A Mechanism of Unidirectional Transformation, Leading to Antibiotic Resistance, Occurs within Nasopharyngeal Pneumococcal Biofilm Consortia

ABSTRACT Streptococcus pneumoniae acquires genes for resistance to antibiotics such as streptomycin (Str) or trimethoprim (Tmp) by recombination via transformation of DNA released by other pneumococci and closely related species. Using naturally transformable pneumococci, including strain D39 serotype 2 (S2) and TIGR4 (S4), we studied whether pneumococcal nasopharyngeal transformation was symmetrical, asymmetrical, or unidirectional. Incubation of S2Tet and S4Str in a bioreactor simulating the human nasopharynx led to the generation of SpnTet/Str recombinants. Double-resistant pneumococci emerged soon after 4 h postinoculation at a recombination frequency (rF) of 2.5 × 10−4 while peaking after 8 h at a rF of 1.1 × 10−3. Acquisition of antibiotic resistance genes by transformation was confirmed by treatment with DNase I. A high-throughput serotyping method demonstrated that all double-resistant pneumococci belonged to one serotype lineage (S2Tet/Str) and therefore that unidirectional transformation had occurred. Neither heterolysis nor availability of DNA for transformation was a factor for unidirectional transformation given that the density of each strain and extracellular DNA (eDNA) released from both strains were similar. Unidirectional transformation occurred regardless of the antibiotic-resistant gene carried by donors or acquired by recipients and regardless of whether competence-stimulating peptide-receptor cross talk was allowed. Moreover, unidirectional transformation occurred when two donor strains (e.g., S4Str and S19FTmp) were incubated together, leading to S19FStr/Tmp but at a rF 3 orders of magnitude lower (4.9 × 10−6). We finally demonstrated that the mechanism leading to unidirectional transformation was due to inhibition of transformation of the donor by the recipient.

It is therefore expected, although to the best of our knowledge not experimentally demonstrated, that homologous recombination of genes occurs symmetrically between naturally competent pneumococcal strains. In this study, we demonstrated that recombination between two transformable pneumococcal strains was unidirectional. The other strain, besides being competent for DNA uptake, acted as the donor.

RESULTS
Pneumococcal strains D39 (S2) and TIGR4 (S4) cohabit within biofilm consortia on human pharyngeal cells. Recombination events leading to antibiotic resistance and capsule switching occur during pneumococcal nasopharyngeal carriage (22,25,26). We first investigated whether pneumococcal strains D39 (serotype 2 [for simplicity referred to as S2]) and TIGR4 (S4) can cocolonize in a nasopharyngeal biofilm consortium, and its ultrastructure was then imaged by confocal microscopy. To achieve this, we simulated a nasopharyngeal environment in a bioreactor where human pharyngeal cells were incubated for 8 h with a mixture of the two strains. After 8 h of incubation, the relative densities of S2 and S4 within the biofilm consortium were similar (Fig. 1A). , after which bacteria were counted and the rF was calculated. (E) Extracellular DNA (eDNA) was purified from supernatants collected from bioreactor chambers at the indicated time. The DNA was used as a template in serotype-specific qPCRs amplifying eDNA from either S2 Tet or S4 Str . In panels A and C to E, error bars represent the standard errors of the means calculated using data from at least three independent experiments. Unidirectional Pneumococcal Transformation ® To further visualize the localization of strains within the biofilm consortium, we conducted confocal studies, staining both strains with fluorescently labeled, serotypespecific, anti-S2 and anti-S4 antibodies, while the DNA was stained with DAPI (4=,6diamidino-2-phenylindole). Figure 1B clearly shows S2 and S4 bacteria, both expressing their own capsule and forming aggregates of pneumococci, consistent with a biofilm consortium (Fig. 1B, top panels). Optical sections taken from the top of colonized pharyngeal cells, going down through the bottom, further revealed that the biofilm consortium was made of both strains integrated into a single structure with points of physical contact across the consortial biofilm (Fig. 1B, bottom panel). We hypothesize that the observed close proximity allows exchange of genetic material via transformation.
