Comparative Expression Profiling of the Chlamydia trachomatis pmp Gene Family for Clinical and Reference Strains

Background Chlamydia trachomatis, an obligate intracellular pathogen, is a leading worldwide cause of ocular and urogenital diseases. Advances have been made in our understanding of the nine-member polymorphic membrane protein (Pmp) gene (pmp) family of C. trachomatis. However, there is only limited information on their biologic role, especially for biological variants (biovar) and clinical strains. Methodology/Principal Findings We evaluated expression for pmps throughout development for reference strains E/Bour and L2/434, representing different biovars, and for clinical E and L2 strains. Immunoreactivity of patient sera to recombinant (r)Pmps was also determined. All pmps were expressed at two hours. pmpA had the lowest expression but was up-regulated at 12 h for all strains, indicating involvement in reticulate body development. For pmpD, expression peaked at 36 h. Additionally, 57.7% of sera from infected and 0% from uninfected adolescents were reactive to rPmpD (p = 0.001), suggesting a role in immunogenicity. pmpF had the highest expression levels for all clinical strains and L2/434 with differential expression of the pmpFE operon for the same strains. Sera were nonreactive to rPmpF despite immunoreactivity to rMOMP and rPmpD, suggesting that PmpF is not associated with humoral immune responses. pmpFE sequences for clinical strains were identical to those of the respective reference strains. We identified the putative pmpFE promoter, which was, surprisingly, 100% conserved for all strains. Analyses of ribosomal binding sites, RNase E, and hairpin structures suggested complex regulatory mechanism(s) for this >6 Kb operon. Conclusions/Significance The dissimilar expression of the same pmp for different C. trachomatis strains may explain different strain-specific needs and phenotypic distinctions. This is further supported by the differential immunoreactivity to rPmpD and rPmpF of sera from patients infected with different strains. Furthermore, clinical E strains did not correlate with the E reference strain at the gene expression level, reinforcing the need for expansive studies of clinical strains.


INTRODUCTION
Chlamydia trachomatis is an obligate intracellular pathogen that is responsible for significant worldwide morbidity associated with ocular and sexually transmitted diseases (STD). The developmental cycle of the organism is biphasic beginning with the adhesion of the elementary body (EB), an infectious and metabolically inert form, to the host cell. After endocytosis, the EB differentiates ,2 to 12 h post infection (p.i.) into a larger, non-infectious and metabolically active reticulate body (RB), which initiates intracellular replication by binary fission within a vacuole called an inclusion body. At ,30 to 36 h p.i., RBs reorganize into new EBs, which are released by host cell lysis or exocytosis at 48 to 72 h p.i. that initiates another infectious cycle [1,2].
The disease spectrum of C. trachomatis ranges from conjunctivitis and ocular trachoma to tubal factor infertility, ectopic pregnancy and infant pneumonitis [3,4]. C. trachomatis serological variants (serovars) are grouped within two human biological variants (biovars) according to characteristics of disease presentation: the trachoma biovar, including serovars A to C and Ba, which cause conjunctivitis and trachoma, and serovars D to K and Ba, Da, Ia and Ja, which cause conjunctivitis, urogenital infections and infant pneumonitis, and the lymphogranuloma venereum (LGV) biovar (serovars L 1 to L 3 and L 2a ). The latter serovars are more invasive, causing genital ulceration, lymphadenitis and proctitis [3,5]. However, serotyping of the major outer membrane protein (MOMP), and phylogenetic reconstructions of this protein and the corresponding gene (ompA) [6,7] do not group serovars by trachoma, non-invasive urogenital or invasive LGV disease groups.
The molecular mechanisms behind these biological differences among serovars (or strains) are not well understood. Genome sequences of reference strains D/UW-3 [8] and A/Har-13 [9], as well as ongoing C. trachomatis genomic sequencing are providing information on specific genes and proteins that may explain tissue tropism and virulence differences for the three disease groups. C. trachomatis contains a partial tryptophan operon (trpRBA) where urogenital strains, but not trachoma strains, can synthesize tryptophan from mucosal substrates [10]. The toxin gene possesses an intact N-terminal region that encodes an active enzymatic domain for the urogenital strains but not for trachoma or LGV strains [11].
