Table 1 describes the strains used in this study. All strains were individually propagated in McCoy cells prior to density gradient purification as we previously described (Read et al., 2013; Joseph et al., 2016). McCoy cells were screened for Mycoplasma contamination before use (Universal Mycoplasma Detection Kit, ATCC^® 30-1012K^TM, Manassas, VA, USA).

The in vitro tetracycline susceptibility was determined as the minimal inhibitory concentration (MIC) according to Suchland et al. (2003) with minor changes. Briefly, each chlamydial strain was inoculated onto 20 wells of a 48-well plate (E & K Scientific, Santa Clara, CA, USA) seeded with McCoy cells at a multiplicity of infection (MOI) of 0.5 or 1 depending on the infectivity of the strain. After inoculation of cells, the plate was centrifuged at 1500 RPM (Sorvall LegendXTR) for 1 h at 37°C. A tetracycline (Sigma-Aldrich, St. Louis, MO, USA) stock solution (10 mg/mL) in ddH₂O was used for a twofold dilution in propagation medium consisting of 450 mL Minimal Essential Medium alpha (MEMa, Life Technologies, Carlsbad, CA, USA), 10% Fetal Bovine Serum (FBS, JR Scientific, Woodland, CA, USA), 15 ml sodium bicarbonate (2.8%) (Thermo Fisher Scientific, Waltham, MA, USA), 10 ml glucose (45%) (Fisher Scientific), 10 mM HEPES (Life Technologies) and 1.4 μg/mL Cycloheximide (Sigma-Aldrich) with final concentrations ranging of 0.002 to 256 μg/mL (18 concentrations). After centrifugation, the chlamydial inocula were aspirated and replaced with the serial tetracycline dilutions. Two infected and two uninfected wells received media without tetracycline and served as positive and negative controls, respectively. Cells were fixed with methanol (-20°C) for 10 min after 24–36 h of incubation, depending on the developmental cycle of the strain, at 37°C in 5% CO2. Chlamydial inclusions were detected by direct immunofluorescence using Chlamydia Confirmation Pathfinder (Bio-Rad, Hercules, CA, USA). The MICs were evaluated by analyzing size and morphology as well as the number of inclusions (200 X, Nikon Eclipse Microscope and SPOT imaging software; Diagnostic Instruments, Inc., Sterling Heights, MI, USA). We determined the MIC transition point (MIC_TP) to be the tetracycline concentrations where 90% or more of the inclusions displayed alterations in size and morphology. Furthermore, the actual MIC was set at twofold higher concentrations (two times the MIC_TP) as defined by Suchland et al. (2003).

Supplementary Figure S1 shows the schematic and timeline for generation of recombinants. Two different recombination protocols were applied for co-infections. For Protocol 1 (recipient-first before co-infection with donor), confluent monolayers grown in shell vials were inoculated with tet^S S45 at an MOI of 4, centrifuged and incubated for 24 h before the addition of tet^R parental strains Rogers132, R19, or R27 at an MOI of 0.5. Tetracycline challenge was at 2 μg/mL (1/2 MIC_TP Rogers132) as described below. For protocol 2 (simultaneous co-infection), tet^S S45 was first grown in McCoy cell monolayers in shell vials to reach 100% infection; 25–50 μL of the infected culture were transferred to a new shell vial with a 80–100% confluent McCoy cell monolayer and simultaneously or immediately consecutively co-infected with tet^R R19, tet^R R27, or tet^R Rogers132. Inoculation was followed by centrifugation for 1 h at 1500 RPM. Tetracycline challenge was with 0.25 μg/mL (8x MIC S45) as described below (Supplementary Figure S1).

