Chlamydia muridarum Infection Elicits a Beta Interferon Response in Murine Oviduct Epithelial Cells Dependent on Interferon Regulatory Factor 3 and TRIF

ABSTRACT Chlamydia trachomatis is the most common sexually transmitted bacterial infection in the United States. Utilizing cloned murine oviduct epithelial cell lines, we previously identified Toll-like receptor 2 (TLR2) as the principal epithelial pattern recognition receptor (PRR) for infection-triggered release of the acute inflammatory cytokines interleukin-6 and granulocyte-macrophage colony-stimulating factor. The infected oviduct epithelial cell lines also secreted the immunomodulatory cytokine beta interferon (IFN-β) in a largely MyD88-independent manner. Although TLR3 was the only IFN-β production-capable TLR expressed by the oviduct cell lines, we were not able to determine whether TLR3 was responsible for IFN-β production because the epithelial cells were unresponsive to the TLR3 ligand poly(I-C), and small interfering RNA (siRNA) techniques were ineffective at knocking down TLR3 expression. To further investigate the potential role of TLR3 in the infected epithelial cell secretion of IFN-β, we examined the roles of its downstream signaling molecules TRIF and IFN regulatory factor 3 (IRF-3) using a dominant-negative TRIF molecule and siRNA specific for TRIF and IRF-3. Antagonism of either IRF-3 or TRIF signaling significantly decreased IFN-β production. These data implicate TLR3, or an unknown PRR utilizing TRIF, as the source of IFN-β production by Chlamydia-infected oviduct epithelial cells.

Chlamydia trachomatis is a gram-negative obligate-intracellular bacterium. Serovars D to K, referred to as the urogenital serovars, cause genital tract infections including urethritis, salpingitis, epididymitis, prostatitis, and pelvic inflammatory disease. In women, Fallopian tube scarring associated with the infection causes infertility and ectopic pregnancies. In 2004 929,462 C. trachomatis infections were reported to the Centers for Disease Control and Prevention, a 5.9% increase compared to 2003 (6). Ongoing surveillance of 16-to 24-year-old women entering the U.S. National Job Training Program shows recent stabilization of C. trachomatis prevalence at roughly 10% after previous declines (22). Recently, the rates of C. trachomatis infection in Canada have begun to rebound in spite of, or possibly because of, public health measures taken to control the infection (4).
Urogenital serovars of C. trachomatis replicate predominantly in reproductive tract epithelial cells. Chlamydia muridarum is a rodent pathogen closely related to the C. trachomatis urogenital serovars. We derived murine oviduct epithelial cell lines to study host-pathogen interactions within the reproductive tract epithelium. Murine oviduct epithelial cells lines infected by C. muridarum secrete a plethora of inflammatory cytokines (24). We previously identified Toll-like receptor 2 (TLR2) as the major principal epithelial pattern recognition receptor (PRR) responsible for the epithelial secretion of interleukin-6 (IL-6) and granulocyte-macrophage colony-stimu-lating factor (8). These initial studies did not identify the epithelial PRR responsible for secretion of beta interferon (IFN-␤), an immunomodulatory cytokine that plays an important role in innate and adaptive immunity (33).
TLRs capable of triggering IFN-␤ production include TLR3 (48,64,66), TLR4 (49,65), TLR5 in a TLR4-dependent manner (44), and TLR7 to TLR9 (59). Members of the caspase activation and recruitment domain (CARD) family, including RIG-I (68) and MDA5 (3), also trigger IFN-␤ production (23). A putative cytosol-resident DNA-activated PRR that triggers IFN-␤ production has recently been reported (21,56). Nod1 and Nod2, members of the leucine-rich repeat protein family, signal through RIP2 to activate NF-B (20). Nod1 has recently been shown to contribute to the activation of IL-6 transcription in murine embryonic fibroblasts infected with C. muridarum (61); however, Nod1 and Nod 2 are not associated with IFN-␤ production (58). We have previously shown that murine oviduct epithelial cell lines do not express TLR4 and TLR7 to TLR9 by as determined by reverse transcription-PCR (RT-PCR) analysis and by measuring the responses to specific TLR agonist stimulation (8). By process of elimination, TLR3, RIG-I, MDA5, and the unidentified PRR recognizing cytosolic DNA were the remaining candidate epithelial PRRs for triggering IFN-␤ secretion in infected oviduct epithelial cells. The expression of RIG-I and MDA5 has not been investigated in oviduct epithelial cell lines. Murine oviduct epithelial cells expressed TLR3 and upregulated TLR3 mRNA with infection but did not secrete IFN-␤ in response to externally delivered poly(I-C), a potent TLR3 agonist (8), presumably because TLR3 was not on the cell surface.
