Biology and engineering of integrative and conjugative elements: Construction and analyses of hybrid ICEs reveal element functions that affect species-specific efficiencies

Integrative and conjugative elements (ICEs) are mobile genetic elements that reside in a bacterial host chromosome and are prominent drivers of bacterial evolution. They are also powerful tools for genetic analyses and engineering. Transfer of an ICE to a new host involves many steps, including excision from the chromosome, DNA processing and replication, transfer across the envelope of the donor and recipient, processing of the DNA, and eventual integration into the chromosome of the new host (now a stable transconjugant). Interactions between an ICE and its hosts throughout the life cycle likely influence the efficiencies of acquisition by new hosts. Here, we investigated how different functional modules of two ICEs, Tn916 and ICEBs1, affect the transfer efficiencies into different host bacteria. We constructed hybrid elements that utilize the high-efficiency regulatory and excision modules of ICEBs1 and the conjugation genes of Tn916. These elements produced more transconjugants than Tn916, likely due to increased excision frequencies. We also found that several Tn916 and ICEBs1 components can substitute for one other. Using B. subtilis donors and three Enterococcus species as recipients, we found that different hybrid elements were more readily acquired by some species than others, demonstrating species-specific interactions in steps of the ICE life cycle. This work demonstrates that hybrid elements utilizing the efficient regulatory functions of ICEBs1 can be built to enable efficient transfer into and engineering of a variety of other species. Author summary (non-technical) Horizontal gene transfer helps drive microbial evolution, enabling bacteria to rapidly acquire new genes and traits. Integrative and conjugative elements (ICEs) are mobile genetic elements that reside in a bacterial host chromosome and are prominent drivers of horizontal gene transfer. They are also powerful tools for genetic analyses and engineering. Some ICEs carry genes that confer obvious properties to host bacteria, including antibiotic resistances, symbiosis, and pathogenesis. When activated, an ICE-encoded machine is made that can transfer the element to other cells, where it then integrates into the chromosome of the new host. Specific ICEs transfer more effectively into some bacterial species compared to others, yet little is known about the determinants of the efficiencies and specificity of acquisition by different bacterial species. We made and utilized hybrid ICEs, composed of parts of two different elements, to investigate determinants of transfer efficiencies. Our findings demonstrate that there are species-specific interactions that help determine efficiencies of stable acquisition, and that this explains, in part, the efficiencies of different ICEs. These hybrid elements are also useful in genetic engineering and synthetic biology to move genes and pathways into different bacterial species with greater efficiencies than can be achieved with naturally occurring ICEs.


Introduction 68
Integrative and conjugative elements (ICEs), also called conjugative transposons, are mobile 69 genetic elements that are major drivers of bacterial evolution (Bellanger et  IPTG for Tn916 and ICEBs1, respectively, for one hour. We used qPCR to measure excision 160 from the respective integration sites and normalized to a nearby chromosomal locus (yddN for 161 ICEBs1, and mrpG for Tn916) Wright and Grossman, 2016 As noted above, there were different frequencies of excision of each element. Because only 195 the excised circular form of either element is competent for transfer, we normalized the number 196 of transconjugants to the excision frequency of each element. After normalization, we found that 197 the conjugation efficiencies of Tn916 and ICEBs1 were 0.11% and 6.7%, respectively, per donor 198 with an excised element. Based on this analysis, we conclude that acquisition of ICEBs1 by 199 recipient cells during mating was approximately 50-fold more efficient than that of Tn916. This 200 could be due to differences in any of several steps of the ICE life cycle, including DNA nicking, 201 unwinding, replication, association with the conjugation machinery, transfer, second strand DNA 202 synthesis in the recipients, and integration. Based on these considerations, we decided to build a 203 hybrid ICE that uses the regulation, excision, and integration components of ICEBs1 and the 204 DNA processing and conjugation components of Tn916. 205

