Development of mass allelic exchange, a technique to enable sexual genetics in Escherichia coli

Significance Sexual genetics is a powerful strategy for dissecting any biological process; however, it is not applicable to most bacteria. We have invented mass allelic exchange (MAE), which enables direct sexual hybridization between arbitrary strains of Escherichia coli (including pathogenic clinical isolates). Like natural sexual hybrids in other organisms, MAE hybrids have no markers or scars. The combination of wild-type strains as well as scarless recombinants differentiates MAE from other genome-engineering technologies. By enabling access to the extant species variation, MAE facilitates a general method for gain-of-function screens for phenotypes of interest, a valuable complement to commonly used loss-of-function strategies. MAE will likely be feasible to implement in other Enterobacteriaceae, which are of strong medical and biotechnological interest.

Following overnight incubation, the donor library for MAE was screened for mCherry positive 115 clones using a fluorescence microscope (MVX10; Olympus, Singapore), followed by one round 116 of purification (streaking to single colonies on LB-kanamycin supplemented with 2% glucose). 117 The whole procedure was carried out three times to obtain a final donor library of 147 individual 118 clones. The pRK24 helper plasmid was conjugated into each clone individually following  Identifying clones with stable transposon insertions 130 Individual donor and recipient colonies were grown in LB with 2% glucose, kanamycin, and 131 tetracycline (for donor clones) or chloramphenicol (for recipient clones) overnight at 30°C or 132 37°C, respectively. The cultures were diluted in PBS with 10-fold serial dilutions and plated on 133 non-restrictive (LB with 2% glucose and antibiotics, for total CFU) and restrictive (M9 with 134 rhamnose) media; these plates were incubated at 37°C for 16 hours (non-restrictive agar) or 48-135 72 hours (restrictive agar). The loss rate of the custom transposon insertions was calculated as 136 the ratio of the titer under restrictive conditions to that under non-restrictive conditions. Clones 137 with a loss rate >=10 -5 were classified as "unstable", while those with loss rates < 10 -5 were 138 classified as "stable". Stable clones were pooled together to create the donor and recipient 139 libraries for MAE. 140 To identify stable and unstable regions of the UTI89 genome, 20 μl of the UTI89 Tn5 library 141 was thawed, diluted in PBS so as to get ~300 colonies on LB agar with chloramphenicol and 2% 142 glucose. 1000 individual colonies were picked and resuspended in 50 μl PBS each, diluted with 143 10-fold serial dilutions in PBS, and plated on non-restrictive (LB with chloramphenicol and 2% 144 glucose) and restrictive (M9 with rhamnose) solid media; these plates were incubated at 37°C for 145 16 hours (non-restrictive agar) to 48-72 hours (restrictive agar). The loss rate of the Tn5 146 insertions was calculated as the ratio of the titer under restrictive conditions to that under non-147 restrictive conditions. Clones with a loss rate >=10 -5 were classified as "unstable", while those 148 with loss rates <10 -5 were classified as "stable". For sequencing, all 1000 individual clones were 149 grown overnight in 500 μl LB broth with chloramphenicol and 2% glucose at 37 o C. 200 μl of 150 each clone was pooled together into an "input" library (containing all 1000 clones). 200 μl of 151 each clone was then also pooled together into a "stable" or "unstable" library, according to its grown to OD600 = 1.0 at appropriate temperatures with shaking. Cells were centrifuged at 5000 161 rpm for 8-10 mins, washed twice with 10 ml LB (no antibiotics) and resuspended to a final 162 OD600 = 20-25 in LB. The donor and recipient were mixed in a 4:1 ratio (total volume = 100 μl). 163 This mating mixture was spotted on a pre-warmed LB agar plate and incubated at 30°C for 2.5-3 164 hrs. The mating mixture was then scraped off and resuspended in 750 μl of PBS. Another 750 μl 165 of PBS was added to this culture, which was then centrifuged at 13,500g for 1 min. The pellets 166 were resuspended in 200 μl PBS. Ten-fold serial dilutions were plated on LB with antibiotics 167 (control plates for total counts) and M9 plates with rhamnose (without niacin) for negative 168 selection. All plates were incubated at 37°C for 16-20 hrs (LB plates) or 60-72 hrs (M9 plates).

169
As controls, donor cells (80 μl) and recipient cells (20 μl) were also treated exactly the same way 170 as conjugation mixtures and plated on control plates as well as test plates. Transconjugants were 171 confirmed by phenotypic and genetic testing (PCR and Sanger sequencing).

