The bacterial strains used in this study are listed in Table 1. E. coli strains were cultured under agitation in LB medium (tryptone 10 g l^-1, yeast extract 5 g l^-1, NaCl 5 g l^-1). The same medium containing 1.5% agar was used for growing bacteria on plates. Selection of E. coli after transformation with pBAD and pMPM vectors was performed using ampicillin 100 μg ml^-1 and tetracycline 15 μg ml^-1, respectively. Kanamycin 25 μg ml^-1 and chloramphenicol 12 μg ml^-1 was used for selection of chromosomal insertion of kan and cat genes.

Expression plasmids used in this study are listed in Table 1 and were constructed as follows. The nucleotide sequence of ycgZ, ymgA, ymgB, and ymgC was amplified by polymerase chain reaction (PCR) using the primers listed in Supplementary Table S1 and E. coli AG100 genomic DNA as template. The fragments were cloned into the pBAD/HisA and pMPM vectors using restriction sites indicated in Table 1. We constructed the pDVMBluR by PCR amplification of bluR nucleotide sequence, the 200 bases upstream of start codon GTG, and the 100 bases downstream of its stop codon TAA using the primers listed in Supplementary Table S1. The PCR fragment was then ligated into the pMPM plasmid using the restriction sites EcoRI and XhoI, resulting in plasmid pDVMBluR. All nucleotide sequences were verified at the Tufts University Core Facility (Tufts University School of Medicine, Boston, MA, United States).

Targeted deletion of bluR and ycgZ-ymgABC, and subsequent marker removal were made using the λRed recombinase method previously described (Datsenko and Wanner, 2000). The Flp recombination target (FRT)-flanked chloramphenicol resistance gene (cat) has been amplified by PCR from plasmid pKD3 using primers listed in Supplementary Table S1. bluR-PA/bluR-PB and ycgZ-PA/ymgC-PB primers contain sequences upstream and downstream of bluR and of ycgZ-ymgABC operon, respectively. The PCR product was gel-purified and concentrated by ethanol precipitation. Transformants carrying the Red helper plasmid pKD46 were then grown in LB medium with 100 μg ml^-1 ampicillin and 10 mM L-arabinose at 30°C to an optical density at 600 nm (OD₆₀₀) of 0.6 and then made electro-competent. Electroporation was done using 200 ng of PCR product. Chloramphenicol resistant clones were selected on LB agar plates containing chloramphenicol 12 μg ml^-1. Correct integration of the cat gene in the targeted genes was verified by PCR using the primers listed in Supplementary Table S1. Appropriate chloramphenicol resistant clones were subsequently transformed with the pCP20 plasmid and ampicillin resistant clones were selected at 30°C on LB agar plates containing ampicillin 100 μg ml^-1. The transformants were then colony-purified non-selectively at 42°C on LB agar and then tested for loss of ampicillin and chloramphenicol resistance. Deletions were further verified by PCR using the primers listed in Supplementary Table S1.

To construct the plasmid pDV415O, amplification of ompF promoter (PompF) was carried out by PCR using chromosomal DNA from strain E. coli AG100 as template and the primers ompF1 and ompF2 listed in Supplementary Table S1. The PompF fragment was 273 bp long (from -273 to +1 relative to the transcription start) and was cloned into the pGEM-T Easy vector (Promega) following the manufacturer instruction. The resulting plasmid was digested with EcoRI and BamHI and the fragment corresponding to PompF was ligated to the similarly cut vector pRS415 yielding the plasmids pDV415O. The sequence of the PompF-lacZ fusion in pDV415O was then verified at the Tufts University Core Facility. Insertion of PompF-lacZ into E. coli chromosome was realized as followed. Recombination between the pDV415O and λRZ5 (Silhavy et al., 1984; Simons et al., 1987) resulted in a lysate bearing λRZ5 (PompF-lacZ). This was used to infect strains BW25113, a λ- and lac- strain of E. coli. Amp^R Lac+ lysogens were selected and purified on LB agar containing ampicillin 20 μg ml^-1 and X-Gal (5-bromo-4-chloro-3-indolyl-β D-galactopyranoside) 40 μg ml^-1. Lysates from these lysogens were then used to infect at low multiplicity of infection (MOI = 0.005) strain BW25113 and derivative mutants. Amp^R Lac+ lysogens were again isolated and the resulting strains were confirmed by PCR, as previously described (Powell et al., 1994), to have a single copy of the transcriptional fusion located in the λatt site on the chromosome. To assess the β-galactosidase (LacZ) activity, overnight cultures of fresh colonies of E. coli carrying a chromosomal λPompF-lacZ fusion were grown in LB medium containing ampicillin 20 μg ml^-1 and were subsequently diluted to an OD₆₀₀ of 0.05 in identical medium for growth with no antibiotic added. When the OD₆₀₀ reached 0.6, β-galactosidase (LacZ) activity was assayed by first rendered the cells permeable with 0.005% sodium dodecyl sulfate and 0.05% chloroform. LacZ activity was expressed in Miller units as previously described (Miller, 1972). All assays were carried out at least in three independent experiments.

