PU.1 and interferon consensus sequence-binding protein regulate the myeloid expression of the human Toll-like receptor 4 gene.

The protein product of the Toll-like receptor (TLR) 4 gene has been implicated in the signal transduction events induced by lipopolysaccharide (LPS). In mice, destructive mutations of Tlr4 impede the normal response to LPS and cause a high susceptibility to Gram-negative infection. Expression of TLR4 mRNA in humans is restricted to a small number of cell types, including LPS-responsive myeloid cells, B-cells, and endothelial cells. To investigate the molecular basis for TLR4 expression in cells of myeloid origin, we cloned the human TLR4 gene and analyzed its putative 5'-proximal promoter. In transient transfections a region of only 75 base pairs upstream of the major transcription initiation site was sufficient to induce maximal luciferase activity in THP-1 cells. The sequence of this region is similar in human and mouse TLR4 genes and lacks a TATA box, typical Sp1-sites or CCAAT box sequences. Instead, it contains consensus-binding sites for Ets family transcription factors, octamer-binding factors, and a composite interferon response factor/Ets motif. The activity of the promoter in macrophages was strictly dependent on the integrity of both half sites of the composite interferon response factor/Ets motif, which was constitutively bound by the myeloid and B-cell-specific transcription factor PU.1 and interferon consensus sequence-binding protein. These results indicate that the two tissue-restricted transcription factors PU.1 and interferon consensus sequence-binding protein participate in the basal regulation of human TLR4 in myeloid cells. Cloning of the human TLR4 gene provides a basis for further investigation of the possible impact of genetic variations on the susceptibility to infection and sepsis.

teria, is an important function of innate immune cells. Detection of LPS is exquisitely sensitive. Minute quantities can activate mammalian immune cells (in particular, monocytes and tissue macrophages) that leads to rapid secretion of cytokines (e.g. tumor necrosis factor, IL-1, IL-6, and IL-10) and enhanced microbicidal capacities. Failure to contain Gramnegative infection and to control pro-inflammatory cytokine secretion can result in septic shock and death (1)(2)(3).
The interaction of LPS with monocytes and neutrophils is known to occur via CD14 (4) and is strongly enhanced by LPS-binding protein (5,6). Multiple signal transduction pathways are activated in macrophages upon LPS stimulation, including the activation of transcription factor NF-B, nonreceptor tyrosine kinases, protein kinase C, and several members of the mitogen-activated protein kinase family (7). CD14, however, is a glycosyl-phosphatidylinositol-anchored membrane protein that is unable to transduce the LPS signal through the membrane.
The prototypic Toll protein of Drosophila controls dorsalventral patterning in the embryo and is required for antifungal responses in adult flies (8). A plasma membrane receptor, Toll has a leucine-rich ectodomain (9,10) and a cytoplasmic domain with sequence homology to several mammalian proteins, including both chains of the IL-1 receptor (11), the IL-18 receptor (12), SIGRR (13), MyD88 (14), and all members of the newly described Toll-like receptor (TLR) family (15)(16)(17). At least two of these proteins (IL-1R and TLR4) are known to signal via activation of MyD88, IRAK, TRAF6, and NF-B (18 -22).
Positional cloning analysis recently revealed that mice of the C3H/HeJ and C57BL/10ScCr strains, which are profoundly unresponsive to LPS (23,24), have separate genetic defects of Tlr4 (25). This finding suggests that TLR4 is, in fact, the transducing subunit of the LPS receptor. Direct measurements of TLR4 activity in macrophages sustain the conclusion, in that modest overexpression of a mutant form of Tlr4, corresponding to that encoded by the codominant Tlr4 Lps-d allele present in C3H/HeJ mice, causes a dramatic increase in the LPS EC 50 for tumor necrosis factor production. Overexpression of the native protein depresses the LPS EC 50 . 2 Interestingly, TLR4 appears to exist in association with MD-2, an exteriorized protein expressed by monocytic cells (27). It may be that co-expression is required for the LPS transducing activity of TLR4.
Historically, a number of mysteries have beset the field of LPS signal transduction. Among these, there has been no molecular explanation for the large interspecific differences in LPS response (28), suppression of LPS response by corticoste-roids (29), augmentation of the response by various bacterial infections (30 -33) or by interferons (34 -39), and tolerance to LPS elicited by pretreatment with LPS itself (40). TLR4 has thus become a focus of inquiry aimed at understanding each of these phenomena.
