Abrupt Expression of TLR4 in TLR4-Deficient Macrophages Imposes a Selective Disadvantage: Genetic Evidence for TLR4-Dependent Responses to Endogenous, Nonmicrobial Stimuli¹

Lipopolysaccharide is an integral structural component of the outer membrane of Gram-negative bacteria and one of the most potent microbial initiators of inflammation. Picogram per milliliter concentrations can stimulate macrophages to secrete a battery of inflammatory products, including TNF, IL-1, IL-6, and NO. A serum transferase, LPS- binding protein, transfers LPS to a GPI-linked cell surface or secreted protein, CD14 (1, 2, 3). The presence of LPS is signaled through the TLR4-MD-2 complex in macrophages (4, 5, 6, 7, 8, 9).

Endogenous ligands have also been reported to activate macrophages through TLRs, including heat shock proteins (10, 11), heparan sulfate (12), fibronectin (13), fibrinogen (14), β-defensins (15), surfactant protein-A, (16), fragments of hyaluronic acid (17), high mobility group box 1 (HMGB1)³ protein (18), and mRNA (19). In addition, nonstimulated bone marrow-derived macrophages cultured in vitro in the apparent absence of microbial products have a basal level of cytokine expression dependent upon MyD88, a protein that transduces TLR4 signals (20).

In this study, we restored TLR4 expression to TLR4-deficient macrophages, initially with the goal of structure-function analysis. As parental cells for transfection, two TLR4^null macrophage cell lines were created from the bone marrow macrophages of C57BL/10ScNCr and TLR4 knockout mice. TLR4 was stably transfected into these cell lines and individual cell clones were isolated and examined. Unexpectedly, abrupt restoration of LPS responsiveness was only possible when TLR4-expressing, stably transfected cells were selected in media conditioned by wild-type macrophages. In the absence of the conditioned media, restoration of TLR4 surface expression but not LPS response was only possible in the context of an incidental mutation in MyD88 that precluded TLR-dependent signaling. These findings strongly suggest that transfected TLR4 exerted a signal, despite the strict absence of microbial products, and that when this signal could be propagated, it was disadvantageous to clonal expansion.

Stably transfected macrophage cell lines and the phenotypes of their parental cells are tabulated in Tables I and II.

C57BL/10ScNCr mice were purchased from the National Institutes of Health. C57BL/10ScSn mice were from The Jackson Laboratory. Forebears of TLR4ko mice and the wild-type littermates mice were a gift from K. Takeda and S. Akira (Osaka University, Osaka, Japan) (6). Bone marrow-derived transformed macrophage cell lines were derived as described (21, 22). Macrophage cell lines and primary macrophages were cultured in DMEM with 4.5 g/L glucose (Invitrogen Life Technologies) supplemented with 10% heat inactivated FBS (endotoxin free; HyClone), 2 mM l-glutamine, 1 mM pyruvate, 200 U/ml penicillin, and 200 μg/ml streptomycin in 7.5% CO₂/92.5% air at 37°C. Primary bone marrow-derived cells were further supplemented with 20% L929 cell conditioned media and cultured for 7 days to induce differentiation into macrophages. All media were monitored for LPS contamination by the chromogenic Limulus amebocyte lysate assay (BioWhittaker) and found to contain <25 pg LPS/ml.

Cells (7.5 × 10⁶ cells) were mixed with 10 μg of plasmid DNA and electroporated at 300 volts and 975 ohms resistance in a sterile 0.4-cm cuvette. Cells were resuspended in complete media and plated in 10-cm dishes. After 48 h, medium was replaced with fresh medium containing 350 μg/ml G418 (Calbiochem). After 13–14 days, for the first transfection, or 8–9 days, for subsequent transfections, individual clones were selected either through limiting dilution or using filter discs. The clones were maintained in the presence of 350 μg/ml G418.

The cDNA for mouse TLR4 was generated through RT-PCR for the entire open reading frame of TLR4 using primers 5′-ATGATGCCTCCCTGGCTCCTGGCT-3′ (forward) and 5′-TCAGGTCCAAGTTGCCGTTTCTTGT-3′ (reverse) from mouse macrophage RNA. The product was subcloned into the expression vectors pIRES-neo3 (BD Clontech) and pTriEx-neo3 (EMD Biosciences) and their sequence verified. Endotoxin-free plasmid DNA for transfection was isolated using the Endofree Maxiprep kit from Qiagen. Endotoxin in plasmid preparations was below 10 pg/ml, the limit of detection.