Design and function of a hybrid conjugative element 206
Like many ICEs, both Tn916 and ICEBs1 have a modular organization (Burrus and Waldor,207 2004; Toussaint and Merlin, 2002), with genes involved in different parts of the life cycle 208 clustered ( Figure 1). Because of this modularity, it was relatively straightforward to use the 209 regulatory architecture of ICEBs1 (the genes and sequences at the left and right ends) and to 210 replace the DNA processing and conjugation genes with those from Tn916 (Figure 1c). Such a 211 construct leaves intact the ICEBs1 genes that are required for regulation (immA, immR), and 212 recombination (int, xis). In addition, key DNA sites are also present, including the promoter Pxis 213 that drives transcription of xis and genes downstream, and the left and right ends of the element 214 that contain the recombination sites (attL and attR). Several genes (yddJ, spbK, and yddM) near 215 the right attachment site are not required for ICEBs1 transfer and were omitted from the hybrid 216 element ( Figure 1c). The hybrid element, (ICEBs1-Tn916)-H1, or H1 for short, contains a 217 kanamycin resistance gene (kan) and H1 is integrated in the genome at the normal ICEBs1 218 attachment site in trnS-leu2 (Figure 1c). 219 We found that cells containing (ICEBs1-Tn916)-H1 (ELC1213) exhibited ICE excision 220 frequencies similar to those containing ICEBs1. Cells were grown in LB medium to exponential 221 phase. Element excision was induced through the addition of 1mM IPTG for one hour during 222 exponential growth. At this time, both ICEBs1 and H1 had excised in ~40% of donor cells (Table  223 1). These results indicate that, as designed, the element H1 is activated at levels similar to 224

ICEBs1. 225
Under the same mating conditions as described above, H1 expectedly produced more 226 transconjugants/donor (0.5%) than WT Tn916 (0.0015%), likely due to its increased activation 227 frequency (Table 1). This is consistent with the increased conjugation efficiencies observed for 228 Tn916 mutants with increased excision frequencies due to mutations upstream of tetM, a region 229 critical for Tn916 regulation (Beabout et al., 2015). However, H1 consistently produced 230 approximately 5-fold fewer transconjugants/donor than ICEBs1 (2.7%). This result indicates that 231 some step(s) other than excision from the chromosome of the donor and integration into the 232 chromosome of the transconjugant has a different efficiency than that of ICEBs1. 233 The genes yddJ, spbK, and yddM from ICEBs1 are not present in (ICEBs1-Tn916)-H1 and  234 we reasoned they might contribute to the different conjugation efficiencies. We deleted these 235 genes in ICEBs1 (∆yddJ-yddM::kan) and compared excision and conjugation efficiencies 236 relative to ICEBs1 (∆rapIphrI::kan). We found that ICEBs1 (∆yddJ-yddM::kan) and ICEBs1 237 (∆rapIphrI::kan) behaved similarly ( ICEBs1 and (ICEBs1-Tn916)-H1. In addition, because the recombinase and element ends are the 242 same in ICEBs1 and H1, integration in the recipient cannot be causing the observed differences. 243 The difference in the conjugation efficiency of each element indicates that some aspect of the 244 conjugation functions encoded by Tn916 are less efficient than those encoded by ICEBs1. 245

Hybrid elements can combine different functional components required for conjugation 246