Research on the nine member polymorphic membrane protein (Pmp) gene (pmp) family has revealed phylogenetic reconstructions where six pmps (pmpB, pmpC, pmpF, pmpG, pmpH and pmpI) form clades that correspond to the three disease groups [12][13][14]. At the gene expression level, previous RT-PCR analyses of reference strains D/UW-3 and L 2 /434 [15], and microarray analysis of D/ UW-3 [16] found that all nine pmps were transcribed starting at 8 h p.i. Yet, based on real-time quantitative (k)RT-PCR, we found expression as early as 2 h p.i. for pmpC for reference strains Ba/ Apache-2, G/UW-57 and L 2 /434, and a differential expression profile with earlier up-regulation of pmpC for L 2 /434 [17]. Also, another study based on kRT-PCR, Kiselev et al. [18] detected pmpD expression as early as 2h p.i. for L 2 /434. Proteomics analyses have also shown that all Pmps of L 2 /434 are detected as outer membrane constituents [19][20][21][22]. There is also evidence that some Pmps are antigenic for human sera [15,23]. We observed a heterogeneous immunoreactivity to recombinant (r)PmpC using sera from patients infected with different C. trachomatis strains, suggesting a role for PmpC in antigenic variation [17]. More recently, Pmps have been considered autotransporters based on bioinformatics analyses [24,25]. Wehrl et al. [26] has experimentally demonstrated the autotransporter model for the C. pneumoniae ortholog of C. trachomatis PmpD, Pmp21. Further, using immunofluorescence microscopy, Western blotting and penicillin treatment, the results of Kiselev et al. [18] for L 2 /434 PmpD are in general agreement with the autotransporter model for this protein.
PmpD has also been shown to be a species-common neutralizing antigen [27], while PmpF has been implicated as a potential target of the host immune response as it contains several predicted major histocompatibility (MHC) epitopes within the N-terminal domain [9].
Despite the potential importance of Pmps in chlamydial biology, there is a lack of expression data for the pmp genes as well as an insufficient understanding of the host immune response to their proteins. Here, we profile the expression of all pmps throughout development for reference strains E/Bour and L 2 /434, representing the two C. trachomatis biovars. We chose E/Bour because it is the most prevalent strain worldwide, although the mechanisms of its ecological success are not yet understood. L 2 /434 was selected as it has been widely studied with a plethora of biological information for comparative analyses. The biological uniqueness of these two strains in vivo is reflected in their differential tissue tropism, virulence and disease presentation. In light of our recent findings that reference strains do not represent the same genetic composition of clinical strains that are circulating among human populations today [28], we also compared the nine pmp expression levels for four C. trachomatis clinical strains, representing ompA genotypes of E and L 2 . Further, we examined the immunoreactivity of sera from adolescents with and without C. trachomatis urogenital infections against rPmps to further define their potential importance in human disease.
MATERIALS AND METHODS C. trachomatis cell culture of reference strains and clinical strains C. trachomatis reference strains E/Bour and L 2 /434, three clinical strains belonging to ompA genotype E (designated as E/537C-05, E/S-141 and E/CS-500-96) and one clinical strain belonging to ompA genotype L 2 were evaluated in this study. E/537C-05 and E/ S-141 were collected from patients with vaginal discharge, E/CS-500-96 from a patient with cervicitis, and L 2 from a patient with proctitis. Each was propagated in HeLa 229 cell monolayers using standard techniques as previously described [4,29]. EBs were harvested at 48-72 h p.i. and purified by discontinuous density centrifugation in Renografin [30].
Confluent HeLa cells were either mock-infected or infected with a multiplicity of infection of one for each reference strain or clinical strain in SPG prior to incubation with culture medium [4,29]. Eight T 25 flasks (one for each time point of 2, 6, 12, 18, 24, 36, and 48 h and mock-infected) per strain were inoculated and placed at 37uC in 5% CO 2 [17]. Cultured cells were harvested at each time point, and total RNA was extracted as previously described [17].