Recombination of antibiotic resistance genes occurs early during the formation of biofilm consortia. Given that antibiotics can be used to select for recombinant pneumococci, a time course study was conducted to investigate the timing of pneumococcal expression of resistance to two antibiotics within nasopharyngeal biofilm consortia. For these experiments, strains S2 and S4 were engineered to encode, in the chromosome, resistance to tetracycline (S2 Tet ), or streptomycin (S4 Str ). We also selected these strains because they produce different competence pheromones, CSP1 (27) or CSP2 (28), respectively, thus avoiding CSP-ComD (i.e., receptor) cross talk. Figure 1C shows that recombinant bacteria (Spn Tet/Str ), i.e., resistant to both tetracycline and streptomycin, appeared soon after 4 h of incubation, reaching a maximum recombination frequency (rF) of 1.1 ϫ 10 Ϫ3 at 8 h postinoculation, after which the rF remained similar for up to 24 h (median rF, 2.0 ϫ 10 Ϫ3 ). Confirming that recombinant pneumococci emerged from transformation, Spn Tet/Str bacteria were not obtained in bioreactor chambers incubated with DNase I (Fig. 1D). Moreover, double-resistant bacteria arose from recombination events, rather than from spontaneous mutations, since we did not obtain double-resistant Spn Tet/Str pneumococci in bioreactor control chambers containing only S2 Tet or S4 Str (rF, Ͻ4.3 ϫ 10 Ϫ7 or Ͻ3.6 ϫ 10 Ϫ8 , respectively). Sequencing confirmed the transference of streptomycin resistance-associated mutations (29) within the rpsL gene encoding ribosomal protein S12 in Spn Tet/Str recombinants (not shown).
Transformation leading to unidirectional acquisition of resistance occurs within pneumococcal biofilm consortia. Both strains S2 (D39) and S4 (TIGR4) are transformable under standard transformation conditions, (27,28). Accordingly, we obtained a similar transformation frequency (tF) when they were transformed with 2.5 g/ml of their own DNA (see Table S1 in the supplemental material) or each other's DNA (i.e., S2 plus DNA from S4 [3.1 ϫ 10 Ϫ7 ] and S4 plus DNA from S2 [3.1 ϫ 10 Ϫ6 ]). We therefore hypothesized that recombinant Spn Tet/Str bacteria would have arisen from both parents, whereby double-antibiotic-resistant pneumococci should belong to both serotype lineages S2 Tet/Str and S4 Tet/Str . To test this hypothesis, 50 Spn Tet/Str colonies were serotyped by conventional PCR (30) and Quellung reactions. All 50 recombinant bacteria, however, belonged to serotype 2 (i.e., S2 Tet/Str ). To screen for a larger number of recombinants, we designed a high-throughput assay utilizing serotype-specific quantitative PCRs (qPCRs). These reactions have a calculated limit of detection (LOD) of~2 genome equivalents (22). To this end, we pooled all isolated colonies obtained in blood agar plates containing Tet and Str (~500 Spn Tet/Str colonies from each plate), and DNA was extracted and utilized as the template in serotypespecific reactions. Using DNA template obtained from recombinant pneumococci, harvested from three independent experiments, serotype 2-specific reactions yielded a threshold cycle of detection (C T ) value corresponding to~7.8 ϫ 10 9 genome equivalents, whereas in serotype 4-specific reactions a C T value was undetectable, confirming that recombinants were all S2 Tet/Str (see Table S2 in the supplemental material). S2 Tet/Str recombinants originated whether recombination took place on living cultures of human pharyngeal cells, immobilized human pharyngeal cells, or abiotic surfaces (Table S2). Altogether, this evidence suggested that unidirectional recombination occurred within pneumococcal biofilm consortia.
The above results prompted us to test additional strains, including a genomesequenced strain, GA13499 (serotype 19F), which, like D39, produces CSP1, thus allowing for cross talk between CSP1 pheromones and ComD receptors. GA13499 is naturally resistant to trimethoprim (Tmp r [S19F Tmp ]). Tmp resistance in S. pneumoniae has been associated with mutations within the folA gene, encoding dehydrofolate reductase, with a key mutation leading to an amino acid substitution at position 100: isoleucine to leucine (I¡L) (31). Sequencing revealed that GA13499 contains mutations within folA, including the Tmp r -associated leucine substitution, whereas Tmpsusceptible (Tmp s ) D39 Tet has an isoleucine ( Fig. 2A). Strain S2 Tet was then incubated in the nasopharyngeal environment along with S19F Tmp for 24 h, at which point Spn Tet/Tmp recombinants were obtained at an rF of 1.5 ϫ 10 Ϫ4 . Recombinants from three different experiments (~500 Spn Tet/Tmp colonies from each) belonged to serotype 2 (i.e., S2 Tet/Tmp ), indicating that S2 strain acquired resistance to Tmp. We sequenced the folA gene in five of those S2 Tet/Tmp recombinant bacteria and confirmed that recombinants had acquired most mutations within the folA gene from S19F Tmp ( Fig. 2A).