Following excision from the host chromosome, ICEs form circular dsDNA intermediates that 247 are processed into a linear, ssDNA form prior to transfer to a neighboring cell (Wozniak and  We constructed a second hybrid element that is identical to (ICEBs1-Tn916)-H1, except it 261 contains the DNA processing module from ICEBs1 in place of that from Tn916. This element, 262 referred to as (ICEBs1-Tn916)-H2, or H2, encodes the relaxase, helicase processivity factor, and 263 coupling protein from ICEBs1 ( Figure 1). H2 contains the single strand origin of replication from 264 Tn916 (sso916, located downstream of orf19) for priming second-strand synthesis during rolling-265 circle replication (Wright and Grossman, 2016). sso1 of ICEBs1 was not included in H2 (Wright 266 et al., 2015). 267 We found that cells containing H2 (ELC1185) had excision frequencies of ~40%, similar to 268 those of ICEBs1 and H1 (Table 1). The conjugation efficiency of H2 was ~2.0% transconjugants 269 per donor (Table 1). This is similar to the efficiencies observed for ICEBs1 matings, and ~5 fold 270 greater than that of H1. These results indicated that the T4SS encoded by Tn916 can support 271 efficient conjugative transfer between B. subtilis cells and that the ICEBs1-encoded coupling 272 protein can successfully interact with the T4SS encoded by Tn916. They also indicate that the 273 differences in conjugation efficiency between elements using the Tn916 and ICEBs1 machineries 274 are likely due to the relaxase, helicase processivity factor, and/or the coupling protein. 275

ICEBs1 and Tn916 coupling proteins can substitute for each other during conjugation 276
The functional transfer of (ICEBs1-Tn916)-H2 demonstrated that the ICEBs1 coupling 277 protein can successfully interact with the T4SS encoded by Tn916. Coupling proteins must also 278 interact with the substrate that is transferred. In this hybrid element, the coupling protein was 279 interacting with its cognate substrate, the relaxosome (relaxase, oriT, and likely the helicase 280 processivity factor) from ICEBs1. In addition to transferring their own nucleoprotein complexes, 281 both Tn916 and ICEBs1 can recognize and mobilize heterologous plasmid substrates that lack 282 their own conjugation machinery (ICEBs1: pC194, pHP13, and pUB110-based pBS42; Tn916: 283 pC194, pUB110, and probably others) (Lee et al., 2012;Naglich and Andrews, 1988;Showsh 284 and Andrews, 1999). Due to the ability to transfer similar substrates, we suspected that Tn916 285 and ICEBs1 could transfer each other's relaxosome substrates, and that perhaps the encoded 286 coupling proteins could be substituted for one another within the elements. Indeed, preliminary 287 experiments revealed Tn916 and ICEBs1 recognize each other's relaxosomes. Therefore, we 288 investigated the interchangeability of the coupling proteins between ICEBs1 and Tn916. 289 We found that the coupling proteins of Tn916 (Orf21) and ICEBs1 (ConQ) can interact with 290 their non-cognate T4SS and non-cognate relaxosome substrate. We replaced the gene encoding 291 the coupling protein in Tn916, ICEBs1, (ICEBs1-Tn916)-H1, and (ICEBs1-Tn916)-H2 with the 292 homologous gene (conQ or orf21) from the other element (Materials and Methods). In this way, 293 each of these elements encoded a non-cognate coupling protein in association with the DNA 294 processing components. Additionally, for Tn916, ICEBs1, and (ICEBs1-Tn916)-H1 these 295 coupling protein replacements required interactions with the non-cognate T4SS. The H2 296 coupling protein swap (conQ::orf21) produced an element encoding the Tn916 coupling protein 297 and T4SS from Tn916 and the DNA processing components from ICEBs1 (Table 2). 298 Orf21 also functioned in place of ConQ in ICEBs1 and H2 during conjugation. ICEBs1 338 (∆conQ::orf21) (ELC866) transferred with an efficiency ~1%, averaging an approximately 3-339 fold decrease from ICEBs1 (conQ) matings. H2 (∆conQ::orf21) (ELC1450) transferred with an 340 efficiency of ~1%, which was nearly indistinguishable from that of its parent element in side-by-341 side comparisons. 342 Coupling proteins are ATPases from the HerA-FtsK superfamily of ATPases and are 343 responsible for recognizing the transfer substrate and physically delivering it to the rest of the 344 conjugation machinery for export out of the cell. It is note-worthy that these coupling proteins 345 interact with both the non-cognate transfer substrate and non-cognate conjugation machinery. 