Reverse Transcription and Quantitative Real-Time PCR (kPCR)
RNA was quantified by O.D. measured at A 260 . cDNA was generated from 500 ng of each RNA sample using TaqMan RT Reagents (Applied Biosystems, Foster City, CA) and random hexamers, and was quantified by O.D. measured at A 260 .
Quantitation of pmp expression was achieved using the ABI 7000 SDS (Applied Biosystems), SYBR Green chemistry, and the standard curve method for relative quantitation, using reagents and thermocycling as previously described [17]. 16SrRNA was used as the endogenous control since normalizing the data against 16SrRNA provides a control for the number of organisms (EBs and RBs) and, therefore, for the differential growth rate of each strain. ompA was included as a quality control for kRT-PCR results since it has been widely used for gene expression studies [16,17].
Primers for each of the nine pmps (Table 1) were designed using Primer Express (Applied Biosystems). Primers for ompA, 16SrRNA, and pmpC were used as previously designed (Table 1) [17].
Each plate contained two replicates of each sample cDNA, three different negative controls and standard curves for each gene as previously described (17). For all experiments, the amount of target and control gene was determined from the respective standard curve by conversion of the mean threshold cycle values. Normalization was obtained by dividing the quantity of the target gene by the quantity of the control gene. The specificity of the amplified products was verified by analysis of the dissociation curves generated by the ABI7000 software based on the specific melting temperature for each amplicon. The results were based on three independent experiments for reference strains E/Bour and L 2 /434, and for the four clinical strains.

Genetic analysis of the pmpFE operon for C. trachomatis reference and clinical strains
Based on the considerable expression disparities between pmpF and pmpE (which belong to the same operon) for reference strain L 2 / 434 and mostly for the clinical strains (see results below), we sequenced the pmpFE operon as well as the upstream 164 base pair (bp) pmpG/pmpF intergenomic region (IGR) that likely contains the operon regulatory region. In the C. trachomatis chromosome, pmpF and pmpE are located on the minus strand; pmpF is located upstream of pmpE, with a 2 bp IGR . pmpE was sequenced for the six strains (Genbank Accession Numbers EF490370 for E/537C-05, EF490371 for E/CS-500-96, EF490372 for E/S-141, EF490373 for L2, EF490374 for E/Bour, and EF490375 for L 2 /434), while pmpG/pmpF IGR and pmpF were sequenced only for the clinical strains (Genbank Accession Numbers EF490366 for L2, EF490367 for E/CS-500-96, EF490368 for E/S-141, and EF490369 for E/537C-05), as the sequences for the reference strains were available from our previous study (GenBank Accession Number AY887650 for E/Bour and AY887660 for L 2 / 434) [12]. The amplification and sequencing strategies were performed as previously described [12], except for the pmpG/pmpF IGR, where we used primer 59-ACT CGG ATC TCC TAT AAC AG-39 for sequencing.
Since the transcription process can be strongly affected by the structure and sequence variability of promoter regions [31][32][33][34], a putative promoter search for the pmpFE operon was performed using EditSeq software (DNASTAR, Madison, WI) for sequences described in the literature and also by using two promoter prediction programs: http://www.fruitfly.org/seq_tools/promoter.html and http://www.prodoric.de/vfp/vfp_promoter.php. Given the expression dissimilarities obtained in this study for pmpF and pmpE, we searched for putative Shine-Dalgarno ribosome binding sequences (RBS) [35] as well as previously described chlamydial RBS [36][37][38][39] within this operon, since ribosomes can either prolong or shorten the lifetime of mRNA in response to events that occur during translation or termination processes [40]. The existence of putative consensus cleavage sites for RNase E [41][42][43], the major endonuclease that generally initiates mRNA degradation in most bacteria [44], was also examined within the pmpFE operon. These two analyses were performed using EditSeq software (DNASTAR). Putative stemloop structures were searched throughout the pmpFE operon using GeneQuest software (DNASTAR) and RNAstructure software version 4.4 (http://rna.urmc.rochester.edu/rnastructure.html) due to the regulatory or processing role of stem-loop structures in premature transcription termination as well as in mRNA degradation and maturation mechanisms [45][46][47][48][49], respectively.

