Recognition of Borrelia burgdorferi, the Lyme Disease Spirochete, by TLR7 and TLR9 Induces a Type I IFN Response by Human Immune Cells¹

Lyme disease, the most frequently reported tick-borne infection in North America, is caused by the spirochete Borrelia burgdorferi and diagnosed in part by the presence of a distinctive skin rash, or erythema migrans (EM),³ which consists of an influx of immune cells at the site of inoculation (1, 2). In patients who develop more serious long-term sequelae, the spirochete disseminates from the skin via the bloodstream to the joints, heart, and CNS (3). Pathological manifestations of disseminated Lyme disease are characterized by inflammation and include carditis/aortitis, arthritis, and neurological disorders (3). Phagocytic cells of the innate immune system, which can include monocytes/macrophages, dendritic cells, and neutrophils, infiltrate the EM lesion of patients and the infected joints and cardiac tissue of susceptible mouse strains (2, 3, 4, 5, 6). Interaction of these phagocytes with B. burgdorferi can control spirochete burden in infected tissues but simultaneously elicit an array of inflammatory cytokines, which contribute to Lyme disease pathogenesis (7, 8).

Cytokine production is triggered by detection of bacterial pathogen-associated molecular patterns (PAMPs) by cell-associated host pattern recognition receptors, including the TLRs. All TLRs except TLR3 activate the NF-κB signaling cascade through the common adapter molecule primary response gene, MyD88 (9). B. burgdorferi has been documented to induce activation of NF-κB, resulting in prolific expression of an array of inflammatory mediators in human endothelial cells (10, 11). Deficiency of MyD88 completely abolishes B. burgdorferi-stimulated production of these mediators by murine macrophages, although it does not affect arthritis development in infected mice (12, 13, 14). MyD88 has an integral function in host defense; MyD88^−/− mice are unable to effectively control infection (14). Upon recognition by host phagocytes, B. burgdorferi is efficiently ingested and degraded within endocytic vesicles (15, 16). The inability of MyD88^−/− mice to effectively control infection may be explained by the inability of macrophages from these mice to internalize B. burgdorferi (12). The observation that phagocytosis is required for spirochete-elicited production of NF-κB-mediated inflammatory cytokines by human monocytes suggests that B. burgdorferi signals through intracellular receptors which are accessible only following internalization (17, 18).

Several TLRs contribute to the host inflammatory response to B. burgdorferi (19). Spirochetal outer surface lipoproteins signal through TLR1/TLR2 heterodimers in a CD14-dependent manner, generating inflammatory mediators in Lyme disease patients, human monocytic cells, primate glial cells, mice and murine monocytes (12, 20, 21, 22, 23, 24, 25, 26). TLR2 also functions in host defense, as TLR2^−/− mice harbor higher spirochete burdens in target tissues than wild-type controls (27, 28). Recent reports have also established a role for TLR5, a receptor for bacterial flagellin, as a mediator of B. burgdorferi-induced inflammation in murine cells (12, 25). However, deficiency of either TLR2 or TLR5 significantly reduces, but does not completely abrogate, production of inflammatory cytokines by murine macrophages (12).

Compared with mouse models, relatively little is known about the specific innate immune receptors mediating the inflammatory response to B. burgdorferi in humans. Increased surface expression of TLR1, TLR2, and TLR4 has been observed in monocytes/macrophages and dendritic cells taken from the blood and EM lesions of Lyme disease patients with disseminated infection (2). However, only TLR2 has been demonstrated to be directly involved in B. burgdorferi-mediated activation of human immune cells (21, 22, 24, 29).

Host global expression profiling is a powerful technique that has been used to identify both common and unique transcriptional patterns induced in host cells by viral, intracellular, and extracellular bacterial pathogens (30, 31, 32, 33, 34, 35, 36, 37). The resulting transcriptional fingerprints have yielded insights into the interactions between pathogen virulence mechanisms and the host innate immune response. The host response to B. burgdorferi has been examined by expression profiling of mouse joint tissue, RAW264.7 murine macrophages, and human endothelial cells (38, 39, 40). PBMCs provide constant surveillance for invading pathogens and contain phagocytic cells, including macrophages, which infiltrate the EM (2). The interactions of B. burgdorferi with PBMCs may therefore have a decisive effect on the establishment of disseminated infection and the promotion of Lyme disease pathogenesis. In the current study, global transcriptional profiling was used to examine the interactions between a clinical isolate of B. burgdorferi and human PBMCs using an ex vivo coculture system.

Low passage (passage 5 to 7) B. burgdorferi B515, a human clinical isolate, was cultured from a human EM lesion and has been previously described (41). Spirochetes were cultured in modified Barbour-Stoenner-Kelly medium (42) at 34°C until cultures reached the mid- to late-log phase of growth (4 to 8 × 10⁷/ml). Spirochetes were enumerated and assessed for motility by dark-field microscopy. Before addition to PBMCs, spirochetes were centrifuged at 7100 × g, washed twice, and resuspended to a concentration of 5 × 10⁸/ml in HBSS. Spirochetes were used live or were heat-killed by immersion in a 65°C water bath for 40 min.

Venous blood was obtained from each of six healthy volunteers (3 male, 3 female; 35–60 years of age) who had no prior history of Lyme disease, were unvaccinated for Lyme disease and were seronegative for B. burgdorferi infection as confirmed by Western immunoblotting. Written informed consent was obtained from all subjects before blood collection, in accordance with the protocol approved by the Institutional Review Board of New York Medical College.

