The Toll-Like receptor adaptor TRIF contributes to otitis media pathogenesis and recovery

TRIF-/-, TLR2-/- and TLR4-/- mice on a C57BL/6 background (the null alleles were 6 × backcrossed onto C57BL/6, and then intercrossed to establish the homozygous lines) were originally generated by Beutler and colleagues [29, 30], and used with their permission, but were generously supplied by Dr. Timothy Bigby of UCSD. Age-matched C57BL/6J control mice, and C57BL/6J:CB F1 hybrids for gene array analysis, were purchased from Jackson Laboratories (Bar Harbor, ME). All experiments were performed according to NIH guidelines and approved by the Institutional Animal Care and Use Committee of the San Diego VA Medical Center. The mice were housed and maintained in accordance with NIH policy for the human treatment of animal subjects.

Haemophilus influenzae (non-typeable, biotype II; NTHi) strain 3655 (originally isolated from the ME of an otitis media patient in St. Louis; provided by Dr. Asa Melhus, Lund University) was streaked onto chocolate agar overnight. Two colonies were picked and grown in 25 ml of brain heart infusion (BHI) media with 1 ml of Fildes Enrichment (BD Diagnostic Systems). The bacteria were spun down (9,000 rpm, 10 min) and then resuspended to a final titer of 10⁵ – 10⁶/ml [31].

TRIF-/- and C57BL/6J mice were divided into groups of 6 mice for each experimental time point (3 each for histopathology and bacterial culture). For DNA microarrays, 40 C57BL/6J:CB F1 hybrid mice per time point were divided into two groups. All animals were inoculated with NTHi as previously described [28, 31]. Briefly, the mice were deeply anesthetized, and both ME bullae were accessed via a ventral midline incision in the neck. A hole was made in the bullae using the tip of a 21 gauge needle. Both MEs were injected with ~5 μl of NTHi inoculum, after which the incision was closed [28, 31]. Uninoculated (time 0) or sham injected (saline) animals served as controls.

Expression of TRIF mRNA during NTHi-induced OM was assessed in individual ME mucosae from wild-type (WT), TLR2-/- AND TLR4-/- mice (n = at least 6 per time point) by qPCR s previously described [25]. The ME mucosae were dissected, mRNA was extracted and reverse transcribed, and 1 μg/μl of cDNA was amplified using commercial TaqMan qPCR primers (Applied Biosystems, Foster City, CA) for TRIF (Mm01260003_m1) and MyD88 (Mm01351743_g1) in an ABI Prism 7000 Sequence Detection System (Applied Biosystems). Fold induction was calculated using the comparative threshold cycle (Ct) method [32]. Relative expression of each target gene was normalized to levels of GAPDH and compared to uninfected mucosa.

For each time point and condition (NTHi or PBS injection), forty WT (C57BL/6J) ME mucosae were harvested from deeply anesthetized mice at 0, 3 and 6 hours, as well as 1, 2, 3, 5 and 7 days after NTHi inoculation. The tissue from 20 different mice was pooled, to generate two independent samples. The tissue was homogenized in TRIzol™ (Invitrogen, Carlsbad, CA) and total RNA was extracted. Total RNA quality was assessed using the RNA 6000 Labchip Kit on the Agilent 2100 Bioanalyzer for the integrity of 18S and 28S ribosomal RNA. The mRNA was reverse transcribed using a T7-oligodT primer then in vitro transcribed using T7 RNA polymerase to generate biotinylated cRNA probes that were hybridized to Affymetrix MU430 2.0 microarrays. Duplicate arrays were hybridized for each time point, using RNA from pools of different mice to obtain an independent biological replication. Raw intensity data was median normalized and statistical differences in gene transcript expression levels were evaluated using a variance-modeled posterior inference approach (VAMPIRE) [33]. This program uses a Bayesian approach to identify altered genes. Statistical analysis by VAMPIRE requires two distinct steps: (1) modeling of the error structure of sample groups and (2) significance testing with a priori-defined significance thresholds. VAMPIRE models the existing error structure to distinguish signal from noise and identify the coefficients of expression-dependent and expression-independent variance. These models are then used to identify microarray features that are differentially expressed between treatment groups. This method allows the use of small numbers of replicates to evaluate gene expression across a continuum of conditions, down to one array per condition if (as in the present study) many samples are pooled for each array (i.e. the array itself samples the mean value) and if multiple conditions are assessed. We used two arrays per condition, as recommended by the UCSD Microarray Core. We compared mice inoculated with NTHi or PBS for each time point against uninjected (0 hour) controls to generate sets of genes that change over time with NTHi or PBS injection. We also compared mice inoculated with NTHi versus PBS at each time point to generate sets of genes that are differentially regulated between the two conditions. Bonferroni multiple testing correction (α_Bonf < 0.05) was applied to identify only those genes with the most robust changes. Functional gene families were assessed by gene ontology (GO) analysis, and specific genes were overlaid on pathways using Genespring GX 7.3 (Agilent Technologies, Santa Clara, CA).