Immunoreactivity of patient sera against Pmp fusion proteins
We generated fusion proteins for PmpD and PmpF because the latter displayed such high mRNA expression for L 2 /434 and the clinical strains, and the former was expressed late in development for all strains under study, being the last up-regulated protein for four of the six strains analyzed. Also, PmpD has been associated with neutralizing epitopes [27]. The rMOMP fusion protein was available from a previous study [50]. The PET30 expression system (EMD Biosciences, San Diego, CA) was used for cloning PCR products containing pmpD or pmpF generated from strain E/ Bour genomic DNA as we have described [17]. The forward and reverse primers were 59-GACGACGACAAGATGAGTTCCGA-GAAAGATATA-39 and 59-AATGCTGGATTGCGATTGATC-TTTTAACCGGGCTTCTCCTC-39 for pmpD, respectively, and 59-GACGACGACAAGATGATTAAAAGAACTTCTCTA-39 and 59-AATGCAGGAGGAGCTCTGGTCTTTTAACCGGGCTTC-TCCTC-39 for pmpF, respectively. Sequencing confirmed that the insert was in frame with the S-tag and His-tag as we have described previously for rPmpC [17]. The clones were transformed into E. coli BL21 and induced using 0.1mM IPTG during the exponential growth phase. Ni-agarose (Sigma, St. Louis, MO) was used for fusion protein purification according to the package insert. Recombinant proteins were determined to be the correct molecular weight (calculated at ,160.6 kDa for rPmpD and ,112.3 kDa for rPmpF) by immunoblot using AP-conjugated S-protein, which binds to the S-tag peptide with a distinct band at the correct molecular weight for each. Optimal protein concentrations were determined and standardized using BCA (Pierce, Rockford, Ill) before analyzing the clinical sera. The optimal protein concentration for rPmpD was 50 ng and 100 ng for rPmpF. Sera from 39 consented female adolescents 14 to 19 years of age attending clinics in Oakland, CA, were used at a 1:50 dilution for immunoblotting as described previously [17].

Real-time quantitation for pmp expression
The results of specificity assays revealed no non-specific products, and indicated the presence of the expected amplicon for each gene. Standard curves for all 11 genes had slope values between -3.1 and -3.5, which represents efficiencies between 93 and 100%. There were only minor variations in the slope for each standard curve among independent experiments, indicating a highly reproducible kPCR as we have also shown in previous experiments [17]. We defined the gene expression profile as the qualitative gene expression pattern throughout development where quantitative values are not considered. For example, one expression profile would show increasing expression up to a peak with tapering down of the expression after the peak. An expression peak was defined as the time point of the highest relative mRNA value. All quantitative expression comparisons refer to differences between the expression peak of each gene, even those occurring at different time points p.i.
Expression profile of the nine pmp genes throughout development for E/Bour and L 2 /434 L 2 /434 had strikingly different mRNA levels among some pmps and also between different time points for the same pmp (Fig. 1A). pmpF had the highest relative mRNA expression, up to 11.5-fold higher than for pmpA, the least expressed gene. mRNA levels were detected at 2 h p.i. for all pmps, including pmpA where the scale limits visualization of the low mRNA expression, and peaked at different time points. For all pmps except pmpA, mRNA levels decreased consistently after the peak until 48 h.
The pmp expression profiles for E/Bour were more homogeneous than for L 2 /434, and, in some cases, mRNA levels were lower than for the corresponding L 2 /434 pmp (Fig. 1B). Similar to L 2 /434, mRNA levels were detected at 2 h p.i. for all pmp genes. pmpE, pmpF and pmpG showed the highest expression levels for strain E/Bour with up to 3.1-fold higher mRNA levels than for pmpA and pmpI, the least expressed genes. As for L 2 /434, pmpA and pmpD were the earliest and the latest up-regulated genes, respectively. In contrast to L 2 /434, all pmp genes except pmpA and pmpI had stable mRNA levels after the expression peak until 48 h. For this reason, the expression peak for E/Bour pmps was defined as the time point at which a noticeable expression increase occurred.