346 Previous studies determined that some coupling proteins, including TrwB of plasmid R388, can 347 interact with non-cognate conjugation machinery, but not non-cognate transfer substrates (Llosa 348 et al., 2003). It was later determined that many conjugative coupling proteins including TrwB, 349 VirD4 of the canonical Agrobacterium tumefaciens system, and PcfC of pCF10, contain a so-350 called "all-alpha domain" that is responsible for conferring the specificity of substrate 351 recognition to these coupling proteins (Whitaker et al., 2015 replaced the kanamycin resistance gene in ICEBs1 and the hybrid elements with tetM from 368 Tn916. These elements are referred to as ICEBs1-tetM, (ICEBs1-Tn916)-H1-tetM (or H1-tetM), 369 and (ICEBs1-Tn916)-H2-tetM (or H2-tetM). Additionally, donor strains contained a D-alanine 370 auxotrophy (∆alr::cat). Cells bearing this mutation will not grow on or in media without the 371 addition of D-alanine (Wecke et al., 1997) thereby serving as a mechanism for selecting against 372 donors (counter-selecting) in a mating assay without using another antibiotic resistance gene 373 (Brophy et al., 2018). 374 We found that the D-alanine auxotrophy and changing the antibiotic resistance marker in ICE 375 donors did not affect the overall conjugation efficiencies. We compared the mating efficiencies 376 of these donors to the initial donors (alr+, kan rather than tetM in the elements). Donor cells 377 containing Tn916 (ELC1566), ICEBs1-tetM (ELC1795), H1-tetM (ELC1722), or H2-tetM 378 (ELC1725) were grown in LB medium containing 200 µg/ml D-alanine and mated with the B. 379 subtilis recipient CAL419 under standard mating conditions as described above. Conjugation 380 efficiencies were calculated as the number of transconjugants (tetracycline-resistant, D-alanine 381 prototrophs) normalized to the number of donors. These elements exhibited similar conjugation 382 efficiencies as those described for the alr+ donors (Table 1)  When these elements were mated into E. faecalis recipients, H1-tetM, which utilizes Tn916 395 DNA processing and conjugation machinery, consistently produced the most 396 transconjugants/donor (~0.09%) (Figure 2). H2-tetM, which is identical to H1-tetM except for its use of the DNA processing module from ICEBs1, produced ~0.0034% transconjugants/donor, a 398 consistent ~30-fold decrease compared to H1-tetM. These consistent differences indicated that 399 the Tn916 DNA processing module allows for more efficient acquisition by E. faecalis recipients 400 in the context of these hybrid elements. Compared to the hybrid elements, ICEBs1 produced the 401 fewest transconjugants per donor (0.00018%), a nearly 20-fold decrease compared to H2-tetM. 402 H2-tetM differs from ICEBs1 through its use of the Tn916 T4SS and single strand origin 403 (sso916), indicating that one, or both, of these Tn916 features contributed to more efficient 404 acquisition by E. faecalis. Notably, Tn916 produced similar numbers of transconjugants/donor as 405 ICEBs1 (0.00025%). This result is in direct contrast to results observed when these elements 406 were mated into B. subtilis recipients. The lower activation frequency of Tn916 (~0.66 ± 0.08% 407 in donors immediately prior to the start of these matings) compared to that of ICEBs1 (~22 ± 408 3%) did not correspond to lower conjugation efficiencies, indicating that steps downstream of 409 element activation allowed for Tn916 to produce more transconjugants. 410 Similar results were observed when these elements were mated into E. caccae, which was 411 first isolated from human stool samples in 2006 (Carvalho et al., 2006). H1-tetM produced the 412 most transconjugants/donor (0.027%), which was nearly 10-fold more efficient than H2-tetM 413 (0.0033%) (Figure 2). Tn916 produced more transconjugants/donor (0.0011%) than ICEBs1, 414 which produced the fewest transconjugants of the elements tested (2.8 x 10 -5 %). Together, these 415 results indicate that both hybrid elements (H1-tetM; H2-tetM) are more readily acquired by E. 416 faecalis and E. caccae than either parent element (Tn916 and ICEBs1). Notably, Tn916 produced 417 more transconjugants than ICEBs1 into these species, indicating there is a benefit to using Tn916 418 genes required for conjugation. H1-tetM and H2-tetM had the advantage in that they utilize 419 components of Tn916 DNA processing and conjugation machinery, but they activated at higher 420 frequencies in the donor cells (~20 ± 1%, ~22 ± 8%, respectively during these experiments) than 421 Tn916 (~0.66 ± 0.08%) allowing more mating events to occur. 422 In contrast, the relative mating efficiencies were quite different with E. durans as a recipient. 423 E. durans belongs to a distinct phylogenetic Enterococcus group (E. faecium group) than E.