In myeloid cells and human pulmonary epithelial cells, TLR3 is an endosomal PRR (10,12,40,47). In human fibro-blasts TLR3 has been reported to be expressed on the cell surface (41). It is not known where TLR3 localizes in murine reproductive tract epithelial cells, but intracellular localization in pulmonary epithelial cells and the lack of a murine oviduct epithelial response to extracellular poly(I-C) supports an endosomal localization. Known ligands for TLR3 include viral double-stranded RNA (dsRNA) (2) and possibly cellular RNA (27).
CARD family members RIG-I and MDA5 are resident in the cytosol and likely to play a role in host defense against some RNA viruses (3,68). Like TLR3, RIG-I and MDA5 recognize dsRNA (25,60,68). RIG-I and MDA5 signal through an adaptor variously known as IPS-1, MAV, VISA, and Cardif (reviewed in reference 23). IPS-1 serves a TRIFlike role for RIG-I and MDA5, connecting these PRRs to the IKKε/TBK1 activation pathways for IRF3 and to the TRAF6/ IKK␣␤ pathways for the activation of NF-B. Most studies of RIG-I and MDA5 signaling support a TRIF-independent signaling pathway for IFN-␤ production (28,29,43,53,55), although one study did not (63). Mice deficient in IPS-1 are markedly deficient in their IFN-␤ response to RNA viruses or transfected poly(I-C), thereby identifying IPS-1 as the sole adaptor for RIG-I and MDA5 (30). The putative cytosol-resident DNA PRR is also IPS-1 dependent (21).
For our purposes, investigation of the IRF3 and TRIF signaling pathways would potentially identify the PRR responsible for IFN-␤ production. RIG-I, MDA5, the putative cytosolic DNA PRR, and TLR3 are IRF3 dependent for IFN-␤ production. Of these candidate PRRs, only TLR3-mediated production of IFN-␤ would be dependent on the TRIF adaptor. TRIF dependence would implicate TLR3, whereas TRIF independence would implicate RIG-I/MDA5/cytosolic DNA PRR as the PRR responsible for IFN-␤ production by C. muridaruminfected oviduct epithelial cells.

MATERIALS AND METHODS
Reagents. Poly(I-C), product number P-0913, was purchased from Sigma Chemical Co. (St. Louis, MO).
Mycoplasma-free C. muridarum, previously known as C. trachomatis strain MoPn, was grown in McCoy cells (American Type Culture Collection). The titers of mycoplasma-free C. muridarum stocks were determined on McCoy cells with centrifugation as previously described (8).
Transfections. To generate the dominant-negative TRIF-expressing clones (DN)TRIF1 and (DN)TRIF2 or the pcDNA3.1D control cell line, 75% confluent Bm1.11 cells in six-well plates were transfected with 5 g of either the pcDNA-TRIF-TIR plasmid or the pcDNA3.1D His parent vector using Lipofectamine 2000 reagent (Invitrogen). The DNA-Lipofectamine 2000 complexes were incubated for 30 min in serum-free Dulbecco modified Eagle medium at room temperature, and the complexes were added to the Bm1.11 monolayers, followed by incubation 5 h in a 37°C CO 2 incubator. After 5 h of incubation, the transfection medium was replaced with fresh epithelial cell medium and incubated an additional 30 h at 37°C. The dominant-negative TRIF and empty vector control cells were selected in epithelial media supplemented with 800 g of G418/ml. The (DN)MyD88 and (DN)TRIF cell lines were maintained in epithelial-cell media supplemented with 400 g of G418/ml. Transient transfections were performed with 10 g of either the pcDNA-TRIF-TIR plasmid or the pcDNA3.1D-His control vector in Lipofectamine 2000.
