The expression of integron arrays is shaped by the translation rate of cassettes

Integrons are key elements in the rise and spread of multidrug resistance in Gram-negative bacteria. These genetic platforms capture cassettes containing promoterless genes and stockpile them in arrays of variable length. In the current integron model, expression of cassettes is granted by the Pc promoter in the platform and is assumed to decrease as a function of its distance. Here we explored this model using a large collection of 136 antibiotic resistance cassettes and show that the effect of distance is in fact negligible. Instead, cassettes have a strong impact in the expression of downstream genes because their translation rate affects the stability of the whole polycistronic mRNA molecule. Hence, poorly translated cassettes decrease the expression and resistance phenotype of cassettes downstream. Our data puts forward a novel integron model in which expression is contingent on the translation of cassettes upstream, rather than on the distance to the Pc.


INTRODUCTION
Antimicrobial resistance (AMR) is one of the major threats for health globally 1 .
The spread of resistance genes through Horizontal Gene Transfer (HGT) has fostered the rise of AMR during the last decades.Mobile Integrons (MIs) have been key players in this phenomenon through their association with mobile genetic elements [2][3][4] .MIs are genetic platforms that capture and stockpile new genes through site specific recombination.Indeed, 89% of cassettes in MIs encode antimicrobial resistance genes against a variety of antibiotic families (IntegrAll database, 5 ).
Structurally, integrons are genetic elements composed of a stable platform and a variable array of genes embedded in discrete elements called integron cassettes.The stable platform comprises an integrase-coding gene (intI), as well as a recombination site (attI) for cassette integration.Cassettes are generally composed of an open reading frame (ORF) and an attC recombination site.Arrays in MIs can contain from one up to eleven cassettes.In class 1 integrons, the majority (80%) of arrays contain 1 to 3 cassettes (IntegrAll database, 5 ).Importantly cassettes are generally promoterless, and their expression is fostered by the dedicated Pc promoter, located within the platform, upstream of the attI site where cassettes are inserted.This makes of integron arrays an operon-like structure, with a single promoter governing the expression of several cassettes.As a consequence, it is generally accepted that cassettes at first position in the array are more expressed than those further downstream, simply as a function of the distance to the promoter.
Nevertheless, the order of cassettes is not static, since the integrase is able to reshuffle them, excising and integrating them in the first position of the array.Because the integrase is under the control of the host's SOS response 6 , integrons represent a low-cost memory of functions that provides adaptation on demand [7][8][9][10] (Fig. 1A).
The effect of cassette position has been observed in several studies [10][11][12][13] so that the distance-to-Pc gradient of expression along the array is a paradigm that has been central to the working model of integrons.Only some exceptions to this rule have been documented, like the existence of cassettes that contain their own promoters [14][15][16][17][18] or the presence of small ORFs in attC sites that enhance translation of downstream cassettes 13 .However, most of these examples only cover a few integron cassettes and seem to remain anecdotal.Recently, we have also shown the existence of gene-less cassettes in chromosomal integrons that contain promoters modulating the expression of the array 19 .All these mechanisms suggest that expression in integrons does not always fit the distance-to-Pc gradient.This motivated us to revisit the expression model of the integron, and specifically, if and how cassettes shape the expression of the array and influence the resistance conferred by others antibiotic resistance cassettes (ARCs).
In this work, we study the effect of 136 ARCs on the expression of a downstream gfp gene.We show that the impact of a cassette in downstream expression is independent of and more important than the distance to the Pc.This impact is strong enough to decrease resistance levels conferred by a second ARC below clinical breakpoints.To determine the mechanism underlying this phenomenon, we first assessed the influence of known exceptions to the rule of distance, and found an occasional and generally modest contribution of additional promoters or attC sites.Instead, the translation rate of the first ARC influences the expression of the second cassette by altering the mRNA levels.This effect is strong and is pervasive across the collection.Our findings change the working model of the integron, to one where the expression gradient of the array becomes dependent on the identity of the first cassette.Cassettes are generally promoterless but can be expressed when integrated at the attI site, where their expression is controlled by the integron-borne Pc promoter.An expression gradient is then generated: the first cassette displays the highest expression while expression gradually decreases for cassettes further away in the array.(B) 136 different antibiotic resistance cassettes were individually cloned in first position in pMBA vector 20,21 , allowing to study how each ARC affects the expression gradient.