172
Mass allelic exchange. The protocol was similar to the GAE protocol above with minor 173 modifications and larger volumes. 50 μl of the donor library was thawed and grown overnight in 174 5 ml LB with tetracycline, kanamycin, and 2% glucose at 30°C under shaking conditions. 50 ul 175 of the recipient library was grown in 5 ml LB with streptomycin, spectinomycin (for plasmid 176 pSLC-391), chloramphenicol, and 2% glucose at 37°C with shaking. The rest of the procedure 177 remained the same as that described above for GAE. conditions. When the OD600 reached 0.25-0.30, 50 μl of phage lysate (10 9 or 10 10 PFU/ml) and 214 50 μl of 1M CaCl2 was added to the culture, and incubated at 37°C for 3-4 h with shaking. Phage 215 resistance and sensitivity were assessed by the lysis/turbidity of the culture, as well as by titering 216 the culture on LB agar before and after phage treatment.       The resulting cells were examined microscopically to look for intracellular pod formation.   For live imaging, cells were incubated at 37°C with 5% CO2 for 1 hour, after which the plates 317 were transferred to a PerkinElmer Operetta CLS machine maintained under the same conditions 318 (37°C, 5% CO2). Each well was imaged using the non-confocal 40x air objective using 319 brightfield and EGFP filters at 1 hour time intervals over 8-9 hours. Images were analysed using 320 the Harmony high-content image and analysis software (PerkinElmer) and combined using 321 ImageJ to create time lapse videos. transposon (defined as < 10 -6 CFU/ml when titered on rhamnose); these "stable clones" were 336 then pooled to make the final libraries used for conjugation.

337
As a side product of this screening, we also had a pool of "unstable clones", which allowed us to 338 directly identify unstable regions of the E. coli genome. This application was again feasible due 339 to the advance in the selection stringency added by our negative selection system. We carried out found prophages, PAIs, and the plasmid, but also some novel unstable regions that may provide 347 new insights into mechanisms of chromosome plasticity, maintenance, and evolution (Fig. S1B).  366 We noted a clear bias for transferred regions near the origin of replication in the selected hybrids 367 (Fig S2A). To investigate the source of this bias, we pooled clones with stable Tn insertions from 368 the donor and the recipient libraries and carried out TraDIS to map their insertion sites. The 369 donor pool consisted of 106 clones, while the recipient pool consisted of 495 clones. We found 370 that unique insertion locations were biased towards the replication origin (oriC) in both pools 371 (Fig. S2B, S2C). This correlates with the read depth per insertion and is consistent with other 372 reports of transposon insertion bias due to the higher copy number of origin-proximal DNA (13). 373 We observed an approximately 3x and 7x higher maximum insertion frequency in the donor and  To verify this last suggestion, we directly tested the efficiency of hybridization using directed 381 transfers. We selected individual donor and recipient clones with insertions of the oriT and 382 negative selection transposons, respectively, near the origin and the terminus (Fig. S2D). In these 383 control hybridizations, we saw no trend towards lower efficiency in the terminus-proximal 384 transfers (Fig. S2E). This data confirm that there is no general bias against hybridization in the 385 terminus-proximal region of the recipient chromosome.

434
The majority (267, 75.6%) of the transconjugants had only one detectable recombination block.

435
One strain each had 4 and 5 recombination blocks (Fig. S3A). The distribution of the length of 436 the recombined regions was trimodal, with several large (> 1 Mbp) transfers, a second peak at 437 ~400 Kb, and a third peak at ~50 Kb (Fig. S3B). Strains with multiple recombined regions must 438 have had at least one recombination whose location was not determined by the transposon 439 containing the negative selection marker in the recipient. Given that the smaller recombination 440 blocks seem to be occurring in strains that had more than one recombined region, we asked 441 whether recombination blocks in such strains were systematically biased towards shorter lengths. 442 We compared the length distribution of the larger recombination block in strains with 2 transfers 443 with the length distribution of the sole recombination block in strains with one transfer (Fig.   444   S3C). The medians of these distributions were not statistically different (P=0.9242, two-sided 445 Wilcoxon rank sum test). This suggests that in strains where more than one recombination has 446 occurred, at least one of the recombinations (perhaps a "primary" recombination) has the same 447 recombination statistics as strains where only one recombination has occurred, and we suspect 448 that such primary recombination events are directed by the negative selection cassette.  The length of the recombination block is plotted on the x-axis. 463 isolated transconjugants 464 A pooled transconjugant library containing ~12,000 hybrids (VK10-85) was screened for clones 465 that had gained the serum resistance phenotype using human serum. We isolated 60 clones and 466 repeated the serum treatment individually on these clones to confirm the phenotype (Fig. S4A).

