Cotrimoxazole (SXT), the combination of trimethoprim (TMP) and sulfamethoxazole (SMX), is widely used in the treatment of several infections, particularly those affecting the urinary tract. As an effective and inexpensive medication, it is on the list of the World Health Organization of essential medicines (World Health Organization, 2023). Its antibiotic components act as competitive inhibitors of two enzymes of the folic acid (Fol) synthesis pathway, and their action is enhanced when combined. Due to their mode of action, they are generically called antifolates. SMX belongs to the family of sulfonamides, which includes several compounds that inhibit dihydropteroate synthase, the first enzyme in the Fol pathway, and TMP inhibits dihydrofolate reductase, the third and last enzyme required to produce tetrahydrofolate. This latter compound is the active cofactor that performs the essential function of providing one-carbon units to many biosynthetic reactions in cells (Green and Matthews, 2007). Sulfonamides opened the antibiotic era, being the first antibacterial drugs used systemically in medicine, since the 1930s. TMP was developed and employed in the 1960s, and SXT came into use shortly after due to the observed improved effect of the combination of TMP and SMX (Huovinen, 1997).

Over time, resistance to SXT emerged and has steadily increased, limiting its clinical use. Resistance to this combination is due to the sum of resistance to both its components, and is particularly frequent among Gram-negative bacteria. It is mediated by genes belonging to the accessory gene pool encoding variants of the dihydropteroate synthase and dihydrofolate reductase that are resistant to SMX and TMP, respectively. Currently, four SMX resistance genes have been described (sul1, 2, 3, and 4) and several dozen for TMP resistance (dfrA numbered in the order in which they were described, and a dfrB series of eight variants) (Sköld, 2001; Alonso and Gready, 2006; Ambrose and Hall, 2021; de los Santos et al., 2021; Alcock et al., 2023).

In principle, Gram-negative strains resistant to SXT contain at least one sul gene and one dfr gene. These are usually associated with mobile genetic elements that enable them to move within the bacterial genome, such as integrons and transposons, and to be transferred horizontally between bacteria, typically being carried on conjugative plasmids. The sul and dfrA genes are significantly associated with class 1 and 2 integrons. These are genetic elements often referred to as “clinical integrons” because they confer resistance to multiple antibiotics and are commonly found in Gram-negative clinical isolates. These features distinguish them from other integrons found in environmental bacteria (Labbate et al., 2009).

Integrons consist of a conserved platform and a variable region. In the platform they encode a site-specific recombination system that captures and integrates free gene cassettes into an attachment site (attI) so that they become part of the integron and are expressed as an operon from a promoter also provided by the platform. Then, the variable region of an integron is the array of gene cassettes it contains. As free circular molecules, gene cassettes usually contain a single gene without a promoter, followed by its specific attachment site (attC). The integrase recombines the attC and the attI sites to operate the gene cassette integration. The mechanism described above implies that the first cassette found in an array was in fact the last one to be integrated. Also, since it is located immediately after the promoter, this cassette would be the one expressed at a higher level (Hall, 2012; Cury et al., 2016).