Poly(I-C) treatment. Poly(I-C) was added directly to the medium of Bm1.11, (DN)TRIF1, and (DN)TRIF2 cells and the pcDNA3.1D control cells at 50, 75, and 100 g/ml depending on the experiment. Supernatants were assayed for poly(I-C)-induced IFN-␤ responses by enzyme-linked immunosorbent assay (ELISA) 24 h after exposure. Transfections of poly(I-C) were done by complexing either 50 or 75 g/ml to Lipofectamine 2000 reagent in serum-free media as previously described (36). After a 30-min incubation at room temperature, the poly(I-C) complexes were added to Bm1.11, (DN)TRIF1, and (DN)TRIF2 cells and the pcDNA3.1D control cell lines for 5 h as described above. After 5 h, the transfection medium was removed and replaced with fresh epithelial cell media. IFN-␤ secreted into the medium after 24 h was measured by ELISA.
Infections. Bm1.11, (DN)MyD88, (DN)TRIF1, and (DN)TRIF2 cell lines were plated in 48-well tissue culture plates and were used when confluent (10 5 cells/well). For all experiments, the cells were infected with 10 inclusion forming units (IFU) of C. muridarum per cell in 900 l of culture medium. The 48-well plates were centrifuged at 1,000 rpm (300 ϫ g) for 1 h and then incubated at 37°C in a 5% CO 2 humidified incubator without subsequent change of medium for 6 to 30 h, depending upon the assay. Mock-infected wells received an equivalent volume of sucrose-phosphate-glutamic acid buffer lacking C. muridarum.
For experiments with the endosomal acidification inhibitor bafilomycin A (Sigma), epithelial cells were infected with 10 IFU per cell with 1 h of centrifugation followed by 2 h of incubation at 37°C prior to the addition of bafilomycin in the indicated concentrations. The bafilomycin A stock solution was prepared in dimethyl sulfoxide (DMSO). Comparator "untreated" wells received an equivalent amount of the DMSO vehicle.
Western blotting. Control and (DN)TRIF expressing Bm1.11 cells were grown in monolayers in a six-well plate to confluence. After removal of the growth medium, the monolayers were gently washed with phosphate-buffered saline, and cytosolic proteins were recovered in the cell fractionation buffer provided in the PARIS kit (Ambion, Austin, TX). Cytosolic proteins were quantified by using the Micro-BCA protein assay kit (Pierce, Rockford, IL). Then, 25 g of lysate from either the (DN)TRIF1, (DN)TRIF2, or nontransfected Bm1.11 cell line was boiled in 5ϫ Immunopure reducing sample buffer (Pierce) prior to sodium dodecyl sulfate-polyacrylamide gel electrophoresis. After separation by sodium dodecyl sulfate-10% polyacrylamide gel electrophoresis, the proteins were transferred to Immobilon-P (Millipore, Bedford, MA) transfer membranes. Transfer membranes were blocked in 5% nonfat dry milk, and subsequent immunoblotting was performed by using a 1:5,000 dilution of the His tag-specific rabbit polyclonal antibody SC-804 (Santa Cruz Biotechnology, Santa Cruz, CA) with a 1:10,000 dilution of horseradish peroxidase-conjugated goat anti-rabbit polyclonal antibody (Amersham Biosciences, Piscataway, NJ). Proteins were visualized via chemiluminescence using the ECL plus Western blotting detection system (Amersham) as described in the manufacturer's protocol.
