The SGI1/PGI1/PGI2/AGI1 islands were identified with the nucleotide BLAST in the National Center for Biotechnology Information (NCBI) databases (nucleotide collection and whole-genome shotgun contigs databases) available until March 31, 2021¹ first using the integrase gene and second using the complete backbone of the GIs as a query sequence. We searched for SGI1/PGI1/PGI2/AGI1 islands in alpha-, beta- and gamma-proteobacteria and retained only GIs with syntenic genes. Redundant backbone nt sequences found in the strains of the same species were excluded. SGI1-REs with the same int_GI homology but with no or identical backbone variations (insertions, deletions, nt sequence exchanges, or ≥0.2% of nt substitutions) were retained when they were detected in different bacterial species to determine the distribution of the SGI1-REs. All identified uncharacterized GIs were analyzed using BLAST and the G+C content of the backbones was determined. The 18-bp attachment sites (attL and attR) were analyzed and the position of the MDR region, when present, was determined. Putative AcaCD-binding sites were predicted using FIMO (MEME suite) based on the AcaCD logo, as presented in the study² of Carraro et al. (2014).

Multiple alignments of nucleotide sequences, including the 37 fully characterized backbones to date and the uncharacterized GIs retained for this study, was performed using the online MAFFT software version 7 available at the Japanese Computational Biology Research Center (CBRC)³. The 37 known SGI1-REs included 18 SGI1 variants: SGI1 (Boyd et al., 2001), SGI1-K1 (Levings et al., 2007), SGI2 (Levings et al., 2008), SGI1-V (Siebor and Neuwirth, 2011) recently renamed SGI-V (de Curraize et al., 2020a), SGI1-PmMAT (Siebor and Neuwirth, 2013), SGI1-B2 and SGI1-Z (Lei et al., 2015), SGI1-L (Schultz et al., 2017), SGI1-Pm2CHAMA (de Curraize et al., 2018), SGI1-PmJN40 (Bie et al., 2018), SGI1B-Ec1 (Cummins et al., 2019), SGI1-PmSC1111 (Wang et al., 2019), SGI1-PmCA11 (Xiao et al., 2019), SGI0 (de Curraize et al., 2020a), SGI1-LK1 (de Curraize et al., 2020b), SGI1-Vc2CHAMA and VGI (Cummins et al., 2020), and SGI1-XJ9S (Lei et al., 2020a); 3 PGI1 variants: PGI1-PmESC (Siebor et al., 2016), the PGI1 variant of Salmonella enterica Heidelberg SL476 named here PGI1-SL476 (Siebor and Neuwirth, 2014), and PGI1-PmPEL (Girlich et al., 2015); 3 PGI2 variants: PGI2 (Lei et al., 2018), PGI2-Ec-2, and KGI (Cummins et al., 2020); and 12 AGI1 variants belonging to 5 lineages: AGI1 (Hamidian et al., 2015), AGI1-A (renamed AGI2), AGI1-B and AGI1-C (renamed AGI3) (Siebor et al., 2019; Siebor and Neuwirth, 2020), AGI1 variants of Escherichia coli MOD1-EC6520, K. pneumoniae k1781 (AGI4), of 3 Salmonella [S. Agona 24.H.04 (AGI4), S. Cubana 76814 (AGI5) and S. Stanley DMS 1112 (AGI1 variant)], V. cholerae 133-73 (AGI2) and 4874 (AGI3), and Vibrio sp. 2017V-1124 (Siebor and Neuwirth, 2020). The IMESsW3-18-1 of Shewanella sp. W3-18-1 originally described as Shewanella putrefaciens was also included as a reference and was renamed SGI1-RE5SsW3-18-1 (Qiu et al., 2013; Siebor and Neuwirth, 2020; de Curraize et al., 2021) (Supplementary Table 1). The relationships among these GIs were determined based on the phylogeny of their int_GI gene (Boyd et al., 2009). A phylogenetic analysis of the complete GI backbones was also performed. Neighbor-joining phylogenetic trees with 100 bootstrap replicates were generated from the multiple alignments and were viewed on Phylo.io accessed at CBRC webserver. The percent identity matrix of all int_GI genes and of the complete backbones of the SGI1-REs taken as a reference in each cluster was created by Clustal Omega⁴.