Cassette identity modulates expression of the array downstream
To assess if cassettes can modulate the expression of downstream genes in the array (i.e.: if they have polar effects on the array) we took advantage of the recently generated pMBA collection 20,21 .pMBA is a p15a replicon designed to provide cassettes with an appropriate genetic context: it contains a class 1 integron platform encoding the Pc, the Pint, and the attI site where cassettes are cloned in first position mimicking an integrase mediated attI x attC reaction.The integrase gene (intI1) is truncated to avoid its well-known deleterious effect 22 .Immediately downstream each cassette there is a gfp gene mimicking a second cassette (Fig. 1B) 20 .The collection is composed of 136 pMBA variants containing different ARCs in E. coli MG1655.Transcription is driven by the strong variant of the Pc promoter (PcS).Hence, the pMBA collection offers an isogenic setting to study the impact of 136 cassettes on the rest of the array (Fig. 1B).Hereafter, and for the sake of simplicity, we will use the name of the resistance gene encoded in each cassette to refer to the corresponding E. coli-pMBA-ARC variant.
We measured GFP fluorescence in all 136 E. coli strains in the collection using flow cytometry.Data was then normalized to the fluorescence of the E. coli strain containing pMBA, which carries the gfp gene in the first position in the array.
Our results show a large variation in the expression of the second cassette across the collection (Fig. 2A), with fluorescence levels scattered across the 120-fold range between the most and the least repressive ones (blaOXA-20 and aacA43 respectively).
A similarly broad distribution of fluorescence values can be observed for each different antibiotic family represented in the collection, ruling out the influence of the function encoded (Supplementary figure 1).As expected, the vast majority of ARCs decrease the expression of the second cassette, with a mean fluorescence ratio (pMBA-ARC / pMBA) of 0.417 for all ARCs (Fig. 2B).While this value can be interpreted as the average repressive effect of cassettes on downstream genes, our data shows a broad dispersion, with some cassettes even increasing fluorescence mildly (up to 2-fold) (n=6, 4%).
Our data highlights that cassettes can exert very different -even oppositeeffects on the expression of the array.This puts the model of expression of integrons in question.Our experimental setup allows for the verification of the distance-to-Pc model since the distance to the second cassette (the gfp gene) is the length of the ARC in first position.ARC length varies substantially in the pMBA collection, with the lower and upper limits being 384bp-(dfrB2) and 1464bp-long (ereA3) (Fig. 2C).
However, correlation between ARC length and GFP fluorescence ratios was inexistent (r= -0.18, p = 0.03) (Fig. 2D).We conclude that the effect of the distance to the Pc is probably masked by other mechanisms, to the point of becoming negligible.Instead, our data shows that the expression of the second cassette depends on the identity of the first cassette.
To study how this could impact the resistance levels of downstream ARCs, we specifically selected three aminoglycoside resistance cassettes (aacA54, aacA61, and aacA8) displaying distinct levels of polar effects (Fig. 2E).We substituted the gfp gene in these strains with blaVIM-1 cassette and examined their resistance to β-lactams ertapenem and cefaclor (aacA54, aacA61 or aacA8 do not show cross resistance to these antibiotics (Supplementary figure 2)).Notably, aacA54 and aacA61 significantly reduced cefaclor resistance, while aacA8 had no effect (Fig. 2E).All three cassettes strongly reduced ertapenem resistance, with the extent of reduction depending on cassette identity.Of note, aacA54 and aacA61 (but not aacA8) lowered ertapenem resistance below the clinical breakpoint, highlighting the clinical relevance of cassette identity.Moreover, this pattern persisted when we replaced the blaVIM-1 cassette with catB3 and measured chloramphenicol resistance.
Hence, polar effects in integrons arrays are contingent on cassette identity and are clinically relevant.

