Integrons are widely used by Gram-negative bacteria to resist antibiotics. These DNA elements can acquire, exchange and express promoterless coding sequences embedded within gene cassettes (Escudero et al., 2015). The integron functional platform is composed of a gene (intI) that encodes a site-specific recombinase (IntI); a recombination site (attI); and a functional promoter (Pc), divergently oriented to the integrase gene, that allows the expression of gene cassettes (Stokes and Hall, 1989). IntI catalyzes recombination events that lead either to the incorporation of gene cassettes within the integron, or to their excision. Several integron classes can be discriminated on the basis of their IntI sequences (Collis et al., 2002). In clinical settings, five integron classes involved in the expression and dissemination of antibiotic-resistance gene cassettes have been described. Class 1 integrons prevail in most epidemiological studies in human and animals, followed by class 2 integrons (Gillings, 2014).

In class 1 integrons, hundreds of distinct gene cassette arrays have been described (Moura et al., 2009). Class 2 integrons are associated with transposons related to Tn7 (Cambray et al., 2010) and usually carry three resistance-encoding cassettes designated dfrA1, sat2 and aadA1 (encoding resistance to trimethoprim, streptothricin and streptomycin/spectinomycin, respectively), followed by a pseudocassette of unknown function (orfX, also known as ybeA) (Figure 1A) (Hall and Stokes, 1993). Variations in this gene cassette array have rarely been described (Biskri and Mazel, 2003; Ahmed et al., 2005; Ramírez et al., 2005, 2010; Barlow and Gobius, 2006; Dubois et al., 2007; Gassama Sow et al., 2008; Márquez et al., 2008). This low diversity of the gene cassette array is thought to be due to disruption of the integrase gene intI2 by an internal ochre STOP codon (TAA) at position 179, yielding an inactive 178-aa polypeptide (Hansson et al., 2002) (Figure 1A). Few class 2 integrons with a non-disrupted intI2 gene encoding a 325-aa full-length functional integrase have been described (Barlow and Gobius, 2006; Márquez et al., 2008; Rodríguez-Minguela et al., 2009; Wei et al., 2014). A putative promoter for intI2, hereafter named PintI2, has been proposed in the annotation of the R483 plasmid (GenBank accession number L10818, Figure 1B). This PintI2 promoter overlaps a putative LexA repressor operator conserved in many integron classes (Cambray et al., 2011), suggesting that intI2 could be under the control of the SOS response, like the intI1 gene of class 1 integrons and intIA of the chromosomal integron of Vibrio cholerae (Guerin et al., 2009).

Gene cassette expression depends on the Pc promoter which, in class 1 and 3 integrons, is located within the intI coding sequence (Collis and Hall, 1995; Collis et al., 2002). In class 2 integrons, however, no Pc promoter sequence has been found within the intI2 gene, and part of the attI2 region seems sufficient for gene cassette expression (Hansson et al., 2002) (Figure 1A). Four putative Pc promoters, hereafter renamed Pc2A to Pc2D, have been proposed in the sequence between the start codon of the intI2 gene and the first gene cassette (Figure 1) (Simonsen et al., 1983; Hansson et al., 2002 and Genbank accession number AM261760). A transcriptional start site consistent with the Pc2A promoter was recently mapped (da Fonseca et al., 2011). However, none of these potential Pc2 promoters has been experimentally characterized.

The aim of this study was to examine the expression of both the integrase and the gene cassettes of class 2 integrons, and to evaluate the role of the SOS response in class 2 integrons integrase expression. We mapped the PintI2 promoter and found that despite the presence of a potential LexA binding site, PintI2 is not under control of the SOS response. We also found that two promoters, Pc2A and Pc2B, seem to contribute equally to the expression of gene cassettes in class 2 integrons.

This work highlights particularities in the expression of the integrase and gene cassettes of class 2 integrons. After mapping the PintI2 promoter (Figure 1), we showed that despite the presence of a potential LexA box, LexA did not bind this region. Thus, integrase expression is not regulated by the SOS response in class 2 integrons, contrary to what has been shown for class 1 and V. cholerae integrons (Guerin et al., 2009). Interestingly, Cambray et al. (2011) found a correlation between the lack of a LexA binding box and an inactive integrase. In their analysis, four classes of integron, including class 2, did not fit this general scheme, as they displayed both a putative LexA operator and an inactive integrase (Cambray et al., 2011). As we show here that intI2 expression is not SOS-dependent, class 2 integrons also comply with this general model, belonging to integron classes in which the absence of LexA control correlates with an inactive integrase. Closer examination of the class 2 integron putative LexA operator (CTGTATAGGCAGACAG) revealed the presence of 4 C/Gs, with a stretch of 3 consecutive C/Gs, in the 10-bp central variable region, whereas most experimentally validated LexA operators in E. coli display only 3 or fewer C/Gs in this region¹. One can hypothesize that the lack of the usual TA stretch within the class 2 integron putative LexA site may explain why LexA does not bind this site. In the three other integron classes with both a putative LexA operator and an inactive integrase, the putative LexA operator includes 0 or 1 C/G in the central region (Cambray et al., 2011). In class 1 integrons, besides LexA-dependent regulation, expression of the integrase gene is also inhibited by transcriptional interference from the strong Pc variant, due to the face-to-face arrangement of the Pc and PintI1 promoters (Guerin et al., 2009). In class 2 integrons, the Pc promoters (Pc2A and Pc2B) and PintI2 are arranged tail-to-tail, so no such transcriptional interference can exist and intI2 should be constitutively expressed. Consistently, we found no significant difference in β-galactosidase activity with pPintI2-2 (PintI2 cloned alone) and pPintI2-1 (entire attI2 site, includes the Pc2 promoters) (Figure 2). Although intI2 is fully expressed, the encoded integrase is not functional, owing to the ochre codon in position 179. Hansson et al. (2002) suggested that this shortened 178-aa IntI2 peptide might interfere with the attI2 site, preventing any IntI-mediated recombination. This hypothesis could explain the stability of the gene cassette array of class 2 integrons (Hansson et al., 2002). Consistently, the rare class 2 integrons that encode an active IntI2 integrase display a broader range of gene cassettes and gene cassette arrays (Supplementary Figure S1).