Given that both S4 Str and S19F Tmp acted as donors when incubated along with S2, we incubated in the bioreactor both donor strains and scored for resistance to Str and (S19F) trimethoprim-resistant FolA protein are shown. Mutations associated with resistance to trimethoprim are indicated. Extracellular DNA (eDNA) was purified from supernatants collected from bioreactor chambers at the indicated incubation time. The eDNA was used as a template in serotype-specific qPCRs amplifying either (B) S2 Str or S19F Tmp or (C) S4 Str or S19F Tmp . In panels B and C, error bars represent the standard errors of the means calculated using data from at least three independent experiments. *, statistical significance (P Ͻ 0.02).
We then incubated the three strains together (i.e., S2 Tet , S4 Str , and S19F Tmp ). The density of each strain was similar at 8 h postinoculation (Fig. 3A), and extracellular DNA (eDNA) from all three strains was detected in the supernatant (Fig. 3B). Recombinants belonged to S2. The rF of S2 that had acquired Str or Tmp resistance from S4 Str or S19F Tmp , respectively, was similar to that when only two strains were incubated together~10 Ϫ3 (Fig. 3B). The rF, however, significantly decreased (1.4 ϫ 10 5 ) when we scored for the acquisition of the two markers acquired from donors, Str and Tmp (Fig. 3B). Together these data confirmed a mechanism of unidirectional transformation within pneumococcal nasopharyngeal biofilm consortia leading to acquisition, via recombination, of antibiotic resistance.
Both strains, the donor and the recipient, secrete eDNA into the supernatant. A possible explanation for the unidirectional transformation observed within nasopharyngeal consortial biofilms could simply be the absence of spontaneous release of extracellular DNA (eDNA) by the recipient strain or increased release of eDNA by the donor. Therefore, we quantified eDNA, using serotype-specific qPCRs, in the supernatant of the bioreactor chambers inoculated with S2 Tet and S4 Str . To minimize the presence of DNA at inoculation, inocula were prepared under conditions that were not permissive for competence, as described by Moscoso and Claverys (32). Before inoculation, pneumococci were washed three times with sterile culture medium. Even after these procedures, eDNA was still detectable in the supernatant of inocula from either S2 Tet or S4 Str , although this residual DNA was detected at a very low concentration (~100 pg/ml [ Fig. 1E]). Our experiments demonstrated, however, a marked increase in eDNA released from both strains after 1 h of incubation ( Fig. 1E; see Table S3 in the supplemental material). At this time point, eDNA from S2 Tet (mean, 7.90 ϫ 10 4 pg/ml) was significantly higher (P ϭ 0.023) than that from S4 (mean, 3.91 ϫ 10 4 pg/ml). This eDNA was able to transform competent pneumococci (either S2 Tet or S4 Str ); thereby it was permissive for transformation (not shown). The amounts of eDNA from both strains were then similar (P Ͼ 0.12) at 2, 4, 6 and 8 h postinoculation ( Fig. 1E; Table S3). Overall, eDNA in the supernatant increased after 1 h postinoculation, and then its presence decreased-perhaps by degradation-with only 2.80 ϫ 10 2 pg/ml (median) from S2 Tet and 9.06 ϫ 10 2 pg/ml (median) from S4 Str detected in the supernatant at 8 h postinoculation ( Fig. 1E; Table S3). Quantification of eDNA was also performed in supernatants of bioreactor chambers inoculated with S2 Tet and S19F Tmp , S4 Str and S19F Tmp , or , and GA13499 (S19F) were inoculated into a bioreactor and incubated at~35°C for 8 h, after which both (A) the density (CFU per milliliter) of each strain was obtained by culture in BAPs with the appropriate antibiotic and (B) the recombination frequency (rF) of recombinants growing in BAPs with the indicated antibiotics was calculated. (C) Extracellular DNA (eDNA) was purified from supernatants collected from bioreactor chambers at the indicated incubation time and used as a template in serotype-specific qPCRs amplifying eDNA from either S2, S4, or S19F. Error bars represent the standard errors of the means calculated using data from at least three independent experiments. *, statistical significance (P Ͻ 0.05).
Lattar et al. ® S2 Tet , S4 Str , and S19F Tmp , with a similar peak of eDNA released by strains at 2 h postinoculation ( Fig. 2B and C and Fig. 3C). This similar release of eDNA by strains in the biofilm consortium within 2 h of incubation, permissible for transformation (not shown), ruled out the possibility that unidirectional recombination was related to the availability of eDNA and indicated that spontaneous competence occurred.