Different ARCs modulate downstream array expression mostly by impacting mRNA levels
We next sought to investigate the underlying mechanisms modulating cassette expression.An ARC can affect the expression of the second cassette by changing its mRNA and/or protein levels.To assess to what extent each ARC impacts gfp mRNA levels, we performed qRT-PCR of the gfp gene in all 136 pMBA-ARC strains and calculated its fold change relative to pMBA (Fig. 3A).Our results indicate that most ARCs affect negatively the mRNA levels of the gfp gene (mean mRNA fold change = 0.757) and that variation is similar to that observed for GFP fluorescence (Fig. 3B).mRNA levels showed a strong positive correlation (r = 0.73, R 2 = 0.53, p < 0.001) with fluorescence ratios (Fig. 3C).It is of note that the values of gfp mRNA fold change rarely match those of GFP fluorescence ratios, which is somewhat expected due to the intrinsic differences between the two methodologies.
However, we cannot rule out the existence of additional mechanisms affecting gfp translation independently of mRNA levels.Indeed, exceptionally some ARCs do not entail changes in gfp mRNA levels but decrease significantly GFP fluorescence.
Altogether, our results show that polar effects of cassettes are generally observed at the mRNA level.

Negligible role of secondary promoters and attC sites in polar effects
The panoply of polar effects observed in this work could be the result of the interplay between several known mechanisms.For instance, cassettes could all have negative polar effects of similar intensity, but some might alleviate them through additional promoters while others intensify them through the presence of transcriptional terminators.
We asked whether additional transcriptional activity could in part explain the high fluorescence levels of a subset of 12 ARCs showing the highest expression levels of the second cassette.These include qacE, for which the existence of additional promoters has already been documented 18,23 ; and ereA2 and ereA3 that share respectively 94 and 87% identity with ereA1, which also contains its own promoter 14 .We thus deleted the Pc promoter in these 12 variants of pMBA and measured fluorescence to reveal any additional promoters (Fig. 4A).The deletion of the Pc in pMBA decreased fluorescence 400-fold, to levels similar to a strain without pMBA.Deleting the Pc in this subset confirmed promoter activity in qacE and ereA3.Other cassettes, like dfrA1 and dfrA15 also showed a minor (2-fold) but significantly higher level of fluorescence compared to the pMBA∆Pc control.
Nevertheless, for most cassettes the deletion of the Pc led to a decrease in fluorescence similar to the one observed for pMBA∆Pc.We hence conclude that while additional promoters within certain ARCs might influence the expression of integron arrays, they only account for a limited portion of the overall variation observed in our dataset.attC recombination sites are sequences located at the 3' region of each integron cassette.In their recombinogenic form, attCs adopt a hairpin structure which is essential for the recognition by the integrase and its recombination.Due to this inherent highly structured conformation, attC sites were initially proposed to function as Rho-independent transcriptional terminators 11 but later work questioned this view, arguing they rather affect translation of the cassette downstream 13 .We sought to test if attCs play a role in the negative polar effects observed, using the subset of 11 ARCs displaying the lowest GFP fluorescence levels.To rule out the influence of plasmid loss in our observations, a phenomenon that could arise from the fitness cost of ARCs, we measured plasmid copy number (PCN) in this subset and pMBA by qPCR (Supplementary Figure 3).Only pMBA carrying blaOXA-20 showed a significant reduction in PCN and was thus excluded from further analysis.
Deleting attC sites in the remaining cassettes reveals no effect on fluorescence for most cases, increasing only mildly the fluorescence in 3 cassettes (Fig. 4B).
However, for aacA35 the deletion of the attC results in a large (5-fold) increase in GFP fluorescence.Accordingly, mRNA levels were also higher in 3 out of 4 ∆attC strains (Fig. 4B inset).Overall, attC removal impacted downstream expression only in a minority of cases and generally in a subtle manner.This suggests that they are not major determinants of polar effects.