Blood was collected directly into BD-Vacutainer CPT tubes (BD Biosciences), and PBMCs were isolated according to the manufacturer’s instructions. PBMCs were washed twice in HBSS without calcium, magnesium, or phenol red (Invitrogen) and suspended in 50 ml of RPMI 1640 without phenol red containing 10% (v/v) heat-inactivated and endotoxin-free FBS (HyClone). Cells were maintained for 24 h at 37°C in a humidified incubator containing 5% CO₂ to allow equilibration of basal gene expression under in vitro growth conditions before isolation of RNA for microarray analysis.

The viability of PBMCs after 24 h was determined by trypan blue staining and, with the exception of one donor, was found to be >90%. PBMCs were washed twice with RPMI 1640 containing 10% FBS and resuspended to a final concentration of 5 × 10⁶ viable cells/ml in the same medium. 5 × 10⁷ live or heat-killed B. burgdorferi were added to 5 × 10⁶ viable PBMCs (multiplicity of infection (MOI) = 10:1) in 24-well tissue culture plates, at a final volume of 1.1 ml. In some experiments, B. burgdorferi were separated from PBMCs using 10 mm diameter tissue culture inserts containing membranes with 0.2 μm diameter pores (Nalge Nunc International). Triplicate wells were harvested for RNA extraction after 0, 4, or 12 h in a humidified 37°C incubator containing 5% CO₂. Cell-free culture supernatants were prepared by centrifugation at 9000 × g for 10 min and stored at −20°C. In other experiments, cytochalasin D (Sigma-Aldrich) at a final concentration of 5 or 10 μg/ml in DMSO, or DMSO alone, was added to the PBMCs 45 min before stimulation with B. burgdorferi. Cell-free culture supernatants were collected after 12 h. The presence of cytochalasin D did not affect PBMC viability under these experimental conditions, as assessed by trypan blue exclusion staining.

The contents of each well were transferred to a 1.5-ml microcentrifuge tube and centrifuged at 1800 × g for 10 min at 4°C. PBMCs were washed with HBSS and RNA was isolated using Purescript total RNA isolation kit (Gentra), according to the manufacturer’s instructions. Contaminating DNA was removed using DNA-free kit (Ambion). RNA was eluted in 20 μl of RNase/DNase-free water and stored at −80°C after the addition of 0.8 μl of RNase inhibitor (40 U/μl; Promega). RNA integrity was assessed by electrophoresis using an Agilent Bioanalyzer 2100 before cDNA synthesis for microarray hybridization.

Between 5 and 20 ng of total RNA from each PBMC sample was used to generate a high fidelity cDNA using the Ovation RNA amplification system (NuGEN Technologies) according to the manufacturer’s protocol. The amplified cDNA was fragmented to 50–100 nucleotides, labeled with biotin and hybridized to the Human Genome U133 Plus 2.0 high-density oligonucleotide array containing 54,675 human probe sets (Affymetrix). Following hybridization, the arrays were stained with streptavidin-PE and washed in an Affymetrix fluidics module using standard Affymetrix protocols. The detection and quantitation of target hybridization was performed using a GeneArray Scanner 3000 (Affymetrix).

Microarray data were analyzed using GeneSpring GX9.0.2 software (Agilent Technologies). Raw expression values were normalized by robust multiarray analysis (RMA), filtered to include only those with intensity values above the 20th percentile and baseline transformed to the median of the control samples. Samples were grouped according to time and organism (none, live, or killed). Two-way ANOVA was performed with organism set as the main effect and time set as a random effect. The Benjamini-Hochberg multiple testing correction was applied to reduce false positives. Differentially expressed transcripts were defined as those having a value of p < 0.01 and a fold change ≥2.0 relative to the time-matched control group. Microarray data have been deposited in the GEO database ([email protected]) under accession no. GSE17103. Data were independently analyzed using GeneTraffic UNO software (Iobion Informatics).

Real-time quantitative RT-PCR was used to validate the microarray data (Table I) and to measure transcriptional expression of selected IFN-induced genes in subsequent experiments. PCR assays were performed in duplicate or triplicate on 50× diluted cDNA, prepared as described above, using predesigned TaqMan gene expression assays (Applied Biosystems). Assays used for microarray validation are listed in Table I. Assays used for measurement of gene expression in experiments using TLR inhibitors or B18R included SOCS3 (Hs00269575_s1), IFNB (IFN, β; Hs02621180_s1), IFNA2 (IFN, α 2; Hs02621172_s1), MX1 (myxovirus (influenza virus) resistance 1, IFN-inducible protein p78 (mouse); Hs00182073_m1), IFIT2 (IFN-induced protein with tetratricopeptide repeats 2; Hs00533665_m1) and IRF7 (IFN regulatory factor 7; Hs00185375_m1). All assays were performed in 20-μl reaction mixtures containing TaqMan gene expression assay mix, TaqMan universal PCR master mix without AmpErase UNG (Applied Biosystems), and cDNA using the ABI 7900HT SDS sequence detection system (Applied Biosystems) according to the manufacturer’s instructions. For each sample, the expression of GAPDH was quantified by real-time RT-PCR using the TaqMan human GAPDH endogenous control assay (Applied Biosystems), and a ΔΔC_t method was used to estimate the differential gene expression relative to the time-matched control (43).