For both reference strains, ompA had remarkably higher mRNA values at all time points than for the pmps (Fig. 1). In contrast to most pmps, ompA revealed a similar gene expression profile for both reference strains.
Expression profile of the nine pmp genes throughout development for C. trachomatis clinical strains The four clinical strains had a similar pmp expression profile (Fig. 2), which showed decreasing mRNA levels after the expression peak to 48 h. mRNA levels were detected at 2 h p.i. for all pmps, although the scale limits visualization. Overall, pmps peaked at 18h for L 2 ( Fig. 2A) and E/CS-500-96 (Fig. 2D), and at 36h for clinical strains E/537C-05 (Fig. 2B) and E/S-141 (Fig. 2C). Similar to reference strains E and L 2 , pmpA was the first upregulated gene for all clinical strains. In addition, pmpD was the last up-regulated gene for clinical strains L 2 and E/CS-500-96 ( Fig. 2A  and 2D), and was also expressed late in development (together with other pmps) for the other two clinical E strains under study.
pmpF had the highest expression among all of the pmps for the clinical strains (Fig. 2). In fact, there was a 27-fold higher expression of pmpF compared with the least expressed gene (pmpD) for L 2 . For clinical E strains, there was a 19.2-and 22.6-fold higher expression of pmpF compared with the least expressed gene (pmpI) for E/537C-05 and E/CS-500-96 respectively, and a 54.2-fold higher expression than pmpA for E/S-141 Although no relevant dissimilarities were observed for pmpF between clinical L 2 and L 2 / 434, there were considerable expression differences among the clinical strains and E/Bour with up to 11.7-fold higher mRNA values for E/CS-500-96 than for E/Bour.
For ompA, mRNA levels peaked at 36 h for E/537C-05 and E/ S-141, and at 18h for E/CS-500-96 and L 2 , declining thereafter (Fig. 2). The most striking example of differential mRNA levels between ompA and pmps occurred for E/S-141, where ompA had a 232.0-fold higher value compared with the least expressed gene, pmpA. However, all clinical strains except E/CS-500-96 had lower ompA expression levels for all time points compared with the corresponding reference strains.

Genetic analysis of pmpFE operon for C. trachomatis reference strains and clinical strains
The pmpF, pmpE and pmpG/pmpF IGR sequences for the three clinical E strains were 100% similar to the corresponding E/Bour sequences, while L 2 showed 4 nucleotide (nt) differences to L 2 /434 but only for pmpE. Compared to both L 2 strains, the four E strains showed 317 (10.2%) nt and 106 (10.3%) amino acid (aa) differences for pmpF as well as 56 (1.9%) nt and 21 (20 to L 2 ) (2.1%) aa differences for pmpE. For the pmpG/pmpF IGR, which comprises the ,164 bp upstream region of pmpF, there were 5 nt differences between the L 2 and the E strains, although none of them fell within the putative promoter region for the pmpFE operon (Fig. 3). The putative promoter is located within a 100% conserved stretch of the pmpG/pmpF IGR for both reference and all clinical strains (Fig. 3). The -10 promoter element (TAAAAT) identified in this study was identical to the one that was previously characterized for the L 2 /434 and D/UW-3 ltuB promoter, while the -35 region (TTGCAT) was 100% similar to the hctA promoter of the same chlamydial reference strains [32].