Cassette translation plays a key role on the polar effects in integron arrays
The general absence of promoters within cassettes, together with the lack of repressive effects of attC sites suggest that i) these elements only explain a small fraction of polar effects and ii) the underlying mechanism(s) is novel in integrons.We showed above that ARCs significantly affect the mRNA levels of the cassette downstream.Given that the transcription initiation rate originated at the Pc is the same across our collection, we hypothesized that elements within cassettes might either promote transcriptional termination or negatively affect the stability of the polycistronic mRNA 24 .Considering the critical role of translation in mRNA stabilitywhere translating ribosomes can protect mRNA from degradation [25][26][27][28][29] -we examined the translation initiation (TI) rates of each ARC and their correlation with downstream gfp mRNA levels (Fig. 5A).We predicted TI rates using the web software RBS Calculator v2.1, which employs a highly accurate thermodynamic model of ribosome-mRNA interactions to estimate the TI rates for a given mRNA sequence 30 .Our analysis revealed variable TI rates among ARCs and a significant positive correlation between TI rates and gfp mRNA levels (ρ = 0.39, p-value < 0.0001) (Fig. 5A), highlighting cassette translation as a global determinant in downstream gene expression.Contrarily, we did not observe similar correlations with the codon adaptation index (CAI) of cassettes (ρ = 0.10), or with the overall complexity of secondary structures in their mRNAs (ρ = 0.15), features that have also been shown to be determinants of gene expression and mRNA stability in bacteria 25,26,[31][32][33][34][35] .Interestingly, we observed a negative correlation between the GC content of cassettes and both the TI rates (ρ= -0.37, p-value < 0.0001) and gfp mRNA levels (ρ= -0.32, p-value = 0.001).A low GC content can play a role in gene expression in different ways.When located near the 5' of ORFs, it can lead to a less structured mRNA, favoring translation initiation 32,34,[36][37][38] , while along the coding sequence, it may lead to spurious transcription due to AT-rich tracts 39 .Notably, the correlation between translation rates and the GC content in the 60bp around the start codon was clearly higher than for the whole gene (ρ= -0.49 vs. -0.37,p-value < 0.0001), suggesting that the impact that GC content has on the polar effects of an ARC is due, at least partly, to its influence on translation initiation rates.
As our data points to the translation of the first cassette as a key driver of the polar effects, we sought to prove it experimentally.We hypothesized that preventing translation initiation of the first cassette, or stopping translation prematurely, should result in a lower expression of the second cassette.To test this, we selected a subset of ARCs from distinct antibiotic families, covering different levels of GFP expression.
We hindered translation in these cassettes by i) replacing the ATG initiation codon by a TAG stop codon (met::stop) or ii) by introducing a TAG stop codon in the middle of the CDS (50%::stop).As expected, we observed a general reduction in GFP fluorescence in all untranslated mutants (Fig. 5B).Preventing translation initiation (met::stop mutants) lead to a 2-to 10-fold reduction in downstream cassette expression, while introducing a stop codon in the middle of the CDS (50%::stop) also reduced downstream cassette expression but to a lower extent, suggesting that the sooner translation is prevented, the higher the reduction in downstream expression.Indeed, we find that the length of the untranslated segments in these mutants (i.e., the distance between each stop codon and the ATG of the gfp gene), correlates positively with the decrease in fluorescence (Fig. 5C).Notably, in all cases, qRT-PCR of the gfp gene shows that downstream cassette mRNA levels also decrease with these mutations, in line with the view that translation interruption affects mRNA levels (Fig. 5B).
To verify that mRNA stability is indeed affected by translation in integrons we sought to prove that stabilization of mRNAs results in an increase in expression of the 2nd cassette.Notably, it has been demonstrated that some translation inhibitors, such as chloramphenicol (Cm), can stabilize bacterial mRNAs [40][41][42][43][44] .However, because Cm has this dual activity, -inhibiting translation elongation and stabilizing mRNAs-, we hypothesized that the later effect could only be observed in cassettes with low translation levels and low mRNA stability.To test this we measured growth and normalized fluorescence (fluorescence/OD) over time of cultures treated with 0.5, 1 and 2 µg/mL chloramphenicol, (12.5%, 25% and 50% of the MIC, respectively) 20 (Fig. 5D).The antibiotic effect of Cm was observed on the growth of all strains (Supplementary figure 4) except on the one containing catB6 which confers highlevel resistance to Cm 20 .As expected, Cm decreased fluorescence in pMBA, but increased fluorescence of aacA54, arr7 and dfrA21 cassettes, which are those displaying the lowest fluorescence.In the rest of ARCs, the effect of Cm on fluorescence was intermediate and its sign depended on the initial GFP levels, supporting that there is a balance between the inhibition of translation and the stabilization of unstable mRNAs (Supplementary figure 5 and blaVIM-1 in Fig. 5D).
Additionally, when we prevent translation initiation (met::stop), we observe that Cm treatment results in a higher increase in fluorescence in all cassettes except one (Fig. 5D and Supplementary figure 5).Altogether, these results show that, in general, ARC translation strongly impacts integron array expression by impacting the mRNA levels of downstream cassettes, which are likely influenced by the effect of translation on the stability of the whole mRNA molecule.