The concentration of human IFN-α in cell-free culture supernatants was quantitated using the Human Verikine IFN-α ELISA kit (PBL Biomedical Laboratories).

B18R (suspended in PBS, pH 7.2, 150 mM NaCl, 1% BSA), a vaccinia virus-encoded neutralizing type I IFN receptor, was purchased from eBioscience and certified to have an endotoxin level of <0.01 ng/μg, as determined using the Limulus amebocyte lysis assay (LAL) (Lonza). A final concentration of 0.1 μg/ml B18R was added to 5 × 10⁶ PBMCs, cultured as described above, 1 h before the addition of B. burgdorferi. Cocultures were incubated for 12 h, and RNA and cell-free culture supernatants were prepared as described above.

Imiquimod (R837), a ligand specific for human TLR7, was purchased from InvivoGen and added to 5 × 10⁶ PBMC at a concentration of 5 μg/ml. TLR7 and TLR9 inhibitors and control oligodeoxyribonucleotide (ODN) sequences were synthesized on a phosphorothioate backbone by Integrated DNA Technologies and purified by ion-exchange HPLC (IE-HPLC) (44, 45, 46). Endotoxin levels of all ODNs were <6 U/mg ODN, as determined by LAL. The control ODN and IRS661, an inhibitor of human TLR7, were used at a concentration of 5.6 μM. IRS869, an inhibitor of human TLR9, was used at a concentration of 1.4 μM. These concentrations have been shown to result in optimal effectiveness and specificity of the inhibitory sequences (44).

Cell populations expressing IFN-α were identified by phenotypic markers and intracellular cytokine staining. PBMCs were cultured with B. burgdorferi as described above. Two hours into the incubation, GolgiPlug (BD Biosciences) was added at 2.5 μg per 5 × 10⁶ cells to prevent protein transport. After 10 h of incubation in the presence of GolgiPlug, PBMCs were washed twice with HBSS, and Fc receptors were blocked with 10 μg/ml purified human IgG (Sigma-Aldrich). Cell viability was assessed by trypan blue staining and found to be 80–90%. Cells were stained with Ab conjugates to specific surface Ags or with the relevant conjugated isotype-matched control Ab, fixed, permeabilized and stained for the presence of intracellular IFN-α using the BD Cytofix/Cytoperm Plus kit (BD Biosciences). Ab conjugates used included anti-human CD11c-PE-Cy5, anti-human IFN-α-PE and anti-human lineage mixture 1 (CD3, CD14, CD16, CD19, CD20, CD56)-FITC (BD Bioscience), anti-human CD14-APC and anti-human BDCA2-FITC (Miltenyi Biotec). Isotype control conjugates were purchased from eBiosciences. Cells were sorted by multiparameter flow cytometry using a MACSQuant analyzer (Miltenyi Biotec). For each sample, 300,000 events were collected. Data analysis was performed using software from Miltenyi Biotec, Inc.

B. burgdorferi encounters the host immune system initially upon deposition into the skin and subsequently in the bloodstream en route to target tissues. Cellular components of PBMCs are found in the inflammatory infiltrate of EM lesions and thus comprise a population of innate immune cells that mediate the initial interactions between host and pathogen (2). As PBMCs also constitute a readily accessible source of human immune cells, an ex vivo coculture model of infection using PBMCs was used to investigate the host transcriptional response to B. burgdorferi. To examine the global transcriptional response of human leukocytes to B. burgdorferi, PBMCs from 5 healthy donors were cultured for 0, 4, or 12 h with B. burgdorferi B515, a human clinical isolate which also disseminates in mice (41). Spirochetes were used live or were heat killed by immersion in a 65°C water bath for 40 min, which renders the spirochetes incapable of growth in BSK-S medium (data not shown). Microscopic examination revealed that spirochetes heat killed under these conditions retained spirochetal morphology, although some membrane blebbing was observed (data not shown).

Amplified cDNA was hybridized to the Affymetrix Human Genome U133 Plus 2.0 oligonucleotide array which contains 54,675 human probe sets representing 38,500 well-characterized genes. Gene array data were analyzed using GeneSpring GX software and filtered to remove transcripts with intensity values below the 20th percentile, or which were not present in at least one sample. The resulting 53,538 entities were analyzed by two-way ANOVA with parameters set on time and treatment of organism (live or killed). A total of 2196 (5.12%) transcripts exhibited expression levels which were significantly changed at least 2-fold (p ≤ 0.01) by either live or heat-killed B. burgdorferi after 4 or 12 h relative to the time matched control. After 4 h of coculture, 783 transcripts (439 induced, 344 repressed) were differentially expressed (≥2-fold, p ≤ 0.01) in PBMCs exposed to live B. burgdorferi, while 838 transcripts (550 induced, 288 repressed) were differentially expressed by heat-killed B. burgdorferi (Fig. 1). After 12 h, the number of differentially expressed transcripts increased to 1761 (1195 induced, 566 repressed) and 1183 (867 induced, 236 repressed) in response to live or heat-killed B. burgdorferi, respectively (Fig. 1). The data were separately analyzed by multi-class ANOVA using GeneTraffic software, and the same set of transcripts was identified as differentially expressed.