RNA interference. MyD88-specific and IRF3-specific small interfering RNA (siRNA), MyD88-and IRF3-specific primers, scrambled control siRNA, siRNA transfection reagent, and siRNA transfection medium were all purchased from Santa Cruz Biotechnology as proprietary reagents. Custom-designed TRIF-specific siRNA and primers were purchased from Ambion (Table 1). siRNA transfections targeting endogenous MyD88, TRIF, IRF3, and scrambled controls were carried out in 48-well plates seeded with 4 ϫ 10 4 Bm1.11 cells per well using the manufacturer's protocol. The siRNA cocktail was mixed with the transfection reagent mix and allowed to form liposome-siRNA complexes for 30 min at room temperature before being added to the cells. After an initial 5 h of incubation, the transfection medium was replaced with fresh epithelial medium, and the cells

RT-PCR.
Total RNA was isolated from dominant-negative cell lines and the Bm1.11 oviduct epithelial control cells by using RNeasy minicolumns (QIAGEN, Valencia, CA). During purification, all RNA samples were treated with RNasefree DNase I (QIAGEN) to remove genomic-DNA contamination. The RNA was quantified by spectrophotometric analysis. For TLR analyses, RNA integrity was confirmed by agarose gel electrophoresis. Optimized primer pairs were designed by using the Vector NTI Suite (Infomax, Inc., Frederick, MD). The specific primer pairs (Amitof, Alston, MA) are listed in Table 1. Using 1 g of total RNA as the template for each reaction, RT-PCR was accomplished by using a single-tube avian myeloblastosis virus RT-Tfl polymerase kit (Access RT-PCR; Promega, Madison, WI). The cycling conditions were as follows: 1 min and 30 s of initial denaturation at 95°C, followed by eight cycles of 30 s at 95°C, 15 s at 56°C, and 30 s at 72°C. After the initial 8 cycles, the 30-s 72°C extension cycle was increased 3 s per cycle for 31 cycles. During the 40th cycle, the 72°C extension was 3 min to complete the RT-PCR. Reactions were also amplified in the absence of reverse transcriptase as negative controls.
Real-time RT-PCR. Cytoplasmic RNA was purified from the siRNA transfected Bm1.11 cells by using the PARIS kit (Ambion). The RNA was quantified by spectrophotometric analysis, and RNA integrity was confirmed by agarose gel electrophoresis. cDNA synthesis was performed using 1 g of the cytoplasmic RNA with the iScript cDNA synthesis kit (Bio-Rad, Hercules, CA). The cDNA product was diluted 1:50, whereas the MyD88-specific, IRF3-specific, TRIFspecific, and ␤-actin control primers (Table 1) were adjusted to 2 pmol/l working stock. Real-time PCR was conducted with the diluted cDNA and primers in accordance with the protocol outlined in the iTaq SYBR green Supermix with ROX kit (Bio-Rad). Real-time PCR was performed with an ABI Prism 7700 machine (Applied Biosystems, Foster City, CA): 2 min at 50°C and 10 min at 95°C, followed by 40 cycles of 15 s at 95°C and 2 min at 60°C. Cycle threshold (C T ) values were determined by automated threshold analysis with ABI Prism version 1.0 software. The amplification efficiencies were determined by serial dilution and calculated as E ϭ expϪ1/m, where E is the amplification efficiency and m is the slope of the dilution curve. Dissociation curves were recorded after each run to ensure primer specificity.