Human pharyngeal cells trigger spontaneous competence within pneumococcal biofilm consortia. Given that we did not add synthetic CSP to trigger competence, but recombination occurred within hours, we hypothesized that spontaneous competence occurred in the bioreactor. To investigate if pharyngeal cells, the culture medium, or the mammal serum used would trigger spontaneous competence, we inoculated strains S2 Tet and S4 Str into bioreactor chambers containing (i) living cultures of human pharyngeal cells, (ii) pharyngeal cells that had been immobilized with paraformaldehyde, and (iii) abiotic surfaces, all incubated with cell culture medium containing serum or (iv) living cultures of pharyngeal cells incubated with cell culture medium lacking serum. Our experiments demonstrated a similar rF when S2 Tet and S4 Str were incubated in the bioreactor with living cultures of pharyngeal cells (rF, 4.3 ϫ 10 Ϫ4 ), in pharyngeal cells that had been immobilized (rF, 1.2 ϫ 10 Ϫ4 ), and in cultures of pharyngeal cells incubated with medium without serum (rF, 6.2 ϫ 10 Ϫ4 ) ( Table 1). In contrast, the rF was 3 orders of magnitude lower (rF, 3.3 ϫ 10 Ϫ7 ) when pneumococci were incubated on an abiotic surface (Table 1). Together these experiments demonstrated that spontaneous competence within pneumococcal biofilm consortia, leading to recombination of antibiotic resistance genes, is triggered upon contact between pneumococci and the host cell.
Transformation of the donor is inhibited by the recipient strain by a mechanism different from Com. Strains S2, S4, and S19F are transformable in vitro (Table S1); however, in consortial biofilms only S2 acquired resistant determinants (19,20). A possible explanation for the observed unidirectional transformation is that the recipient may have an increased transformation phenotype, or transformation of the donor is inhibited. If the first hypothesis is true, then removing CSP signaling in the recipient, and therefore competence for transformation, would now allow transformation of S4. To test this hypothesis, we incubated an S2 Ery ΔcomC mutant, which is not transformable under standard transformation conditions, with S4 Str in the bioreactor. S2 Ery/Str recombinants were obtained when the S2 Ery wild type and S4 Str were incubated together for 8 h (Fig. 4A). In contrast, recombinants with resistance to both erythromycin and streptomycin were not harvested from bioreactor chambers incubated with S2 Ery ΔcomC and S4 Str (Fig. 4A). This experiment revealed, as opposed to our original hypothesis, that reducing transformation of the recipient was not enough to allow transformation of a strain acting as a DNA donor.
Since we did not obtain recombinants, but we demonstrated that a similar amount of eDNA permissible for transformation from both strains was available during incubation in the bioreactor (Fig. 4B), we hypothesized that S2 inhibits transformation of S4. To investigate this hypothesis, we first asked whether S4 would become competent for DNA uptake in the simulated nasopharyngeal microenvironment by incubating strain S4 Tmp in the bioreactor along with 10 ng/ml of S2 Str DNA. Our experiments demonstrated a tF of 4.4 ϫ 10 Ϫ3 (Fig. 4C), which was 3 orders of magnitude different from the tF obtained in the conventional transformation assay done in a test tube, where  4C; Table S1). Therefore, S4 Tmp became naturally competent in the bioreactor. S4 was then incubated in the presence of S2 Str DNA (10 ng/ml), but this time we also inoculated the nontransformable S2 Ery ΔcomC strain. Surprisingly, S4 transformants were not obtained, indicating that the recipient, S2, inhibited transformation of the donor. Several attempts were made to conclude that the supernatant did not inhibit transformation of the donor (not shown). Therefore, the potential mechanism appears to be independent of CSP production and contact mediated.

DISCUSSION
In this study, we recreated nasopharyngeal recombination of antibiotic resistance genes between pneumococcal strains and we demonstrated that it occurs very rapidlywithin 4 h postcontact of pneumococci with human pharyngeal cells. Our experiments also found, as revealed by antibiotic selection, that unidirectional transformation occurred within pneumococcal biofilm consortia, leading to the rise of a single lineage of highly transformable pneumococci. Antibiotic selection was used as a surrogate for the pressure that, in the pneumococcal vaccine age, is challenging pneumococcal strains.
