The identification in Streptomyces genomes of ICEs and IMEs was achieved using ICEScreen (unpublished), an automated and improved version of the two-way detection method developed by Ambroset et al. (2016) and Lao et al. (2020). Briefly, signature proteins corresponding to the recombination module (tyrosine or serine integrases) or the conjugation module (i.e., relaxase, coupling protein, VirB4 translocase of T4SS) are firstly identified with hidden Markov model profiles (HMM) from Lao et al. (2020) and secondly with BlastP searches (NCBI default parameters on nr database; Altschul et al., 1990) using protein tags from a curated database (Ambroset et al., 2016; Lao et al., 2020). In a similar strategy, AICEs were identified by the detection of signature proteins corresponding to the site-specific integrase (tyrosine or serine integrases), the replication function (Rep) and the transfer protein (TraB) with HMMsearch (Eddy, 2009) from the HMMER server (Finn et al., 2011) combined with blast search of known signature proteins (Ghinet et al., 2011) and HMM profiles (Mistry et al., 2021). HMM profiles (Supplementary Table S1) were either retrieved from Pfam 35.0 database (Mistry et al., 2021) or built from protein alignments for Streptomyces TraB (MUSCLE; Edgar, 2004) using HMMbuild from HMMERv3.3 with default parameters in Ugene platform (Okonechnikov et al., 2012). Identification of direct repeats flanking the element was manually performed by searching almost identical sequence patterns at the borders of each identified element. Cargo genes were assigned to functional categories (metabolism, DNA metabolism, resistance, regulation, phage, toxin/antitoxin, signaling, transposase and hypothetical proteins) using BlastP searches (E-value <10⁻⁶, coverage >80%).

All sequence analyses were performed with Ugene platform (Okonechnikov et al., 2012). MUSCLE alignments (Edgar, 2004) of protein sequences were performed with default settings and 3 iterations. Identity matrices were calculated based on previous alignments. A Neighbor-Joining tree was built with Mega X (Kumar et al., 2018). Synteny breaks between similar elements were identified by pairwise comparison with MAUVE (Darling et al., 2004). Comparison and visualization of the nucleotide identities along the integrated elements were performed using Artemis Comparison Tool (Carver et al., 2008) after BlastN alignment.

The Streptomyces strains used in this study were isolated from mm-scale soil aggregates collected from a centimeter-scale rhizosphere population (Tidjani et al., 2019a,b; Supplementary Table S2). Strain storage, culture conditions and DNA extraction procedures were performed as described in Kieser et al. (2000). Briefly strains were grown on Mannitol Soya Flour medium (SFM) agar plates or grown in Hickey-Tresner (HT) medium liquid culture and mating experiments were performed on R2YE medium plates at 30°C. Kanamycin, apramycin and hygromycin antibiotics were used at 50 μg/ml (Kieser et al., 2000). Mitomycin treatments (125 ng/ml, corresponding to the quarter of the minimal inhibitory concentration) to induce ICE excision were performed in HT medium (OD 600 nm = 0.2) for 3 h (Volff et al., 1993; Bellanger et al., 2007).

To select recipient strains after mating, an antibiotic resistant gene (either neo for kanamycin or aac (3) IV for apramycin resistance) was inserted by homologous recombination at the chromosomal locus. The recombinant plasmid consisted of a pWED2 (Karray et al., 2007) derived suicide vector (this work) containing upstream and downstream homologous regions to the targeted locus flanking the hygromycin resistance gene (hyg). Labelled strains were obtained following a double crossover between plasmid-borne and chromosome homologous sequences and selection for resistance. Constructs were obtained by Overlap Extension PCR (Hilgarth and Lanigan, 2020) with primers listed in Supplementary Table S3. Constructs were introduced into Streptomyces sp. by conjugation with Escherichia coli ET12567/pUZ8002 (Kieser et al., 2000).

Pock assays enabling directly visualization of conjugal AICE transfer were performed as described in (Huang et al., 2003). Briefly, 50–100 spores of the donor strain were spread on a lawn of the recipient strain (10⁶ spores) on R2YE medium plates. Pocks resulting from a growth delay of the recipient cells were observed by naked eye after 5–7 days of incubation at 30°C. At least three replicates were performed for each couple of donor and recipient strains. AICE flux between pairs of strains were represented using the CIRCOS tool (Krzywinski et al., 2009). To assess AICE transfer, clones were picked from pocks and streaked using sterile toothpicks on SFM media with kanamycin, allowing the selection of recipient strains. AICE transfers in streaks were confirmed by PCR using AICE-specific primers and primers targeting specific genomic sequences discriminating recipient and donor strains (Supplementary Table S3). To determine whether multiple AICE transfers resulted from the acquisition of multiple AICEs in a genome or from single transfers in different genomes, independent recipient clones were isolated from a streak in a second round of isolation and tested by PCR for AICE transfer.