The transcriptional regulation of TLR4 is a starting point for such an inquiry. Expression of human TLR4 is restricted to a small number of cell types, including endothelial cells, B-cells, and predominantly myeloid cells (monocytes, macrophages, dendritic cells, and granulocytes) (15,16,21,41). Both the basal level of TLR4 expression and its regulation in myeloid cells may influence responses to LPS and, hence, Gram-negative infection. Accordingly, we have sought to define the TLR4 promoter and to analyze those factors that govern TLR4 gene expression in human macrophages, which are the main LPSresponsive cell type. This analysis has entailed determination of the complete sequence of TLR4 and flanking genomic DNA.

EXPERIMENTAL PROCEDURES
Chemicals-All chemical reagents used were purchased from Sigma-Aldrich unless otherwise noted. Protease inhibitors are from Roche Molecular Biochemicals. Oligonucleotides were synthesized by TIB Molbiol (Berlin, Germany). Antisera for supershift analyses were purchased from Santa Cruz.
Cells-Peripheral blood mononuclear cells were separated by leukapheresis of healthy donors, followed by density gradient centrifugation over Ficoll/Hypaque. Monocytes were isolated from mononuclear cells by countercurrent centrifugal elutriation in a J6M-E centrifuge (Beckman, Mü nchen, Germany) as described previously (42). Monocytes were Ͼ90% pure as determined by morphology and expression of CD14 antigen. Isolated monocytes were cultured in RPMI 1640 medium (Biochrom KG, Berlin, Germany) supplemented with vitamins, antibiotics, pyruvate, nonessential amino acids (all from Life Technologies, Inc.), 5 ϫ 10 Ϫ8 M ␤-mercaptoethanol, 2% human pooled AB-group serum on Teflon foils for a period of 7 days. The human monocytic cell line THP-1 was grown in RPMI 1640 medium supplemented with 10% fetal calf serum (Life Technologies, Inc.). The human cervical carcinoma cell line HeLa and the human melanoma cell line Mel Im (a gift from A. Bosserhof) were maintained in Dulbecco's modified Eagle's medium plus 10% fetal calf serum. Where indicated, cells were treated with human recombinant IFN-␥ (200 units/ml; Roche Molecular Biochemicals).
Gene Cloning-The human TLR4 gene was identified by hybridization screening in BAC 110p15 (Genome Systems human BAC library). The mouse Tlr4 gene, as previously reported (43), resides in BAC 152c16 (Research Genetics mouse BAC library, 129 strain). Both BACs were fragmented by ultrasonic disruption, repaired using the Klenow fragment of DNA polymerase I, and shotgun cloned into the vector pBluescript(KS). Bidirectional sequencing was carried out on clones picked at random from the library. To increase the density of reads over the gene itself, both BACs were amplified using long range PCR to yield a 12-kb fragment containing all of the human exons and a 16-kb fragment containing all of the mouse exons. These fragments were also shotgun cloned and sequenced.
Assembly was accomplished using the programs phred and Phrap, and contiguous sequences were visualized using the program consed_ alpha. All of these programs, produced by the University of Washington Genome Center, were run under a UNIX operating system, on a DECalpha computer. A 4.3-kb genomic fragment of the hTLR promoter was amplified from the BAC clone containing the human TLR4 gene using the Expand High Fidelity PCR system (Roche Biochemicals) and the primers 5Ј-CTC CAT GGC ACA TTC TGC AGT AAA CTT GGA GGC-3Ј (sense) and 5Ј-CAC GCA GGA GAG GAA GGC CAT GGC TG-3Ј (antisense).
RNA Preparation and RT-PCR Analysis-Total RNA was isolated from different cell types by the guanidine thiocyanate/acid phenol method (44). 2.0 g of total RNA from either cell type was reverse transcribed using oligo(dT) primer and Superscript II (Life Technologies, Inc.). Primer positions and sizes for the amplified fragments of TLR4 and ␤-actin are indicated in Fig. 1. PCR conditions were optimized to assure that the amplification was still exponential at the indicated cycle numbers. The amplified products were sequenced to confirm their identity.