Class 1 integrons (Int1) are most prevalent in clinical isolates of Enterobacteriaceae and other Gram-negative families. They usually contain one or two genes in their variable region: a dfrA gene and/or an aadA gene, conferring TMP and streptomycin resistance, respectively. The presence of other gene cassettes has been repeatedly reported, but is less frequent (Domingues et al., 2015). In the Int1, the platform is divided by the variable region into two parts: (i) a 5′ conserved segment (5′ CS), containing the site-specific integrase gene intI1, the Pc promoter that directs transcription of the integrated gene cassettes, and the attachment site attI; and (ii) a 3′ conserved segment (3′ CS), containing a truncated gene for an export pump (qacEΔ) and the sul1 gene, conferring resistance to SMX (Figure 1A) (Labbate et al., 2009). As to the Pc promoter, several variants with different strengths have been recognized, being most frequent Pc_W (weak) and Pc_H1 (hybrid 1). In addition, some integrons contain a second promoter downstream of Pc, called P₂, which usually accompanies the Pc_W variant, enhancing its strength (Pc_W-P₂) (Poey and Laviña, 2014). It is worth mentioning that sul1, which is not a gene cassette, is the only resistance gene residing in the conserved platform of Int1s (Labbate et al., 2009). In a previous study we showed that even when sul1 is deleted, Int1⁺ strains remain resistant to SMX thanks to the presence of an unlinked sul2 gene or, less frequently, of a sul3 gene. This led us to propose that the presence of an SMX resistance gene would be a requirement for Int1⁺ strains (de los Santos et al., 2021). Therefore, in principle, every strain bearing an Int1 contains at least one sul gene, and in the vast majority, a dfrA gene in the variable region. The result is that all Int1⁺ strains are SMX resistant and most of them are also SXT resistant.

Class 2 integrons (Int2) have a structure quite similar to that of Int1, but lack the 3′segment of the platform and have a nonsense mutation that inactivates the integrase gene intI2. Therefore, its structure is fixed and, in principle, its variable region is no longer variable. Nevertheless, it appears that the Int1 integrase would be able to trans-act on the Int2 attachment site, generating some variability; in fact, it is not uncommon to find both integrons 1 and 2 in the same strain. In its classical fixed structure, Int2 carries three antibiotic resistance genes in the form of gene cassettes: dfrA1, sat2, and aadA1, conferring resistance to TMP, streptothricin and streptomycin, respectively. This integron exemplifies very well the association of these types of elements with transposons. Int2 is located inside Tn7, so it moves while being carried by this transposon. Regarding its relationship with the resistance to antifolates, Int2 contains the dfrA1 gene cassette, and frequently resides in bacteria containing the sul2 gene elsewhere (Deng et al., 2015; de los Santos et al., 2021). Then, most Int2⁺ strains have a SXT resistant phenotype.

In this communication we address the genetic basis of cotrimoxazole resistance in a collection of 101 uropathogenic E. coli (UPEC) strains, gathering present with previous results (Poey and Laviña, 2014, 2018; de los Santos et al., 2021). These strains belong to a broader collection of 230 clinical isolates which had been gathered following an epidemiological design conducted in 2007–2009 in a hospital in Montevideo, Uruguay (Poey et al., 2012). Therefore, this collection is representative of the UPEC population circulating at that place and time. Since then, it has been studied from different points of view, including antibiotic resistance, presence of sul genes and of clinical integrons. Regarding antifolate resistance, more than half of the strains are resistant to one or both SMX and TMP, with a clear predominance of SXT resistance. These latter strains are 101, being the subject of interest of the present study. As to the genetic basis of their SXT^R phenotype, they all contain at least one type of sul gene: sul1 (n = 27), sul2 (n = 46), sul1 and sul2 (n = 27), and sul1 and sul3 (n = 1), what explains their SMX resistance. Regarding the basis of their TMP^R, we only had information on 54 strains, specifically those that contain an Int1 whose variable region could be amplified and sequenced, and which proved to have a dfrA gene cassette. These are: dfrA17 (n = 37), dfrA1 (n = 9), dfrA5 (n = 3), dfrA7 (n = 3), dfrA14 (n = 1), and dfrA15 (n = 1). In addition, the Pc promoter is also identified, thus completing the information on the central expression unit in these Int1s (Figures 1, 2) (Poey and Laviña, 2014, 2018; de los Santos et al., 2021). Therefore, the genetic basis of the SXT^R of 54 strains had been previously elucidated and their TMP^R proved to be due to dfrA gene cassettes included in class 1 integrons. In this work we devoted to complete the analysis of the genetic bases underlying TMP^R in the 47 remaining SXT^R strains and succeeded in identifying them in 40. Many were associated with Int1 remnants or with Int2, while others were found in other apparently less frequent genetic structures, including a novel gene arrangement.