The IFNα's comprise a large family of highly similar genes, clustered on the same chromosome along with some pseudogenes, that are not well represented on the Affymetrix MU430 2.0 microarray: Affymetrix identifies two probes for IFNα2, two for IFNα11, three for IFNα12 and four for IFNα13. However, for human Affymetrix gene arrays, oligonucleotides for the IFNα family are reported to be incorrectly identified [34]. We therefore evaluated the IFNα probes present on the MU430 2.0 microarray by nBLAST against the mouse genome, and found similar problems with gene identification. Several IFNα probes aligned to IFNα isoforms different from those specified by Affymetrix, while other IFNα probes cross-aligned with several different IFNα's [see Additional File 1]. We therefore considered only those probes that matched a single IFNα gene. Probes were found to align uniquely to the IFNα2, 4, 5, 9, 11 and 14 genes, as well as to an unassigned IFNα gene similar to IFNα7.

As a separate evaluation of IFN signaling, we assessed the expression of IFN-responsive genes that have been identified as preferentially activated by type I IFNs (that is, IFNα's and IFNβ as opposed to IFNγ) by Baechler et al. [35], choosing the 9 genes shown in their study to be most strongly up-regulated in leukocytes by IFNβ.

As a negative control, we assessed the expression of several genes whose EST profiles on Genbank show expression highly restricted to the testis, reasoning that these genes are highly unlikely to play a role in OM: Cyct (testis cytochrome c); Tct3 (t-complex-associated testis-expressed 3); Dnajb3 (HSP40 homolog B3); Adam4 (a disintegrin and metaloprotease 4); Tcp1b (t-complex protein B) and Nkx2-6 (NK2 transcription factor-related, locus 6). None of these genes showed significant changes during OM.

As a positive control, we assessed genes reported in the literature to be strongly up-regulated in OM. All were significantly regulated in OM in the array data, at levels roughly similar to those previously reported: Muc4 (mucin 4, 5-fold at 48 and 72 hours) [36]; Muc5ac (10-fold at 24 and 48 hours) [37, 38]; Muc5b (5-fold at 24 hours) [36]; S100a9 (36-fold at 24 and 48 hours), S100a8 (20-fold at 24 and 48 hours), Fth1 (ferritin heavy polypeptide 1, 3-fold at 48 hours) [39].

Deeply anesthetized mice were perfused intracardially with 4% paraformaldehyde. MEs were harvested at 0, 6, 12 hours and 1, 2, 3, 5, 10, 14 or 21 days post inoculation, processed and sectioned. Sections were stained with haematoxylin and eosin (H&E), and micrographs were digitally recorded. Mucosal thickness and percent area of the ME lumen occupied by inflammatory cells were measured at standardized areas of the ME, and computer-averaged as described previously [28].

Bacterial presence was evaluated and analyzed from at least six WT, and TRIF-/- MEs per time point, by obtaining a sample from each for NTHi culture, as described previously [24]. Briefly, a 1 μl culture loop was used to sample the ME contents. If no fluid was present, the loop was used to scrape the ME mucosa to recover any adherent bacteria. Each loop was then used to separately streak four quadrants of a chocolate agar plate, which was incubated for 24 hours prior to evaluation of bacterial colonies. Positivity was judged based upon the presence of multiple NTHi colonies (>1) on one or more quadrants of the plate. The presence of NTHi was verified by Gram staining of colony-forming units, and by negative cultures on blood agar plates versus chocolate agar plates.