Whole-genome sequences of 11, very closely related clones isolated from a mm-cm scale rhizosphere population were mined for the presence of conjugative elements. First, we screened for four markers of ICEs using the tool ICEscreen, namely the recombination module (tyrosine or serine integrases), the conjugation module (relaxase), the coupling protein (CP) and the VirB4 translocase of T4SS complex. The presence of all four functions was used to detect and classify ICEs, whereas elements that lacked T4SS were classified as IMEs. ICEs that have lost the integrase and/or the coupling protein and/or the relaxase were classified as defective ICEs (DICEs). Three putative ICEs were identified in the population (Figure 1; Supplementary Table S4). Two were found on the chromosomes of the closely related strains RLB1-8 and RLB3-5 (99.9% ANI) and not in their closest conspecific strains, which could be consistent with horizontal inheritance. Both encode 87 CDSs, share the same insertion site (tRNA-Thr) and produce the same terminal direct repeats upon integration into the chromosome (93 nucleotides). The third complete ICE was found on plasmid pRLB3-6.1 (present in the strain of the same name). It had an integrase distinct from that of the chromosomal element (no amino acid identity) and exhibited limited nucleotide similarity to the coupling proteins (48%), the relaxases (42%) and the virB4-like proteins (33%) of the chromosomal element. We identified three DICEs. One was close to the middle of the chromosome and present in all strains. One was shared by strains RLB1-9, S1A1-3, S1A1-8, and RLB3-17, this latter being located in the opposite arm compared to others. Similarly, the third DICE is shared by S1D4-23, S1A1-7, and RLB3-6, but in the opposite arm in RLB3-6. Finally, an IME was identified in all strains but RLB3-6, S1D4-23, and S1A1-7, with a single loss event in that clade being the most parsimonious scenario (Figure 1).

Second, we screened for AICEs using transfer protein (TraB) as a signature (Ghinet et al., 2011; Bordeleau et al., 2012) with HMM profiles and tBlastn searches. When elements thus identified possessed the integrase (int) and replicase (rep) signature genes, they were considered putatively functional and classed as an AICE. When the replicase and/or integrase was missing, the element was classed as a defective AICE (DAICE). A total of 51 traB genes flanked by direct repeats, were retrieved in the 11 chromosomes. TraB proteins showed a great disparity in sequence and in size (ranging from 139 to 738 amino acids). The four smallest versions of TraB proteins were associated with genes encoding an ATP/GTP-binding protein gene and a partitioning protein (ParM). Blast research systematically attributes the tripartite system ATP/GTP binding protein; short TraB; ParM protein to Streptomyces extrachromosomal elements. This could be an original chromosomal conjugation system derived from the integration of a conjugative plasmid. A phylogenetic analysis (Supplementary Figure S1) showed that the TraB proteins cluster with different Streptomyces TraB families (Ghinet et al., 2011), suggesting diverse evolutionary origins. We used a 95% amino acid identity threshold consistent with this clustering to group TraB proteins into 19 families, 7 of which were represented by only a single TraB protein (Supplementary Figure S1). All 51 traB genes identified on the chromosomes were flanked by direct repeats (2–94 nt in length) and delineated regions from 11.9 to 36.2 kb (Supplementary Table S5). Forty-three traB genes were associated both with a replicase and an integrase gene, suggesting a functional AICE. Eight remaining traB genes were not associated with functional replication or integration genes (six lacked replication signature genes, one lacked a canonical integrase gene and one harbored a transposon interrupting the integrase gene; Figure 1; Supplementary Table S5).

Four additional TraB ranging from 139 to 789 amino acids were found on extrachromosomal elements (data not shown). Because of their plasmid location, their divergence with chromosomal traB and the absence of direct repeats, they were discarded from further analyses although we cannot exclude the possibility that, these elements are functionally conjugative. Overall, each strain of the population harbored 5–10 elements (7.2 per genome on average), mostly represented by AICEs (Figure 1). In terms of size, IMEs/ICEs (86.0–99.7 kb) far exceed the size of the AICEs/DAICEs (11.9–36.3 kb), mostly depending on the presence of accessory genes (Supplementary Tables S4, S5).