In addition to the 37 fully characterized SGI1-REs, we detected 89 new and non-repetitive SGI1-REs that were identified among 7 bacterial orders belonging to gamma-proteobacteria (Supplementary Table 1). These new SGI1-REs were all inserted at the 3′-end of the trmE gene, as were all SGI1-REs described to date, except for PGI1-SL476 and two other SGI1-REs of Salmonella Heidelberg that were located at an alternative site in the intergenic region between the superoxide dismutase (sodB) gene and the transcriptional repressor (purR) gene. They were found in Enterobacteriales (n = 42): Salmonella enterica (33), E. coli (2), K. pneumoniae (2), P. mirabilis (2), Cronobacter sakazakii (1), Morganella morganii (1), and Providencia stuartii (1); Vibrionales (27): V. cholerae (19), V. mimicus (3), Vibrio parahaemolyticus (3), Vibrio fluvialis (1), and Vibrio navarrensis (1); Alteromonadales (15): Shewanella sp. (3), Shewanella algae (3), Shewanella vesiculosa (2), Shewanella fodinae (1), Shewanella frigidimarina (1), Idiomarina tyrosinivorans (1), Idiomarina sp. (1), Marinobacter adhaerens (1), Marinobacter lutaoensis (1), Marinobacter sp. (1); unclassified gamma-proteobacteria (2): Pseudohongiella nitratireducens (2); Pseudomonadales (1): Pseudomonas aeruginosa (1); Chromatiales (1): Rheinheimera nanhaiensis (1); and Oceanospirillales (1): Halomonas meridiana (1). The length and G+C content of all backbones ranged from 16,505 to 30,708 bp and from 42.90 to 50.51%, respectively (Supplementary Table 2).

Phylogenetic analysis based on the nt sequences of the 126 integrase genes revealed the presence of 7 clusters. Between the members of each cluster, the int_GI gene homology was greater than 97% (Supplementary Tables 3B–G). Of the 89 new GIs retrieved, 32 SGI1-REs belonged to the SGI1 cluster (cluster 1), 23 to the PGI1 cluster (cluster 2), 11 to the PGI2 cluster (cluster 3), and 7 to the AGI1 cluster (cluster 4). The results revealed 3 additional GI clusters (Figure 1). The first cluster (cluster 5) composed of 8 GIs included SGI1-RE5SsW3-18-1 from Shewanella sp. W3-18-1. The second cluster (cluster 6) consisting of 6 GIs included a GI found in I. tyrosinivorans CC-PW-9, a species isolated from marine environments, named here as SGI1-RE6ItCC-PW-9 and used as a reference for this group since it was the only one harboring an MDR region. The last group (cluster 7) included 2 GIs from Marinobacter and the GI of M. adhaerens KG14 was named SGI1-RE7MaKG14. The nt identity of the int_GI gene of the 7 SGI1-REs used as a reference is presented in Supplementary Table 3A. Int_PGI1 and int_PGI2 were the closest (94.98% nt identity), as were int_SGI1 and int_{SGI1–RE5SsW3–18–1} (94.99%). Int_{SGI1–RE6ItCC–PW–9} and int_{SGI1–RE7MaKG14} shared only 71–76% with int_{SGI1/PGI1/PGI2}. Int_AGI1 was the most distant from the other 6 int_GI (≤64.82% nt identity). The Int_GI integrase catalytic residue was conserved for each detected SGI1-RE (Supplementary Figure 1).

New SGI1-REs from the SGI1 cluster were detected only in Enterobacteria and in the members of the genus Vibrio. SGI1-REs from 23 additional Salmonella of various serovars [including SGI1-D and SGI1-F (Levings et al., 2005) and an SGI1-K variant (Hawkey et al., 2019)] were included in this cluster due to variations in their backbone (int_GI ≥ 99.92% nt identity). SGI1-REs belonging to the SGI1 cluster were also found in K. pneumoniae (1) and P. stuartii (1) (99.92 and 99.59% of nt identity with int_SGI1, respectively). It should be noticed that among the published SGI1 variants, int_SGI–V and int_SGI0 harbored by P. mirabilis have only 99.59 and 99.10% of nt identity with int_SGI1, respectively. SGI1-REs were also recovered from 6 strains of V. cholerae and one strain of V. parahaemolyticus (int_GI ≥ 99.67% of nt identity with int_SGI1) (Supplementary Table 3B).

