Innate Immune Responses and Modified Extracellular Matrix Regulation Characterize Bacterial Infection and Cellular/Connective Tissue Changes in Scarring Trachoma

ABSTRACT Trachoma is the most common infectious cause of blindness and a major public health problem in many developing countries. It is caused by recurrent ocular infection with Chlamydia trachomatis in childhood, with conjunctival scarring seen later in life. The pathogenesis of trachomatous scarring, however, is poorly understood, and this study was carried out to investigate the immunofibrogenic correlates of trachomatous conjunctival scarring. A case-control study of 363 cases with conjunctival scarring and 363 control participants was conducted. Investigations included in vivo confocal microscopy (IVCM) assessment, quantitative real-time PCR gene expression, C. trachomatis detection, and nonchlamydial bacterial culture. Trachomatous scarring was found to be strongly associated with a proinflammatory, innate immune response with increased expression of psoriasin, interleukin-1β, tumor necrosis factor alpha, defensin-β4A, chemokine ligand 5, and serum amyloid A1. There was also differential expression of various modifiers of the extracellular matrix, including metalloproteinases 7, 9, 10, and 12, tissue inhibitor of matrix metalloproteinase 1, and secreted protein acidic cystein-rich-like 1. The expression of many of these genes was also significantly associated with the presence of nonchlamydial bacterial infection. These infections had a marked effect on conjunctival immune processes, including an increased inflammatory infiltrate and edema seen with IVCM. This study supports the possibility that the immunofibrogenic response in scarring trachoma is partly stimulated by nonchlamydial bacterial infection, which is characterized by the expression of innate factors.

T rachoma is a major cause of blindness, especially in developing countries. Children suffer recurrent episodes of ocular infection with Chlamydia trachomatis, which stimulates a follicular and papillary conjunctivitis. These children are subsequently at risk of developing conjunctival scarring, which can progress to entropion (in-turning of the eyelid) and trichiasis (eyelashes rubbing against the eyeball). Corneal opacity and blindness can result, as well as severe and persistent discomfort. Trachoma is the most common infectious cause of blindness, with at least 1.3 million people estimated to be blind from the disease, 8.2 million to have trichiasis, and 40 million to have active disease (40,52).
The pathogenesis of trachoma, and especially of the scarring process, is poorly understood. Histological studies of conjunctival biopsy specimens from scarred individuals show a chronic inflammatory cell infiltrate, especially in the substantia propria, with large numbers of lymphocytes (1,4,25,51). T cells, both CD4 ϩ and CD8 ϩ , tend to outnumber B cells. The conjunctival stroma is replaced by thick, mostly avascular scar tissue. Earlier studies examining lymphoproliferative responses to chlamydial antigens in trachomatous subjects were consistent with findings from animal models of infection, which suggested that a strong Th1 response with production of gamma interferon (IFN-␥) was important in clearing chlamydial infection and may be protective against the development of scarring (5,27,28). These studies also provided limited evidence that Th2 responses were associated with conjunctival fibrosis, which may have parallels with other infectious diseases, such as schistosomiasis (59). However, recent studies using microarray analysis of conjunctival swab samples from subjects with trachomatous trichiasis have not found evidence of Th1/Th2 polarization. Instead, proinflammatory mediators, particularly suggestive of innate immune responses and factors affecting extracellular matrix (ECM) remodeling, were prominent (14,29). A number of studies have also provided compelling evidence supporting the importance of conjunctival inflammation in the development of scarring and blinding complications (11,18,43,55,57).
Two models have been proposed to account for the pathogenesis of C. trachomatis-induced scarring in the eye or genital tract: the "immunological" and the "cellular" paradigms, which are not necessarily mutually exclusive (7,16,53). The immunological paradigm argues that cellular immune responses, especially those involving T cells, against specific chlamydial antigens are important in causing disease. The cellular paradigm proposes that host epithelial cells act as a key innate responder cell and are central in driving tissue damage.
In this case-control study we investigated the pathophysiology of trachomatous conjunctival scarring by measuring conjunctival gene expression. We examined the hypotheses that conjunctival scarring is associated with (i) a predominantly Th2 (interleukin-13 [IL-13] mediated) rather than Th1 response, (ii) various profibrotic mediators, such as matrix metalloproteinases, and (iii) markers of innate immunity. In addition, we investigated the relationship between the expression of these various factors and both C. trachomatis and other bacterial infections. Finally, we related the gene expression profile to the microscopic tissue morphology changes and inflammatory cell infiltrate observed by in vivo confocal microscopy (IVCM). IVCM is a relatively new technique to examine the ocular surface, which provides highresolution images down to the cellular level (31,32).