Genomic DNA was isolated using the DNAeasy kit from Qiagen according to manufacturer’s protocol. Ten micrograms of DNA was digested overnight at 37°C with HindIII (New England Biolabs), separated on 0.8% agarose gel and transferred overnight to GeneScreen membranes (NEN), which were hybridized with ³²P-labeled oligonucleotide probes (Prime-a-Gene kit; Promega), from exon 1 or exon 5 of MyD88.

RNA was isolated from cells using Tri-Reagent (Molecular Research Center, Cincinnati, OH). RT-PCR was performed using the Gene-Amp RNA PCR Core kit from Applied Biosystems. RNA was reverse-transcribed using a random hexamer primer. This product was amplified with Platinum Taq polymerase (Invitrogen Life Technologies) using forward primer (bp 1079–1096) 5′-GAACAAAGGGTCTATCAG-3′ and reverse primer (bp 1991–2009) 5′-GCTTTCTCCTCTGCTGTAC-3′ to determine TLR4 transcription in the clones. To establish that the TLR4 construct was stably integrated into the genome of the clones, PCR amplification of TLR4 genomic DNA was done using forward primer (bp 1386–1405) 5′-GGTATATTTCTTGGCTTGAC-3′ and reverse primer (bp 1991–2009) 5′-GCTTTCTCCTCTGCTGTAC-3′.

The SMART RACE kit (BD Clontech) was used to determine the sequence downstream of exon 3 of MyD88. After the first strand of cDNA was synthesized from the RNA using a poly dT primer, a PCR amplified the cDNA using a MyD88-specific primer and a universal downstream primer provided in the kit. The sequence of forward primer from exon 3 of MyD88 is 5′-ACAAACGCCGGAACTTTTCGATGCC-3′. The amplified cDNA was sequenced.

Ten million resting T4Cr cells and the wild-type control ScSn cells were lysed with TRIzol (Invitrogen Life Technologies) and total RNA was isolated. After treatment with DNase I (Ambion) and purification (RNAeasy; Qiagen), 2–3 μg of RNA was used to synthesize double-stranded cDNA without second round amplification. Biotin-labeled and fragmented cRNA was then generated in vitro and hybridized to the Affymetrix whole mouse genome expression array 430 2.0 at 45°C overnight. Arrays were washed, stained with streptavidin-PE (Molecular Probes), and scanned by gene scanner 3000 according to the Affymetrix protocol.

Primary image was generated using GeneChip Operating Software 1.1 (GCOS; Affymetrix) to extract the signal with the target intensity value for each chip scaled to 250. All parameters, including background value, noise, percent gene present, scaling factor, internal controls such as β-actin and GAPDH, sensitivity control (bio B, C, D, and cre), and the 3′/5′ ratios of internal controls met the requirements, indicating high quality transcribed cDNAs and completeness of in vitro transcription reactions. Data normalization, log transformation, fold change, and statistical analysis were obtained with GeneSpring software version 7.0 (Agilent Technologies). Duplicate arrays were performed for each of two independent experiments. Statistical comparison between the T4Cr and WtSn was conducted using Welch t test with log transformed data. The p-value cutoff was 0.05.

Ampli-Taq polymerase was used for the detection of PCR-amplified nucleic acids. Probes were synthesized by Biosearch Technologies and labeled with the reporter dye FAM at the 5′ end and Black Hole Quencher at the 3′ end. A total of 100 ng of RNA was transcribed into cDNA with oligo dT primer in 20 μl using 50 U of murine leukemia virus reverse transcriptase (PerkinElmer). PCR was performed in a volume of 15 μl on the ABI PRISM 7900HT sequence detection system (PerkinElmer). For analysis of TGFβi, the sequence of the probe is 5′-CCTGGTTAGCGGAGGCATCGGG-3′, the sequence of the forward primer is 5′-TGAAGTACCACATTGGTGATGAAA-3′, and the reverse primer is 5′-TTCAGCCGCACCAGGG-3′.

Cells were lysed in a Triton X-100 lysis buffer (50 mM Tris (pH 7.4), 150 mM NaCl, 2 mM MgCl₂, 2 mM CaCl₂, and 1% Triton X-100 with protease inhibitors). Equal amounts of protein (50 μg) were fractionated on a 10% SDS-PAGE gel, transferred to nitrocellulose membranes (Schleicher & Schuell) and immunoblotted with Abs against MyD88.