Salmonella genomic island 1-related elements of the PGI1 cluster were found in more diverse bacterial species, including SGI1-REs in Enterobacteriales: E. coli (1) and Salmonella (9); in Vibrionaceae: V. cholerae (6), V. mimicus (1) and V. navarrensis (1); in Shewanella: S. algae (2), S. fodinae (1) and Shewanella sp. (1); and also in R. nanhaiensis (1), a Gram-negative rod-shaped bacterium isolated from deep-sea sediment in the South China Sea at a water depth of 1,800 m (Zhang et al., 2012). The int_GI showed ≥98.68% of nt identity with int_PGI1–PmESC (Supplementary Table 3C).

Eleven SGI1-REs belonged to the PGI2 cluster: SGI1-REs in C. sakazakii (1), E. coli (1), K. pneumoniae (1), M. morganii (1), in V. cholerae (3), V. mimicus (2), V. parahaemolyticus (1), and for the first time, a GI of the SGI1-RE family was detected in P. aeruginosa (1). Their int_GI showed ≥98.94% of nt identity with int_PGI2 (Supplementary Table 3D).

Only 7 new SGI1-REs were found in the AGI1 cluster, a group gathering recently discovered AGI1 variants in the GenBank^® (Siebor and Neuwirth, 2020). They were detected in V. cholerae (3) and V. fluvialis (1). For the first time, SGI1-REs belonging to the AGI1 cluster were found in 2 strains of P. mirabilis (int_GI = 99.65% of nt identity with int_AGI1) (Supplementary Table 3E). The SGI1-RE of Vibrio sp. 2017V-1124 whose int_GI shared only 96.37% of nt identity with int_AGI1 was classified in this cluster (98.70–99.22% AA identity with the Int_GI of this cluster).

Salmonella genomic island 1-related elements not belonging to the SGI1/PGI1/PGI2/AGI1 clusters were identified in 16 strains isolated mainly from water or marine samples. These SGI1-REs belonged to 3 new clusters. The first cluster (cluster 5) found only in Shewanella and Vibrio species included SGI1-RE5SsW3-18-1 from Shewanella sp., W3-18-1 and other Shewanella sp. (2), S. vesiculosa (2), S. algae (1), S. frigidimarina (1), V. cholerae (1), and V. parahaemolyticus (1). Compared to the int_GI used as a reference in each cluster, int_{SGI1–RE5SsW3–18–1} and int_SGI1 were the closest (94.99% nt identity), and int_{SGI1–RE5SsW3–18–1} and int_AGI1 were the farthest (64.34% of nt identity) (Supplementary Table 3F). The int_GI of Shewanella and Vibrio shared ≥98.28% and ≥97.54% nt identity with int_{SGI1–RE5SsW3–18–1}, respectively.

The second cluster (cluster 6) included 6 SGI1-REs found in I. tyrosinivorans (1), a Gram-negative curved rod-shaped marine bacterium isolated from the estuarine water off Pingtung in Taiwan (Hameed et al., 2016) and Idiomarina sp. (1); in M. lutaoensis (1), a marine bacterium isolated from a hot spring on the coast of Lutao in Taiwan (Shieh et al., 2003); in H. meridiana (1), isolated from the hydrothermal plume seawater in the Northeast Pacific Ocean at a depth of 2,000 m (Kurihara et al., 2020); and P. nitratireducens (2) isolated from seawater in the northern South China Sea at a 450 m water depth (Xu et al., 2016). Their int_GI showed ≥99.84% of nt identity with int_{SGI1–RE6IdCC–PW–9} (Supplementary Table 3G).