ELISA determination of cytokine production. Confluent Bm1.11, (DN)TRIF, and (DN)MyD88 monolayers grown in 48-well tissue culture-treated plates were either infected with 10 IFU of C. muridarum per cell or pre-treated with siRNA transfection prior to infection. Supernatants were harvested at 18 h or 24 h (depending upon experiment) and analyzed by ELISA for IFN-␤ as previously described (8). All standards and experimental samples were analyzed in triplicate. The lower range of assay sensitivity for IFN-␤ was 10 pg/ml. Statistical analyses. Summary figures for each experimental investigation are presented as a "pooled" means and with their associated standard deviations. Figure legends indicate the number of independent experiments pooled to generate each figure. Analysis of variance models with one or two fixed effect factors, including the two-way interaction if significant, considered the time of the experiment or the run as a random block effect to account for the correlation of observations measured within the same experimental investigation. The tests of lack of fit and model assumptions of homogeneity of variance and normality suggested a log transformation for the Chlamydia growth response (see Fig. 7). Each fixed factor or group effect was tested, and the group means were com-pared. The Tukey-Kramer adjustment method for multiple comparisons was used to control the type I error. Statistical analyses were performed by using the statistical software packages SAS version 9.1 (SAS Institute, Cary, NC).

C. muridarum infection induces an IFN-␤ response by
Bm1.11 oviduct epithelial cells. We previously reported that C. muridarum infection induced IFN-␤ mRNA and IFN-␤ protein synthesis in Bm1.11 oviduct epithelial cells (8,24). Consistent with our previous reports, IFN-␤ secretion by infected Bm1.11 cells was detectable by 12 h postinfection and increased steadily through the 24 h time point (Fig. 1A). Bm1.11 cells at 18 h postinfection produced a level of IFN-␤ intermediate between that seen at the 12-and 24-h time points (data not shown). IFN-␤ secretion was dependent on active Chlamydia replication since Bm1.11 cells exposed to heat-killed C. muridarum did not secrete levels of IFN-␤ statistically different from mock-infected cells (Fig. 1B).

IFN-␤ secreted by the oviduct epithelial cells infected with
Chlamydia requires IRF3. The transcription factor IRF3 is required for the induction of IFN-␤ through TLR and CARD pathways. We utilized siRNA technology to examine the effect of lowering IRF3 mRNA levels on IFN-␤ secretion by C. muridarum-infected Bm1.11 cells. Transfection of Bm1.11 cells with IRF3-specific siRNA in individual experiments resulted in a 3.23-to 3.84-fold reduction in IRF3 mRNA 24 to 30 h posttransfection compared to scrambled siRNA-transfected controls (si-SCR) as determined by real-time RT-PCR (Fig.  3A). As shown in Fig. 3B, lowering IRF3 mRNA levels with si-IRF3 resulted in a reduction of ca. 60% in IFN-␤ secretion 24 h post C. muridarum infection compared to si-SCR controls. Control si-SCR RNA augmented IFN-␤ production by infected Bm1.11 cells, a finding consistent with the reported nonspecific siRNA induction of IFN-␤ pathways (26). si-IRF3treated Bm1.11 cells also produced significantly less IFN-␤ than untreated Bm1.11 cells, although the appropriate experimental comparison is between the si-SCR control and si-IRF3. Partial knockdown of IRF3 levels using siRNA caused a significant decrease in IFN-␤ production by C. muridaruminfected Bm1.11 oviduct epithelial cells, thereby identifying IRF3 as a component of the PRR signaling pathway responsible for IFN-␤ production. IFN-␤ secretion by infected Bm1.11 oviduct epithelial cells is dependent on TRIF. TLR3 signaling for IFN-␤ production is dependent on TRIF, whereas RIG-I/MDA5/cytosolic DNA PRR signaling for IFN-␤ production is TRIF independent. We used siRNA to determine the role of TRIF in IFN-␤ production by infected Bm1.11 cells. Transfection of Bm1.11 cells with TRIF-specific siRNA resulted in a 2.89-to 4.56-fold reduction in TRIF mRNA 24 h posttransfection compared to si-SCR transfected cells as determined by real-time RT-PCR (Fig. 4A). Figure 4B shows the results for experiments using siRNA to decrease levels of TRIF mRNA prior to (B) Bm1.11 cells were mock infected or exposed to 10 IFU of heat-killed C. muridarum per cell. Supernatants were harvested at 18 h, and the levels of IFN-␤ were determined. Bm1.11 cells exposed to heat-killed C. muridarum were compared to mock-infected Bm1.11 cells. Pooled data from two independent experiments are shown. NS, not significant; ***** , P Ͻ 0.0001.   Fig. 4E, the level of IFN-␤ secreted into the medium of pcDNA-TRIF-TIR transiently transfected cells was significantly lower than control transfected and untransfected Bm1.11 cells. These data are consistent with the results obtained with the TRIF-specific siRNA and dominant-negative TRIF cell lines and therefore identify a major role for TRIFmediated signaling in the epithelial IFN-␤ response to C. muridarum infection.