For each co-infection, three conditions and two controls (single infection with each parental strain) were used. Condition A did not contain any tetracycline (no tet); Condition B contained sub-inhibitory concentrations of tetracycline (1/2 MIC_TP for S45; 0.004 (μg/mL); and Condition C contained two times the MIC of S45 (0.032 μg/mL). Co-infected cultures were propagated for 72 h, sonicated once (20% amplitude, Sonic Dismembrator Ultrasonic Processor, Fisher Scientific), and new shell vials were infected to produce 100% infection. Each condition was either directly challenged with tetracycline (2 or 0.25 μg/mL depending on Protocol 1 or 2 described above) or passaged once in propagation medium without antibiotics prior to the tetracycline challenge. Following the challenge for 36–72 h depending on the developmental cycle of the strain, infected cultures were sonicated, and the inoculum was used to perform a harvest assay derived from the classic plaque assay, PCR and sequencing of PCR products to identify putative recombinants (see below, Supplementary Figure S1).

Isolation of clonally pure putative recombinants was accomplished using a modified cell culture harvest assay protocol (harvest assay), which derives from the classic plaque assay and is closely related to the shotgun cell culture harvest assay described by Somboonna et al. (2008). Briefly, the first well of a 6-well plate with 60% confluent McCoy monolayers was inoculated with the desired culture; seven serial 10-fold dilutions were performed, of which dilutions 2–7 were applied to wells 1–6. After 24 h, 2 mL of agarose gel (0.5% agarose, Lonza, Rockland, ME, USA) in phenol-red free MEM (Gibco), 10% FBS and 1 μg/ml Cycloheximide was added and topped with propagation medium. After incubation for 4–16 h, the well with detectable but low-level infection was chosen to select individual inclusions (no neighboring inclusions) at 200x magnification. Individual inclusions were picked with a sterile transfer pipet (Fisher Scientific) by punching a hole of 1–2 mm in diameter through the agarose; the plug was then sonicated in propagation medium and used to inoculate one shell vial per picked inclusion containing 500 μL propagation medium. The clones were propagated until the infection reached 100% (3–5 passages). Material from these clones was sonicated and frozen at -80°C, and a paired vial was collected for PCR and sequencing (Somboonna et al., 2011).

DNA from collected clones was extracted using the Roche High Pure PCR Product Purification Kit (Roche, Pleasanton, CA, USA), and PCR was performed as previously described (Somboonna et al., 2008). All primers are listed in Supplementary Table S1. Clones were considered putative recombinants if they had the following characteristics by PCR: positive for the correct size band for the tet(C) gene; positive for the intergenic region (IGR) between the polymorphic membrane protein gene (pmp)B and pmpC using primers specific for S45; negative for the pmpC region using primers specific for Rogers132; and positive for the major outer membrane protein A gene (ompA) with confirmation of the S45 ompA genotype by Sanger sequencing.

Putative clonal recombinants were then propagated in 0.063 μg/mL tetracycline (4x MIC of S45) to grow stocks for whole-genome sequencing, MIC determination, and the tet(C) stability assay (see below). After the first passage, the harvest assay was performed a second time to ensure clonal purity. Picked clones were either grown as described above or directly picked and inoculated into 100 μL HBSS (Gibco) prior to DNA extraction, PCR, and ompA sequencing.

Stocks of clonally purified putative recombinants were treated with DNase prior to gDNA purification as described previously (Somboonna et al., 2011). Libraries for sequencing were prepared from 0.5 to 1 μg of genomic DNA. Illumina MiSeq libraries were constructed using the Nextera kit and sequenced using the 250 bp paired end protocol on a MiSeq instrument. The resulting sequence data was assembled de novo using the SPAdes (Bankevich et al., 2012) software. The CONTIGuator web service (Galardini et al., 2011) was used to map assembled contigs against the Rogers132 strain. We visualized the structure of the assemblies aligned against individual regions of the chromosome using the Bandage graph visualization (Wick et al., 2015) software and its integrated BLAST tool. Querying of short sequences against the raw sequence reads in FASTQ format was performed using the R ShortRead (Morgan et al., 2009) package.

The raw genome data generated for this study are deposited in the SRA database: accession no. SRP096281.