Analysis of the pmpFE operon sequence revealed several putative hairpin loop structures although the actual RNA folding in those regions functioning as a classic rho-independent type transcriptional terminator [49] cannot be assumed. At least 41 putative RNase E cleavage sites were identified throughout the pmpFE operon, 13 of which were not conserved between L 2 /434 (and L 2 ) and the four E strains (Fig. 4). One of these non-conserved sites involved the pmpF/pmpE IGR, and is specific for the E strains. The search for an RBS revealed a perfect prokaryotic Shine-Dalgarno sequence (AGGAGG) located 17 nts upstream of the start codon of pmpE, which is approximately 3000 bp below the last bp in Figure 3 and, therefore, is not shown. This RBS is in close proximity to the above-described putative RNase cleavage site shared only by the four E strains. However, the best approach for a putative RBS sequence for pmpF has two mismatches when compared with the ones described in the literature, and is unusually distant from the start codon (Fig. 3). Two additional putative RNase E cleavage sites, one of which was in close proximity to this RBS, were identified within the pmpF regulatory region (Fig. 3). Table 2 shows the clinical and microbiologic characteristics of the 39 adolescents enrolled in the study and the results of their serum immunoreactivity to rPmpD and rPmpF. All sera from patients infected with chlamydial clinical strains Ba, E, F and K (n = 15; 57.7%), but none with D, Ia, J or G (n = 11; 42.3%), were reactive to rPmpD while sera from uninfected patients were nonreactive with rPmpD (p = 0.001). Figure 5 shows the immunoblot results of representative sera from patients infected with Ba, D, E, F, G, Ia, J and K to rPmpD. Because pmpD is highly conserved among all reference strains [12], constructing rPmpD using the pmpD sequence of reference strain E/Bour should not have contributed to the observed differences in immunoreactivity. Further, crossreactivity between strains was unlikely since the patients were infected with only a single strain, and sera that were reactive to rPmpD were not reactive to rPmpF. In our previous study, sera form the same individuals infected with clinical strains D, E and G reacted with rPmpC [17]. Surprisingly, none of the sera reacted to rPmpF ( Figure 5), not even sera from the eight patients infected with strain E, although all sera from infected patients and one uninfected patient reacted with rMOMP as previously shown [17].

DISCUSSION
In this study, we determined the gene expression profile of the nine pmps throughout development for reference strains L 2 /434 and E/ Bour, and four clinical strains belonging to ompA genotypes E and L 2 . The reference strains had significant gene expression differences where E/Bour had relatively lower mRNA levels and generally sustained expression from 24 to 48 h compared with L 2 / 434 (Fig. 1). Surprisingly, in contrast to clinical L 2 , the three clinical E strains showed a dissimilar pmp expression profile to E/ Bour (Fig. 2). These remarkable expression dissimilarities are generally supported by our recent comparative genomics findings where the laboratory adapted reference strains did not reflect the same genetic make-up of strains that are circulating among human populations today and currently exposed to immune selection [28].
It is well known that the developmental stages for reference E strains occur at later time points than for reference L 2 strains [51]. This is supported by the ompA expression for E/Bour, which is shifted ,6 hours later than for L 2 /434. However, the differential growth rate between these two reference strains does not explain the dissimilar pmp expression, as most pmps were up-regulated at the same time point for both (Fig. 1). For E/Bour, almost all of the pmps had increased expression during the second half of development with comparable mRNA levels at these stages, suggesting a similar involvement in RB division and RB to EB transformation. For L 2 /434, although most pmps showed a general up-regulation of transcription at the exponential growth phase of RB division when new membranes are being formed, pmpC, pmpE and pmpF appeared to play a more important role during this phase. Thus, the gene expression results of both E/Bour and L 2 / 434 suggest their potential importance in membrane integrity. However, some Pmps may have more specific functions than others, depending on the chlamydial strain. In support of this, a proteomics study by Shaw et al. [21] detected five Pmps among reference strains A/HAR-13, D/UW-3, and L 2 /434, where PmpF was only detected for L 2 /434. In another proteomics study, only pmpE, pmpG, and pmpH were detected for L 2 /434 [22]. However, it is possible that these studies reflect a lack of sensitivity in detecting Pmps since a recent study was able to detect all Pmps for L 2 /434 [19].