Cells were plated in 96-well plates at 5 × 10⁴ cells/well in 100 μl of complete medium and treated with stimuli at the indicated concentrations. Nitrite was assayed by mixing 48-h conditioned medium with an equal volume of Griess reagent (1% sulfanilamide, 0.1% naphthyl-ethylenediamine dihydrochloride, 2.5% H₃PO₄). Absorbance at 550 nm was recorded. Nitrite content of similarly incubated cell-free medium was subtracted as background. Mouse TNF and TGFβ contents in culture supernatants were measured by ELISA (R&D Systems) according to the manufacturer’s instructions.

Cells (10⁶) were washed with PBS and preincubated with 2.4G2 anti-FcR Ab (BD Pharmingen) to eliminate background FcR binding. Cells were incubated with anti-TLR4 or anti-FLAG Ab conjugated to PE for detection on a FACScan flow cytometer (BD Biosciences).

Triacylated peptide Pam3Cys-Ser-(Lys)4 (pam3Cys)-SKKKK was purchased from EMC Microcollections. The Ab against a MyD88 peptide was from Chemicon International. The Ab against the complex of TLR4 and MD2 was TLR4/MD2-PE clone MTS510 from eBioscience. All other reagents including LPS derived from Escherichia coli (0111:B4) were from Sigma-Aldrich.

E. coli LPS was repurified according to Manthey et al. (23). Briefly, LPS was resuspended in water with 0.2% triethylamine and 0.5% sodium deoxycholate at 5 mg/ml and phenol-extracted twice to remove any remaining protein. The aqueous phases containing LPS were collected and adjusted to 75% ethanol and 30 mM sodium acetate to pellet the purified LPS. Recovery was estimated at 80%. Operational definition of purity was demonstrated by the ability of purified LPS to stimulate TLR2-deficient macrophages but not TLR4-deficient macrophages, as judged by secretion of TNF and nitrite.

A bone marrow-derived macrophage cell line named T4Cr from TLR4-deficient C57BL/10ScNCr mice and a cell line named WTSn from wild-type C57BL/10ScSn mice (Table I) were established by expression of the viral genes, v-myc and v-raf (21, 24).

WTSn cells and T4Cr cells were rounder in culture than primary bone marrow macrophages, but the cell lines looked similar to each other (Fig. 1,A). Both cell lines expressed F4/80, FcγR (CD16/CD32), CD11b, and CD18 as did their parental primary macrophages (Fig. 1,B). WTSn responded to LPS by producing TNF at a level comparable to that from bone marrow-derived primary macrophages from C57BL/10ScSn mice (Fig. 1,C). The T4Cr cells produced no detectable TNF in response to LPS, like the C57BL/10ScCr bone marrow-derived macrophages from which they were derived (Fig. 1 C).

As expected, the macrophages and cell lines from both wild-type ScSn mice and TLR4-deficient ScCr mice responded to other microbial stimuli: peptidoglycan, lipoteichoic acid (LTA), hypomethylated bacterial DNA (CpG) and a synthetic TLR2 ligand pam3Cys (Fig. 1 D). Thus, the cell lines resemble primary macrophages in surface marker expression as well as their ability to respond to LPS and other stimuli.

As is typical with macrophages, transient transfection of full-length TLR4 could be accomplished only with very low efficiency (data not shown). Therefore, we turned to a stable transfection system (25). Briefly, we created two different expression vectors, pIRESneo3 and pTriEx3-neo, both encoding TLR4 cDNA and a neomycin-resistance selection marker. Endotoxin-free expression plasmids were electroporated into T4Cr, the macrophage cell line derived from TLR4^null C57BL/10ScNCr mice. Two days after transfection, the selection reagent was added to the culture. After 14 days, surviving cells underwent limiting dilution to establish individual clones, which were expanded. A total of 48 clones in a set named T4Cr-TLR4-I were selected, including 24 clones each using either of two vectors, the pIRESneo3 vector in a subset termed “P” and the pTriEx3-neo vector in a subset termed “Q” (Table II). Control cell lines, termed T4Cr-V, were generated similarly with the empty vectors.

All of the 48 clones tested, derived with both P and Q vectors, expressed the TLR4/MD-2 complex on the cell surface, and at a level similar to wild-type macrophages, while the untransfected T4Cr cell line was negative (Fig. 2 A). Because MD-2 was not cotransfected into the T4Cr cells, exogenous TLR4 must have associated with endogenous MD-2, and the availability of endogenous MD-2 may have confined the expression of TLR4 to a physiologic level.