This study first concentrated on completing the characterization of the genetic basis of the TMP^R in the 101 SXT^R UPEC collection, i.e., in the remaining 47 strains for which we lacked this information. Whenever possible, we sought to identify the genes responsible for the TMP^R phenotype as well as their genetic context. The final goal was to gather previous and present results to reach a general view on the genetics of cotrimoxazole resistance in E. coli and also on the mobile genetic elements involved.

The search for genes conferring TMP^R first focused on strains containing class 1 or class 2 integrons, and then addressed the study of the remaining SXT^R strains.

In this work we present the genetic bases of cotrimoxazole resistance in a collection of 101 uropathogenic E. coli strains. The study integrates previous and present results to provide the most complete understanding possible of the joint resistance to both antifolate components of the SXT formulation. In addition, considering the systematic strategy followed to collect the strains, the view provided here acquires epidemiological significance.

A generalized observation is that resistance to antifolates is strongly related to the presence of clinical integrons in many Gram-negative isolates from specimens of infected material. SXT is one of the treatments of choice for urinary tract infections, which are most frequently caused by E. coli strains, so testing for SXT resistance in UPEC isolates is routinely performed in the clinics. In a recent article, a 100% association was found between the phenotype of SXT resistance and the detection of class 1, 2, and 3 integrase genes when PCR reactions were applied directly to urine samples (Elias et al., 2022). In the present work we also found an important, albeit lower association, around 80%, between SXT^R and class 1 and 2 integrons. To reach this figure we had to survey many portions of integrons, particularly of Int1, in a significant number of UPEC isolates.

Int1 or remnants of it were thoroughly searched for and were found in 69 of the 101 SXT^R strains. In all but three, a dfrA gene cassette was found in the variable region, which explained the TMP^R phenotype of the strains (Figure 1). Although more than 40 dfrA variants have been described (Ambrose and Hall, 2021; Alcock et al., 2023), only six were found in the Int1 sequences of our SXT^R collection, and these were among the most frequently found in multiple surveys worldwide. As observed by other authors, the most commonly found cassette arrays were dfrA17-aadA5 and dfrA1-aadA1, while dfrA5 and dfrA7 appeared as unique cassettes (Figure 1) (Domingues et al., 2015). One of the strains that lacked a dfrA gene in its Int1 variable region also had an Int2, which provided its dfrA1 gene. The other two were positive for the ISCR1 element, a marker of a complex Int1 structure that contains non-cassette resistance genes outside the Int1 variable region, dfrA among them. These latter dfrA genes are quite unusual and are specifically found in ISCR1 complex integrons, e.g., dfrA10, 11, 19, among others (Toleman et al., 2006; Ambrose and Hall, 2021). It seems very likely that the two ISCR1⁺ strains in our collection possess one of these dfrA genes.

Two apparently abundant dfrA variants, as found in databases and in multiple reports, are dfrA12 and dfrA27. However, these were not found in our collection, a result that raises the question of whether there could be differences in gene prevalence between geographic regions and time of strain collection. This topic has been addressed by some groups with varying results. In a survey of integrons and genes conferring SXT^R performed on UPEC strains from Europe and Canada in 1999 and 2000, no regional dependence was observed (Blahna et al., 2006). On the contrary, other authors found differences in the prevalence of dfrA variants and of gene cassette arrays in surveys conducted in other regions as well as in follow-up studies over several years (Yu et al., 2003; Su et al., 2006; Dahmen et al., 2010; Manyahi et al., 2017). Perhaps the dfrA set of variants found in our study may indeed reveal a certain bias with respect to other parts of the world and to the time period in which the UPEC isolates were collected (2007–2009).