467
MG1655 and a randomly selected serum sensitive clone (green "S") were used as negative 468 controls, and UTI89 was used as a positive control. Since the serum and P1 phage sensitivity is 469 known to be a single locus phenotype in MG1655, we also confirmed the P1 resistance of these 470 clones using phage streak assays (Fig. S4B). As expected, all 60 clones that were serum resistant 471 were also phage resistant, while the serum sensitive clone was sensitive to P1 phage. To confirm the serum and phage resistance phenotype observed using MAE, we verified that a 495 targeted transfer of the entire rfb locus from UTI89 to MG1655 indeed confers these phenotypes. isolated and screened for resistance to human serum and P1 phage. 13 transconjugants were 500 confirmed to be resistant to both serum and P1 phage ( Fig. S5A and S5B). Illumina whole 501 genome sequencing was carried out on 4 serum R phage R and 6 serum S phage S clones of this  as area in pixels and shape parameters). As previously reported, wt UTI89 formed large 526 intracellular pods (> 10,000 pixels at 400x magnification) (Fig. 3D-E); a video showing multiple 527 developing pods is shown in Movie S1 (still frames in Fig. S6A). Wild-type MG1655 formed no 528 such large pods (Fig. 3D-E, Movie S1, Fig. S6A). We noted that both of these controls strains 529 also formed smaller collections of bacteria, some of which appeared intracellular but did not 530 expand like typical pods formed by UTI89; such structures could be readily distinguished by 531 manual observation of fixed cells or by time lapse video (red boxes in Movie S1). However, to 532 avoid potential bias, we developed an automated method for distinguishing these. 533 We first extracted GFP features that were surrounded by WGA staining (i.e. intracellular 534 bacteria). We initially attempted to set a simple area threshold for identifying large pods; 535 however, we encountered two issues: (i) the saponin-mediated pod formation assay was highly 536 variable between biological replicates (Fig. S6B) and (ii) there seemed to be some large 537 structures that seemed "loosely packed". The variability was not due to cell passage number, as 538 we tested multiple (independently purchased) stocks of 5637 cells from ATCC, and this behavior 539 was not correlated with lot numbers for any reagents we used. The observation of high 540 variability, leading to occasional "jackpot" experiments where large pod formation was 541 observed, was consistent with the experience of the original discovery of this phenotype; similar 542 experiments attempting to correlate "jackpots" with cell passage number or media conditions 543 were also unsuccessful in the original reporting lab. Of note, IBC formation in vivo also can be 544 highly variable, with variation exceeding 2 orders of magnitude between mice in the same 545 experiment (14).

546
To quantify the observation that some structures seemed "loosely packed", we used compactness 547 (the ratio of the area of an object to the area of a circle with the same perimeter as the object) and 548 solidity (the total GFP area of the object divided by the area of the convex hull of the object 549 (imagined as wrapping a tight rubber band around the object)). Plotting intracellular structures 550 based on these two parameters showed a clear separation between large pods formed by UTI89 551 and smaller, more common intracellular structures formed by both strains (Fig. S6C). We set a 552 threshold established from data from 7 experiments for the two control strains, UTI89 and 553 MG1655 (gray line, Fig. S6C). This threshold further includes a size cutoff of 10,000 pixels.

554
These thresholds were then used to identify pods from imaging data for the other strains ( Figure   555 S6D). 556 These analyses verified that introduction of the chu operon into wt MG1655 (but not the empty 557 vector) conferred the ability to form large pods to MG1655 (a representative video is shown in 558 Movie S1, right). Importantly, in VK9-57-11, a knockout of the chu operon abolished pod 559 formation, a phenotype that was restored by complementation (Movie S2, Fig. S6D-F). 560 Consistent with previous reports, a knockout of the chu operon in UTI89 decreased the size of 561 the pods formed but did not affect formation of these large intracellular structures; in addition, by 562 video analysis, these appeared to be smaller due to delayed kinetics of pod formation (Movie S2 563 and Fig. S6E). Furthermore, we found that the presence of the chu operon had no significant 564 effect on the number of intracellular structures identified (considering all identified structures 565 regardless of the thresholds described above; Fig. S6B) but was associated with larger 566 intracellular structures (Fig. 3E). 567 Of note, the variability in this assay resulted in one experiment where the UTI89Δchu/VC 568 formed 2 pods while no other UTI89 strains formed any. As a qualitative phenotype, all UTI89