WTSn cells, derived from wild-type C57BL/10ScSn mice, responded to LPS stimulation by secreting TNF and NO. However, none of the 20 T4Cr-TLR4-I clones that were tested responded to LPS by secreting TNF or nitrite (Fig. 2,B). Moreover, none of the T4Cr-TLR4-I clones produced nitrite or TNF in response to pam3Cys, peptidoglycan, or CpG bacterial DNA, although the parental T4Cr cells did so (Fig. 2 C). This result indicated that a common member of different TLR signaling pathways might be affected.

Indeed, MyD88 expression was not detected in 15 T4Cr-TLR4-I clones tested with an Ab against a peptide from exon 4 of MyD88, while WTSn and T4Cr cells and T4Cr cells transfected with an empty vector expressed MyD88 normally (Fig. 3 A).

Southern blots using a probe from exon 1 from MyD88 revealed an altered sized band in the T4Cr-TLR4-I clones, compared with WTSn cells, T4Cr cells, or T4Cr cells transfected with an empty vector (Fig. 3,B, top). A probe from exon 5 of MyD88 did not hybridize to DNA from the T4Cr-TLR4-I clones in the same membrane (Fig. 3 B, bottom).

RT-PCR on individual exons of MyD88 showed that exons 1, 2, and 3 were expressed normally, but there was no product for exon 4 or 5 in any of the 10 T4Cr-TLR4-I clones tested (data not shown). To identify the mutation in MyD88, a 3′ RACE (amplification of cDNA ends) was performed using a primer from the 5′ end of exon 3 of MyD88. We found that T4Cr-TLR4-I clones contained a deletion encompassing 4 bp of exon3, the entirety of exons 4 and 5 and 19 bp of the 3′UTR (Fig. 3,C). The deletion removed one HinDIII site, accounting for the different sized Southern blot band (Fig. 3 B). We presume that the deletion caused a truncated MyD88 protein to be produced, with a dominant-negative effect on signaling through TLR4 and other MyD88-dependent receptors. Thus, we had to consider two possibilities. Either the transfection procedure was consistently mutagenic to MyD88 in precisely the same site in multiple clones, or more likely, it selected for a pre-existent, spontaneously occurring MyD88 mutant in the TLR4-deficient parental population. The latter explanation would require a corollary, that the MyD88 mutant clone acquired an enormous selective advantage over all the other members of the parental cell population when the population was suddenly required to express TLR4. To test the first idea, we produced another round of TLR4 transfectants.

New clones, termed T4Cr-TLR4-II, were established in a second round of transfection with either of the TLR4-containing vectors, P or Q, in TLR4-deficient T4Cr cells. Individual clones were selected 8–9 days after addition of G418 to preserve clonal variation. None of 81 clones was positive for TLR4 surface staining, in contrast to the T4Cr-TLR4-I clones (Fig. 4 A).

Genomic PCR revealed that the TLR4 vector was integrated into the genome of most of the T4Cr-TLR4-II clones tested (Fig. 4,B, left panel), consistent with a resistant phenotype to G418 during their selection. However, RT-PCR analysis of 14 clones showed no TLR4 mRNA, although expression of a transfected mutant TLR4 was normal in other T4Cr cells (Fig. 4,B, right panel). Thus, wild-type TLR4 was integrated but the gene was silenced. In contrast to the T4Cr-TLR4-I clones, Western blots demonstrated that all of the T4Cr-TLR4-II clones tested expressed full-length MyD88 (data not shown). Consistent with their failure to express TLR4, all T4Cr-TLR4-II clones tested failed to respond to LPS, as judged by production of TNF and nitrite (Fig. 4,C). However, the T4Cr-TLR4-II clones were able to respond to TLR2 ligands, pam3Cys and zymosan, secreting nitrite and TNF at levels similar to those of parental T4Cr cells (Fig. 4 D), albeit with clonal variations.

Expression of TLR4 truncated to remove either its intracellular domain or extracellular domain, could be achieved in T4Cr cells (Fig. 4 E). These cells, as expected, did not regain responsiveness to LPS (not shown). Thus, in TLR4-deficient T4Cr cells, expression of nonfunctional TLR4, including mutated or truncated TLR4, was possible, but expression of functional TLR4 was not achieved.

To rule out the possibility that an unknown mutation in the C57BL/10ScCr background interfered with expression of TLR4, a bone marrow macrophage-derived cell line was created from TLR4 knockout mice on the C57BL/6 background, termed T4ko, and the corresponding wild-type cell line from C57BL/6 mice, termed WTB6. Characterization of these cell lines was similar to that described for T4Cr, as documented in Fig. 5.