An important focus in this work was on the Int1 central expression unit, composed of a Pc promoter and the gene cassettes under its transcriptional influence. A close relationship was found between these elements: the most frequent cassette array, dfrA17-aadA5, was invariably under the Pc_H1 promoter, while Pc_W alone or reinforced by P₂ generally preceded the other types of arrays (Figure 1). In fact, the number of different combinations of all these central elements was quite limited. The Int1 central expression unit analyzed here was found in canonical integrons, with their entire structure (40), but also in remnants lacking the intI1 gene from the 5′ CS (15), or lacking sul1 from the 3′ CS (8), or lacking both (6). Considering these results and our previous analyses on the whole UPEC collection, it should be noted that no strain was found with only the 5′ CS and only one just conserved the 3' CS (qacEΔ-sul1) (Poey and Laviña, 2014, 2018; de los Santos et al., 2021). This suggests that the Int1 variable region would paradoxically be the most “conserved region” of the integron, being particularly selected over time.

Gene cassettes are the mobile genetic elements that characterize integrons, and each is a gene followed by its cognate attC site (Patridge et al., 2009; Hall, 2012; Cury et al., 2016). In most cases, they can be simply recognized as such when located in the variable region of an integron, but they must be analyzed at the sequence level when they are outside their normal integration site. This is what happened with two of the SXT^R strains in our collection, whose dfrA genes could only be detected by PCR using primers internal to these genes, being one dfrA14 and the other dfrA1 (Figure 3). They were carried by plasmids that could be transferred by conjugation to an E. coli K12 and then were sequenced. Both dfrA genes proved to be true gene cassettes, each one with its specific attC site, but located outside an integron; that is, they were single gene cassettes.

The dfrA14 cassette, carried by a large conjugative plasmid, was inserted within a strA gene. This same location has been found repeatedly in small plasmids from different bacterial genera in many countries, and the gene arrangement described is sul2-strA'-dfrA14-'strA-strB. The question arose as to how it had been transferred horizontally since it was carried by small plasmids that apparently could not be mobilized by conjugation (Ojo et al., 2002; Anantham and Hall, 2012; Miranda et al., 2016). However, the same insertion was also detected in E. coli isolates from food animals in Lithuania and could be transferred by conjugation, although the plasmid containing it was not further analyzed (Šeputiene et al., 2010). In the present work, a variant of the described arrangement was found in a conjugative plasmid, p31_CH-1, which included the addition of a Tn3 insertion in strB: sul2-strA'-dfrA14-'strA-strB'-Tn3-'strB. As shown in Figure 3A, small and large plasmids contain the region of reference, with or without the dfrA14 insertion, but both large plasmids share the same Tn3 insertion. In fact, p31_CH-1 and pME20 are highly homologous to each other in all their extension, and their main difference is the presence or not of the dfrA14 gene cassette into strA. Therefore, pME20 would be representative of a p31_CH-1 precursor, and their comparison indicates that p31_CH-1 could have acquired the dfrA14 insertion by exchange of a DNA segment, for example with a small plasmid carrying the sul2-strA'-dfrA14-'strA-strB arrangement. This would most likely be operated by homologous recombination occurring in sequences flanking the dfrA14 cassette in a situation in which both the large conjugative plasmid and a small plasmid with the cluster sul2-strA'-dfrA14-'strA-strB coexisted in the same bacterium. Then, the presence of the arrangement in conjugative plasmids explains its horizontal transfer between strains, species and related genera. Curiously, the UPEC carrying p31_CH-1 was isolated from a child with urinary tract infection in 2008, a date that coincides with an outbreak of shigellosis in Chile caused by a strain of Shigella sonnei carrying the same sul2-strA'-dfrA14-'strA-strB arrangement in a small plasmid (Miranda et al., 2016). At this point, it should be recognized that this genetic arrangement is undoubtedly successful given its wide dissemination among bacteria all over the world, and would represent a gene combination different from integrons that confers resistance to cotrimoxazole.