The IFN-␤ response to C. muridarum infection of the oviduct epithelial cells is largely independent of MyD88. We previously demonstrated major roles for TLR2 and MyD88 in infected oviduct epithelial secretion of the inflammatory cytokines IL-6 and granulocyte-macrophage colony-stimulating factor and showed that the majority of IFN-␤ secretion by infected Bm1.11 cells was MyD88 independent (8). Real-time PCR analyses of MyD88-specific siRNA treatment showed a 7.5-to 9.6-fold reduction in the MyD88 mRNA (Fig. 5A). Consistent with our previously published data, MyD88-specific siRNA suppression of MyD88 mRNA levels caused a significant 15 to 20% decrease in IFN-␤ secretion by infected Bm1.11 cells (Fig. 5B). In Fig. 5C Fig. 2 showed that Bm1.11 oviduct epithelial cells express RIG-I and MDA5. To test whether RIG-I/MDA5 were functional in Bm1.11 epithelial cells, poly(I-C) was delivered externally or transfected into the cytoplasm, and the IFN-␤ response was measured (Fig.  6A). Consistent with our previously published data, extracellular poly(I-C) did not trigger detectable IFN-␤ secretion by Bm1.11 cells even though TLR3 mRNA can be detected by RT-PCR analysis. Conversely, transfected poly(I-C) caused marked IFN-␤ section by Bm1.11. These data are consistent with functional RIG-I/MDA5 activation and signaling in oviduct epithelial cells. To test whether this signaling was TRIF dependent, Bm1.11, (DN)TRIF1, (DN)TRIF2, and pcDNA3.1D cells were treated with extracellular poly(I-C), transfected with poly(I-C), or exposed to transfection vehicle alone (Fig. 6B). (DN)TRIF1 and (DN)TRIF2 cell lines responded to transfected poly(I-C) similar to Bm1.11 cells and the pcDNA3.1D control cell line. There was no IFN-␤ response to extracellular poly(I-C). The straightforward interpretation of these results is that Bm1.11 cells have functional RIG-I/MDA5 receptors in their cytoplasm and that the RIG-I/MDA5 signaling pathway to IFN-␤ production is independent of TRIF.
Bafilomycin A inhibits C. muridarum replication in infected Bm1.11 cells more profoundly than IFN-␤ production. Activation of endosomal TLR3 and TLR9 by their microbial ligands requires acidification that can be blocked with bafilomycin A, an endosomal H ϩ -proton pump inhibitor (1,7,38). Bafilomycin A does not interfere with C. trachomatis serovar L2 replication in Vero cells (17). Because Bm1.11 cells do not express TLR7 to TLR9, IFN-␤ production susceptibility to bafilomycin inhibition during C. muridarum infection would implicate TLR3 as the relevant PRR. We therefore infected Bm1.11 cells and subsequently exposed them to bafilomycin A or control medium containing the bafilomycin A vehicle DMSO. While bafilomycin had a minimal inhibitory effect on IL-6 secretion (Fig. 7A), it potently suppressed IFN-␤ production whether added 1 h postinfection (data not shown) or 3 h postinfection (Fig. 7B). However, bafilomycin A blocked C. muridarum replication more profoundly than it suppressed IFN-␤ production (Fig. 7C). The organic solvents used to dissolve bafilomycin, DMSO (Fig. 7) and ethanol (data not shown) had no significant effect on C. muridarum replication or IL-6 production. For reasons that are not clear, the DMSO (Fig. 7) and ethanol vehicles (data not shown) augmented IFN-␤ by Chlamydia-infected epithelial cells. It is unclear whether bafilomycin inhibition of endosomal acidification blocked IFN-␤ production directly through preventing TLR3 activation or secondarily by inhibiting C. muridarum replication.