Fresh colonies of E. coli strains were grown overnight (14–16 h) in 3 ml of LB medium. The cultures were then diluted to an OD₆₀₀ of 0.05 in 3 ml of identical fresh medium and cells were grown until an OD₆₀₀ of ∼0.6. The total RNAs were isolated from 0.5 ml of culture using the Trizol Reagent (Thermo Fisher Scientific) following the manufacturer protocol. Directly after the preparation of the RNAs, the integrity of the RNAs was evaluated on a bleach gel stained with ethidium bromide (Aranda et al., 2012). The RNAs amount was quantified by absorbance at 260 nm (A260) and purity was evaluated by a A260/A280 ratio >1.9 and a A260/A230 ratio >1.7 using a Nanodrop ND-1000 spectrophotometer (Thermo Fisher Scientific). RNAs were stored for no more than 3 days at -80°C before DNase treatment and cDNA synthesis (see Reverse Transcription and Quantitative PCR Analysis).

Two micrograms of purified RNAs were treated using the Turbo^TM DNase (Thermo Fisher Scientific) following the manufacturer instructions. A total of 500 ng of RNAs was then used to synthesize cDNA by reverse transcription using the Quantitect Reverse Transcription kit (Qiagen) and following the manufacturer instructions. To control for chromosomal DNA contamination, the reverse transcription step was also performed with reaction mixtures containing no reverse transcriptase and was used as a negative control in subsequent quantitative PCR (qPCR) reactions. Primers used for the qPCR were designed using the online PrimerQuest tool from Integrated DNA Technologies and are listed in Supplementary Table S1. Amplification efficiency and specificity for each set of primers are reported in Supplementary Figure S1. All qPCR reactions were performed using a Roche LightCycler 480 instrument II. The following experimental run protocol was used: UDP activation (50°C for 2 min), denaturation (95°C for 2 min), quantification program repeated 40 times (denaturation at 95°C for 5 s, anneal/extend 60°C for 30 s with a single fluorescence measurement), melting curve program (60–95°C with a heating rate of 0.15°C s^-1 and a continuous fluorescence measurement). Reactions were carried out in 20 μl with 2 μl of diluted cDNA, 0.4 μl of 10 μM forward primer (0.2 μM final concentration), 0.4 μl of 10 μM reverse primer (0.2 μM final concentration), 7.2 μl H₂0 and 10 μl of the Power Up SYBR Green Master Mix (Applied Biosystems by Life Technologies). After the reverse transcription step, the cDNA samples were diluted fivefold and used as templates for qPCR amplification of gapA, ompF, and ycgZ. A mix with no cDNA was also prepared (non-template control, NTC) and was run in parallel. NTC wells did either gave a C_T > 35 or no C_T at all. We quantified the relative expression of ompF and ycgZ transcripts to the reference gene gapA. Using the Pfaffl method (Pfaffl, 2001), we determined the ratio R of a transcript expressed in a sample versus that expressed in the wild type strain grown at 37°C (see C_T and calculations in Supplementary Tables S2–S7). gapA encodes the glyceraldehydes 3-phosphate deshydrogenase-A.