Detecting stable acquisition of conjugative elements in transconjugants 441
We found that ICEBs1 and the hybrid elements did not consistently integrate into the 442 chromosome of the recipient cells, and thus were not stably maintained. We re-streaked transconjugants obtained from these mating assays non-selectively and subsequently checked for 444 tetracycline resistance. We found that fewer than 10% of the apparent transconjugants had 445 maintained the element (Table S1). In the few apparent transconjugants that retained tetracycline 446 resistance, we used arbitrary PCR to map the insertion sites and frequently detected circular 447 ICEs, indicating that the element was not stably integrated into the chromosome ( Tn916 was stably maintained in the transconjugants following non-selective growth. We 459 mapped the integration sites of three transconjugants of each species and identified three unique, 460 frequently intergenic, AT-rich integration sites in each, as expected (Table S1)  chromosomal site, but instead within an AT-rich region that can be found in many places on a 463 chromosome, Tn916 had the advantage in target site selection and thus stable acquisition during 464 these matings. For this reason, although Tn916 produced fewer apparent transconjugants than 465 H1-tetM or H2-tetM, its attachment site availability can confer an advantage during mating events. By using attachment sites from ICEBs1, the hybrid element is more limited in its ability 467 to integrate into heterologous host chromosomes. 468 Because the hybrid element containing Tn916 DNA processing machinery mated more 469 efficiently into E. faecalis and E. caccae than the hybrid containing ICEBs1 DNA processing 470 machinery, we predict that the DNA processing machinery encoded by Tn916 may be better 471 suited for interaction with these Enterococcus species' host machinery. The hybrid elements we generated demonstrate the practicality and efficiency of using the 488 regulation and excision-integration functions of ICEBs1 for studying the conjugation machinery and various functions encoded by other conjugative elements. The conjugation machinery of 490 other elements could easily be used in place of that encoded by ICEBs1, just as was the case for 491 that from Tn916. The ability to induce ICEBs1 by overproduction of the activator RapI should 492 allow investigation of ICEs that may be difficult to study on the population-level due to limited 493 activation. Additionally, the use of such hybrid elements might allow future host range 494 optimization of elements for purposes of genetic engineering. It may be possible to mix and 495 match DNA processing and conjugation machinery from different elements to be compatible 496 with a desired recipient species. However, it is worthwhile to note that although the ICEBs1 497 DNA processing machinery and coupling protein worked with Tn916 conjugation machinery (in 498 H2), this will not always be the case for other hybrid elements; many elements will likely require 499 the use of cognate transfer substrates and coupling proteins for successful transfer.  (Table 3), except BS49, were all derived from JH642, contain the trpC2 527 pheA1 alleles (Perego et al., 1988;Smith et al., 2014) and were made by natural transformation 528 (Harwood and Cutting, 1990). The construction of various alleles and the hybrid conjugative 529 elements is summarized below. 530 531  reading frames of alrA-ndoA (using the same borders as previously described (Brophy et al.,554 2018)) with the chloramphenicol resistance gene from pC194. 555

Construction of hybrid elements 556