The last cluster (cluster 7) contained 2 SGI1-REs detected in M. adhaerens (1) and Marinobacter sp. (1) isolated from the traditional Korean soy sauce and Mediterranean Sea water, respectively; their int_GI showing only 89.52 and 89.60% of nt identity with int_{SGI1–RE6IdCC–PW–9}, respectively. The clusters 6 and 7 were more distant from the other clusters sharing only 62–75% nt identity with int_AGI1 and int_PGI1, respectively (Supplementary Table 3G). We have chosen to group clusters 6 and 7 together for further analysis.

Phylogenetic analysis of the nt sequences of the complete backbones of the 126 SGI1-REs showed that the SGI1 cluster was the most homogeneous. Greater diversity was observed particularly in the PGI1 and PGI2 clusters (Figure 2A). Moreover, the SGI1-RE of E. coli KK-NP016 harboring int_PGI2 was found in the PGI1 cluster, with a major part of its backbone sharing nt identity with PGI1 regions. The nt identity of the backbones used as a reference showed that PGI1-PmESC and SGI1-RE6ItCC-PW-9 were the farthest from SGI1, PGI2, and AGI1 (≤58.25 and ≤53.55% nt identity) (Supplementary Table 4). These results were consistent with the branch length values shown in Figure 2A.

The backbone of the SGI1-REs could be divided into two distinct parts. The first part, from the attL attachment site to the oriT transfer origin, was highly conserved and contained int_GI, xis (S002), rep (S003), S004, traN (S005), sgaC (S006), +/−sgaD (S007), traG (S011), +/−traH (S012), mpsB (S019), and mpsA (S020) genes (de Curraize et al., 2021), and oriT known as a high potential recombination site. The second part of the backbone, from oriT to the 3′-end of the backbone (attR attachment site), varied between SGI1-REs. The phylogenetic trees obtained for the complete backbones and parts from attL to oriT were similar (Figures 2A,B). Compared to the phylogenetic analysis of the int_GI genes, the same clusters were recovered with the first parts of the backbones of SGI1-REs, except for the SGI1-RE of E. coli KK-NP016, which was phylogenetically closer to PGI1, despite the presence of an int_PGI2 gene. The part of the backbone from oriT to attR was generally characterized by the presence of a res gene. When an MDR region containing a class 1 integron was present, it was inserted upstream of res. Phylogenetic analysis of this part of the backbone showed greater dispersion among the clusters (Figure 2C).

The synthetic description of the characteristics of each cluster of SGI1-RE is reported in Figures 3A–F.

All SGI1-REs were integrated at the 3′-end of trmE, except for the 3 PGI1 variants of S. Heidelberg that were inserted into the intergenic region between the sodB and purR genes. The 18-bp attachment sites (attL and attR) were identified (Supplementary Table 2). Based on the analysis of the att sites after conjugation experiments, the last 9-bp of att sites were examined (Siebor and Neuwirth, 2020). As expected, the last 9-bp of attR sites corresponding to the 3′-end of the trmE gene were variable, as trmE genes are quite different among the bacterial species (Doublet et al., 2007) (Supplementary Figure 6). Only two different sequences were detected for the last 9-bp of attL sites corresponding to the attP site of the circular form of SGI1-REs. The 9-bp GGGAAGTAA sequence was found in all SGI1-REs of SGI1 cluster, and GGGAAGTGA in SGI1-REs of the other clusters (PGI1, PGI2, AGI1 clusters, and clusters 5, 6, and 7), except for SGI1-REs of V. mimicus 2011V-1073 and 08-2414 from PGI2 cluster, and that of M. adhaerens KG14 from cluster 7, which were the same as for SGI1 cluster.

Complex GIs consisted of two elements sharing the same attachment site and integrated in tandem array. GI tandems were found in 2 strains, in addition to the one previously described in P. mirabilis JN40. In each case, a classical SGI1-RE (from different cluster) was inserted at the 3′-end of trmE and was followed by a GI of other families that targeted the same attachment site. The GI in tandem (23,741 bp) with SGI1-PmJN40 from P. mirabilis JN40 showed 100% of nt identity with GIs found at the 3′-end of trmE in P. mirabilis strains isolated from wild animal samples (CP073245, CP053681, CP053682, and CP053614). It harbored 17 ORFs and contained genes encoding an integrase, DNA binding proteins (including an excisionase family protein), transcriptional regulators belonging to the AlpA family phage regulatory protein and the XRE family transcriptional regulator, a toxin-antitoxin module, and a DEAD/DEAH box helicase (Bie et al., 2018). The int_GI gene showed 89% of nt identity with that of V. parahaemolyticus strains (CP012950 and CP034305) whose GIs were also integrated at the 3′-end of trmE. The last 9-bp of the attL site were identical to those of the PGI1 cluster.