Primer Extension Analysis-Poly(A) RNA was isolated from THP-1 cells using poly(A)Tract System 1000 (Promega). Determination of the transcription initiation site of human TLR4 was performed as described by Calzone et al. (45). Briefly, 1 and 4 g of poly(A) RNA were annealed to the 32 P end-labeled oligonucleotide 5Ј-GGT GTC TTC TCT TCC TCG AGC-3Ј at 58°C for 20 min and then cooled for 10 min at room temperature. The reverse transcription was performed with AMV RT (Promega) at 41°C for 30 min and then stopped by addition of 8 l of loading buffer. Extended products were separated on a 6% polyacrylamide/8 M urea sequencing gel along with sequencing products of the corresponding region of the TLR4 gene obtained with the above oligonucleotide Plasmid Construction and Purification-PCR fragments were inserted into the plasmid vector pCR2.1-TOPO (TOPO Cloning Kit, Invitrogen) for sequencing and subcloning purposes. The 4.3-kb genomic PCR fragment of the hTLR4 proximal promoter was subcloned into the NcoI restriction site of pGL3-B (Promega) and sequenced. Deletions of this constructs were generated by digestion with either KpnI, ApaI, HindIII, EcoRI, PstI, or XhoI and subsequent religation of the remaining plasmid. Mutations of the PU.1, OCT, and IRF-binding sites were done by PCR-mediated mutagenesis. PCR fragments with correctly introduced mutations were subcloned back into the TLR4 Ϫ75 plasmid. Transient DNA Transfections-THP-1 cells were transfected using DEAE-dextran. 5 ϫ 10 5 THP-1 cells/ml were seeded into tissue culture flasks the day before transfection. On the next day, 6 ml of cell suspension were washed twice with STBS (46) and pelleted. 0.2 g of reporter plasmid and 0.02 g of Renilla luciferase control vector were mixed with DEAE-dextran (400 g/ml) in 140 l of STBS buffer and immediately added to the pelleted THP-1 cells. The cells were incubated at 37°C for 20 min, washed twice with STBS, resuspended, and cultured in complete RPMI medium. The two cell lines HeLa and Mel Im were transfected using LipofectAMINE reagent (Life Technologies, Inc.) according to the manufacturers instructions. Briefly, 4 ϫ 10 5 HeLa cells were transfected using 7 l of LipofectAMINE and 1 g of total DNA (including 0.05 g of Renilla luciferase control vector) and incubated for 5 h before the addition of complete medium. 2.7 ϫ 10 5 Mel Im cells were transfected using 5 l of LipofectAMINE and 1 g of total DNA (including 0.05 g of Renilla control vector) and incubated for 6 h before the addition of complete medium. The transfection mixture was removed after 24 h and substituted with complete medium. The transfected cell lines were cultivated for 48 h and harvested, and cell lysates were assayed for firefly and Renilla luciferase activity using the Dual Luciferase Reporter Assay System (Promega) on a Lumat LB9501 (Berthold). Firefly luciferase activity of individual transfections was normalized against Renilla luciferase activity.
Nuclear Extracts and Electrophoretic Mobility Shift Assay-Nuclear extracts were prepared with a variation of the method of Osborn et al. (47). All buffers used contained 1 mM Na 3 VO 4 and a mixture of protease inhibitors (2 g/ml aprotinin, 0.5 g/ml leupeptin, 1 g/ml pepstatin, 20 g/ml bestatin, 5 g/ml E46, 50 g/ml antipain, 100 g/ml chymostatin). Oligonucleotides were end-labeled with [␥-32 P]ATP using T4 polynucleotide kinase. The binding reaction contained 2.5 g of nuclear extract protein or 0.5-2 l of the in vitro translation reaction, 0.5 g of poly(dI-dC), 20 mM HEPES, pH 7.9, 60 mM KCl, 1 mM dithiothreitol, 1 mM EDTA, pH 8.0, 5% glycerol, and 20 nmol of probe DNA in a final volume of 10 l. Antisera used in supershift analyses were added after 15 min, and samples were loaded onto polyacrylamide gels after standing at room temperature for a total of 30 min. Buffers and running conditions used have been described (48). Gels were fixed in 5% acetic acid, dried, and autoradiographed.