A notable observation was that live and heat-killed B. burgdorferi elicited remarkably similar transcriptional profiles in human PBMCs (Figs. 1 and 2). Of the transcripts differentially expressed upon exposure to either live or killed B. burgdorferi, 55.7% and 48.3% were shared at 4 and 12 h, respectively (Fig. 1). In contrast, length of time of coculture resulted in differential expression of distinct sets of genes; only 29.9% of transcripts differentially expressed in response to live spirochetes were shared at both 4 and 12 h (Fig. 1).

The gene array expression values for a total of 13 genes were validated by real-time RT-PCR. These consisted of the genes encoding CXCL1, PTX3, INHBA, GJB2, TNFSF10, EREG, IFIT5, OAS1, IFN-γ, IL-10, RSAD2 (viperin), TNFRSF9, and SOD2 (Table I). Although the absolute expression values obtained by this method differed from the microarray values, a similar trend in transcriptional induction was observed by both methods for all genes assayed (Table I).

To further examine the relative expression kinetics of genes involved in innate immune signaling and apoptosis induction, these transcripts were subjected to K-means clustering (Fig. 2). Cluster A included inflammatory cytokines, TNF-α, IL1β, and IL1α, which were rapidly and intensely induced in response to B. burgdorferi. A second cluster (Fig. 2, cluster D) of transcripts associated with positive regulation of innate immunity was also induced during the early response to B. burgdorferi. These included TLR2, IL18, TNF superfamily members, and factors involved in the NF-κB-dependent signaling cascade. Elevated levels of these transcripts coincided with induction of the IL1 receptor antagonist (IL1RN) and down-regulation of the inflammatory mediators IFN-γ receptor (IFNGR), TLR4 and other members of the TNF superfamily (Fig. 2, cluster B). A fourth group of transcripts was significantly induced only after 12 h of coculture (Fig. 2, cluster C). This group of transcripts included inflammatory caspases (CASP1, CASP4, CASP7, CASP10), mediators of IL18 signaling (IL18R and IL18RAP), IFN responsive factors involved in TLR-mediated signaling (IRF1, IRF4, IRF7, IRF9) and transducers of the JAK/STAT signaling cascade (JAK2, STAT1, STAT3).

Hierarchical clustering by both entity and condition of the 2196 transcripts which were differentially expressed upon exposure to B. burgdorferi at either time point revealed two clusters of genes that were robustly induced (Fig. 3). The most intensely up-regulated gene cluster consisted of 13 transcripts encoding factors associated with innate immunity and inflammation, including proinflammatory cytokines and chemokines (IL1α, IL1β, IL12β, IL6 and CCL20/MIP-3α) and PTGS2 (prostaglandin-endoperoxide synthase 2). These transcripts were induced after 4 h of coculture and remained up-regulated at 12 h, with mean fold changes between 7.9 and 65.9 relative to unstimulated PBMCs. A second cluster of 61 transcripts, which were significantly up-regulated after 4 h and further induced after 12 h, included additional proinflammatory cytokines and chemokines, immunomodulatory cytokines, a matrix metalloprotease, and genes annotated as involved in the host antiviral response (Fig. 3).

A striking and immediately apparent observation was the predominance of genes within these two clusters which were annotated by GeneSpring as IFN-inducible. In addition, a subsequent literature search revealed that a number of other genes that were significantly induced by B. burgdorferi could be induced by one or more IFNs (47, 48), (http://www.lerner.ccf.org/labs/williams/). Remarkably, of the 100 transcripts which were most strongly induced by B. burgdorferi after 12 h of coculture, 50 (50%) were documented as inducible by IFNs α, β, or γ. A list of B. burgdorferi-up-regulated IFN-inducible genes, categorized by annotated biological function, is presented in Table II. Categories containing five or more transcripts included those with functions in apoptosis, cell cycle, and cell proliferation, inflammation and chemotaxis, innate immune response to virus, and signal transduction and transcription. The majority of the transcripts were significantly induced by both live and killed B. burgdorferi at 4 h and showed increased transcript levels after 12 h.

Many of the up-regulated IFN-inducible transcripts can be induced by both type I and type II IFNs, whereas others (e.g., MX1, OAS1, and OAS2) are preferentially induced by type I IFNs. Transcription of IFN-γ, a type II IFN, was strongly induced by B. burgdorferi (Fig. 2), but no significant increase in transcript for either IFN-β or IFN-α subtypes was detected by oligonucleotide array. To detect transcription of type I IFNs, the more sensitive technique of real-time RT-PCR was performed on RNA isolated from PBMCs which had been cocultured for 12 h. B. burgdorferi was found to significantly induce transcription of both IFN-β and IFN-α, in addition to IFN-γ (Table III).