Except for gene array values as above, data were analyzed with ANOVA using StatView statistics software with Bonferroni correction, as described elsewhere [40]. Differences were considered significant at p ≤ 0.05.

We evaluated TRIF expression in the ME mucosa of WT mice by qPCR (Figure 2). Expression increased after administration of NTHi, reaching approximately 2.5-fold of that seen in the normal ME by 3 days after inoculation, remained elevated for 14 days, and recovered to baseline by 21 days.

To evaluate TRIF expression in mice with a deficit in a major activator of MyD88, we assessed TRIF mRNA expression in TLR2-deficient mice (Figure 2A). Since TLR4 can activate both MyD88 and TRIF, we also measured TRIF mRNA in TLR4-/- mice (Figure 2A). TRIF expression in the absence of NTHi was significantly greater in both knockout strains than in uninfected WT mice. Upon infection with NTHi, TRIF mRNA decreased in these TLR deficient animals. Within 24 hours, expression in the knockout and WT mice converged, and remained similar across the three strains over the course of 21 days.

The expression of genes relevant to type I IFN and TRIF signaling in the ME mucosa (see Figure 1) was evaluated by gene array in C57BL/6J:CB mice (Figure 3A–C, Additional Files 2-S). GO analysis revealed that genes related to type I IFN signaling were significantly regulated during the course of OM. In addition, individual genes encoding signaling molecules associated with TRIF showed up-regulation, which was greatest at 1 day after inoculation (Figure 3A, Additional File 2). Type I IFN receptors were similarly up-regulated. In contrast, the expression of most type I IFNs represented on the array showed relatively little change, with a modest tendency for down-regulation from 3 hours to 1 day after inoculation, and slight up-regulation at 2–3 days. An exception was an unassigned IFNα gene similar to IFNα7, which was sharply up-regulated at 3 and 6 hours. (Figure 3B, see Additional File 1). Genes that are normally induced by type I IFNs were very strongly up-regulated in the ME mucosa in a biphasic manner; with peaks at 6 hours and at 2–3 days after inoculation (Figure 3C, Additional File 3). None of the genes that were differentially regulated by NTHi were significantly altered in sham (saline-injected) MEs.

To assess the functional influence of TRIF, we investigated the ME response to NTHi in TRIF-/- mice. WT mice demonstrated hyperplasia of the ME mucosa, which reached a maximum thickness 2 days after inoculation with NTHi (Figures 4 and 5). Mucosal thickness remained elevated at 3 days, and then recovered to baseline by 5 days in WT mice. In TRIF-/- mice, there was a slightly greater thickness of the mucosa prior to NTHi administration, compared to WT mice. However, the additional increase in mucosal thickness induced by NTHi was more gradual, peaking at 3 days after inoculation, and recovered more slowly, not reaching baseline until 10–14 days. Moreover, peak thickness was less than in WT mice.

The infiltration of leukocytes into the ME lumen of WT mice (Figure 6) was minimal until 1 day after NTHi inoculation, and like mucosal thickness, reached a maximum at 2 days. Thereafter cellular infiltration decreased to reach baseline at 10 days. The infiltration of leukocytes in TRIF-/- animals was substantially reduced and delayed compared to WT animals. However, recovery to baseline had occurred by 5 days.

Bacterial clearance of the ME cavity was examined in TRIF-deficient mice and compared to that observed in WT mice (Table 1). In WT mice, 4 out of 6 (4/6) culture plates were positive at 1 day, increasing to 6/6 at 2 days, and decreasing to 3/6 at 3 days. Thereafter, no bacteria were recovered from WT MEs. TRIF-/- MEs were similar to WT MEs for 1, 2, and 3 days after NTHi administration, with 4/6, 5/6 and 3/6 MEs positive for bacterial colonization respectively. However, in contrast to WTs, bacterial colonies could be still detected at 5 and 7 days, with 1 positive plate out of 6. Cultures were negative thereafter. By comparison, the MEs of MyD88 mice were strongly positive through 21 days (data from Hernandez et al.) [24], and even at 42 days (our unpublished observation) 1/6 MEs remained positive. TLR4-/- MEs were culture-positive until 14 days.