Cloning and Characterization of the Human TLR4 Gene-A
BAC containing the human TLR4 gene was isolated by hybridization screening, shotgun cloned and sequenced as described under "Experimental Procedures." The mouse Tlr4 gene was similarly sequenced from a previously described BAC (43). A sequence 19 kb in length containing the human gene and a sequence 91.7 kb in length containing the mouse gene have been deposited in GenBank TM (accession numbers AF177765 and AF177767, respectively). These sequences are considered to be 99.99% accurate.
Comparison of the two published cDNA sequences (accession numbers NM_003266 and U93091) for human TLR4 revealed the presence of an additional 120-bp region in the sequence reported by Rock et al. (16). A corresponding additional exon, which is not present in the mouse gene, was identified in the human TLR4 genomic sequence (Fig. 1). The presence of this exon in TLR4 transcripts introduces an in-frame stop codon, which would theoretically terminate TLR4 translation after 34 amino acids. To clarify whether this transcript is present in macrophages, we performed RT-PCR with specific primers flanking the region between exons I-IV of the human TLR4 gene. As shown in Fig. 1B, three products were amplified from macrophage and THP-1 cDNA that represent three different splicing variants of human TLR4. Sequencing of the purified fragments confirmed the specificity and nature of the three splicing forms, which is indicated in Fig. 1B. As outlined in Fig.  1C, only the transcript containing exon I, III, and IV yields the proper TLR4 protein including a putative signal peptide. The amino acid sequence as proposed by Rock et al. (16) is proceeded by an in-frame stop codon and does not contain a typical leader sequence (Fig. 1C). It is therefore unlikely that the combination of exons I, II, III, and IV would yield a functional TLR protein.
To determine the transcriptional start site of the TLR4 gene we performed primer extension analysis. Two specific extension products were obtained from THP-1 RNA (Fig. 2). The major start site resides 190 bp upstream of the adenine residue of the first start codon. A second, minor initiation site was detected 74 bp upstream of the major start site.
Myeloid-specific Activity of the Proximal Human TLR4 Promoter-Published tissue Northern analyses suggest a predominant expression of hTLR4 in spleen and peripheral blood leukocytes (15,16). RT-PCR with freshly separated and in vitro cultivated human blood cells indicated that human TLR4 is mainly expressed in myeloid cells, including monocytes, in vitro differentiated macrophages, monocyte-derived dendritic cells, granulocytes, and, to a much lesser extent, mixed lymphocytes (data not shown).
To further analyze the myeloid expression of TLR4, we cloned fragments of the 5Ј-proximal promoter region of the human TLR4 gene, ranging from 4.3 kb to 100 bp upstream of the ATG start codon, into a luciferase reporter plasmid (Fig. 3). Transient transfection analysis was performed in the monocytic cell line THP-1 and two nonmyeloid cell lines, Mel Im (melanoma) and HeLa (cervical carcinoma). Luciferase activities were normalized for transfection efficiency by co-transfection with a Renilla luciferase construct, and results for individual cell lines were compared relative to the activity of the promoter-less pGL3-basic construct. As shown in Fig. 3, the presence of both positive and negative elements can be observed. A negative regulatory region seems to reside between nucleotides Ϫ3228 and Ϫ743. Within this region elements may exist that repress the activity of luciferase constructs specifically in THP-1 cells. Reporter constructs including residues above Ϫ385 showed low but significant reporter activity in transfected HeLa cells. However, further deletion localizes a region directing macrophage-specific reporter gene expression to approximately 75 bp proximal to the major transcriptional start site.
The Proximal Promoter Regions of Human and Mouse TLR4 Are Similar and Contain Several Purine-rich Motifs-Sequence comparison of the proximal promoter regions of both mouse and human TLR4 revealed a high degree of conservation. The 5Ј proximal regions of both TLR4 genes are characterized by the absence of TATA boxes, consensus initiator sequences, or GCrich regions found in "housekeeping" genes that normally determine transcriptional initiation (Fig. 4). They also lack Sp1 or CCAAT box sequences and instead contain several purine-rich elements with a 5Ј-GGAA-3Ј core on either strand, a characteristic feature of many myeloid-specific genes (49 -56). The smallest promoter fragment directing maximal myeloid activity (hTLR4 Ϫ75) contains two purine-rich elements that might be bound by members of the Ets family, which includes the myeloid and B-cell-specific transcription factor PU.1. In addition, a highly conserved consensus-binding site for octamer transcription factors was detected in both human and mouse proximal TLR4 promoters.