Case-specific analysis of a highly repressive cassette confirms importance of translation
To confirm the importance of translation rates in polar effects, we sought to investigate their role in an extreme case, where repression is maximal and additional mechanisms can be involved.Such is the case of aacA54, the most repressive cassette in our collection (it entails 50-and 10-fold decreases of fluorescence and mRNA levels, respectively) (Fig. 2A and 3A).This cassette contains a 25bp 5' UTR (untranslated region), a 555bp coding sequence, and is one of the very few to contain a long 3' UTR (162 bp), encompassing the 70bp-long attC site (Fig. 6A).We sought to determine the relative importance of translation rates in the polar effects of this cassette, and the presence of other mechanisms.
The 5' UTR region of genes contains the Ribosome Binding Site (RBS) and other signals (like mRNA structures or A-rich tracts) that collectively control their translation initiation rate 37,[45][46][47] .In this sense, aacA54 harbors a Shine-Dalgarno (SD) motif, AATCAA, that is far from the canonical AGGAGG.To assess if this weak RBS might result in a low translation rate for aacA54 and a low stability of mRNA, we modified the SD to the canonical AGGAGG.Notably, the resistance levels against four different aminoglycosides increased 2-to 4-fold, confirming an increased translation rate in this mutant (Supplementary figure 6).This change in the RBS of the first cassette increased 3-fold the expression (Fig. 6B) and the mRNA levels of the second cassette (the gfp gene) (Fig. 6C) which confirms the important role of translation rates.
Additionally, we searched for elements within the CDS of aacA54 that could lower the expression of the second cassette, such as stretches of rare codons, known to slow down translation 25,32 .By performing serial 90bp deletions covering the whole length of aacA54, preserving the start and stop codons as well as the reading frame, we found no changes in fluorescence (Fig. 6B), indicating that the CDS does not affect downstream expression.
As mentioned before, aacA54 contains a 98bp long 3' UTR that separates the stop codon from the attC site, contrarily to the majority of the other ARCs where the attC overlaps with the stop codon of the CDS.We hypothesized that the presence of this region could negatively affect the expression levels of the next cassette, either due to the presence of transcriptional terminators 48 or an untranslated stretch of mRNA.Indeed, the deletion of the UTR resulted in a 4-fold increase in fluorescence and a similar increase in gfp mRNA levels (Fig. 6B and 6C).When combined with a canonical RBS, the effect was additive both at the fluorescence and the mRNA levels (Fig. 6B and 6C), suggesting that both elements act independently.Moreover, when we introduced this 3' UTR between the stop codon and the attC site of aacA52, a closely related cassette with high initial fluorescence levels, we observed a 4-fold reduction in fluorescence (Fig. 6D).Instead, replacing only the attCaacA52 site with attCaacA54 caused a minor reduction in fluorescence.We identified a small hairpinlike structure that could be