Interestingly, pmpA had, in general, the lowest expression levels of all pmps at each time point except that it had one of the highest levels at 12 h p.i. (Fig. 1 and 2), suggesting a greater importance of PmpA during early stages of development. This is supported by shotgun proteomics where Skipp et al. [19] identified PmpA exclusively in RBs, whereas all other Pmps were detected in both RBs and EBs for L 2 /434. Additionally, for PmpD, the late upregulation at 36 h corresponds to RB transformation into EBs, suggesting a role in EB outer membrane structure. In support of this, PmpD has a cysteine content considerably higher than any other Pmp [12]. There are 26 conserved cysteine residues in PmpD for all 19 C. trachomatis reference strains, while the mean for all other Pmps is only 13.9 [SE 2.3]. Cysteine residues are responsible for the highly disulfide cross-linked proteins of the outer membrane complex of EBs. Previous studies found that PmpD is surface located and cross-linked in the chlamydial outer  membrane complex through disulfide bonds [52]. Furthermore, the N-terminal domain of C. pneumoniae Pmp21, the C. trachomatis PmpD ortholog, was shown to be non-covalently bound to other components of the EB surface [26]. Additionally, PmpD has shown species-specific neutralizing activity [27]. These collective data are supported by our findings that sera from C. trachomatis infected patients were reactive to rPmpD (Fig. 5). Our results were remarkably consistent for sera from patients infected with the same strain. For example, sera from all eight patients infected with strain E were reactive as were sera from three patients infected with strain Ba and three infected with strain K, although only one of the five patients with strain F were reactive; none of patients with strains D, Ia, J or G were reactive. Additional research is required to determine epitopes on PmpD that may correlate with the differential immune responses we observed.
Overall, considering both reference strains and clinical strains, pmpA and pmpI were the least expressed genes, while pmpF was the most highly expressed, although pmpE and pmpG also had similar expression levels for E/Bour. We previously found that PmpF is the most polymorphic protein among the C. trachomatis Pmps for both reference and clinical strains [28,12]. Consistent with the observed protein diversity, phylogenetic analyses of PmpF grouped C. trachomatis strains by tissue tropism properties [12]. Further, comparative analyses of PmpF reveal distinct domains that may be associated with a specific disease group.
The outer membrane exposure of the N-terminus has been experimentally demonstrated for some C. pneumoniae Pmps [26,53], suggesting that these proteins may be subjected to host immune pressure. The N-terminal half for C. trachomatis PmpF also contains numerous non-synonymous amino acid changes at locations of predicted MHC epitopes [9], indicating that it may be involved in eliciting a cellular immune response. Our findings that none of the sera from infected patients reacted with rPmpF suggest that this protein is not associated with the humeral immune response. Strain origin (E/Bour) of rPmpF did not seem to be an issue as sera from the eight patients infected with strain E were nonreactive. Furthermore, sensitivity was unlikely to be an issue given the immunoreactivity of the same sera with rPmpC and rMOMP, as we have previously described [17], and with rPmpD in this study. The occurrence of highly repeated GGAI motifs in the Nterminus suggests that Pmps may be associated with cell adhesion [54], which has been reported for Pmp21 of C. pneumoniae [26]. These cumulative findings suggest that Pmps are expressed with a differential immune response for patients infected with a specific strain. These findings and the remarkable pmpF expression dissimilarities among L 2 /434, E/Bour and the clinical strains suggest that there may be differential biological functions across strains and within the same strain for PmpF, either as a structural component to maintain membrane integrity, as part of a large pool of polymorphic antigens to elicit cellular immunity, or as an adhesin.