PU.1 Binds to Both Elements within the Proximal hTLR4 Ϫ75 Promoter-The importance of PU.1 recognition motifs for tissue-restricted expression of myeloid genes has been demonstrated in a growing number of cases (50 -56). To test the ability of PU.1 to bind elements within the human TLR4 promoter, we performed EMSA with double-stranded oligonucleotides corresponding to the two proximal purine-rich sequences. Fig. 5A shows an EMSA with in vitro translated PU.1, which specifically bound to both oligonucleotides. In both cases, complex formation was disrupted by the addition of an excess amount of unlabeled wild type oligonucleotide but not by an oligonucleotide with a mutated 5Ј-GGAA-3Ј core sequence. Both complexes were supershifted by the addition of a PU.1specific antiserum. Similar experiments were done using nuclear extracts of THP-1 cells. As shown in Fig. 5B, the outer purine-rich motif (PU-11) formed a specific quickly migrating complex similar to the complex observed with in vitro translated PU.1. The complex was disrupted by the addition of a PU.1-specific antiserum but remained unchanged in the presence of antisera against two other Ets family members Fli-1 and Elf-1. The specific complex observed for the inner purinerich motif (PU-12) was migrating significantly more slowly compared with the complex observed with in vitro translated PU.1. The complex was also supershifted in the presence of PU.1-specific antiserum but not by antisera against Fli-1 and Elf-1. The different mobility of this complex suggested the presence of at least one additional factor that interacts with PU.1 on this site.
Identification of the PU.1 Interacting Factor at the Inner Purine-rich Motif-In gel shift assays nuclear extracts of various cell types, the slower migrating, PU.1 containing complex was only observed in THP.1 cells and human macrophages. In contrast to THP-1 cells, PU.1 was detectable as both slower and faster migrating band in human macrophages (Fig. 6A). To further investigate the nature of the slower migrating, PU.1 containing complex, we prepared THP-1 nuclear extracts either without protease inhibitors or without the phosphatase inhibitor Na 3 VO 4 and performed EMSA with the PU-12 oligonucleotide. Extract preparation without protease inhibitors yielded a decrease in complex formation. Without the phosphatase inhibitor Na 3 VO 4 , the slower migrating complex disintegrated and a faster migrating complex containing PU.1 alone appeared, suggesting that phosphorylation of either factor or both is necessary for complex formation.
Further sequence analysis revealed a possible binding site for members of the family of IRF located next to the inner purine-rich motif PU-12. Interactions between PU.1 and members of the IRF family have been observed before (57-60). Fig.  7A compares the putative PU.1/IRF-binding sites of the human and mouse TLR4 promoter with so far identified sequences of similar binding sites in the human gp91 phox (58), CD20 (59), murine IL-18 promoters (60), and immunoglobulin light chain enhancers E 2-4 B and E 3Ј (57). We next performed EMSAs with specific antisera against the three IRF family members (ICSBP, interferon-stimulated gene factor 3␥, and PIP) to iden-tify the interacting partner of PU.1 in the context of the hTLR4 promoter. As shown in Fig. 7B, a strong supershift was detected with THP-1 extracts and the ICSBP antiserum. A supershift (albeit less intense) was also observed with the PIP antiserum, suggesting the presence of PIP, which is generally thought to be lymphocyte-specific (57), or a PIP cross-reactive protein in THP-1 nuclear proteins. No supershift was detectable with the interferon-stimulated gene factor 3␥ antiserum. In vitro translated ICSBP was unable to bind the PU-12 element independently but was able to interact with PU.1 to form a complex with slightly decreased mobility to that observed in THP-1 nuclear extracts (Fig. 7C). The complex of in vitro translated PU.1 and ICSBP was supershifted with the ICSBP and the PIP antiserum, indicating that the latter is cross-reactive with human ICSBP protein in gel shift assays. These results suggested that ICSBP and PU.1 are the main components of the slower migrating PU-12 complex. A similar band pattern of ICSBP⅐PU.1 and predominantly Pip⅐PU.1 containing complexes was observed in gel shift experiments performed with nuclear extracts from human B cells, which also express PU.1 and both PIP and ICSBP (data not shown). Sequence comparison revealed a single base pair difference in the IRF-binding site between mouse and human promoter (Fig. 3). Binding of the PU.1/ICSBP complex was also detectable using the corresponding mouse sequence (mPU-12; Table I)
A Conserved Element Adjacent to the Composite ICSBP⅐PU.1 Motif Is a Weak Binding Site for Octamer Transcription Factors-A highly conserved octamer element is present in both human and mouse TLR4 promoters, which prompted us to investigate its interaction with POU domain transcription factors. Direct binding of nuclear proteins to the site was almost undetectable in THP-1 cells and macrophages (data not shown); however, competition studies with macrophage nuclear proteins and the consensus octamer motif (OCT consensus) identified the TLR4 octamer element as a weak binding site for Oct-1 and Oct-2 (Fig. 8). Cold oligonucleotide with a 3-bp mutation of the putative binding site (OCT-M) did not affect the binding of octamer factors to the consensus octamer sequence.