(ICEBs1-Tn916)-H1 is a large deletion-insertion that removes the ICEBs1 DNA processing 557 and conjugation genes from the start codon of helP through to 17bp upstream of yddI and inserts 558 Tn916 genes from 121bp upstream of the orf23 start codon to the stop codon of orf13. An 559 additional deletion-insertion removes ICEBs1 genes yddJ-yddM starting 7bp downstream of 560 yddI's stop codon through to the yddM stop codon and inserts a kanamycin resistance gene from 561 pGK67 that is codirectional with the genes in the Pxis operon. These fragments were fused by 562 isothermal assembly (Gibson et al., 2009) with homology arms and transformed into AG174, 563 which contains a copy of ICEBs1 at trnS-leu2, selecting for acquisition of kanamycin resistance. 564 ICEBs1 ∆yddJ-yddM::kan is an insertion-deletion of ICEBs1 identical to that contained in 565 (ICEBs1-Tn916)-H1. 566 (ICEBs1-Tn916)-H2 is an insertion-deletion of the Tn916 DNA processing machinery in 567 (ICEBs1-Tn916)-H1 removing from 121bp upstream of the orf23 start codon to 42bp upstream 568 of the orf19 start codon and inserting ICEBs1 from the helP start codon to the nicK stop codon 569 (such that this element was identical to WT ICEBs1 from attL-nicK). Two 1kb DNA fragments 570 containing DNA flanking this region in (ICEBs1-Tn916)-H1 were amplified and fused with the 571 insert from ICEBs1 and inserted into pCAL1422, a plasmid containing E. coli lacZ (Thomas et 572 al., 2013), cut with BamHI and EcoRI, via isothermal assembly to generate pELC1091. The 573 resulting plasmid was integrated by single crossover into a strain containing (ICEBs1-Tn916)-574 H1. Transformants were screened for loss of lacZ, indicating loss of the integrated plasmid and 575 checked by PCR for replacement of the DNA processing genes, thereby generating (ICEBs1-576 Tn916)-H2. In generating this functional hybrid element, we determined the Tn916 gene orf19, a 577 predicted integral membrane protein is required for conjugative transfer. with ~1kb homology arms from pLW805 [oriT(916), Pspank-orf20-orf22-orf23], a plasmid that 589 contains the upstream genes (orf23, orf22) fused to the downstream orf20 (Wright and 590 Grossman, 2016). This amplification was inserted into pCAL1422 cut with BamHI and EcoRI to 591 generate pELC1068. ∆orf17-orf13 fuses the ninth codon of orf17 to the 302 nd codon of orf13 592 within WT Tn916. Approximately 1kb homology arms were amplified and inserted into 593 pCAL1422 cut with BamHI and EcoRI via isothermal assembly to generate pELC421. These 594 plasmids were used to make deletions as described above. LDW981 (Tn916 [∆orf21]) was 595 generated first and then used to generate ELC423 (Tn916[∆orf21, ∆orf17-13]).

Mating assays 606
Mating assays were performed essentially as previously described (Auchtung et al., 2005). 607 Briefly, donor cells contained an ICE marked with either a kanamycin or tetracycline antibiotic 608 resistance cassette. Donor cells containing either ICEBs1 or a hybrid element were grown in LB 609 medium in the presence of its respective antibiotic (2.5 µg/ml kanamycin or 3 µg/ml 610 tetracycline) to maintain ICE. Donor cells containing Tn916 were grown non-selectively in LB 611 medium. All donors were grown with D-alanine (200 µg/ml), as needed. B. subtilis recipient cells (typically CAL419, unless otherwise indicated) were also grown in LB medium, did not 613 possess any ICE, were resistant to streptomycin (str-84), and were defective in competence 614 (comK) . Enterococcus recipient cells were grown in BHI medium and 615 were D-alanine prototrophs. 616 B. subtilis donor strains were grown for at least three generations in LB medium to an 617 OD600~0.2 before stimulating ICEBs1 or Tn916 activation with either 1 mM IPTG or 2.5 µg/ml 618 tetracycline, as appropriate. After 1h, when donor and recipient cultures were at an OD600~1.0, 619 they were mixed in a 1:1 ratio (5 total ODs of cells) and applied to a nitrocellulose filter for a 620 solid-surface mating. At this point, cells were also harvested for DNA isolation to determine via 621 qPCR the percentage of donor cells with excised ICEs (see below). Filters were incubated on a 622 1.5% agar plate containing 1X Spizizen salts (Harwood and Cutting, 1990) at 37°C for 1-3h. 623 Cells were harvested off the filters and dilutions were plated on LB or BHI plates containing the 624 appropriate selections to enumerate the number of transconjugants. The conjugation efficiency 625 was calculated as the number of transconjugants present at the end of the mating (CFU/ml) 626 divided by the number of donor cells applied to the mating (CFU/ml) (pre-mating donor counts 627 were used to prevent an overestimation of efficiency due to a drop in viability of donor cells 628 during the course of the mating). Where indicated, mating efficiencies were also normalized to 629 the number of donor cells from which ICEBs1 or Tn916 had excised at the start of the mating 630 (see below). 631