Functional chromosomal conjugative elements are exclusively distributed in the central region of the chromosome (Figure 1), i.e., in the region containing the 971 genes composing the core genome of the Streptomyces genus in between the left- (2.4 Mb) and right- (1.4 Mb) chromosomal arms which are devoid of conserved genes (Lorenzi et al., 2019). The insertion sites are located mostly in tRNA genes (n = 36) and also in protein-coding genes (n = 14) or intergenic space (n = 1; Supplementary Table S5). This preference of integration in such conserved targets could contribute to the high prevalence of elements in the core region. Only defective elements (one IME and two DICEs) were encountered in the arms. Their decay could result from intense recombination in these regions (Fischer et al., 1998; Choulet et al., 2006; Lorenzi et al., 2019). The subsequent formation of numerous insertions/deletions could make these elements non-functional until they eventually disappear from the genome.

Closely related strains show insertion sites that harbor the same element suggesting its vertical inheritance (Figure 1; Supplementary Table S5). For example, AICE family 13 is only shared by the strains of the same clade (RLB1-9, S1A1-3 and S1A1-8) and could have been acquired by their most recent common ancestor. On the other hand, some distributions (family 03, 11, 17, and 18) clearly suggest that the population is subjected to an intense flux of AICEs resulting from HGT and/or differential gene loss. For instance, AICE family 11 is present in several closely related strains (S1A1-3, S1A1-8, RLB3-17, S1D4-23, and RLB3-6) and is probably lost in strain RLB1-9. About one third of the elements (10 AICEs and 4 DAICEs) were present in only one strain, suggesting recent acquisition from unsampled donor strains. The fact that the insertion sites remain empty in some strains shows that they probably did not encounter the above elements and remain available for further gene acquisition.

In one case, several elements are integrated within the same insertion site (tRNA-Arg, Figure 2). One to three elements, AICEs or DAICEs belonging to three different families, are flanked by identical direct repeats. An additional trace of an ancient element, devoid of traB, is also present at this site. This phenomenon is called accretion (Bellanger et al., 2014) and corresponds to the successive integration of different elements having the same integration specificity. Notably, the different AICE families integrated at a single site (for instance tRNA-Arg) harbor different integrases (having 39%–88% of amino acid identity). Thus, there is no correlation between integrase sequence and integration site.

In some cases, elements belonging to the same family can be found at different insertion sites, as illustrated by a family 3 element (Figure 3). The two AICE03 in strains S1A1-7 and S1D4-23 that share the same integrase are inserted in a tRNA-Val gene while AICE03 in RLB3-6 possesses a divergent integrase and excisionase and is found inserted in rlmN, a 23S methyltransferase encoding gene. The direct repeats flanking these elements are different in sizes and in sequences (46 nucleotides for S1A1.7 and S1D4-23 vs. 11 nucleotides for RLB3-6; Supplementary Table S5). Switching the integrase thus may provide a way to diversify the integration site of an element. In addition, the comparison of the association of the three signature genes (traB, rep and int) within a same family showed that exchanges of functional modules can occur between elements (Supplementary Table S6; Supplementary Figure S2), leading to changes in integration specificity as previously seen in mobile genetic elements (Garriss et al., 2009). While the integrase can be considered the driver of integration specificity, the same site can host elements with different integrases as demonstrated by the integration at the tRNA-Thr gene of family 13 AICE and the chromosomal ICE (Figure 1; Supplementary Table S5).

Most AICEs exhibited high diversity in both cargo and mobility genes (Figure 3; Supplementary Figures S2, S3). Syntenic AICEs with identical gene content were only identified in very closely related strains (for instance for the family 01 of the cluster RLB1-8; RLB3-5; S1D4-14; S1D4-20 or family 14 of the RLB3-17; RLB1-9; S1A1-3; S1A1-8; Figure 1). Most of the diversity results from deletion and/or gene replacement occurring within the element. Recombination events resulting in losses of integrase or replicase genes are at the origin of DAICEs (Figure 4; Supplementary Figure S4). Some strains exhibit an empty integration site (RLB1-9, S1A1-3, and S1A1-8), when strain S1D4-23 had a genuine AICE (Figure 4). In other strains, only a partial integrase, cargo genes and two direct repeats [both related to those of the AICE06 (S1D4-23)] are suggestive of a former presence of AICEs at this site. It has to be noted that these latter could not be detected by our in silico approach and were here identified by the comparison of the genomes at the integration site of the AICE06 (S1D4-23). It is possible that the presence of flanking direct repeats could result in the future accretion of an AICE and a possible mobilization in cis of these remaining genes.