Mutational Analysis of the Identified Binding Sites and Effect of IFN-␥-To analyze the functional significance of the four identified binding sites in reporter assays, we used site-di-rected mutagenesis to abolish each of the four putative binding sequences. The oligonucleotides carrying the mutated binding sites (PU-11M, PU-12M, IRF-M, and OCT-M; for sequences see Table I) did not compete with complex formation of wild type oligonucleotides in gel shift assays (data not shown). In monocytic THP-1 cells the reporter constructs with mutations of either half-side of the PU.1/ICSBP motif showed a markedly reduced activity compared with the wild type promoter (Fig. 9). The effect of the PU.1 mutation was most distinct, and reporter activity was almost completely abolished. In contrast, disruption of the octamer-binding site showed a trend toward a slightly increased activity, although statistical analysis revealed that the difference was not significant. Mutation of the outer PU.1 motif also had no significant impact on the activity of the hTLR4 Ϫ75 promoter (Fig. 9).
To investigate whether the activity of the TLR4 promoter could be induced in a nonmyeloid cell line that does not express either PU.1 or ICSBP, we co-transfected HeLa (cervical carcinoma) cells with the hTLR4 Ϫ75 reporter plasmid and expression plasmids for PU.1 (PU-pECE) and/or ICSBP (hICSBP-pTarget). We only observed a slight up-regulation of relative luciferase activity (1.5-fold) when PU.1 was co-transfected alone (data not shown). Co-transfection of PU.1 and ICSBP expression plasmids had no cooperative effect on the promoter activity in HeLa cells with or without the addition of IFN-␥. Similar results have been reported for the functional ICSBP⅐PU.1-binding motif of the CYBB gene, which encodes for gp91 phox (58). This might indicate that additional factors, which are not present in HeLa cells, are required for the induction of the TLR4 promoter by PU.1/ICSBP or that inhibitory proteins are present that prevent PU.1/ICSBP binding and transactivation.
IFN-␥, a cytokine released by activated T lymphocytes, is a major activator of macrophage microbicidal functions and enhances both sensitivity and magnitude of their response to LPS (36,37,39,40,61). In mouse macrophages, priming with IFN-␥ drastically up-regulates expression of the transcription factor ICSBP (62), which has been implicated in IFN-␥-mediated gene regulation (63)(64)(65). These observations prompted us to investigate a possible effect of IFN-␥ on PU.1⅐ICSBP complex formation in macrophages. Constitutive binding of the ICSBP⅐PU.1 complex observed in untreated cells was only slightly increased during IFN-␥ stimulation for 3, 6, or 18 h. No additional bands were observed upon IFN-␥ treatment, and the ICSBP⅐PU.1 complex was completely supershifted by the ICSBP antiserum (data not shown). Whereas high levels of IRF-1 transcripts were detectable after IFN-␥ treatment, suggesting efficient IFN-␥ priming (66), only a slight induction (approximately 2-fold) of all three TLR4 splicing forms by IFN-␥ treatment was detectable by RT-PCR using total RNA from in vitro differentiated macrophages or THP-1 cells (data not shown). IFN-␥ treatment of THP-1 cells did not affect the intensity of the ICSBP⅐PU.1-DNA complex in gel shift assays, and the relative activities of TLR4 promoter constructs in transient transfections of THP-1 cells were not affected by INF-␥ treatment either 24 or 4 h before cell harvest (data not shown).