A GI not belonging to the SGI1-RE family was located at the right end of SGI1-REs in V. cholerae 555448 of the SGI1 cluster, and in V. cholerae N2775 of the PGI2 cluster. The former (19,970 bp) harbored 17 ORFs including genes encoding an integrase, a DNA-binding protein (excisionase family), transcriptional regulators (AlpA and XRE families), a methyltransferase, a cysteine desulfurase, and a PAPS sulfotransferase. The backbone shared approximatively 90% of nt identity (84% coverage) with the GIs of V. parahaemolyticus strains (CP034305 and CP051111) also located at the 3′-end of trmE. However, the int_GI gene was only 93.25% identical to that of V. parahaemolyticus 2012V-1165 (CP051111).

The GI in tandem V. cholerae N2775 (17,347 bp) harbored 15 ORFs and was closer to that found in S. algae 2NE11 (CP055159) at the 3′-end of trmE (96.98% of nt identity with 89% coverage). It carried genes encoding an integrase, a DNA-binding protein (excisionase family), a helicase, transcriptional regulators (AlpA and XRE families), a virulence RhuM family protein, a restriction endonuclease, N-6 DNA methylases, and a resolvase. Its int_GI gene showed 99.91, 94.96, and 94.19% of nt identity with the int_GI of a GI found downstream of trmE in V. cholerae FORC_073 (CP024082), S. algae 2NE11 (CP055159), and GIPmi1 (MF490433), respectively. We noticed that for both GIs in the tandem of V. cholerae, the last 9-bp of the attL site was identical to those of the SGI1 cluster.

As usual, MDR regions were inserted upstream of the res gene and generated a 5-bp target site duplication. In some cases, insertion of the MDR region resulted in larger or smaller deletions of res, or on the 3′-side of the backbone, or on both sides. The insertion site was then determined on the intact side when present. Of the 89 uncharacterized SGI1-REs, 41 of them lacked an MDR region. These MDR-free SGI1-REs were detected mainly in bacteria not belonging to the Enterobacteriales, particularly Vibrio (n = 18), Shewanella (9), and marine bacteria (8): Rheinheimera, Halomonas, Idiomarina, Marinobacter, and Pseudohongiella. Furthermore, 4 of them (belonging to the PGI2 cluster and to clusters 5 and 7) did not harbor a res gene (Supplementary Table 2).

Except in the case of SGI2 for which the MDR region was not inserted upstream of the res gene, but in S023, the MDR regions of SGI1-REs of the SGI1 cluster were inserted at the res site in 3 different positions (de Curraize et al., 2021). Most of them were at the ACTTG site. The backbones of SGI1-Pm2CHAMA and VGI-Vc2CHAMA were identical and their MDR region was inserted at position AACTT, as for the SGI1-RE of Salmonella Saintpaul FNE0134. This position was shifted one bp to the left relative to ACTTG. For SGI-V and SGI1-RE from P. stuartii FDAARGOS_294, whose backbone was identical to that of SGI-V, their MDR region was inserted in another position (AAATT). It is noteworthy that this position was also observed for the insertion of the MDR region into PGI1. Indeed, res_SGI–V showed only 97.26% of nt identity with res_SGI1, but 99.32% of nt identity with res_PGI1. As with SGI1-XJ9S, the insertion site of the MDR region of 7 SGI1-REs could not be determined due to the deletion of parts of res and S044 after insertion of the MDR region flanked by IS elements. Finally, in addition to SGI0 and VGI, 4 SGI1-REs from Vibrio did not contain an MDR region.