Forty-seven T4ko-TLR4 clones, generated with TLR4 containing vectors P or Q, were isolated and stained for TLR4/MD2 surface expression. None of the clones tested showed any positive surface staining for TLR4 (Fig. 6,A). Genomic PCR confirmed integration of TLR4 DNA in most of the T4ko-TLR4 clones tested (Fig. 6,B, left panel). However, none of the clones produced any detectable TLR4 transcripts (Fig. 6,B, right panel). Full-length MyD88 protein expression was normal in all T4ko-TLR4-I clones tested by Western blots (data not shown). None of the T4ko-TLR4 clones tested responded to LPS (Fig. 6,C), but they did respond to other TLR ligands such as pam3Cys and LTA (Fig. 6 D).

To test whether stable TLR4-expressing clones could be generated from wild-type macrophages, RAW cells were electroporated with pTriEx3-TLR4 and individual clones selected in the presence of G418. Expression of the exogenous TLR4 in some of these clones was detected (Fig. 6,E). The surface expression of TLR4 of these cells, however, did not differ from that of their parental cells, nor did they respond to LPS differently (Fig. 6 E). This may reflect the constraint imposed by the physiologic level of MD2.

Restoration of TLR4 to TLR4^null macrophages appeared to be detrimental to clonal selection. However, expression of exogenous TLR4 in matched wild-type cell lines exerted no detectable adverse influence. To explain this paradox, we considered the following hypothesis. Cells expressing TLR4 from the onset of differentiation to macrophages may coexpress factors that tend to damp potentially adverse signals from TLR4, such as those reported to lead to apoptosis when TLR4 is strongly engaged (18, 26, 27, 28, 29, 30, 31, 32). Cells that differentiate into macrophages without TLR4 may not express such putative counterregulatory mechanisms. When TLR4 is abruptly expressed in the latter cells, even a low level of signal delivered in response to endogenous ligands may lead to an exaggerated response, blocking clonal expansion. To test this hypothesis, we used microarray analysis of global gene expression to compare WTSn and T4Cr cells, both in the resting state (that is, with no exogenous stimuli).

One gene stood out as being differentially expressed and possibly functionally relevant, termed TGFβ induced (TGFβi), also known as β ig-h3 (BIGH3) and keratoepithelin (33). TGFβi transcripts were detected on the array with three different probe sets and were 110-, 90-, and 84-fold more abundant in WTSn cells than T4Cr cells in microarray analysis (Table III). Quantitative real-time RT-PCR revealed a 10-fold difference in the expression of TGFβi between WTSn and T4Cr cells (Fig. 7 A).

As TGFβ induces expression of TGFβi (33), we suspected that the difference in TGFβi hinted at a deficiency of TGFβ expression in TLR4-deficient macrophages. Because TGFβ is largely regulated at a posttranscriptional level, such a difference could be inapparent on the microarray. TGFβ is a ubiquitously expressed cytokine with profound anti-inflammatory effects (34, 35) and an ability to deactivate macrophages (36). TGFβ can inhibit ligand induced cytokine production through TLR2, TLR4, and TLR5 by promoting degradation of MyD88 (37).

Indeed, a TGFβ1 ELISA on culture supernatants from resting macrophage cell lines revealed that the level of secreted TGFβ1 was >10-fold higher in WTSn cells than T4Cr cells (Fig. 7 A). These levels refer to the total of latent and active TGFβ, because active TGFβ was below the limit of detection. Differences in active TGFβ, which may have been more pronounced, are reported to be functionally relevant at concentrations of 1–10 pg/ml (36), below the level of sensitivity of the ELISA.

To test whether the relative deficiency in TGFβ or other host factors contributed to the inability of TLR4-deficient macrophages to express TLR4, a new set of T4Cr clones, termed T4Cr-TLR4-III, was generated in media alone (control) or in the presence of 10 ng/ml exogenous TGFβ1 or conditioned media (mixed 1:1 with fresh media) from wild-type macrophages. Clones from each condition (52 altogether) were assayed for expression of TLR4 on the cell surface. T4Cr clones that were prepared in the presence of exogenous TGFβ1 or conditioned media from wild-type macrophages showed a small increase in TLR4 surface expression, but not to the level of wild type (Fig. 7,B). The clones raised in the presence of TGFβ1 did not acquire LPS responsiveness, but 13 of 15 T4Cr clones tested that expressed TLR4 as a result of selection in macrophage-conditioned medium did secrete TNF in response to LPS (Fig. 7 C). These results suggested that secreted factors from wild-type macrophages play an important role in the regulation of endogenous ligand signaling through TLR4 and are required for reconstitution of TLR4 in TLR4-deficient macrophages with intact downstream signaling pathways.