DISCUSSION
In this study we investigated the transcriptional regulation of TLR4 in human macrophages, the main LPS-responsive cell type. We cloned and sequenced 19 kb of the human TLR4 gene, sequenced 91.7 kb of the mouse Tlr4 gene, determined the transcriptional start sites, and identified three alternative splicing forms of human TLR4 in human myeloid cells. Furthermore, we performed an initial characterization of regulatory elements controlling the expression of human TLR4 and defined a minimal proximal promoter that confers full reporter activity in human monocytic THP-1 cells.
Gene cloning and sequence comparison revealed the existence of an additional exon (exon II) in the human TLR4 gene, which is not present in the mouse gene. Its integration into TLR4 mRNA introduces an in-frame stop codon, which prompted us to investigate the presence and ration of alternative splicing forms of TLR4. By RT-PCR, we were able to detect three alternative splicing variants of hTLR4; however, only one of them encodes the proper TLR4 protein. Although we did not observe major changes in the relative abundance of the three transcripts in myeloid cells, alternative splicing could represent an additional level of TLR4 regulation in humans. Disruptions of alternative splicing have been correlated with some human genetic diseases (67); mutations that affect the splicing of TLR4 mRNA could favor the expression of noncoding transcripts. It is also possible that alternative splicing events are regulated in response to developmental or physiological cues (67). As a consequence, the expression of TLR4 protein could, for example, be switched off, if the second exon is included in all transcripts. However, further investigations will be needed to clarify these issues.
Expression of human TLR4 transcripts has primarily been detected in myeloid cells (monocytes, macrophages, and dendritic cells) and some B-cell lines (15,16,27). In accordance with the observed cell type restricted expression, the TLR4 gene is controlled by a typical myeloid type promoter in macrophages. The 5Ј-proximal region lacks a TATA box, consensus initiator sequences, or GC-rich regions found in "housekeeping" genes and instead contains multiple purine-rich sequence motifs that are recognized by transcription factors of the Ets family, including PU.1. The transcription factor PU.1 was shown to be required for the optimal expression of a growing number of myeloid-specific genes (49 -56, 68 -70). Our data suggest that PU.1 also plays an important role in the myeloid expression of TLR4. In human macrophages, PU.1 binds an essential motif in the proximal promoter and recruits ICSBP to an adjacent binding site that is also required for the optimal activity of the promoter.
ICSBP is a member of the IRF family of transcription factors and is expressed mainly in cells of hematopoietic origin (71). Its mouse homologue was originally cloned as interferon-␥-regulated protein that bound an interferon-inducible enhancer element of major histocompatibility complex class I genes (71). Because of the phosphorylation of tyrosine residues in vivo, ICSBP does not bind to DNA by itself but has been shown to interact with two other IRF family members (IRF-1 and IRF-2) (72,73). Cooperative DNA binding of ICSBP with one of the other IRF family members results in an increased binding activity for IFN-stimulated response elements and transcriptional repression of genes containing the IFN-stimulated response elements (57,(63)(64)(65)74). In addition to its interaction with members of its own gene family, ICSBP and the related lymphoid-restricted factor Pip (ICSAT/IRF-4) are able to form complexes with the Ets family member PU.1 (57). ICSBP (or Pip) protein bind a composite IRF/Ets motif only in the presence of PU.1, and complex formation requires the phosphorylation of PU.1 at serine 148 (57,75). Composite IRF/Ets-binding sites have been implicated in cell type-specific gene expression in myeloid and B-cells. The CD20 promoter and immunoglobulin light chain enhancers E 2-4 B and E 3Ј contain IRF/Etsbinding sites that seem to be important for their B-cell-specific regulation and are bound by PU.1 and Pip (57,59). An IRF/Ets motif, which is bound by PU.1 and ICSBP, seems important for the myeloid-restricted expression of gp91 phox (58). Kim et al. (60) recently described an ICSBP-binding site that was critical for the activity of the murine 5Ј-flanking IL-18 promoter. Although ICSBP usually does not bind DNA alone, the authors did not investigate the presence of additional factors in the observed ICSBP-DNA complex. The binding sequence of the IL-18 promoter is similar to other IRF/Ets motifs (Fig. 7A); it is therefore likely that Ets factors (most likely PU.1) also participate in the regulation of the 5Ј-flanking IL-18 promoter.