MATERIALS AND METHODS
Ethical approval. This study adhered to the tenets of the Declaration of Helsinki. It was approved by the ethics committees of the Tanzanian National Institute for Medical Research, the Kilimanjaro Christian Medical Centre, and the London School of Hygiene and Tropical Medicine. The study was explained to potential study subjects and written, informed consent was obtained before enrollment.
Participant recruitment. The recruitment of participants into this population-based study has been previously described (30,32). Briefly, adults with trachomatous conjunctival scarring, but without trichiasis, were recruited from an area of trachoma endemicity in the Siha district of northern Tanzania. Control subjects without scarring from the same community, frequency matched for ethnicity, were also recruited.
Clinical examination and sample collection. All participants were examined clinically, as previously described, using the 1981 World Health Organization trachoma grading system with some modifications, including a more detailed grading for conjunctival scarring (17,30,32). The left conjunctiva was anesthetized with preservative-free proxymetacaine 0.5% eye drops (Minims; Chauvin Pharmaceuticals) and a swab collected for microbiological analysis from the inferior fornix. Two upper tarsal conjunctival swabs were also collected (Dacron polyester-tipped; Hardwood Products Company, Guildford, ME). The first was for RNA isolation (described later) and was placed directly into a tube containing 0.2 ml of RNA stabilizer (RNAlater; Life Technologies, United Kingdom). The second was for C. trachomatis detection and put into a dry tube. Samples were kept on ice packs until frozen later the same day at Ϫ80°C.
Confocal microscopy assessment. IVCM examination of the upper tarsal conjunctiva was performed using a Heidelberg Retina Tomograph 3 (HRT3) in combination with the Rostock Corneal Module (RCM) (Heidelberg Engineering GmbH, Dossenheim, Germany) according to previously described examination and image grading protocols (31). IVCM images were graded for inflammatory features and subepithelial connective tissue organization/scarring ( Fig. 1 and 2).
Total RNA was extracted from the swab sample using an RNeasy microkit (Qiagen, Hilden, Germany) and reverse transcribed to cDNA using the QuantiTect reverse transcription kit (Qiagen) according to the manufacturer's instructions. Multiplex real-time quantitative PCR was performed on a Rotor-Gene 6000 (Corbett Research, Cambridge, United Kingdom) using the QuantiTect Multiplex NoROX kit (Qiagen) according to the manufacturer's instructions. Probes and primers for multiplex assays, which included up to four separate targets (including HPRT-1, which was used as a reference gene), were designed and synthesized by Sigma Life Science (www.sigma.com/designmyprobe; available on request). The samples were tested in duplicate, in a total reaction mixture volume of 25 l, which contained 2 l of sample or standard. The thermal cycle protocol used the following conditions: 95°C for 15 min, followed by 45 to 50 cycles of denaturation at 94°C for 30 s and then annealing and extension at 60°C for 30 s. Fluorescence data was acquired at the end of each cycle. The relative efficiency of the component reactions was assessed using standards containing all targets in a sequence of 10-fold serial dilutions.
Microbiology samples and analysis. C. trachomatis DNA was detected using a PCR-based assay (Amplicor CT/NG Test; Roche) with previously described modifications (12). Samples for culture were inoculated onto blood and chocolate agar later on the day of collection (rarely more than 6 h) and incubated at 37°C for 48 h. Culture isolates were identified by standard microbiological techniques.
Data analysis. Data were entered into Access 2007 (Microsoft) and analyzed using STATA 11.0 (StataCorp LP, TX). The transcript abundances for the genes of interest were standardized relative to that of HPRT1 in the same reaction mixture using the ⌬⌬C T method and were normalized by log 10 transformation (37). Linear regression models were used to estimate the fold changes (age and sex adjusted) in the levels of gene expression in those with clinically visible scarring compared to controls, subdivided into those with no/minimal inflammation (TS [inflammation grades P0 and P1]) and those with inflammation (TSI [inflammation grades P2 and P3]), where P0 is no inflammation, P1 is minimal inflammation with individual papillae prominent, P2 is moderate inflammation in which normal vessels appear hazy, and P3 is pronounced inflammation in which normal vessels on the tarsus are hidden over more than half the surface) (17). Similar fold changes were estimated for the presence of clinically significant inflammation (P2 or P3), adjusted for the level of clinical scarring as well as age and sex. Age-and sex-adjusted fold changes in gene expression by grade of IVCM connective tissue organization/scarring and for the presence of dendritiform cells were also estimated. Correlation coefficients were estimated between the increase in gene expression (log 10 transformed) and the IVCM inflammatory infiltrate, with the significance tested using age-and sex-adjusted linear regression models. Likelihood ratio tests were used to assess the strength of association of each factor with the gene expression fold change. P values for all associations are shown in Results. For reference, the P values for the critical significance thresholds determined by a Bonferroni correction for multiple comparisons are shown in the table footnotes. However, the choice of targets for gene expression was not selected at random, as is assumed by the Bonferroni method, but contain related groups of genes which might be expected to act in a similar manner. In addition, using such a conservative correction increases the likelihood of a type 2 error.