We used the parsnp rapid genome alignment tool, which is part of the Harvest suite (Treangen et al., 2014) to identify regions of genome exchange in the recombinants. De novo assembled contigs of the recombinants were mapped against the S45 recipient genome reference and compared to the pattern of SNPs obtained when the Rogers132 donor was mapped against S45. Mosaic regions introduced by homologous recombination events appeared as clusters of ‘Rogers132-like’ SNPs in the background of the S45 genome. The approximate boundaries of recombination events were mapped as the edges of continuous runs of inserted SNPs.

We noted from preliminary analysis that recombinant DNA preparations in some cases also contained residual DNA from donor and/or recipient strains. To visualize the proportion of reads containing the donor and recipient backbone, we used a technique we called ‘SNP painting.’ We extracted two 20mer DNA sequences centered on each SNP identified between the donor and recipient reference sequences that had in their 11th position either the donor or recipient base. We challenged this reference SNP library against a 20mer database created from the FASTQ of the post-mating mixture using Jellyfish software (Marçais and Kingsford, 2011). From the counts of the donor and recipient 20mers at each base, we were able to (1) map recombinant boundaries and (2) ascertain the extent of mixed populations using R software (R Development Core Team, 2015).

Genes were aligned using PRANK (Löytynoja, 2014). The R package Phangorn (Schliep, 2010) was used to calculate maximum likelihood phylogeny using the Symmetric+GI model (chosen based on best fit). Trees were bootstrapped 100 times.

The MIC of all confirmed recombinants (by PCR and ompA genotyping) was tested as described above. To further test the stability of the recombinants regarding the presence of tet(C), recombinants were grown at an MOI of 5. Cultures were passaged five times before they were challenged or not with tetracycline using three different conditions: (A) no tetracycline; (B) ½ MIC_TP of the recombinant; and (C) 4x MIC of the recombinant. Following challenge, we propagated individual recombinants once under each condition and determined the MIC. Condition A was grown for another five passages (10 total) and was treated similarly as for passage five. The MIC was determined again after one additional passage. Additionally, DNA was extracted from the samples of the recombinants obtained after every passage to confirm that the ompA genotype was identical to the recombinant prior to the start of the assay and to test for the presence of tet(C) by PCR (Table 2).

Previous studies have shown that co-infection of the tet^R C. suis strain R19 with a tet^S C. trachomatis L₂ strain results in tet^R C. trachomatis recombinants (Suchland et al., 2009; Jeffrey et al., 2013). We aimed to obtain tet(C)-positive C. suis S45 recombinants by co-infecting S45 with three tet^R C. suis strains representing three of the four tet(C)-containing cassette Classes: I (strain R19; complete cassette with all three segments); II (strain Rogers132; without Segment 3); and IV (strain R27; without Segment 1) (Figure 1). Eight co-infections with R19 were performed using various protocols, including infection with R19 at 8 or 24 h pi with S45, and pre-treatment of R19 with high doses of tetracycline prior to co-infection. These co-infections tested negative for S45 by PCR using primers specific for the IGR of pmpB/C and by ompA genotyping (Supplementary Table S1) after tetracycline challenge, and were not further investigated. S45 was further grown in shell vials prior to co-infection with R19 (MOI 0.5) at 9 h pi and challenged with 0.25 μg/mL tetracycline. These infections were mixed; however, the 21 inclusions picked via harvest assay either resulted in cultures only positive for R19 by PCR and ompA genotyping or in cultures positive for only R19 after two passages (data not shown). One R27 co-infection with S45 was performed and yielded no recombinants.

Rogers132 (donor strain) and S45 (recipient strain) recombinants were successfully generated by simultaneous or sequential co-infection (see Materials and Methods). The success rate of producing recombinants was low. We obtained seven putative recombinants (Rec 2–8), which were propagated for whole-genome sequencing following a second harvest assay (Table 2). In genome analyses described below, several ‘recombinants’ turned out to be mixtures of strains or only parental strains, but the original term is used to describe Rec 2–8. In detail, for co-infection 1, 21 clones were Rogers132; 3 were S45 survivors or mixed infections (including Rec 2); and 2 were true recombinants (Rec 3 and 4). For co-infection 2, 11 were Rogers132; 18 were S45 or mixed infection (including Rec 6–8); and one was a true recombinant (Rec 5). Rec 2 and 8 were tetracycline sensitive after propagation while Rec 6 was positive for both S45 and Rogers132 by ompA genotyping and for Rogers132 by pmpC PCR. Figure 2 shows the PCR results for tet(C), pmpC and pmpB/C-IGR for the seven recombinants (Supplementary Figure S2: original gel images). We only obtained putative recombinants from co-infections performed in media containing sub-inhibitory (1/2 MIC_TP) or inhibitory (4x MIC) concentrations of tetracycline for S45; no recombinants were obtained without tetracycline. Only Rec 3, 4 and 5, all originally grown in sub-inhibitory tetracycline concentrations, were later confirmed as true recombinants by genomic analyses (see below).

