proinflammatory response to MRP8 at the genome-wide level in human monocytes. Linking our data to prior knowledge by GOEA combined with GO term network construction corroborated the strong inflammatory signal induced by MRP8. Additionally, this approach has revealed an antiapoptotic role for MRP8 in monocytes as indicated by the network visualization of GOEA. We could experimentally confirm that MRP8 modulates the mitochondria-dependent apoptotic pathway in monocytes. Expression of genes involved in this antiapoptotic mechanism was significantly upregulated at the mRNA and protein levels. MRP8-stimulated monocytes were likewise significantly protected from staurosporine-induced apoptosis. These findings are in contrast to the proapoptotic effect of MRP8/MRP14 described in endothelial cells and several tumor cell lines (25, 42). The molecular pathways responsible for these cell-specific differences are, however, not yet clear.

To elucidate whether TLR4 is the main receptor for MRP8 on human monocytes, we have performed a comparative genomic analysis, again utilizing the transcriptome to determine overlaps or differences between various stimuli in comparison with MRP8. We postulated that transcriptional changes induced via TLR4 should closely mimic those induced by LPS, whereas downstream effector cytokines such as IL-1β or TNF-α should induce alternative responses (Fig. 3). Utilizing a set of previously developed techniques, including CRA, PCA, HC, and PCCM, we have demonstrated on a global level that MRP8 stimulation is most closely related to LPS stimulation, strongly indicating that TLR4 is a prime candidate for the MRP8 receptor. Single-transcript analysis using ratio plots further corroborated these findings. Our data did show some quantitative differences between MRP8 and LPS stimulation, which are most likely explained by different dosing effects. Owing to the low expression level of RAGE in human monocytes, we cannot completely rule out the possibility of an additional MRP8/RAGE axis, which might be relevant in other cell types. However, using a HEK293 cell line–based system where RAGE was overexpressed to a high level, we neither could detect direct nor costimulatory or antagonistic effects of MRP8 on these cells, indicating that RAGE has an inferior relevance for MRP- mediated effects.

To experimentally validate this computationally derived model indicating a role for TLR4 in MRP8 stimulation in human monocytes, we analyzed downstream elements of TLR4 signaling in MRP8-treated monocytes. We observed that expression of MyD88-dependent genes such as TNF-α, IL-6, and IL-8 as well as MyD88-independent genes such as CD80, IP-10, and MIG is induced by MRP8, indicating that both TLR4 signaling pathways are activated on MRP8 stimulation. Analysis of HEK293 cells stably transfected with either TLR4/CD14/MD2 or RAGE confirmed signaling of MRP8 through TLR4, and blocking of this receptor in human monocytes abolishes TNF-α secretion. Finally, we confirmed that these signaling pathways are identical in humans and mice. Collectively, these data unequivocally confirm TLR4 as the primary receptor for MRP8 in human monocytes under inflammatory conditions as previously described (9).

Because the MRP8/MRP14 heterodimer is one of the most significantly upregulated molecules in many inflammatory diseases (7, 14), the mechanism described in the present study is likely to be highly clinically relevant. Targeted deletion of these genes in mice results in a significant phenotype under inflammatory conditions such as autoimmune diseases, arthritis, infections, LPS-induced shock, and allergies (12, 43). Expression of these proteins in humans correlates very well with disease activity, especially in predicting the response to therapy for arthritis and inflammatory bowel disease, as well as the risk of flares in remitting–relapsing disease courses (7, 8, 11). These molecules may therefore be attractive targets for novel therapeutic approaches involving the prevention of the binding of these molecules to their receptor. Taken together, our results clearly indicate that data obtained in murine inflammatory models can be transferred to the human system in the case of MRP8-dependent signaling in monocytes.

We thank H. Berheide, H. Hater, D. Lagemann, and U. Nordhues for excellent technical support.

The authors have no financial conflicts of interest.