Protein levels of IFN-α were quantitated in culture supernatants of PBMC/B. burgdorferi cocultures. As shown in Fig. 4,a, IFN-α protein was detectable after 4 h exposure to B. burgdorferi and significantly increased after 12 h exposure to either 5 × 10⁷ live (145.2 pg/ml, p < 0.0001) or killed (45.5 pg/ml, p < 0.0001) spirochetes. Although B. burgdorferi has not been documented to produce toxins, PBMCs and B. burgdorferi were separated by culture chamber inserts containing semipermeable membranes (0.2-μm pore) which enabled passage of small molecules but not of intact B. burgdorferi to determine whether IFN-α production was elicited by means of soluble secreted factors or through direct contact. IFN-α production was absolutely dependent upon direct contact between spirochetes and PBMCs, as separation of host cells and pathogen by semipermeable membranes completely eliminated the production of IFN-α (Fig. 4,a). To determine whether surface or intracellular/phagosomal receptors were mediating the type I IFN response, cocultures were incubated in the presence of cytochalasin D, an inhibitor of phagocytosis. The addition of 5 μg/ml cytochalasin D significantly reduced IFN-α secretion, and incubation in the presence of 10 μg/ml cytochalasin D, a concentration which effectively inhibits internalization of B. burgdorferi by human monocytes (18), reduced IFN-α levels to those observed with unstimulated PBMCs (Fig. 4 b). These results establish phagocytosis of spirochetes as a requirement for IFN-α production.

Type I and type II IFNs promote expression of IFN-inducible genes by autocrine/paracrine signaling transduced by two classes of high-affinity receptors, the IFN-α/β (IFNAR) and IFN-γ (IFNG) receptors, respectively (49). As many IFN-inducible genes are responsive to both types of IFNs, the contribution of type I IFNs to the B. burgdorferi-elicited IFN-inducible profile was determined by inhibiting type I IFN-mediated autocrine and paracrine signaling. B18R is a soluble vaccinia virus-encoded type I IFN-binding protein which serves to modify host anti-viral responses by neutralizing secreted type I IFN (50, 51, 52, 53). PBMCs were cultured in the presence or absence of 0.1 μg/ml recombinant B18R for 1 h before addition of viable B. burgdorferi. After 12 h, supernatants from cocultures incubated without B18R contained 731.3 pg/ml IFN-α, whereas IFN-α protein could not be detected in supernatants from unstimulated PBMCs or from cocultures to which B18R had been added (Fig. 5,a). This established that the concentration of B18R was sufficient to completely neutralize secreted IFN-α. The effect of type I IFN neutralization on the transcriptional expression of IRF7, OAS1, MX1, IFIT2 and SOCS3 was evaluated by real-time RT-PCR. OAS1 and MX1 are canonical type I IFN-induced genes, IRF7 is induced predominantly by type I IFNs, and IFIT2 and SOCS3 are responsive to both type I and type II IFNs. As shown in Fig. 5 b, transcription of all genes was markedly and significantly induced in PBMCs by B. burgdorferi relative to unstimulated PBMCs after 12 h coculture. Addition of 0.1 μg/ml B18R significantly reduced transcription of IRF7, MX1, IFIT2 and SOCS3. Importantly, although this concentration of B18R completely neutralized soluble IFN-α, transcript levels of all five IFN-inducible genes were still significantly elevated relative to levels in unstimulated PBMCs. The presence or absence of B18R did not affect the transcript levels of these genes in unstimulated PBMCs (data not shown). These results demonstrate that B. burgdorferi-elicited expression of these IFN-inducible genes can occur in the absence of IFNAR signaling.

IRF7 is a transcription factor which has a critical role in the MyD88-dependent induction of type I IFNs through TLR7- and TLR9-mediated signaling (54, 55, 56). Moreover, IRF7 is expressed at high basal levels by pDCs, a human cell type which preferentially expresses both TLR7 and TLR9 relative to other PBMC subsets (57, 58, 59). The observation that B. burgdorferi induces transcription of IRF7 (Fig. 2 and Table III) suggested that TLR7 and TLR9 may function as recognition molecules for components of this pathogen. This hypothesis was tested using ODN-based inhibitors of TLR signaling (Fig. 6). IRS869 is a specific inhibitor of TLR9, whereas IRS661 specifically inhibits TLR7 activation (44, 45). PBMCs were cultured in the presence of IRS661 (5.6 μM), IRS869 (1.4 μM) or a nonspecific control ODN (5.6 μM) for 1 h before the addition of a stimulus. Culture supernatants were collected after 12 h and assessed for the presence of secreted IFN-α protein (Fig. 6,a). As a control for receptor specificity, PBMCs were stimulated with the TLR7-specific ligand R837 (60), which elicited substantial levels of IFN-α (584.8 ± 72 pg/ml). Addition of either the control ODN or IRS869 did not significantly reduce R837-induced IFN-α production, while the presence of the TLR7 inhibitor IRS661 completely abrogated IFN-α production, as expected. Stimulation of PBMCs with B. burgdorferi resulted in IFN-α at a level (707.6 pg/ml) comparable to that observed after addition of R837. Addition of IRS869 or IRS661 to PBMC/B. burgdorferi cocultures significantly reduced IFN-α production (63.4 or 100.2 pg/ml, respectively; p < 0.01) relative to IFN-α levels in the presence of the control ODN. Simultaneous addition of both inhibitors resulted in complete abrogation of B. burgdorferi-induced IFN-α production (Fig. 6 a).