acting as an intrinsic transcriptional terminator (Supplementary figure 7).However, removing 8 nucleotides essential for this structure did not affect fluorescence levels, ruling out the role of this sequence as a terminator (Fig. 6B).We then mutated the canonical stop codon in aacA54 so that translation is maintained in frame throughout all the 3' region until the attC, where it terminates at a new stop codon.This led to a 2-fold increase in fluorescence, which agrees, in part, with the view that lack of translation at the 3' UTR lowers downstream expression (Fig. 6B).
In summary, translation rates are a key element in polar effects even in extremely repressive cassettes.Nevertheless, translation is likely not the only factor producing polar effects in integron arrays.Indeed, we have found some exceptions in our findings that support the occurrence of other phenomena, such as cases of known and novel promoter activity in ereA3, dfrA15, dfrA1 and qacE; and the negative effect of the attC site of aacA35.As for the mechanism through which some attCs lower downstream expression, one can hypothesize that the highly structured form of attCs may be facilitated by the lack of ribosome trafficking in cassettes with low translation rates, which could potentially leave the 5' region of the gfp "cassette" temporarily more exposed to ribonucleolytic attack.
Our finding that the identity of cassettes ultimately controls polar effects in integrons may have important consequences in the clinical context.A limitation of our study is that we used a multicopy vector, which likely results in higher absolute MIC values than those in natural environments, as discussed in 20 .Nonetheless, we believe the described polar effects should exist independently of copy number or Pc variant in natural integrons.The fact that a given ARC may confer different levels of resistance, depending on the preceding cassette has important consequences in coselection phenomena.Indeed, antibiotic combination therapies may select for arrays whose cassette order allows for the expression of the entire array.Moreover, considering the fitness costs entailed by the majority of antibiotic resistance genes, and how it relates to the expression levels of the AR gene 51,52 , cassette order can optimize the cost of the array.Interestingly, some resistance genes -such as trimethoprim resistance dfrs-have an almost digital phenotype, conferring high resistance even at very low expression levels.Hence, one could imagine a situation in which a cassette represses the expression of the next ARC in the array, decreasing its cost, but maintaining its resistance phenotype.In other words, in the light of our findings, integrons can optimize the tradeoff between fitness cost and function of antibiotic resistance genes.This work changes the paradigm of expression in integrons, and -together with other works showing the presence of promoters in cassettes-puts forward a more complex scenario.In this new model, inferring the levels of expression of a given cassette is extremely challenging and needs a case-by-case assessment, especially if preceded by cassettes not found in our collection.