We evaluated whether confirmed recombinants Rec 3, 4, and 5 remained stable over 10 passages without sub-inhibitory concentrations of tetracycline. After five passages, recombinants were either left drug-free (A) or challenged with low (B) and high (C) doses of tetracycline with subsequent MIC analyses (see Materials and Methods). For each recombinant, every passage was collected and tested for the presence of tet(C). At passage 10, ompA genotyping was performed to exclude the possibility of a small population of Rogers132 survivors. As expected, Rec 3, 4, and 5 remained positive for tet(C) with the identical ompA genotype throughout the stability assay (data not shown). These recombinants continued to be tetracycline resistant after passages 5 and 10 with relatively high MICs (Table 3; 8–16 μg/mL).

In contrast, putative recombinant Rec 7 was tet(C) positive by PCR in the first five passages but negative in passages 6–10 with an S45 ompA sequence, while Rec 6 was tet(C) positive throughout the assay and tetracycline resistant (Table 3).

All recombinants required propagation in 0.064 μg/mL tetracycline (4x MIC S45) in order to produce tet(C)-positive chlamydial stocks in sucrose phosphate glutamate (SPG) as described (see Materials and Methods). Attempts to propagate recombinants in the absence of tetracycline led to the loss of the tet(C) cassette following serial passages. However, once SPG stocks were produced, the confirmed recombinants were consistently positive for tet(C) throughout the 10 passages in the absence of tetracycline for the stability assay, and were phenotypically resistant in vitro (MIC > 4 μg/ml).

One co-infection experiment was performed for each of the two protocols under three different tetracycline Conditions: A–C (see Table 2). For protocol 1, co-infections were used to infect one 6-well plate per tetracycline Condition. Six to eleven clones were picked (26 total) per plate, resulting in two confirmed recombinants, Rec 3 and 4, from Condition B. For protocol 2, 30 clones were picked with 10 clones each per Condition, resulting in one confirmed recombinant, Rec 5, from Condition B. Taken together, only seven (12.5%) of 56 clones were identified as putative recombinants, and three were confirmed (5.4%). For Condition B, the success rate was 15.0% (3/20).

Altogether, we identified a number of Rogers132 clones (21/26; 80.8%), few S45 survivors and/or mixed infection (3/26 including Rec 2; 11.5%) and two confirmed recombinants (Rec 3 and 4; 7.7%) using Protocol 1 (Supplementary Figure S1). This can be explained by the lack of counter-selection used in this study as we wished to observe the outcome of natural co-infection with tetracycline as the only selection method. In contrast, the second co-infection using Protocol 2 (Supplementary Figure S1) yielded more S45 survivors or mixed infection (18/30 including Rec 6–8; 60%), several Rogers132 clones (11/30; 36.7%) and only one true recombinant (Rec 5; 3.3%), which can be explained by the abundance of S45 growth in shell vials used for this co-infection experiment as well as the lower dose of tetracycline applied to eliminate S45 survivors (0.25 μg/ml).