Regarding the PGI1 cluster, the position of the MDR region was only detected at a single position (AAATT) for the published SGI1-REs. Only the insertion site of the MDR region of the SGI1-RE from S. fodinae 74A (ATGAC) was different from that of PGI1. Indeed, the P022-res region and P025 located on either side of the MDR region were different, including the genes encoding a BsuBI/PstI restriction endonuclease and an N-6 DNA methylase downstream of res (Supplementary Figures 2B, 5B). Its res gene shared from 99.00 to 99.34% of nt identity with res from 6 other SGI1-REs (3 from PGI1 clusters and 3 from PGI2 cluster) that also harbored these same ORFs downstream of res, but they were free from MDR region. Of the SGI1-REs recovered, 16/23 did not harbor an MDR region; among them were E. coli (n = 1), Salmonella (3), but mostly Vibrio (8), Shewanella (3), and R. nanhaiensis E407-8.

The insertion site of the MDR region in PGI2 could not be determined due to the absence of the res gene and a truncated S044-like element at the right end of the backbone. In contrast to PGI2, a res gene and an MDR region were present in SGI1-REs of C. sakazakii CFSAN019572, V. cholerae 984-81, and N2775. This res gene showed 99.66–99.83% of nt identity with res_AGI1. The insertion sites of MDR region for these SGI1-REs were determined (CATAG) and were the same as some insertion sites of the MDR region in the AGI1 cluster. This insertion site was previously found for the AGI1 variants of S. enterica Derby 2014LSAL02547 and V. cholerae 4874. As for KGI, 6/11 new SGI1-REs were exempt from MDR region, among them were E. coli (n = 1), K. pneumoniae (1), and Vibrio (4). Moreover, the SGI1-RE of V. parahaemolyticus V67, which had a unique PGI2-023-PGI2-068 region, did not harbor a res gene.

Five insertion sites of the MDR region were previously described (ATAGG, CATAG, CCATA, TAGGT, and TGCAC) for SGI1-REs of the AGI1 cluster (Siebor and Neuwirth, 2020). The ATAGG position found in AGI1 was the most common. The other positions were shifted 1-bp, 2-bp to the left and 1-bp, 32-bp to the right relative to ATAGG, respectively. Moreover, a particular situation was observed for the SGI1-RE of P. mirabilis PmBR607, which was characterized by the absence of a target site duplication generated by the insertion of the MDR region. This feature was a consequence of the backbone composition. Indeed, the res gene showed 100% of nt identity with resPGI1 and the 5-bp near the 5′-end of the class 1 integron were AAATT, as for PGI1; at the right end of the backbone was A028 (100% nt identity) and the 5-bp near the 3′-end of the integron were ATAGG, as for AGI1. Finally, the MDR region of the SGI1-RE from V. cholerae 495747 made of a Tn3-like element was inserted into A015 and created the duplication of the ATCGT site.

In clusters 5 and 7, none of the SGI1-REs carried an MDR region. It should be noted that 3 SGI1-REs among them did not harbor a res gene (V. cholerae VN-2808 from cluster 5, M. adhaerens KG14, and Marinobacter sp. SAT109 from cluster 7). In cluster 6, only the SGI1-RE from I. tyrosinivorans CC-PW-9 carried an MDR region. Its res gene showed 97.61% of nt identity with the res of SGI1-REs without MDR region from H. meridiana Eplume2 and P. nitratireducens. Insertion of the MDR region generated a 5-bp duplication site (AAATG).

Only SGI1-REs with a complete MDR region obtained on a single contig caught our attention. In most cases, the MDR regions contained a class 1 integron that was inserted by a transposition event upstream of the res gene, resulting in a duplication of the target site. Only genes conferring resistance to the main antibiotics used in the clinical practice or a particular resistance that has never been described in an MDR region will be presented.

Salmonella genomic island 1-related elements contributed to the diffusion of multidrug resistance genes, including extended spectrum beta-lactamase (ESBL) or carbapenemase genes, sometimes associated with 16S rRNA methyltransferase genes or quinolone resistance genes. Coexistence of the bla_VEB–6 ESBL gene and the qnrA1 quinolone resistance gene was found in the MDR region of the SGI1-RE harbored by P. stuartii FDAARGOS_294, as previously described in SGI-V (Siebor and Neuwirth, 2011).