Bacterial strains, plasmids, and culture conditions
Escherichia coli MG1655 strains and plasmids used in this study are based on the previously published pMBA collection 20 and are listed in Supplementary Table 1.All strains were cultured in liquid Müeller Hinton (MH; Oxoid, UK) media, at 37°C with agitation at 200 rpm or solid lysogeny broth (LB) agar (1.5%) (BD, France).Zeocin (Zeo) (Invitrogen, USA) was added at 100µg/mL for plasmid maintenance.
All pMBA collection-derived mutants were obtained by backbone amplification using primers designed to achieve the desire modifications followed by Gibson assembly 53 (Supplementary Table 1).

Flow cytometry analysis of fluorescence
Strains containing pMBA ARC-gfp plasmids were previously streaked on LBsolid medium with 100µg/mL Zeo.Three independent colonies were then inoculated in 200µL MH liquid medium supplemented with 100µg/mL Zeocin and incubated at 37°C with agitation for 20h.The next day, cultures were diluted 1:100 in 200µL MH+Zeo and incubated for 2h to reach exponential phase.At this point, cultures were diluted 1:20 in filtered saline solution (NaCl 0.9%) and fluorescent intensity was measured by flow cytometry using a CytoFLEX-S cytometer (Beckman Coulter, USA).The measurement of each biological replicate is the result of the mean of the fluorescence intensity of 30 000 events per sample.Data processing was performed with Cytexpert software.

Quantitative Reverse Transcription PCR of gfp gene
For RNA extraction, overnight cultures of three biological replicates of each strain were diluted 1:100 in MH medium supplemented with zeocin (100 µg/mL) and grown in 96 well plates with agitation at 37°C for approximately 2 -3 hours, until reaching exponential phase.Total RNA was extracted using a KingFisher Flex automated system with the MagMAX mirVana Total RNA Isolation Kit from Applied Biosystems, according to manufacturer's instructions.The purity and concentration of total RNA was determined by spectrophotometry (BioSpectrometer).For cDNA synthesis, 200ng of RNA was used for reverse transcription using the QuantiTect Reverse Transcription Kit (Qiagen) according to manufacturer's instructions.A negative control without Reverse Transcriptase was included to exclude possible gDNA carryover.cDNA was diluted 1:100 and 1µL of this dilution was used for quantitative PCR (qPCR).qPCR was performed in QuantStudio 3 Real-Time PCR system (Applied Biosystems) with the QuantiTect Multiplex PCR kit (Qiagen) according to manufacturer's instructions.Primers and Probes used in the mix are listed in SUP TABLE.PCR thermocycling conditions contained an initial stage of 2 min at 50°C and 15 min at 95°C followed by 42 cycles of amplification (1 min at 94°C and 1 min at 60°C).Negative controls included both a reaction containing water instead of template and a reverse transcriptase-free reaction.The relative abundance of gfp transcripts was normalized to that of the housekeeping gene rssA 54 using the 2 -∆∆Ct method 55 .

Plasmid copy number (PCN) assessment
PCN was assessed as in 56 .Briefly, overnight cultures of three biological replicates of each strain were diluted 1:100 in MH medium supplemented with zeocin (100 µg/mL) and grown in 96 well plates with agitation at 37°C for approximately 2 hours, until reaching exponential phase.50µL were then collected and briefly centrifuged for 2 minutes.The supernatant was then discarded, the pellet resuspended in molecular biology grade water and then boiled at 98°C for 10 minutes.After a brief centrifugation, 30µL were collected and then diluted 1:10. 1 µL of this dilution was then used as template for the exact same qPCR reaction described above.

Growth curves and GFP measurement of chloramphenicol-treated cultures
Three independent colonies of each strain were inoculated in MH+Zeo and incubated at 37°C with agitation for 20h.Cultures were then diluted 1:1000 in fresh MH+Zeo media containing or not subinhibitory concentrations of chloramphenicol (0.5, 1 and 2µg/mL).Growth (OD600) and GFP fluorescence (measured at 488nm wavelength) were followed for 20h using a Biotek Synergy HTX plate reader.
Measures were taken every 20 minutes with prior shaking.

Figure 1 .
Figure 1.(A) Schematic representation of the current integron model.When expressed, the integrase (encoded by the intI gene) is able to integrate, excise and re-shuffle discrete elements called cassettes.These are composed of a gene and an attC recombination site, and are arranged in arrays that contain multiple cassettes.