To determine the effect of TLR7/TLR9 inhibition on transcription of selected IFN-inducible genes, total RNA was prepared from the PBMCs cultured with B. burgdorferi alone or in the presence of the oligonucleotide inhibitors, and transcript levels of OAS1, MX1, IRF7 and IFIT2 were measured by real-time RT-PCR (Fig. 6 b). Addition of the control ODN did not significantly affect expression of any of the selected genes. In contrast, IRS661 significantly inhibited transcription of MX1 and IRF7, and IRS869 significantly inhibited transcription of OAS1, MX1 and IRF7, relative to transcription in the presence of the control ODN. Inhibition by either IRS661 or IRS869 alone was not absolute, as transcription of all genes was still significantly elevated relative to unstimulated PBMCs. However, as observed with IFN-α protein production, addition of a combination of IRS661 and IRS869 reduced transcription of all four genes to levels which were not significantly different from those in unstimulated PBMCs. These data demonstrate that B. burgdorferi stimulates production of type I IFNs and IFN-inducible genes via TLR7- and TLR9-mediated signaling.

Although pDCs have been classically described as high IFN-α-producers, other cellular components of PBMCs have been documented to produce type I IFN (61, 62, 63, 64, 65). Flow cytometric analysis was used to identify the cell population(s) expressing IFN-α in response to B. burgdorferi. PBMCs were stimulated with B. burgdorferi for 12 h in the presence of a protein transport inhibitor, and then stained for the cell surface markers CD14, CD11c and BDCA2, a specific marker of human pDCs (66), and for the presence of intracellular IFN-α. BDCA2⁺ cells expressing IFN-α accounted for 0.05% of the population of unstimulated PBMCs (Fig. 7,a). After stimulation with B. burgdorferi, IFN-α-expressing BDCA2⁺ cells increased nearly three-fold to 0.14% of the total cell population (Fig. 7,a). However, IFN-α was also observed to be expressed by a substantial population of BDCA2⁻ cells. This population increased from 0.2% to 2.31% of PBMCs upon stimulation with B. burgdorferi (Fig. 7,a). To identify this IFN-α-expressing BDCA2⁻ population, cells were gated for IFN-α expression and assessed for the presence of CD11c and CD14. 97.23% of the IFN-α-secreting cells stained double positive for both CD11c and CD14 (Fig. 7,b). Background expression of IFN-α was observed in unstimulated PBMCs. Of these cells, 90% were CD14⁺CD11c⁺. However, the total number of CD14⁺CD11c⁺ cells expressing IFN-α increased 10-fold after B. burgdorferi stimulation (459 vs 4461 cells, Fig. 7,b). To validate the identity of the BDCA2⁺ cells as pDCs, PBMCs were stained in a separate panel with anti-human lineage mixture, anti-BDCA2 and anti-CD11c. The lineage⁻CD11c⁻ population was analyzed for expression of BDCA2 and IFN-α. Lineage⁻CD11c⁻BDCA2⁺ pDCs did not express IFN-α in the unstimulated PBMCs (Fig. 7 c). In contrast, 9.97% of this cell population produced IFN-α upon stimulation with B. burgdorferi. These results demonstrate that multiple PBMC subsets, including pDCs and CD14⁺CD11c⁺ cells, contribute to the IFN-α response to B. burgdorferi.

Inflammation is induced upon recognition of PAMPs by host cellular receptors, including TLRs, leading to the production of inflammatory mediators including chemokines and cytokines. Lyme disease is characterized by inflammation which may either lead to eradication of infection or promote pathogenesis, depending on the kinetics of induction. Although genetically manipulated mice and murine strains with inherent differences in their susceptibility to Lyme disease have yielded crucial insights into the mechanisms of B. burgdorferi-mediated inflammation, relatively little information exists about this process in the human system. We therefore used an ex vivo coculture model to examine the global transcriptome elicited in human PBMCs upon interaction with a clinical isolate of B. burgdorferi.

Live or heat-killed spirochetes were found to induce qualitatively similar profiles in PBMCs after 4 or 12 h of culture, including the intense up-regulation of components of the NF-κB signaling cascade and NF-κB-dependent inflammatory and immunomodulatory chemokines and cytokines, including IL10, IFN-γ, IL12, IL6, IL1, IL8 and TNF-α. Interestingly, viable spirochetes elicited significantly higher levels of IFN-α protein than did killed spirochetes. The method used for heat-killing resulted in morphologically intact spirochetes as determined by microscopy, but such organisms were not cultivable. A previous study of B. burgdorferi/human PBMC coculture under conditions similar to those used in the current investigation demonstrated that significantly less TNF-α and IL-1β were induced by heat-killed B. burgdorferi relative to live organisms; by contrast, lysed spirochetes induced negligible amounts of these cytokines (17). Although the conditions for spirochetal killing used in the two studies were different (48°C for 30 min. vs 65°C for 40 min.), both demonstrate that morphologically intact, non-viable B. burgdorferi (as opposed to cell lysates) can induce cytokine production, albeit at a reduced level relative to live organisms.

Genes annotated in the literature and by GeneSpring as ‘IFN-inducible’ or ‘IFN-responsive’ were prominently represented among the transcripts up-regulated by exposure to B. burgdorferi. Many of these transcripts have been documented to be responsive to both IFN-γ and type I IFNs. Although IFN-γ was the only IFN shown by microarray to be induced by B. burgdorferi, we subsequently confirmed B. burgdorferi-induced expression of type I IFNs at both the transcript and protein levels. Type I IFNs were initially identified as potent effectors of innate immunity to viruses, and later to intracellular bacteria (67). Within the past several years, however, evidence has accumulated linking the production of type I IFN to the host innate immune response to extracellular bacterial pathogens as well (68, 69, 70, 71, 72). Type I IFNs can be produced through diverse pathways, including activation of TLR3, TLR4 or TLR7, TLR8 or TLR9, as well as through recognition of microbial components by intracellular cytosolic receptors (73).