During analysis of 196 stably transfected clones selected in the absence of secreted factors from wild-type cells, we found that forced expression of full-length TLR4 in TLR4^null macrophages either selected for silencing of TLR4 expression or for outgrowth of cells with a mutation in MyD88 that rendered the TLR signaling pathway inactive. It is possible that the oncogenes used to generate the TLR4^null cell lines might interact with functional TLR4 and induce a TLR4-dependent signal. However, wild-type cell lines generated with the same oncogenes did not display activation of TLR4. It is more likely that in macrophages that normally express TLR4, a regulatory mechanism may be in place to suppress excessive activation. We hypothesized that without such a regulatory mechanism, overstimulation could occur, leading to a selective disadvantage in macrophages that are proliferating. We used microarray analysis to search for differences in transcription between wild-type and TLR4-deficient macrophages in the absence of exogenous stimuli. Transcription of TGFβi was at a lower level in TLR4-deficient macrophages. A lack of TGFβi could be due to a deficiency in TGFβ secretion, and TGFβ deficiency was confirmed by ELISA. New clones created in the presence of TGFβ1 displayed a small increase in the level of TLR4 on the cell surface. This partial restoration of TLR4 expression by TGFβ1 did not lead to a recovery of LPS responsiveness, perhaps due to selection of clones with other defects in the TLR4 signaling pathway. However, cells treated with conditioned media from wild-type macrophages did regain LPS responsiveness, albeit to a lesser degree than wild-type macrophages. Thus, our study suggests that TLR4 expression in signaling competent, TLR4-deficient macrophages is potentially harmful to cell survival or proliferation.

In addition to the multitude of ligands from foreign, pathogenic sources, endogenous ligands for TLRs have been proposed, including three heat shock proteins, hsp60, hsp70, and gp96 (10, 38, 39, 40), fibrinogen, (14), fibronectin (13, 41), β-defensins (15), surfactant protein-A (16), heparan sulfate (12), fragments of hyaluronic acid (17), HMGB1 (18), and mRNA (19). The heat shock proteins and HMGB1 appear to be endogenous ligands for TLR2 and TLR4, mRNA signals through TLR3, and all the other endogenous ligands listed signal through TLR4.

It is clear that TLR signaling may induce cell death. Yersinia activation of macrophages through TLR2 and TLR4 induced apoptosis (26, 27, 28). The Mycobacterium tuberculosis glycolipoprotein p19 induced apoptosis in a TLR2-dependent manner (29). Mycoplasma signaled proapoptotic pathways via TLR2 and TLR6 (30, 31). Bacillus anthracis induced TLR4-dependent apoptosis (42). However, it is not known whether endogenous ligands for TLRs can induce cell death.

In this study, we showed that expression of TLR4 in a TLR4-deficient background conferred a survival or proliferative disadvantage in the absence of regulation. In our system, microbial contaminants were scrupulously monitored and avoided. The genetic evidence presented here for an adverse effect of functional TLR4 expression in a previously TLR4^null background therefore argues that endogenous ligands may be activating the TLR4 pathway in our cell system. The precise nature of the adverse effect is unknown, but it appeared to prevent completely the expression of normal levels of TLR4 except in a signal-incompetent background. TLR4 restitution in a null background has been accomplished in mice by transgenesis (43, 44). We speculate that the putative regulatory factors, including TGFβ, were supplied to macrophages in trans in vivo, in contrast to our system in vitro, in which the starting population contained pure populations of a macrophage cell line.

This study provided interesting biologic insights, but we did not succeed in performing the work we set out to accomplish. Indeed, the analysis presented here may help account for the gap in current work that we were trying to fill: the lack of biochemical studies of truncated, point-mutant, and chimeric forms of TLR4 expressed at physiologic levels in cells that normally express it, whose endogenous TLR4 has been genetically deleted or disrupted. What we have discovered suggests that such studies will be feasible by selecting stably transfected, TLR4-deficient macrophages in the presence of conditioned medium from wild-type macrophages.

We thank S. K. Lee and N. Chan for excellent technical help, and K. Takeda and S. Akira for TLR4 knockout mice.

The authors have no financial conflict of interest.