DISCUSSION
IFN-␤ plays an important role in innate and adaptive immunity via IFN-stimulated genes containing IFN-stimulated response elements in their promoters (34). IFN-␣/␤ has been shown to inhibit C. trachomatis replication in vitro (5,14,16,52), and therefore epithelial secretion of IFN-␤ may help establish a less permissive environment for Chlamydia replication early during infection. The dominant-negative TRIF cell line with the greatest inhibition of IFN-␤ secretion, (DN)TRIF1, had a significantly higher level of C. muridarum replication, possibly reflecting compromise of its innate response and/or resistance to infection.
We and others have shown that Chlamydia species trigger IFN-␤ production in epithelial cells and macrophages (8,13,24,31,46).  tor studies and siRNA experiments showed that C. muridaruminfected Bm1.11 epithelial cell production of IFN-␤ was dependent on viable C. muridarum and was largely TRIF dependent and MyD88 independent. Partial knockdown of IRF3 with IRF3-specific siRNA supported the expected role for IRF3 in IFN-␤ production. We showed that cytosolic dsRNA [liposomal transfection of poly(I-C)] triggered IFN-␤ production through RIG-I/MDA5 in a TRIF-independent fashion (29,43,53,55), a finding consistent with TRIF-independent signaling by RIG-I and MDA5 in murine oviduct epithelial cells (Fig. 6B). The dependence of C. muridarum-infected Bm1.11 epithelial IFN-␤ production on the TRIF adaptor rules out significant roles for RIG-I, MDA5, and the putative cytosolic DNA PRR in infection-triggered IFN-␤ production by Bm1.11 oviduct epithelial cells. TRIF dependence implicates TLR3 or an unknown TRIF-dependent PRR. Our oviduct epithelial data complement the peritoneal macrophage data of Nagarajan et al. (46). IFN-␤ responses of oviduct epithelial cells and murine macrophages to C. muridarum infection differ in their dependence on MyD88, likely reflecting the differences in TLR molecules expressed by each cell type. C. muridarum-infected peritoneal macrophages from MyD88-deficient mice showed a 70 to 80% decrease in IFN-␤ mRNA compared to wild-type mice. Balfilomycin inhibition of endosomal acidification significantly blocked IFN-␤ mRNA induction in infected macrophages, implicating an endosomal TLR as the PRR responsible for IFN-␤ induction. TLR7 to TLR9 are localized within endosomes and require the MyD88 adaptor for IFN-␤ induction, leading us to postulate a major role for these TLRs in IFN-␤ production. Our oviduct epithelial cell lines do not express TLR7 to TLR9 by RT-PCR and do not respond to the TLR9 agonist ODN1826 (8). The existing literature supports TLR7 to TLR9 expression by macrophages, and peritoneal macrophages express TLR7 to TLR9 (data not shown), leading Nagarajan et al. (46) to postulate that infected primary murine lung fibroblasts from MyD88-deficient mice had decreased Cxcl10 (IP-10) production compared to wildtype controls. Cxcl10 transcription is upregulated by IFN-␤ and can serve as an indirect marker for IFN-␤ production. Because fibroblast and epithelial cells are derived from different embryonic tissues (mesoderm versus ectoderm) and the TLR expression pattern of the lung fibroblasts was not characterized, it is not possible at this point to make direct comparisons of our oviduct epithelial cells and the primary lung fibroblasts with respect to IFN-␤ production. IFN-␤ production by murine peritoneal macrophages, bone marrow-derived macrophages (46), and oviduct epithelial cells (the present study) required viable C. muridarum, suggesting that the relevant Chlamydia ligand is either absent from the elementary body or inaccessible in that form. It may be that PRR triggering of IFN-␤ production requires