As to the single dfrA1 gene cassette in the small mobilizable plasmid p61_CH-1m, it was found to be included in the arrangement sul2-strA-strB-dfrA1, being inserted very close to the end of strB, but not interrupting any gene (Figure 3B). Therefore, dfrA1 must be expressed together with strA and strB as an operon. In this case, only four entries in GenBank presented this gene cluster and, notably, two plasmids from racehorses in Hong Kong (in releases dated 2022) were quasi-identical to p61_CH-1m in extension and content, a fact that could be related to the export of this type of horses from our region to Hong Kong. Apparently, the genetic arrangement sul2-strA-strB-dfrA1 in p61_CH-1m has not been widely spread, so that it appears to be a relatively new gene arrangement conferring SXT resistance. In fact, it could be considered that this locus might have been formed in the very context in which the UPEC carrying this plasmid was isolated from the infected urine of a child. Although the isolate was collected some 15 years ago, it should be kept in mind that the horizontal transfer of the arrangement would depend on the chance of coexistence of the mobilizable small plasmid carrying it with a compatible conjugative one, and this requirement could reduce the probability of its spreading. In this sense, a putative ColE1 type oriT sequence was detected in p61_CH-1m. This would make it susceptible to mobilization by conjugative plasmids belonging to several incompatibility groups. For example, these include members of the broad host range group IncP, which is known to spread antibiotic resistance genes among Gram-negative bacteria (Francia et al., 2004; Popowska and Krawczyk-Balska, 2013). Therefore, it could be expected that the wide dissemination of the sul2-strA-strB-dfrA1 locus found in p61_CH-1m is only a matter of time.

As to the 15 strains bearing an Int2, they all contributed to the SXT^R with a dfrA1 gene, in accordance with their fixed structure. Therefore, summing up the 78 strains with dfrA genes inside class 1 and 2 integrons, the two strains with complex Int1 containing an ISCR1, plus two with the single gene cassettes dfrA1 and dfrA14, there were still a good number missing to reach the total of 101 SXT^R strains. Of them, 14 resulted to have a dfrA8 gene, known to be part of the composite transposon Tn5091, as corroborated in the three sequenced dfrA8-carrying plasmids. These also contained a sul2 gene, which together with the dfrA8 gene conferred the SXT^R phenotype. It should also be mentioned that in the three plasmids sul2 was found in a locus sul2-strA-strB, a few kilobases apart from the Tn5091.

In sum, in this work we elucidated the genetic basis of TMP resistance in 40 of the 47 SXT^R strains of the collection that remained to be analyzed. Combining these results with previous findings, four main genetic contexts responsible for the TMP^R phenotype were identified: class 1 integrons, class 2 integrons, transposon Tn5091, and dfrA single gene-cassettes integrated into the sul2-strA-strB locus (Figure 4). These data were then correlated with the already known sul content of the strains, so that the genetic basis of the cotrimoxazole resistance were determined in 94 of the 101 SXT^R strains of the collection (Poey and Laviña, 2018; de los Santos et al., 2021). In these strains, the genetic linkage between the sul and dfrA genes varied depending on the genetic element involved. The loosest linkage would be that of the Int2 with sul2: Tn7, which contains the Int2, is most frequently located in the chromosome at its attachment site and we found no reference of its proximity to a sul gene (Peters, 2014). Similarly, dfrA8, contained in Tn5091, would not be closely linked to a sul gene, although the 14 strains carrying it were positive for sul2. The three plasmids dfrA8⁺ that were sequenced in this work contained the entire transposon and, quite near it, a sul2 gene included in the sul2 strA strB locus; therefore, sul2 and dfrA8 were in the same replicon and had been co-transferred by conjugation. Class 1 integrons and the two sul2 strA strB derivative structures containing dfrA gene cassettes exhibited the most compact arrangements bringing together the two types of antifolate resistance genes. It could be thought that the evolutionary trend is toward the co-carriage of sul and dfrA genes in the same mobile genetic element so that they can be transferred together to new cellular hosts. The pressure exerted by the clinical use of cotrimoxazole should have strongly influenced in this direction. Since the use of sulfonamides preceded that of TMP, the insertion of dfrA genes close to a preexisting sul-containing genetic structure must have been selectively favored. The two most commonly found sul genes are sul1, which is part of class 1 integrons, and sul2, which is often found adjacent to the strA strB genes (Yau et al., 2010; de los Santos et al., 2021). These are precisely the most compact structures encoding SXT resistance.