We identified a similar site in the human TLR4 promoter and showed that both PU.1 and ICSBP bind to adjacent elements that are indispensable for the full activity of the TLR4 promoter. As expected, formation of the ternary complex is dependent on the presence of PU.1 and its state of phosphorylation. These data suggest that the interaction between the two tissue-specific transcription factors PU.1 and ICSBP is impor-tant for the basal activity of the TLR4 promoter in human myeloid cells. Preliminary gel shift analyses using B-cell extracts indicate the presence of a similar complex. 3 The composite IRF/Ets motif may therefore also contribute to the expression of TLR4 in human B-cells. Using in vitro translated proteins, we observed that the complex between PU.1 and ICSBP migrates slightly faster than the native THP-1 complex. This difference could be due to altered post-translational modifications or an effect of the (5-10-fold) higher total protein concentrations in the preparations of in vitro translated proteins. However, it is also possible that the native THP-1 complex contains another, as yet unidentified protein.
We identified constitutive nuclear ICSBP protein in the human monocytic cell line THP-1 and in vitro differentiated human macrophages, which is consistent with the previously described constitutive expression of ICSBP in the human premonocytic cell line U937. This indicates a more constitutive expression in human macrophages rather than a strictly inducible expression pattern, which is described for mouse macrophages, where constitutive expression of ICSBP is undetectable, and both mRNA and protein levels are strongly induced after stimulation with IFN-␥ (62). If at all, ICSBP and the IRF/Ets motif in the TLR4 promoter only weakly confer IFN-␥ responsiveness. The ICSBP⅐PU.1 complex was slightly induced by IFN-␥ in human macrophages, and in both THP-1 cells and human macrophages, TLR4 mRNA expression was slightly enhanced by IFN-␥ priming. However, IFN-␥ stimulation did not significantly affect the activity of TLR4 promoter constructs in THP-1 cells. It remains to be clarified whether the observed increase in TLR4 message is regulated on the level of gene transcription or whether other mechanisms (e.g. mRNA stability) mediate the IFN-␥ induced increase in TLR4 mRNA.
ICSBP-deficient mice show altered antiviral and antibacterial responses and develop a chronic myelogenous leukemialike syndrome (76). An observed failure to develop Th-1-driven immune responses has been correlated with a defect in IL-12 p40 production by cells of myeloid origin (26,77). The response of ICSBP-deficient mice to LPS is described as normal (77). However, it is not clear whether the murine Tlr4 gene is similarly dependent on the IRF/Ets motif, which seems to form a less stable complex with ICSBP and PU.1. Analysis of the mouse promoter will clarify whether ICSBP plays a role in the transcriptional regulation of Tlr4 in mice.
In conclusion, our observations suggest that a functional cooperation between PU.1 and ICSBP regulates the myeloid 3 M. Rehli, unpublished observation. FIG. 9. Effect of site-directed mutations on the hTLR4 ؊75 promoter activity in myeloid cells. Site-directed mutations, were introduced in the hTLR4 Ϫ75 promoter construct using oligonucleotides listed in Table I. A schematic representation of the reporter gene constructs is shown. Four putative binding sites are indicated as boxes, P 1 and P 2 (outer and inner Ets motif), O (octamer motif), and I (IFR factor-binding site). Mutations are indicated by crosses. Each site-directed mutant was transiently transfected into the myeloid cell line THP-1 as described under "Experimental Procedures." Luciferase activity is relative to the wild type hTLR4 Ϫ75 promoter (100%), and values are the means ϩ S.D. obtained from at least four independent experiments. expression of TLR4 in humans. Additional elements and mechanisms are likely to contribute to the regulation of TLR4 in health and disease and are the subject to further studies. Transcriptional regulation, alternative splicing events, or function of TLR4 protein could be altered because of mutations or polymorphisms of the TLR4 gene. This study provides the basis for further investigating a possible correlation of yet unknown genetic variations with a higher susceptibility for infection and septic complications.