Evidence presented in this study, and elsewhere (26), indicates that pherotype cross talk is not involved in unidirectional transformation. For example, we demonstrated here that regardless of whether CSP-ComD receptor cross talk was allowed, unidirectional transformation occurred. This evidence correlated with the absence of transformation of the donor even when it was incubated with a CPS knockout recipient. Moreover, observations from whole-genome sequence studies conducted on prevaccine pneumococcal isolates demonstrated that pherotypes CSP1 and CSP2 or their variants do not account for recombination differences among strains, suggesting that the mechanism leading to unidirectional transformation occurs irrespective of CSP cross talk (26).
Genomic studies have demonstrated that resistance to antibiotics among pneumococcal strains is driven by expansion of clones that have acquired resistance phenotypes by HGT, more than by de novo mutations (10). The mechanism is very efficient given the high resistance rates, and except for transposon-mediated resistance that will  be discussed below, it includes resistance to ␤-lactams, trimethoprim, and some of the last resort antibiotics, such as linezolid and carbapenems (7). The DNA taken by transformation is in general small since it appears to be cleaved into~6.6-kb fragments (33). These fragments undergo homologous recombination to integrate, into the genome, DNA pieces of~4.4 kb, calculated using data from species-wide multilocus sequence typing studies (34), or~6.3 kb, as demonstrated by sequencing (35). A more recent study demonstrated heterogeneity in pneumococcal recombination with microrecombination fragments of~0.03 to 0.6 kb and macrorecombination fragments ranging from 9 to 10 kb (36). Secondary, nonselective, recombination events via unidirectional transformation were not investigated in our study but have been calculated by Croucher et al. to be~2.3 kb (37).
It has been well documented that mobile genetic elements (MGEs [e.g., transposons]) carry genes for resistance to tetracyclines and macrolides. These MGEs are usually~20 kb or larger and are therefore not simply transferred by transformation (38). More recently, capsular switch events were linked to mobilization of the whole capsule locus from a nonvaccine type donor, 35B/ST558, to a vaccine type recipient, 9V/ST156, leading to a new lineage of capsule switch variants belonging to serotype 35B/ST156 (39). Evidence of in vitro transference of Tn-encoded resistance, or capsule genes, in pneumococcal strains is not available yet, but such transference is currently being investigated in our laboratories.
Another important contribution in this article refers to the development of spontaneous competence. Our experiments demonstrated a burst of eDNA released into the supernatant as soon as 1 h postinoculation of human pharyngeal cells, in all different mixtures of pneumococcal strains tested. Release of eDNA has been linked to the development of competence, in studies conducted by Moscoso and Claverys (32), and also to production of bacteriocins and other fratricide factors, whose secretion causes heterolysis and therefore the release of DNA (40)(41)(42). Release of abundant eDNA in the bioreactor was also observed when strains were inoculated alone (data not shown); thereby, it is unlikely that such amounts of DNA were generated by heterolysis but rather by a mechanism coupled to spontaneous competence. Release of DNA early during colonization may also help pneumococci to attach to host cells and/or form bacterial aggregates leading to nasopharyngeal biofilms.
We, and others, have clearly demonstrated that under static culture conditions irreversible autolysis occurs in monostrain pneumococcal biofilms (40,43). Heterolysis also happened in consortial biofilms inoculated with two different strains incubated under static conditions (42,44). All of this evidence correlates with the finding that release of DNA in the supernatants is highest in stationary cultures because it comes from lysis of pneumococci (32). This, however, did not occur in our bioreactor model, where the sharpest peak of eDNA in the supernatants was detected during the log phase, within 2 h postinoculation. Studies by Wholey et al. demonstrated that heterolysis in consortial biofilms is linked to production of the BlpC bacteriocin, whose synthesis and release are controlled by the competence system (42). In the abovementioned study, inoculation of strains at a 1:1 ratio was enough to completely eradicate susceptible pneumococci by a BlpC-producing strain (42). A decrease in density, but by no means eradication of a pneumococcal strain, was observed in recent studies in which two strains were inoculated at a similar density and cultured under static conditions (44). Moreover, when inoculated at similar densities in the bioreactor the density of strains was similar throughout the incubation period. Therefore, unidirectional transformation leading to a rapid acquisition of genes by recombination does not appear to be simply the eradication of one of the strains in the consortial biofilms. We cannot rule out the possibility that killing of a fraction of the population of the recipient or the donor took place before the burst of released eDNA and the appearance of recombinants.