The seven putative recombinants described above were genome sequenced (Supplementary Table S2). We made a database of 4,864 SNPs between Rogers132 and S45 using parsnp (Treangen et al., 2014) (Datasheet S1) and used them as markers to distinguish potential recombination junctions in raw sequence data (Figure 3, ‘SNP painting’: see Materials and Methods). Aside from Rec 6, which was a mixture of donor and recipient, the recombinants were predominantly comprised of the S45 genetic backbone, evincing our success in screening for the recipient following co-infection. Recombinant regions were determined at the approximate junctions between regions of ‘S45-like’ and ‘Rogers-like’ SNPs (Figure 3A). While Rec 2 was PCR-positive for tet(C), there was undetectable recombination based on genome sequence analysis. Similarly, Rec 7 had no evidence for recombination. Rec 3 and 4 were confirmed recombinants with identical tet(C) cassette insertion sites at Rogers132 coordinates of ∼705,300–760,600 (upstream and downstream cross-overs varied in length because of sparse SNPs) in the S45 backbone, constituting ∼55,300 nucleotides (5.3% of the genome) including the rrn operons of 16S, 23S, and 5S rRNAs. The insertion spans from the Na(+)-translocating NADH-quinone reductase subunit F gene (nqrF) to the phosphoenolpyruvate carboxykinase [GTP] gene (pckG).

Rec 5 had a much larger insertion containing the tet(C) cassette at Rogers132 coordinates ∼638,000–813,000 (Figures 3A,B, right plot), spanning the 2-dehydro-3-deoxyphosphooctonate aldolase gene (kdsA) to the 3′(2′), 5′-bisphosphate nucleotidase gene (cysQ) in the S45 backbone. This region constitutes 175,000 nucleotides (16.6% of the genome). There were also second and third insertions within Rogers132 coordinates 529,000–538,000 in the S45 backbone (Figure 3B, left plot), including the putative general secretion pathway protein D gene (gspD) to the secretion system effector C gene (sseC) family. The presence of a subpopulation of S45-like SNPs across the 638,000–813,000 recombinant region revealed that Rec 5 was a mixed culture, consisting of a majority recombinant population and a minority of S45 recipient containing no Rogers132 DNA.

Although Rec 6 was a mixture of donor and recipient (Figure 3A), we identified a recombinant region of ∼40,000 nucleotides between Rogers132 coordinates ∼856,780–884,340 (Figure 3C). These coordinates are in the pmpD and phenylalanine t-RNA ligase (pheS) genes (Figure 3C). Rec 8 had no evidence of recombinants, consistent with the PCR results (Figure 2), and was considered an S45 survivor.

Figure 4 shows the recombinant locations of the cassette and other insertion sites within the context of the entire circular genome for Rec 3 and 4, and for Rec 5. While the cassette was confirmed to reside within ilp and rrn operons for all 11 previously genome sequenced C. suis strains (Joseph et al., 2016), the insertion in Rec 3 and 4 included nqrF, extending to pckG beyond the downstream rrn operon. The upstream crossover is located within ∼2.2 kb, either in pckG or its neighboring gene downstream, encoding a hypothetical protein, and in proximity to ribosomal bindings sites (RBSs), whereas the downstream crossover is found within nqrF (Figure 5A).

For Rec 5, the cassette insertion spanned cysQ to kdsA, which included the rrn operons and the ompA gene of Rogers132. The upstream crossover is located within ∼750 bp between the downstream end of kdsA or tRNA-Arg (tct) and a hypothetical protein-encoding gene, while the downstream crossover is found in cysQ, spanning 230 bp (Figures 4 and 5B). Both crossovers are in proximity to RBSs, and palindromes are evident. We also identified two shorter recombinant regions located adjacent to each other between kdsA and pmpBC (Figure 4). The larger of the two crossovers was found upstream within gpsD (∼380 bp) and downstream, either in mutL or ipgC, spanning ∼3 kb (Figure 6A), and in proximity to RBSs with evidence for palindromes where one is at the site of the SNP at nucleotide position 2230 (Figure 6A1). The complete crossover located within the sseC-like gene family spanned 1500 bp (Figure 6B). Figure 6C shows a schematic of the two Rogers132 insertions.

In examining the plasmids, all except Rec 7 had the S45 plasmid. Rec 6 was a mixture of both S45 and Rogers132 plasmids. None of the plasmids exhibited any genetic exchange.