Study participants.
We recruited 363 cases with trachomatous conjunctival scarring and 363 control participants without scarring. The demographic, clinical, IVCM, and nonchlamydial bacterial infection data have previously been reported in detail (30,32). In summary, controls were younger than the cases (mean 31.9 versus 50.3 years, P Ͻ 0.001), and all analyses were age and sex adjusted. Conjunctival scarring in the cases was mostly mild to moderate and cases had significantly more clinical inflammation than controls. IVCM analysis showed that the cases had more inflammatory cells and a higher connective tissue scarring score. Nonchlamydial pathogenic organisms were cultured more frequently from cases (25 [6.9%]) than from controls (7 [1.9%], P ϭ 0.002). The most frequently cultured pathogenic organisms were Haemophilus influenzae type b and Streptococcus pneumoniae, while coagulase-negative staphylococci, Corynebacterium spp., Streptococcus viridans, and Bacillus spp. were designated as commensal organisms in this analysis. Five (1.4%) of the cases and none of the controls had C. trachomatis infection detected.
Conjunctival gene expression. Quantitative, real-time RT-PCR was performed for 23 gene expression targets for all 363 cases and 363 controls. There was inadequate sample for detection in four samples (three cases and one control), and amplification failed in a number of experiments for technical reasons related to laboratory power supply problems. The number of samples suc- cessfully detected for each target ranged from 313 to 359 for cases and 316 to 362 for controls.
Gene expression levels in relation to clinical findings. Cases with scarring had enriched gene expression compared to controls for the majority of transcripts measured, even after adjusting for age and sex (Table 1). This tended to be more marked in those cases with inflammation (increased: INDO, TNFA, IL1B, S100A7, DEFB4A, CXCL5, SAA1, ARG, NOS2, MMP7, MMP9, MMP12, TIMP1, and CD83; decreased: MMP10, CFH, and SPARCL1). There was also some evidence, but with less robust P values, that the expression of IFNG and IL13RA2 were increased in cases with scarring.
After adjusting for age, sex, and the level of clinical scarring, a number of genes were also differentially expressed in those with clinical inflammation compared to those without inflammation Gene expression in relation to IVCM findings. The relationships between the gene expression results and the IVCM findings are presented in Table 2. Increasing IVCM connective tissue grade was associated with increased expression of S100A7 (which also had the largest fold changes), TIMP1, and CFH and with decreased expression of SPARCL1. There was also some evidence, but with less-robust P values, that the connective tissue grade was associated with increased levels of INDO, IL1B, SAA1, ARG1, MMP12, and CD83.
The expression of a number of genes was correlated with the IVCM inflammatory infiltrate density. The correlation coefficient for IL1B, S100A7, and DEFB4 was greater than ϩ0.40 clearly defined bands of tissue which constitute less than 50% of the area of the scan. (D) Grade 3: clearly defined bands or sheets of tissue which constitute 50% or more of the area of the scan and in which striations are present. If different grades of scarring are seen within a particular volume scan then the highest grade is recorded. The connective tissue which is graded needs to be separate from that associated with the vascular tissue; if this is not possible then the scan is considered ungradable. (Fig. 3). For SAA1, MMP9, and MMP12 it was between ϩ0.25 and ϩ0.35, and for MMP10 and SPARCL1 it was between Ϫ0.20 and Ϫ0. 30. The presence of DFCs was associated with increased expression of INDO and S100A7 and with decreased expression of SPARCL1.
Gene expression levels in relation to microbiology findings. Infection with C. trachomatis was associated with upregulated IFNG, INDO, and NOS2 and with downregulated MMP10 and SPARCL1 (Table 3; note that only five participants had detectable C. trachomatis DNA). There was also some evidence, but with less-robust P values, that C. trachomatis infection was associated with increased IL12B, MMP1, MMP7, and MMP10. Nonchlamydial bacterial infection was associated with increased expression of INDO, S100A7, DEFB4A, and MMP12 and with decreased expression of MMP10, SPARCL1, and CFH (Table 3). There was also some evidence that nonchlamydial bacterial infection was associated with increased IL1B, IL-10, CXCL5, SAA1, MMP7, and MMP9.
IVCM findings in relation to nonchlamydial bacterial infection status. A new combined analysis of previously published IVCM and microbiology data found that pathogenic organisms were associated with significantly more inflammatory cells (P Ͻ 0.0006) and with the presence of tissue edema (P ϭ 0.02) (Table 4) (30,32).