We have demonstrated that B. burgdorferi-induced IFN-α secretion by human PBMCs is completely dependent on recognition of the bacterium by TLR7 and TLR9. To our knowledge, this is the first description of the role of these TLRs in Lyme disease pathogenesis. These receptors traffic to the phagolysosome and recognize single-stranded RNA and small nucleoside analogues, and unmethylated CpG, respectively (60, 74, 75, 76). Bacterial nucleic acids, released following degradation of the organism, may be recognized by TLR7 and TLR9 and, in fact, extracellular bacterial RNA stimulates type I IFN production via phagolysosomal TLR7 recognition in murine conventional DCs (68). During the course of the studies described here, Salazar et al. reported a global transcriptome analysis of human PBMCs cocultured with B. burgdorferi (77). Although there were several methodological differences between the two studies, the results of both are largely in concordance; Salazar et al. noted transcriptional induction of inflammatory cytokines, type I IFNs and type I IFN-responsive genes that was dependent on phagocytosis (77). The current study establishes that IFN-α protein production and IFN-responsive gene transcription are mediated by TLR7/9.

B. burgdorferi is avidly phagocytosed by murine and human phagocytes, including mDCs, pDCs and monocytes/macrophages, resulting in spirochetal degradation within phagolysosomes and the subsequent induction of innate immune mediators and monocyte apoptosis (12, 15, 17, 18, 78, 79, 80). Here, we have demonstrated that IFN-α induction by B. burgdorferi is completely dependent on phagocytosis, consistent with detection by phagolysosomal receptors. A schematic representation of TLR7/9-mediated induction of type I IFNs by phagocytosed B. burgdorferi, based on reviews of signaling pathways and adapted to include the current data, is depicted in Fig. 8 (73, 81, 82, 83). Detection of spirochetal lipoproteins and flagellin by TLR2 and TLR5, respectively, contributes to the inflammatory cytokine profile elicited by B. burgdorferi in murine macrophages (12). However, ablation of TLR2- or TLR5-mediated signaling does not result in the dramatic reduction in cytokine levels observed in MyD88^−/− macrophages, implicating the involvement of other MyD88-dependent receptors in the response to B. burgdorferi (12). The suggestion that these receptors are intracellular is supported by the inability of MyD88^−/− macrophages to internalize B. burgdorferi, as well as by the fact that stimulation of TLR2^−/− macrophages in the presence of an inhibitor of phagocytosis reduces cytokine expression to that observed in MyD88^−/− macrophages (12). In contrast to the present findings, Shin et al. reported that TLR9 did not contribute to B. burgdorferi-elicited cytokine expression, but it should be noted that the five cytokines assessed in that study did not include type I IFNs (12). Transcription of IFN-β is not diminished in TLR2^−/− murine macrophages and occurs in human monocytes in response to a flagellin-deficient strain of B. burgdorferi, thus demonstrating that expression of type I IFNs does not require TLR2- or TLR5-dependent signaling (77).

The identification of B. burgdorferi-elicited signaling via TLRs 7 and 9 has several implications for understanding the development of Lyme disease pathogenesis. The simultaneous activation of both IRF7 and the NF-κB signaling cascade by TLR7 and TLR9 ligands is predicted to result in the production of both type I IFN and NF-κB-dependent proinflammatory and immunomodulatory cytokines (Fig. 8). This implicates a potentially broader contribution of TLR7/9-mediated signaling to the overall cytokine profile elicited by B. burgdorferi in human PBMCs, particularly in light of recent reports that synergistic interactions between multiple receptors are required for full activation of the proinflammatory response of phagocytic cells to Streptococcus pyogenes, Francisella tularensis and bacterial LPS or lipotechoic acid (84, 85, 86). In addition, the direct actions of type I IFNs may have a determining role in the establishment of Lyme disease. Type I IFNs have been shown to enhance susceptibility of mice to infection with Chlamydia muridarum and L. monocytogenes by promoting the apoptosis of monocytes and lymphocytes, a pathogen immune evasion strategy (87, 88, 89, 90, 91). We speculate that the expression of type I IFN may induce the observed B. burgdorferi-elicited apoptosis of phagocytic cells (17), thereby facilitating bacterial dissemination. A similar mechanism has been proposed by Cruz et al. (17). Additionally, as type I IFN is an effective substitute for IL-12 in driving the production of IFN-γ by T and NK cells during microbial infections (62, 69, 92, 93), a synergistic interaction between type I and type II IFNs in promoting a potent proinflammatory response to B. burgdorferi is possible.