This work also highlights two specificities of gene cassette expression in class 2 integrons: (i) the Pc2 promoter is located in the attI region and not within the intI gene as in other integrons (Collis and Hall, 1995; Collis et al., 2002; Baharoglu et al., 2012), and (ii) at least two promoters, Pc2A and Pc2B, are involved in gene cassette expression. Although we observed weak activity of the putative Pc2C promoter when Pc2A and Pc2B were removed (Figure 3B), Pc2C inactivation within the entire attI2 region had no effect on β-galactosidase activity (Figure 4B). This indicated that the contribution of Pc2C to gene cassette expression is negligible which is consistent with its structure, since it displays a suboptimal 19-bp long spacer between the -35 and -10 elements (Mitchell et al., 2003).

The permanent coexistence of two functional Pc promoters within the integron attI region is a unique feature of class 2 integrons. In class 1 integrons, two gene cassette promoters namely Pc and the P2 have been also described but in only 10% of the integrons (Collis and Hall, 1995; Jové et al., 2010). The biological reason for the presence of two active Pc promoters in class 2 integrons is unclear. Further studies are needed to determine whether the presence of these two functional gene cassette promoters in class 2 integrons might be linked to differential regulation of these promoters. One can hypothesize that this peculiar organization might be linked to specific activation of Pc2A and/or Pc2B, either under specific conditions, e.g., in response to distinct lifestyle conditions, or by specific partner proteins.

As in class 1 and 3 integrons (Collis and Hall, 1995; Correia et al., 2003; Jové et al., 2010), there are several Pc variants of different strengths in class 2 integrons. We show here that the class 2 integron promoters Pc2A-V2 and Pc2B-V2 variants are less efficient than Pc2A and Pc2B for gene cassette expression (Figure 5). Interestingly, the Pc2A-V2 and Pc2B-V2 variants have so far always been found together, in association with class 2 integrons that encode a functional IntI2 integrase (Barlow and Gobius, 2006; Márquez et al., 2008; Rodríguez-Minguela et al., 2009; Wei et al., 2014). This observed inverse correlation between the level of gene cassette expression and integrase activity raises another analogy with the class 1 integron model in which the weaker the Pc variant, the more efficient the IntI1 integrase (Jové et al., 2010).

Taken together, our data reveal the existence of two categories of class 2 integrons. The most prevalent category efficiently expresses a limited pool of gene cassettes from promoters Pc2A and Pc2B but is unable to modify its gene cassette array (non-functional IntI2). The rarer category produces an active integrase that permits gene cassette acquisition/rearrangement but expresses the gene cassettes more weakly (twofold; Figure 5). The rarity of functional class 2 integrons may be due to a high fitness cost associated with constitutive intI2 expression. Indeed, even though constitutive intI2 expression might represent an advantage because integrase production is not conditional on an environmental stimulus, it can also be a drawback, given the potential biological cost of intI2 expression or activity. Indeed, it is known than in Acinetobacter and in E. coli, expression of the class 1 integron integrase gene is deleterious, with a high fitness cost, and this can lead to inactivation of the integrase (Starikova et al., 2012; Lacotte et al., 2017).

We propose an evolutionary model for class 2 integrons in which the expression of the ancestral class 2 integrons integrase was under control of the SOS response (Figure 6A). Then, mutation in the LexA operator would have led to a constitutive expression of the integrase gene encoding a fully active IntI2 (Figure 6B). The resulting high fitness cost would have lead to the introduction of either (i) a nonsense mutation in the intI2 gene that inactivated IntI2 but maintained an array of highly expressed gene cassettes; this would have given rise to the currently prevailing class 2 integrons (Figure 6C), or (ii) mutations within the Pc promoters, generating variants of weaker activity, in order to reduce gene cassette expression and the potentially associated fitness cost (Figure 6D). This latter group of class 2 integrons constitutively express a functional IntI2 whose high fitness cost may explain their rarity.

M-CP and TJ conceived the study. M-CP coordinated the study. TJ, AT, and SDR performed the experiments. TJ, M-CP, and SDR analyzed the data and wrote the manuscript. AG-S revised the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.