Remarkably, rapid development of spontaneous competence caused nearly 1 in 1,000 pneumococci to acquire antibiotic resistance within 4 h. High rF (e.g.,~10 Ϫ3 ) was consistently obtained in experiments where the resistance determinant included spe-cific mutations (i.e., folA and rpsL), whereas the transference of whole resistance genes, such as tetM or ermB was observed at a lower rF (usually Ͼ10 Ϫ4 ). These observations were similar to those described in a mouse model of colonization (45). In this mouse model, recombinants were harvested 48 h postinoculation at an rF similar to that obtained in the bioreactor (45). Together this evidence suggests that acquisition of DNA among pneumococci and the development of competence occur soon after pneumococci colonize the host.
What stimulated the rapid development of competence? To the best of our knowledge, the only molecule from the host that has been identified as a trigger for recombination is chitin, an oligosaccharide found in the exoskeletons of crustaceans, the natural habitat of Vibrio cholerae strains (46). Chitin induced an upregulated production of the V. cholerae T4P (47), used for DNA uptake, and proteins of the transformation machinery (48). More recent studies showed that chitin induces production of a type 6 secretion system (T6SS), which is utilized by V. cholerae to kill its neighbors and thus allow release of DNA, which is then taken by the transformation machinery, for recombination (49). While pneumococcus utilizes a T4P to take up DNA, a T6SS has not yet been reported, and a source for chitin in the bioreactor, the mouse model, or the human host, is unlikely.
Attempts were made in this study to begin to understand the development of such early "spontaneous" competence leading to nasopharyngeal recombination. The outcome of our extensive experimentation was that recombination of antibiotic resistance determinants, whose mechanism included release of eDNA and uptake (e.g., spontaneous competence), only required contact with host cells. Experiments by Marks et al. also demonstrated recombination in a biofilm model utilizing paraformaldehyde (PFA)fixed pharyngeal cells in comparison to abiotic surfaces (45). We ruled out the possibility that a secreted product from human cells ignited pneumococcal recombination, given that the rF obtained in experiments using immobilized cells was similar to that obtained in living cultures of pharyngeal cells. Neither the cell culture medium nor the mammal serum utilized had an effect on the rapid pneumococcal recombination observed. As such, pneumococcal strains inoculated in the bloodstream of mice were not able to acquire DNA that was concomitantly inoculated in the seminal experiments published by Griffith in 1928 (12). Likewise, a high recombination frequency was not observed when two pneumococcal strains were inoculated in a mouse model of sepsis in more modern studies of pneumococcal recombination by the group of A. Hakansson (21).
So far, we have not been able to obtain recombinants within 8 h of incubation in a static plate model, whether or not it contains human pharyngeal cells (40,43). The absence of recombination may be because of the heterolysis phenotype reported using static incubation conditions (42,44) or the accumulation of DNases in static cultures, which may degrade available eDNA for transformation (32). In summary, we have demonstrated in this study that unidirectional transformation occurred within pneumococcal biofilm consortia and that unidirectional transformation is mediated by inhibition of transformation within pneumococcal nasopharyngeal biofilms.

MATERIALS AND METHODS
Bacterial strains, culture media, and antibiotics. The S. pneumoniae strains used in the present study are listed in Table 2. Strains were routinely cultured on blood agar plates (BAPs) or grown in Todd-Hewitt broth containing 0.5% (wt/vol) yeast extract (THY) at 37°C with a 5% CO 2 atmosphere. Where indicated, streptomycin (Str; 200 g/ml), trimethoprim (Tmp; 10 g/ml), tetracycline (Tet; 1 g/ ml), and/or erythromycin (Ery; 1 g/ml) was added to the BAP. All antibiotics were purchased from Millipore-Sigma (Saint Louis, MO).
Preparation of the inoculum to produce pneumococcal biofilm consortia. The inoculum was prepared as previously described (43). Briefly, an overnight BAP culture was used to prepare a cell suspension in THY broth to an optical density at 600 nm (OD 600 ) of 0.05. This suspension was incubated at 37°C in a 5% CO 2 atmosphere until the culture reached an OD 600 of~0.2 (early log phase), and then glycerol was added to a final concentration of 10% (vol/vol) and stored at Ϫ80°C until used. A frozen aliquot from each batch was removed to obtain the density (CFU per milliliter) by dilution and plating.