Figure 2 .
Figure 2. Cassette identity dictates expression and the resistance level of downstream ARCs (A) GFP fluorescence of all 136 pMBA-ARC strains measured by flow cytometry normalized to fluorescence of pMBA control strain.Bars represent the mean and SD from three biological replicates.(B) Distribution of the GFP fluorescence ratios (pMBA-ARC / pMBA) with indication of the mean value and SD.(C) Violin plot depicting cassette length distribution of all 136 ARCs.Median and quartiles are represented.(D) Correlation between relative GFP fluorescence levels and ARC length.r indicates Pearson's correlation coefficient.(E) Relative fluorescence of aacA54, aacA61 and aacA8 cassettes.Resistance levels conferred by pMBAØ or pMBA carrying the indicated arrays to cefaclor, ertapenem and chloramphenicol.Bars depict the mean and SD of the MIC values of three independent biological replicates.A red dotted line indicates the clinical breakpoint (EUCAST) for E. coli against the respective antibiotic.

Figure 3 .
Figure 3. gfp mRNA levels of all 136 ARCs.(A) gfp mRNA levels of all 136 pMBA-ARC strains measured by RT-qPCR normalized to the gfp mRNA levels of pMBA control strain (ARCs in the X-axis follow the same order as in Fig. 2A).Bars represent the mean and SD from two to three biological replicates.(B) Distribution of the gfp mRNA ratios (pMBA-ARC / pMBA) with indication of the mean value and SD.(C) Correlation between GFP fluorescence and gfp mRNA ratios.r indicates the Spearman's correlation coefficient.

Figure 4 .
Figure 4. Role of secondary promoters and attCs in polar effects.(A) Detection of Pc-independent transcriptional activity in a subset of cassettes displaying high GFP expression.Bars depict the GFP fluorescence (arbitrary units) of WT and ∆Pc mutants.Dotted line marks the mean value of the pMBA ∆Pc control strain.Statistically significant differences compared to pMBA ∆Pc control were determined using unpaired t-test, *** P<0.001.(B) Role of attC sites in a subset of cassettes displaying low GFP expression.Bars depict GFP fluorescence (arbitrary units) of WT and ∆attC mutants.Statistically significant differences between ∆attC and WT strains were determined using unpaired t-test, *** P<0.001; **** P<0.0001.Only statistically significant comparisons are shown; (inset) gfp mRNA fold change of those ∆attC mutants displaying significantly different GFP fluorescence levels.

Figure 5 .
Figure 5. Impact of translation on downstream cassette expression (A) Translation initiation rates of cassettes were predicted using the web tool RBS Calculator 2.1.Correlations between gfp mRNA levels of all ARCs and their translation initiation rates (TI); codon adaptation index (CAI) and minimum fold energy of mRNA.Correlations between the GC content of all ARCs and their gfp mRNA levels and predicted TI.Spearman's rank correlation coefficient (ρ) and p-values are indicated (B) GFP fluorescence and gfp mRNA levels (fold change) of a subset of ARCs and their respective translation mutants.met::stop represents the indicated ARC with the methionine initiation codon replaced by a stop codon.50%::stop represents the indicated ARC with a stop codon introduced at the 50% of the CDS.Bars represent the mean and SD from at least three biological replicates.(C) Schematic representation of mutants used to study the impact of translation on polar effects in ARCs.Correlation between the length of the untranslated segment and the respective decrease in GFP fluorescence.r indicates the Spearman's correlation coefficient.(D) Effect of subinhibitory doses of chloramphenicol on downstream cassette expression.For each indicated strain the left panels display the GFP/OD, with line and shading representing the mean and SD of three biological replicates, respectively.The right panels display the area under the curve (AUC) of the Cm-treated cultures over the non-treated control with bars representing the mean and SD from three biological replicates.

Figure 6 .
Figure 6.RBS and 3' UTR of aacA54 independently govern gfp expression.(A) Representation of aacA54 array with each analyzed segment depicted.(B) Bar graphics represent the fold change in GFP fluorescence of each modified strain relative to the original aacA54 strain.(C) gfp mRNA fold change of the indicated aacA54 mutants.Bars represent the mean and SD from at least three independent biological replicates.(D) Representation of aacA52 modified strain now including the 98bp segment from aacA54.Bar graphics represent GFP fluorescence levels (arbitrary units) of original or modified aacA52 strains.