Excision assays 632
We used qPCR to quantify excision of these elements. Genomic DNA of donor cultures was 633 harvested using the Qiagen DNeasy kit with 40 µg/ml lysozyme. The following primers were used to amplify the empty ICE attachment site in the chromosome (only present in cells with 635 excised ICE) and a nearby chromosomal locus for normalization (present in every cell). 636 For ICEBs1 and hybrid elements (integrated at trnS-leu2), oMA198 (5'-GCCTACTAAA 637 CCAGCACAAC) and oMA199 (5'-AAGGTGGTTA AACCCTTGG) amplified the empty 638 chromosomal attachment site (attB). oMA200 (5'-GCAAGCGATC ACATAAGGTT C) and 639 oMA201 (5'-AGCGGAAATT GCTGCAAAG) amplified a region within the nearby gene, 640

yddN. 641
For Tn916 (integrated between yufK and yufL) we used previously described primers (Wright 642 and Grossman, 2016). oLW542 (5'-GCAATGCGAT TAATACAACG ATAC) and oLW543 643 (5'-TCGAGCATTC CATCATACAT TC) amplified the empty chromosomal attachment site 644 (att1). oLW544 (5'-CCTGCTTGGG ATTCTCTTTA TC) and oLW545 (5'-GTCATCTTGC 645 ACACTTCTCT C) amplified a region within the nearby gene mrpG. 646 qPCR was performed using SSoAdvanced SYBR master mix and the CFX96 Touch Real-647 Time PCR system (Bio-Rad). Excision frequencies were calculated as the number of copies of 648 the empty chromosomal attachment sites (as indicated by the Cp values measured through 649 qPCR) divided by the number of copies of the nearby chromosomal locus. Standard curves for 650 these qPCRs were generated using B. subtilis genomic DNA that contained empty ICE 651 attachment sites and a copy of the nearby gene (yddN or mrpG). DNA for the standard curves 652 was harvested when these strains were in late stationary phase and had an oriC/terC ratio of ~1, 653 indicating that the copy numbers of these targets were in ~1:1 ratios. 654

Mapping ICE integration sites 655
Arbitrary PCR was used to map ICE integration sites, as previously described (Brophy et al.,656 2018; Das et al., 2005). Following the initial post-mating selection step, transconjugant colonies were re-streaked non-selectively onto a solid medium and subsequently checked for ICE 658 presence by patching for tetracycline resistance. Tetracycline-resistant colonies were used as a 659 template in a PCR reaction containing one arbitrary primer paired with an ICE-specific primer:  Tn916 are indicated by "T"  917 shapes. The current model of transcriptional regulation of Tn916 (A) is adapted from (Roberts 918 and Mullany, 2009). Previously determined origins of transfer (oriT) and single strand origins of 919 replication (sso) are indicated by a "-" above the genetic map (Jaworski and Clewell, 1995 donors also contained amyE::[(Pspank(hy)-rapI) spc] for IPTG-inducible overproduction of RapI to stimulate element excision (activation).
b The genetic stability of these elements in transconjugants was evaluated. One hundred transconjugants were re-streaked nonselectively and then patched to check for tetracycline resistance (indicative of ICE presence). Arbitrary PCR was used to map ICE insertion sites in three isolates that had maintained their elements from each mating pair (TC #1-3 sites). "Circular" indicates that circular ICEs were detected by PCR rather than a chromosomal insertion site. For the detected integration sites, a genomic context of 26bp (Tn916 insertions) or 17bp (ICEBs1, H1, H2 insertions) is shown. The genes or predicted gene products of insertion sites are indicated, unless the site is in a predicted intergenic region. Gene key: HK = histidine kinase, DAHP = DAHP synthase, AAP = amino acid permease, SDH = serine dehydratase, HP = hypothetical protein. *indicates last 10% of ORF. **indicates this site was also identified by (Brophy et al., 2018).