Preparation of antibiotic-resistant, D39 derivative, and TIGR4 derivative pneumococcal strains. SPJV17 and SPJV28 were constructed by transforming D39 or TIGR4, respectively, with integrative Lattar et al. ® plasmid pPP2, which targeted tetM to the nonessential bgaA gene (50). SPJV22 and SPJV23 were transformed with DNA from strain R6Ami9 encoding resistance to streptomycin (44). Strains SPJV27 and SPJV29 were prepared by transformation of TIGR4 or D39, respectively, with DNA from GA13499 encoding resistance to trimethoprim. Chromosomal integration of the tetM gene was confirmed by PCR with primers JVS101L and JVS102R. Mutations within folA or rpsL, conferring resistance to Tmp or Str, respectively, were confirmed by sequencing with primer JVS99L or JVS100R for Tmp and with primer JVS103L or JVS104R for Str. Transformation was done following standard methods (51,52).
Inoculation of the bioreactor with pneumococcal strains. Detroit 562 cells (ATCC CCL-138) were grown on Snapwell filters (Corning, Corning, NY); these filters have a polyester membrane (0.4 m) supported by a detachable ring. Once polarized (4 to 5 days), Snapwell filters containing pharyngeal cells were immediately placed in a sterile vertical diffusion chamber (bioreactor) (43). Where specified, a set of pharyngeal cells were prefixed with 2% paraformaldehyde (Millipore-Sigma, Saint Louis, MO) for 15 min, followed by extensive washing with sterile phosphate-buffered saline (PBS), prior to installation in the bioreactor chambers. To create an abiotic surface, some bioreactor chambers were installed with Thermanox coverslips (Thermo Fisher Scientific, Grand Island, NY). The bioreactor chamber has an inlet from which the apical side (inner chamber) was perfused at a low flow rate of~0.20 ml/min with sterile EMEM, which contained 5% FBS but no antibiotics, using a Master Flex L/S precision pump system (Cole-Parmer, Vernon Hills, IL). Where indicated, DNase I was added to a final concentration of 50 U/ml. Perfused culture medium and planktonic cells exit the bioreactor chamber by a parallel outlet located on top of the chamber.
Bioreactor chambers were then inoculated with~1 ϫ 10 6 CFU/ml of each pneumococcal strain and incubated at~35°C under a sterile environment. At the end of the incubation period, Snapwell inserts or Thermanox was removed, and biofilm consortia were analyzed as follows. Biofilm consortia or monostrain biofilms (control) were harvested by sonication for 15 s in a Bransonic ultrasonic water bath (Branson, Danbury, CT), followed by extensive pipetting to remove all attached bacteria. An aliquot was used to obtain the density of each strain, by dilution and plating in BAP containing the appropriate antibiotic, and another aliquot was directly plated onto BAP containing two or three antibiotics to recover recombinant pneumococci.
Calculation of rF and tF. The density of parent strains was counted in BAPs containing one antibiotic, while recombinants were counted on BAPs with two or three antibiotics. The recombination frequency (rF) was the density of pneumococci with dual or triple markers divided by the density of the parent strain. The transformation frequency (tF) was the number of transformants relative to the total pneumococci recovered in the transformation reaction.
Confocal micrographs of pneumococcal biofilm consortia. To visualize biofilm consortia by confocal microscopy, we installed a glass coverslip inside the Snapwell filters prior to seeding with human pharyngeal cells. Once pharyngeal cells became polarized, the Snapwell filter was installed in the bioreactor and inoculated as described above. At the end of the incubation, the coverslips containing pharyngeal cells with pneumococcal consortial biofilms were washed twice with PBS and fixed with 2% PFA for 15 min at room temperature. Once the fixative agent was removed, cells were washed with PBS and blocked with 2% bovine serum albumin (BSA) for 1 h at room temperature. These cells containing consortial biofilms were then incubated with serotype-specific polyclonal antibodies (~40 g/ml; Statens Serum Institute, Copenhagen, Denmark) for 1 h at room temperature. Antibodies had been previously labeled with Alexa 488 (anti-S4) or Alexa 555 (anti-S2) following the manufacturer's recommendations (Molecular Probes, Thermo Fisher Scientific, Grand Island, NY) (44). Stained preparations were finally washed two times with PBS and were mounted with ProLong Diamond antifade mountant with DAPI High-throughput assay for pneumococcal serotyping. Recombinant pneumococci obtained in BAPs containing two or three antibiotics were pooled in 200 l of sterile PBS, and DNA from this population was purified as detailed below. This DNA was utilized as the template for serotype-specific quantitative PCRs with primers and probes listed in Table 3. Reactions targeted serotype-specific sequences within the capsule polysaccharide (cps) locus of each serotype (22,53) and were run along serially diluted DNA standards corresponding to 4.29 ϫ 10 5 , 4.29 ϫ 10 4 , 4.29 ϫ 10 3 , 4.29 ϫ 10 2 , 4.29 ϫ 10 1 , and 2.14 ϫ 10 1 genome equivalents per reaction (54). Reactions were carried out using a Bio-Rad CFX96 Touch real-time PCR detection system (Bio-Rad, Hercules, CA) with the following cycling parameters: 50°C for 2 min, 95°C for 2 min, and 40 cycles of 95°C for 15 s and 60°C for 1 min. The standard curve and regression equation obtained were then used to calculate final genome equivalents per milliliter using the CFX software (Bio-Rad, Hercules, CA). DNA extraction. DNA was extracted from 200 l of a fresh suspension of pneumococcal strains with the QIAamp DNA minikit (Qiagen, Valencia CA) according to the manufacturer's instructions. Final elution was done with 100 l of elution buffer. DNA preps were quantified using a NanoDrop spectrophotometer and stored at Ϫ80°C until used.