Chlamydia structural subunits (protein or otherwise) that are specific to the reticulate body or injection of Chlamydia molecules into the cytosol via the putative type III secretion apparatus during replication. Inhibition of endosomal acidification did not clarify a role for endosomal PRR recognition of C. muridarum in infected Bm1.11 epithelial cells. Bafilomycin A had a major inhibitory effect on infected Bm1.11 cell secretion of IFN-␤ but had an even greater effect on C. muridarum replication (Fig. 7). Nagarajan et al. reported that C. muridarum replication was not significantly effected by bafilomycin A in peritoneal macrophages (46). C. trachomatis serovar L2 replicating in Vero cells is also indifferent to bafilomycin A (17). The sensitivity of C. muridarum replication to bafilomycin A potentially identifies a major difference in the endosomal contributions to Chlamydia replication in macrophages versus epithelial cells. Alternatively, the endosomal H ϩ -proton pump inhibitor bafilomycin A is an antibiotic with known activity against gram-positive bacteria (62). Detailed studies of Chlamydia susceptibility to bafilomycins have not been performed. It is possible that C. muridarum has greater exposure to bafilomycin A within epithelial cells than it does within peritoneal macrophages.
Our data implicate TLR3, or an unknown TRIF-dependent PRR, as the PRR responsible for IFN-␤ production by C. muridarum-infected oviduct epithelial cells. TLR3 is not a particularly appealing candidate PRR for IFN-␤ production since Chlamydia lacks dsRNA structural subunits or replication intermediates and therefore has no obvious TLR3 ligand. It is possible that there are multiple alternative ligands for TLR3, as was the case for TLR2 (15,35). Interestingly, Stockinger et al. (57) identified a TLR/Nod2-independent, IRF3-dependent pathway for IFN-␤ production in Listeria-infected BMDM. These authors found that TRIF Ϫ/Ϫ bone marrow-derived macrophages (BMDM) had a marked decrease in IFN-␤ mRNA and Stat-1 phosphorylation (signaling event for IFN-␣/␤ receptor) at 2 h postinfection but that IFN-␤ mRNA levels 6 h postinfection approximated those of the wild-type BMDM. The data were interpreted as showing a TRIF-independent pathway for IFN-␤ induction. It is unclear whether the delayed recovery of IFN-␤ at 6 h in Listeria-infected TRIF Ϫ/Ϫ BMDM occurs via the same pathway responsible for induction of IFN-␤ at 2 h in wild-type BMDM. The delayed recovery of IFN-␤ could occur via a slower alternative pathway or via a redundant pathway triggered at a later point during the course of Listeria infection. It is possible that the TRIF-dependent PRR IFN-␤ pathway utilized by C. muridarum-infected Bm1.11 oviduct epithelial cells is the same TLR/Nod2-independent pathway identified in Listeria-infected BMDM. The PRR for the TLR/Nod2-independent pathway in Listeria infection has not been identified.
In summary, we showed that IFN-␤ secretion by C. muridarum-infected oviduct epithelial cells required C. muridarum replication and was dependent on IRF3 and TRIF. TLR expression analysis and TRIF dependence implicate TLR3 as the PRR responsible for infected epithelial IFN-␤ secretion. However, because TLR3 recognizes dsRNA, it would be an unexpected Chlamydia PRR unless there is an uncharacterized non-dsRNA Chlamydia TLR3 ligand. Neither inhibitor studies with bafilomycin ( Fig. 7) nor TLR3 siRNA studies (data not shown) have generated definitive TLR3 data. Our bias is that there is a non-TLR3 TRIF-dependent pathway for IFN-␤ secretion in infected oviduct epithelial cells. Studies to definitively address this issue through derivation of TLR3 Ϫ/Ϫ epithelial cell lines are under way in our laboratory.