Apart from genes mediating excision, integration, replication and intra-and inter-mycelial transfer, ICEs and AICEs also harbor cargo genes that can affect the fitness of the host cell. In silico analyses identified 541 unique genes in AICEs of which 445 are cargo genes. About 50% (223) of them were shorter than 150 amino acids, the threshold applied in our genome annotation process to get rid of probable pseudogenes (Tidjani et al., 2019a). The high percentage of putative pseudogenes in comparison to the rest of the genomes (ca. 7%), indicated that many AICE cargo genes are undergoing pseudogenization (Liu et al., 2004a). Furthermore, in order to identify as many functions as possible, a manual analysis of the shortest CDS was performed and enabled to identify the function of few conserved short proteins. The 222 proteins (ca. 50%) of size greater than or equal to 150 amino acids were defined as cargo gene products (Supplementary Table S7). Figure 5 shows the distribution of the 222 considered cargo genes in functional categories. For a subset of these genes, it is possible to assign (hypothetical) function. Genes that could be beneficial for the host cell included those implicated in antibiotic resistance (e.g., aminoglycoside phosphotransferase), cell signaling and metabolism (e.g., flavoprotein). Moreover, genes involved in illegitimate (encoding ligase) or homologous recombination (e.g., radA) which can provide additional or altered specific repair capacities or allow integration of foreign DNA (Supplementary Table S7) were identified. In addition, predicted protection systems against the entry of foreign DNA such as phages (e.g., Restriction and Modification or AIPR abortive phage infection systems) were detected. Toxin/antitoxin systems carried by some AICEs are likely addiction systems involved in the persistence of the mobile element, not expected to be beneficial to the cell (Song and Wood, 2020). The same analysis was performed on the 87 CDS of the chromosomal ICE which harbored 81 cargo genes of which 23% (n = 19) were shorter than 150 amino acids. Their assignment into functional categories was similar to that of AICE cargo genes (Figure 5; Supplementary Table S8).

To assess the mobility of AICEs, we tested their conjugative ability by performing all pairwise mating combinations between the 11 strains (i.e., each strain was used as both a potential donor and a potential recipient). Conjugation events were assessed by direct visualization of pocks on mating plates. Pocks (2–5 mm) were observed in 15 out of the 110 tested pairwise combinations in seven strains that acted as donors, recipients or both (Figure 6). Horizontal gene transfer within pocks was screened by PCR for four selected mating couples (Figure 7; Supplementary Figure S5). These different couples represented situations with the potential to transfer one to six autonomous AICEs. The RLB3-6 donor strain harbored the plasmid pRLB3-6.1 on which no spd gene was identified, ruling out its participation to pock formation in our experiments. In a first step, we screened a mixture of cells picked from independent pocks. In one case, only one element was found to have transferred from the donor to the recipient in three out of three pocks (S1A1-3 × RLB1-9; Figure 7A) whereas in three other mating pairs several elements were found to have horizontally transferred within individual pocks (25 pocks for each couple; Figures 7B–D).

As a pock is composed of many different recipient cells, we could not conclude at this stage whether this phenomenon was the result of the concomitant transfer of several elements in a single cell, or of multiple transfers of single elements in independent cells. We therefore selected an independent pock showing evidence of transfer of multiple elements in three couples (RLB3-6 × RLB1-9; RLB3-6 × S1A1-8 and S1D4-23 × RLB1-9) and isolated independent recipient clones that undergone a sporulation cycle ensuring a unigenomic stage of the analyzed clones. Each of them was screened by PCR for each previously transferred element. In all three pairings, concomitant transfers from two to three AICEs were observed in independent recipient cells (Figure 7). Depending on the donor-recipient couple, multiple transfers occur in 13%, 33%, and 61% of the exconjugant clones.

As ICE transfer is not accompanied by pock formation, the potential transfer of ICE01 was tested by genetically marking this element in strain RLB1-8 with apramycin resistance (see MM section). Mating of RLB1-8 with nine other strains marked with kanamycin resistance were performed, but no evidence of ICE transfer could be proved under our experimental conditions. As the excision of many chromosomally integrated elements such as phages or ICEs is repressed under lab conditions, their excision is often stimulated by genotoxic agents that induce the SOS response (Beaber et al., 2004). After exposure of the donor strain to the genotoxic antibiotic mitomycin C (125 ng/ml), the circularized ICE form could be detected by PCR analysis with primers targeting the attP sequence (data not shown) and the circularization was confirmed by sequencing the predicted attP sequence of 93 pb (Supplementary Table S4). Interestingly, a gene encoding a Helix-Turn-Helix DNA binding domain and a nudix hydrolase domain which could be related to the cI repressor family was identified on ICE01 (data not shown). Altogether these results suggest that this ICE element is functional and mobile under other conditions, as demonstrated for the TR ICEs in S. turgidiscabies (TR for toxicogenic region; Chapleau et al., 2016).