In our study, the pmpF sequences for the three clinical E strains were found to be 100% similar to the E/Bour sequence as was L 2 to L 2 /434. Since it is highly unlikely that identical proteins have diverse functions, we hypothesized that there may be differential regulation at the promoter level or regulation involving variations in mRNA processing and/or degradation, which would yield distinct mRNA amounts according to strain-specific needs. It is well known that point mutations in regulatory regions, such as promoter regions and RBS, can affect transcription and translation levels. However, analysis of the putative promoter region and RBS for pmpF showed that they are 100% conserved for both reference and the clinical strains (Fig. 3), suggesting that the observed pmpF expression heterogeneity may result from variations   . Dot-Blot of serum immunoreactivity against recombinant (r)PmpD and rPmpF. Sera was obtained from adolescents singly infected and uninfected with a different C. trachomatis clinical strain as described previously [17] (see also methods). rPmpD and rPmpF concentrations were standardized for use on the blots. Immunoreactivity to each fusion protein for sera from patients infected with strain Ba (n = 3), D (n = 3), E (n = 8), F (n = 5), G (n = 1), Ia (n = 1), J (n = 2) or K (n = 3) are shown. Of note is that immunoreactivity was consistent for sera from patients infected with the same clinical strain except for strain F ( in mRNA processing and/or degradation. In fact, regulatory systems of gene expression acting at both the transcriptional and translational levels are well represented in the chlamydial genome, including homologues of endoribonucleases E, III, G and P, exoribonucleases II and PNPase, and oligoribonuclease [8,9]. These are known to control mRNA stability and processing as well as translational efficiency in other bacteria, such as Escherichia coli and Staphylococcus aureus [44,48,[55][56][57]. The susceptibility of mRNA to ribonuclease attack may be influenced by events occurring not only at any stage during ribosome binding, but also during translation elongation or termination [40]. We identified two conserved putative RNase E cleavage sites in the pmpG/pmpF IGR, one of which is in close proximity to the putative RBS (Fig. 3). It is known that RBS sequence variability and sequestering by competitive regulatory proteins or conformational impediments can affect ribosome binding/loading and, thus, mRNA lifetime [40]. Considering this, a hypothetical initial cleavage by RNase E could reduce the affinity of the pmpF translation initiation region for ribosomes, thereby allowing subsequent mRNA degradation/processing by endo-and exonucleases, preferentially for E/Bour when compared to the other strains. A similar regulation has already been reported for sodB mRNA of E. coli at low iron concentrations [58]. However, this hypothetical mechanism, although possible, is speculative and lacks experimental evidence.
pmpF and pmpE belong to the same operon, yet had remarkably dissimilar mRNA levels for L 2 /434, and more so for all clinical strains with up to 8.4-fold higher expression for pmpF than for pmpE ( Fig. 1A and 2). This did not occur for the pmpGH operon. We speculated that the expression heterogeneity within the pmpFE operon may be generated by premature termination of transcription, rapid mRNA processing, or mRNA degradation primarily of the downstream gene (pmpE) of this large operon transcript (.6 Kb). Similar regulatory mechanisms have already been suggested to explain the existence of multiple transcripts within other bacterial policistronic operons [55], such as those of Bacillus subtilis ara [59], Nitrosomonas europaea cbb [60], and Borrelia burgdorferi ospAB and bmpAB [61,62].
Although we cannot assume that the putative stem-loop structures found within the pmpFE operon sequence may act as classic rho-independent type transcriptional terminators [49], the possibility of hairpin formation (a common phenomenon in mRNA, mainly on large transcripts) cannot be ignored nor can its hypothetical processing role in mRNA degradation and maturation be discounted. Furthermore, several putative RNase E cleavage sites were identified throughout the pmpFE operon (Fig. 4), which is expected for policistronic operons, although it is well known that RNase E cleaves mRNA only at a limited number of sites [55]. Interestingly, some of the RNase E sites were not conserved between L 2 and E strains, suggesting that targeted mRNA degradation or rapid processing events may occur in this large transcript. Curiously, one of these non-conserved recognition sites involved solely the pmpF/pmpE IGR of the four E strains. Thus, if RNase E uses this cleavage site, subsequent degradation or processing events from this point would only occur for E strains and could hypothetically yield an mRNA decay of pmpE. Yet, as above, this mechanism is speculative and lacks experimental evidence. However, in a previous study, differential transcript quantities were reported for the MMSO genes of E. coli that contained a consensus RNase E cleavage site in the intergenic regions of the operon, suggesting complex mRNA processing [63].
Overall, the heterogeneous expression levels among pmps and among strains highlight the importance of this gene family in chlamydial biology. In particular, the unique expression disparity for the pmpFE operon with relatively high pmpF mRNA levels for 5 of the 6 strains under study, as well as the differential immunoreactivity of patient sera to rPmpD, suggest that some Pmps may explain phenotypic differences among strains for antigenicity, virulence and tissue tropism. Furthermore, our findings that clinical E strains do not correlate with reference strain E/Bour at the gene expression level are supported by our previously reported genomic data [28], reinforcing the need to examine clinical along with reference strains to advance our understanding of the role of pmps in chlamydial biology and disease pathogenesis.