Although the study reported here describes a mechanism for the production of B. burgdorferi-elicited type I IFN protein expression by human cells, a recently published study identified a crucial function for type I IFNs in arthritis development in a murine model of Lyme disease (94). Miller et al. demonstrated that B. burgdorferi elicits up-regulation of an array of IFN-responsive genes, and that this occurs via IFNAR-mediated signaling. These results are in general agreement with our findings. However, the studies diverge in several important respects. First, Miller et al. proposed that induction of IFN-responsive genes occurs via detection of bacterial components by an unidentified cytosolic receptor, whereas we have identified TLRs 7 and 9 as the mediators of B. burgdorferi-elicited type I IFN. In addition, Miller et al. presented evidence that induction of selected IFN-responsive genes does not occur in IFNAR^−/− mice, whereas the results presented here demonstrate that transcription of these genes can occur in the absence of IFNAR signaling, albeit at significantly reduced levels. The disparities between the present results and the findings of Miller et al. may be attributable to the fact that a number of immunological discrepancies exist between mice and humans, including species-specific expression patterns of TLRs, sequence differences in the extracellular region of TLR9 and production of NO by macrophages (95, 96). A MyD88-independent, IRF3-dependent signaling pathway mediated by an unidentified cytosolic receptor has been implicated in activating type I IFN production by murine macrophages in response to Listeria monocytogenes, as well as to the extracellular Group A and Group B streptococci (70, 72, 97). However, a more recent study conclusively delineated the role of TLR7 in the induction of IFN-α and IFN-β by extracellular bacteria in murine mDCs (68). It is noteworthy that Miller et al. assessed transcriptional induction of four IFN-responsive genes but did not quantitate either transcript or protein levels of IFN-α or IFN-β.

In contrast to the results reported here, studies using murine models consistently pointed to a requirement for IFNAR-dependent signaling for type I IFN production. Notably, macrophages from IFNAR^−/− mice produced significantly lower levels of IFN-γ, NO and TNF-α after exposure to Group B streptococcus, and infected IFNAR^−/− mice died from unrestrained bacteremia (69). Although only scarce information is available detailing the type I IFN response of human cells to bacterial pathogens, one study has demonstrated that the extracellular bacterium, Staphylococcus aureus, elicits IFN-α expression by human pDCs through TLR7 and TLR9, thus establishing a precedent for our findings (71).

Intracellular cytokine staining established that both pDCs and CD14+CD11c+ cells express IFN-α upon stimulation with B. burgdorferi. Previous reports have established that B. burgdorferi spirochetes are avidly phagocytosed by human mDCs and pDCs, as well as by monocytes (18, 78). pDCs are the most potent IFN-α-secreting cell type in human blood and express constitutively high levels of IRF7, a key transcription factor in the positive feedback regulation of type I IFN production (54, 57, 61, 75, 98). In addition to IFNAR-dependent, JAK/STAT-mediated activation, the IRF7 promoter can be directly activated in the absence of type I IFN protein expression via an IFNAR-independent pathway (57, 99). It has been proposed that high constitutive levels of IRF7 drive the direct virus-induced activation of IFN-inducible genes in dendritic cells isolated from IFNAR^−/− mice (100). Additionally, the type I IFN-inducible gene MxA is expressed by human PBMCs in the presence of neutralizing anti-IFN-α/β Abs and in the absence of IFN protein synthesis (101). Taken together, our data demonstrating that transcription of IRF7 is significantly induced by B. burgdorferi in human PBMCs, that IFN-α production is TLR7/9-dependent and that expression of type I IFN-inducible genes can occur in the absence of IFNAR feedback signaling are consistent with the observation that pDCs are producers of B. burgdorferi-elicited IFN-α in the human PBMC coculture model.

In addition to pDCs, CD14+CD11c+ cells were also found to express IFN-α in response to B. burgdorferi stimulation. This phenotype has been used to describe several human blood cell populations, including monocytes, monocytoid dendritic cell (mDC) precursors and immature mDCs (102, 103, 104, 105, 106, 107). Differentiation of monocytes into mature mDCs results in the decline of surface expression of CD14; the CD14+CD11c+ phenotype has been described as a transitional state in this maturation process (103, 104, 106, 108). The results described here are supported by reports that isolated human monocytes and monocyte-derived mDCs, as well as mouse conventional DCs, produce IFN-α in response to microbial PAMPs (61, 64, 65, 68, 109, 110). Although a similar study using isolated human CD14+ cells exposed to B. burgdorferi detected intense up-regulation of IFN-β transcript but only modest transcriptional induction of one IFN-α subtype (IFNA6) (77), it is recognized that pDC-derived type I IFN can prime human monocytes to become IFN-α-producing cells (63, 111). Additionally, although TLR7 and TLR9 are preferentially expressed by human pDCs and B cells relative to other PBMC subsets, there is evidence that tissue expression of TLRs is fluid and changes in response to cell activation (59). Human monocytes significantly increase expression of TLR7 and TLR9 upon exposure to LPS, microbial pathogens or IFN-α (63, 112, 113, 114, 115). These reports support the observation that IFN-α can be produced by CD14⁺CD11c⁺ cells in the context of a mixed cell population which includes pDCs. Moreover, a mixed cell population may more accurately reflect the complexities of the pathogen-immune system interactions than isolated cell subsets.

In summary, the data presented here provide evidence for the first time that the extracellular bacterium, B. burgdorferi, is detected by human TLRs 7 and 9, leading to direct expression of an array of IFN-inducible genes and significant production of IFN-α. The PAMP detected by these receptors, likely nucleic acids, is present within both live and killed spirochetes and induces IFN-α following phagocytosis. Moreover, as signaling through TLR7 and TLR9 is MyD88-dependent and results in the activation of NF-κB, this pathway may have broader contributions to the cytokine profile elicited in response to B. burgdorferi and the subsequent generation of Lyme disease pathogenesis.

The authors gratefully acknowledge the expert phlebotomy services of Diane Holmgren.

The authors have no financial conflict of interest.