DISCUSSION
Previous studies have used gene expression measurement of conjunctival swab samples to study the pathogenesis of trachoma (6,9,14,21,22,29,46). However, this is the first study to systematically compare a large number of cases with trachomatous scarring and control subjects in which the scarring had not yet progressed to trichiasis. Cases with trichiasis are quite distinct from cases with scarring alone since they have a persistent foreign body abrading the globe stimulating inflammation and facilitating secondary infection (13). In the present study we have shown that scarring without trichiasis is associated with large fold changes in gene expression of various cytokines, chemokines, and other immunofibrogenic mediators, even though the scarring was generally mild to moderate.
Serum amyloid A, of which SAA1 is a main isoform, is an acutephase protein which is raised in inflammatory diseases and has been shown to have antibacterial effects (19,39). The upregulation of these various proinflammatory, chemotactic, and antimicrobial mediators suggests an important role for the innate immune response in trachomatous scarring. This is consistent with the cellular paradigm of Chlamydia trachomatis pathogenesis. While there was some evidence of Th1 response in scarring (increased IFNG and INDO expression), there was no significant difference in IL-13 expression, which might have suggested a Th2 response. There was some evidence that the expression of IL13RA2, which appears to act as a decoy receptor for IL-13 and opposes its function, was increased in cases compared to controls (60).
Many of the measured transcripts were upregulated in the presence of clinical inflammation. When scarred cases were compared to controls without scarring, the additional presence of inflammation in the cases tended to lead to more marked increases in gene expression. These observations are consistent with the hypothesis that episodes of inflammation are important in the pathogenesis of scarring and blindness. It is also interesting that a number of genes, for example, S100A7, IL1B, SAA1, and MMP12, showed large fold changes in gene expression even in relatively noninflamed cases with scarring compared to controls. These genes also showed a clearly progressive increase in expression with increasing IVCM connective tissue organization grade. We found infection with C. trachomatis to be infrequent in these adults with conjunctival scarring, with a similar detection rate to previous studies of trachomatous trichiasis (8,11,13). While caution is needed in interpreting these data because of the small numbers, individuals infected with C. trachomatis had gene expression profiles characteristic of C. trachomatis infection in children (9,46). There was a typical Th1-dominated response, with upregulated IFNG and IL12B (P ϭ 0.05) and also INDO and NOS2 (both IFN-␥ regulated) and some acute-phase reactants. Infection with other bacteria was more frequent, and many of the genes described above involved in innate immunity showed moderate to large increases in expression if a pathogenic organism was present. A previous study found that the presence of bacterial infection was associated with increased expression of IL1B, TNF-␣, MMP1, MMP9, and TIMP2 after trichiasis surgery (10).
In the present study, bacterial infection was also found to be associated with an increased inflammatory infiltrate and tissue edema with IVCM. Overall, these studies support the hypothesis that nonchlamydial bacterial infection is important in driving the scarring process. Matrix metalloproteinases are capable of degrading the protein components of the ECM and also have wide-ranging effects on inflammatory and immune responses (50,58). Previous studies have implicated MMPs in the pathogenesis of trachomatous scarring, and this was further confirmed in the present study in a separate population (9,10,14,20,29,44). MMP7, MMP9, MMP12, and TIMP1 were upregulated in scarring and inflammation, whereas MMP10 and SPARCL1 were downregulated. SPARCL1 is a matricellular protein, which regulates the synthesis  and turnover of the ECM (35,54). The gene expression levels by clinical scarring grade were broadly consistent with the levels by IVCM connective tissue grade, and the expression levels by clinical inflammation were consistent with the IVCM inflammatory infiltrate correlation. The differential gene expression of these ECM modifiers was closely related to bacterial infection status as discussed above. Strengths of this population-based study were that we studied a large number of cases with early scarring trachoma to analyze gene expression changes in relation to the clinical, IVCM, and microbiological status. There are, however, a number of limitations. Gene expression levels do not necessarily reflect the level of functional protein, exemplified by transforming growth factor ␤, which has significant posttranscriptional regulation. Conjunctival swabs obtain material from the tissue surface and so have a tendency to limit the observations to events in or near the epithelium. To address these two issues, we are currently comparing gene expression with immunohistochemical analysis of conjunctival biopsy tissue.
This study showed that mild to moderate trachomatous scarring is strongly associated with a proinflammatory, innate immune response and with differential expression of various modifiers of the ECM. We found no evidence for an active role for IL-13/Th2 responses at this stage of the disease. This is consistent with the findings from a microarray transcriptome study from Ethiopia on advanced trachomatous scarring (14). We were also able to use IVCM to show differential transcript levels according to connective tissue morphology and inflammatory cell infiltrate.
A key determinant in the expression of many of these genes appears to be the presence of nonchlamydial bacterial infection which, as well as causing inflammation, may contribute to the scarring process.