In the MDR region of the AGI1 variant of P. mirabilis PmBR607, the class 1 integron was followed by an ISCR1 element and the bla_CTXM–2 ESBL gene. The MDR region of the recently described PGI2 variant of M. morganii zy_m28 (Xiang et al., 2021) contained numerous copies of IS26 elements and antibiotic resistance genes (ARGs), including the bla_CTX–M–3 ESBL gene, the fosfomycin resistance gene fosA3, and the fluoroquinolone acetylating aminoglycoside-(6′)-N-acetyltransferase gene aac(6′)-Ib-cr, as already described in PGI2-C55 (Lei et al., 2020b).

The carbapenemase gene, bla_NDM–1 and the 16S rRNA methyltransferase gene, armA conferring resistance to most beta-lactams, including carbapenems, and most aminoglycosides, respectively, were present in the SGI-NDM-1 of V. cholerae BRV8. The NDM-1 carbapenemase-encoding gene was already described in PGI1-PmPEL (Girlich et al., 2015), while the ArmA-encoding gene had never been encountered on an SGI1-RE until now. In contrast, another 16S rRNA methyltransferase gene (rmtC) has been detected in SGI1-S (Wilson and Hall, 2010).

The MDR region of the PGI1 variant from S. fodinae 74A was a Tn3 family transposon (91% of nt identity with Tn5058: Y17897) that harbored genes encoding MgtC/SapB and MgtA proteins involved in magnesium transport and two other hypothetical proteins of unknown function. It shared 87.61% of nt identity with a transposon from Citrobacter sp. Y3 (CP050009). Genes encoded by mgtABC have been described in Salmonella pathogenicity island 3 (SPI-3) (Fàbrega and Vila, 2013). They are involved in intracellular survival. To our knowledge, the mgtCB operon has never been found on a transposon.

The SGI1-RE of I. tyrosinivorans CC-PW-9 was the only one in cluster 6 to harbor an MDR region that contained a Tn402-like class 1 integron carrying cmlA1 and aadA2. It was acquired via a transposition event as shown by the detection of direct repeats.

Using the motif sequence of AcaCD binding sites, five target sites were detected upstream of xis (S002), S004, traN (S005), traH (S012), and S018 on SGI1, and homologous genes on PGI2, AGI1, and SGI1-RE5SsW3-18-1 (Figures 3A,C–E, Supplementary Figure 7, and Supplementary Table 6). In SGI1-RE6ItCC-PW-9, the target site upstream of traG (I011) was detected in place of that of traH, which was absent in this SGI1-RE (Figure 3F). In PGI1, six target sites were identified upstream of xis (P002), PS004, traN (P004), traG (P009), P012, and P016 (Figure 3B).

In the SGI1-REs of the SGI1, PGI2, AGI1 clusters, and cluster 5, the sequences of AcaCD sites upstream of each gene varied slightly among SGI1-REs (de Curraize et al., 2021). However, the sequences of AcaCD_PGI1, AcaCD_{SGI1–RE6ItCC–PW–9}, and AcaCD_{SGI1–RE7MaKG14} sites were more different from AcaCD_SGI1, AcaCD_PGI2, AcaCD_AGI1, and AcaCD_{SGI1–RE5SsW3–18–1} sites, for each gene. Furthermore, variations in the number of sites were detected in some SGI1-REs. This was a consequence of the changes found in the backbone sequence. The AcaCD_SGI1 site upstream of traN lacked in the SGI1-HKL subgroup (de Curraize et al., 2020b). In the PGI1 cluster, an AcaCD site was found upstream of either traG or traH, depending on the tra genes present. In SGI1-REs of V. cholerae BD21, 01-8_S93 and SAMEA4057619 (from the PGI1 cluster), for backbone parts sharing homology with A. simiae A6, AcaCD sites upstream of xis and PS004 were closer to the sites of AcaCD_PGI2, AcaCD_SGI1, or AcaCD_AGI1. Likewise, for parts of the backbone sharing homology with PGI1 in the PGI2 and AGI1 cluster, the AcaCD sites were similar to those of AcaCD_PGI1.