Taking into account that antifolates are synthetic, resistance to them should have appeared in clinical isolates after the introduction of these drugs in antimicrobial treatments, i.e., after some 80 years ago for sulfonamides and 50 for TMP. In this sense, there is a collection of historical strains (1917–1954) -the Murray collection- of several hundred isolates of mainly Enterobacteriaceae, including many clinical species that are today enriched in antibiotic resistance genes. This collection is being analyzed by some groups in the world and, significantly, no traces of clinical integrons or of antifolate resistance genes have been found (Baker et al., 2015; Sütterlin et al., 2020). Therefore, it can be presumed that antifolate resistance is a relatively recent phenomenon, at least among clinical Gram-negative bacteria, and that it may still be actively evolving. In particular, the pressure exerted by the intensive administration of cotrimoxazole must have been strong enough to select for clones bearing both sul and dfr genes, which predominate over those carrying only one of these genes (de los Santos et al., 2021). As previously mentioned, sul genes integrate or accompany clinical integrons and dfrA genes are mostly located inside them, so that these mobile genetic elements could have reached their current structure and bacterial hosts in the last decades. For instance, dfrA genes are almost always in the first position in integron arrays, which indicates that these genes were the last to be acquired. It should be remembered that soon after the introduction of TMP into medical practice, descriptions appeared of enterobacterial clinical isolates, particularly from infected urines, that were highly resistant to TMP. In one of them, only 3 years after TMP became available for general use in medicine, R factors (resistance plasmids) encoding resistance to SXT were detected through transferring them to an E. coli K12 by conjugation (Fleming et al., 1972). Most probably, these plasmids carried already formed integrons.

In our collection, we found that class 1 integrons, endowed with the resource of a site-specific recombination system, were the main responsible for SXT^R, but we also found that there were other elements contributing to this phenotype. Transposons, represented by Tn7, which contains an already fixed class 2 integron, and Tn5091, carrying the dfrA8 gene, both equally conferred TMP resistance to a good number of additional SXT^R strains, while coexisting with an unlinked sul2 gene. Two further genetic structures were detected, involving dfrA single gene cassettes that associated with a specific locus formed by the sul2 strA strB genes, which may represent the beginnings of well-structured new genetic elements encoding SXT resistance. The recent origin of both arrangements can be deduced from the very limited distribution of one of them (sul2- strA-strB-dfrA1), and from the persistence of the intact sequence of strA, turned into a pseudogene by the insertion of a dfrA14 gene, in the other structure.

Finally, some of all these genetic structures conferring antifolate resistance could be horizontally transferred by conjugation, conjugative mobilization and even transduction, as shown previously (Poey and Laviña, 2018) and in the present communication. In sum, the idea emerges that the genetic structures conferring SXT resistance in Gram-negative bacteria are constantly evolving by recombination, transposition and horizontal gene transfer.

The datasets presented in this study can be found in online repositories. The data has been deposited at https://www.ncbi.nlm.nih.gov/ with accession numbers: PP566047–PP566049.

Ethical approval was not required for the study involving humans in accordance with the local legislation and institutional requirements. Written informed consent to participate in this study was not required from the participants or the participants' legal guardians/next of kin in accordance with the national legislation and the institutional requirements.

MP: Conceptualization, Investigation, Methodology, Supervision, Writing – review & editing. ES: Conceptualization, Investigation, Methodology, Writing – review & editing. DA: Investigation, Writing – review & editing. CG-L: Data curation, Investigation, Methodology, Software, Writing – review & editing. ML: Writing – original draft, Writing – review & editing.