Quantification of eDNA. Supernatants were collected from the outlet of bioreactor chambers, centrifuged for 15 min at 14,000 ϫ g in a refrigerated centrifuge (Eppendorf, Hauppauge, NY), and then sterilized with a 0.45-m-pore syringe filter. This bacterium-free supernatant was DNA extracted using the QIAamp DNA minikit following the manufacturer's instructions. Purified DNA was used as the template in serotype-specific quantitative PCRs (qPCRs) using the primer and probe sets shown in Table 3. Reactions were performed essentially as described as above and in our previous studies (22,55). For eDNA quantification purposes, standards containing 1 ϫ 10 3 , 1 ϫ 10 2 , 1 ϫ 10 1 , 1 ϫ 10°, 1 ϫ 10 Ϫ1 , 5 ϫ 10 Ϫ2 , or 1 ϫ 10 Ϫ3 pg of chromosomal DNA from the appropriate serotype were run in parallel to generate a standard curve. This standard curve was then used to calculate the eDNA concentration using the Bio-Rad CFX Manager software.
Serotype-specific conventional PCRs. Serotype-specific PCRs were performed in 25-l volumes containing~100 ng DNA or 3 l of bacterial lysate, 1 M serotype-specific forward or reverser primer listed in Table 3, and 1ϫ the PCR master mix from the Qiagen Multiplex PCR kit (Qiagen, Valencia CA). Reactions were run using the following cycling parameters: 1 cycle at 95°C for 15 min, followed by 35 cycles of 94°C for 30 s, 54°C for 1 min, and 72°C for 1 min, with a final extension of 72°C for 10 min. Products were run on 2% agarose gels, stained with SYBR Safe DNA gel stain (life technologies, Grand Island, NY), and visualized under a UV transilluminator (Bio-Rad, Hercules, CA).
Transformation reactions. S. pneumoniae strains were made competent using standard procedures and then transformed with 500 ng of pure DNA containing 100 ng of competence-stimulating peptide 1 (CSP1 [EMRLSKFFRDFILQRKK]) or CSP2 (EMRISRIILDFLFLRKK) in a reaction volume of 200 l (15). CSP1 and CSP2 were synthesized at Millipore-Sigma (Saint Louis, MO).
Sequencing reactions. Purified DNA from wild-type strains or recombinant derivatives was used as the template to PCR amplify the folA gene using primers JVS99L and JVS100R listed in Table 3. PCR products were purified using the QIAquick PCR purification kit (Qiagen, Valencia, CA). Both DNA strands

ACKNOWLEDGMENTS
Special thanks goes to Scott Chancey and David Stephens from the Infectious Disease Division of the Department of Medicine at Emory University School of Medicine for providing strain GA13499. We thank Ana Enriquez from the Graduate Program of Microbiology and Molecular Genetics (MMG) at Emory University for assistance in some transformation experiments and David Watson from RSPH, Emory University, for reading and editing the manuscript. We also thank Ilya Nemenman from the Department of Physics, Emory University, and June R. Scott from the Microbiology and Immunology Department, Emory University School of Medicine, for discussion during the preparation of the manuscript.
This study was supported by a grant from the National Institutes of Health (R21AI112768-01A1 to J.E.V.). Confocal studies were supported in part by funds from the Integrated Cellular Imaging (ICI) pediatric core and the EmoryϩChildren's Pediatric Research Center to J.E.V. The content is solely the responsibility of the authors and does not necessarily represent the official view of the National Institutes of Health.