We report the average (mean) and the standard deviation (SD) from at least three experimental values. In Figures 2B and 4B, we report the ratio R3 (±SD3) of two numbers R1 ± SD1 and R2 ± SD2, with R3=R2R1andSD3R3=(SD1R1)2+(SD2R2)2. The statistical significance of differences between two averages was determined by a Student’s t-test (two independent samples, with two-tailed distribution) using GraphPad Prism software.¹

Overnight culture of E. coli wild type (BW25113) and lon (JW0419-1) strains carrying plasmids pBAD/HisA and derivative pDVBZ, pDVBZ-XP, pDVBA, pDVBB, and pDVBC were grown in LB medium in the presence of 100 μg ml^-1 ampicillin. Fresh identical medium supplemented with L-arabinose was then inoculated to an optical density measured at 600 nm of 0.05 and cells were grown at 37 or 25°C to OD₆₀₀ of 1. Whole cell extracts were prepared for analysis of the steady-state levels of protein. To assess the intracellular stability of YcgZ, 150 μg ml^-1 chloramphenicol was added to stop the proteins synthesis and the cultures were kept at the indicated temperatures. Samples to be used for preparing whole cell extracts were removed at indicated times (0–60 min). Cell extracts were prepared as follows: 1 ml of culture was centrifuged and the cells pellet suspended in 200 μl of lysis buffer per OD₆₀₀ of 1 [Tris–HCl 10 mM pH 8.0, EDTA 0.5 mM, CaCl₂ 10 mM, and 1 unit ml^-1 of DNase (Promega)]. Cells were sonicated on ice two times 20 pulses using a Branson Sonifier 250 and the following parameters: output control = 1 and a duty cycle = 50%. Protein concentration was determined using the Pierce 660 nm Protein Assay Reagent and bovine serum albumin as a standard (Thermo Fisher Scientific). Eight micrograms of proteins were separated on a 16% acrylamide gel in denaturing condition (100 mM Tris, 100 mM Tricine, and 0.1% SDS). The gels were stained with Coomassie Brilliant Blue R-250 (Sigma-Aldrich) or used for western blot analysis.

After separation on a 16% acrylamide gel, the proteins were electro-transferred to a nitrocellulose membrane (Millipore, Billerica, MA, United States). The membrane was incubated overnight at 4°C in Tris-borate-saline (TBS) buffer supplemented with 3% milk powder. The membrane was then incubated for 2 h at room temperature with monoclonal anti-XPress antibodies (Thermo Fisher Scientific) diluted in TBS (1/6,000). After three 15 min washes with TTBS (TBS supplemented with 0.05% Tween 20), the membrane was incubated for 2 h with alkaline phosphatase-coupled anti-mouse IgG antibodies (Promega) diluted 1/10,000 in TBS followed by three 15 min washes with TTBS and two 5 min washes with TBS. XPress-YcgZ was visualized by adding 5-bromo-4-chloro-3-indolyl phosphate and nitroblue tetrazolium following the manufacturer’s instructions (Promega).

When grown at 37°C, E. coli expresses low amount of OmpF in the outer membrane, while decreased temperatures lead to increased levels of the porin (Pratt et al., 1996; Nikaido, 2003). A previous study performed in our laboratory demonstrated that E. coli carrying mutations in both lon and bluR loci resulted in significantly lower amounts of the porin when the cells were grown under low temperature conditions (see protein level of OmpF in Duval et al., 2009). To better characterize the effect of lon and bluR on ompF expression, we performed reverse transcription and qPCR (RT-qPCR) and compared the level of OmpF messenger in wild type E. coli AG100 and derivative strains carrying mutations in lon, bluR, and lon bluR. All strains were grown at 37 and 25°C. Our data show that wild type E. coli grown at 25°C induces OmpF expression by approximately sixfold when compared to that of wild type E. coli grown at 37°C (Figures 2A,B). Our data also indicate that a single mutation in bluR has no significant effect on OmpF mRNA level, while a lon mutation slightly decreases the expression of OmpF at both temperatures (Figure 2A). Nevertheless the fold change in OmpF mRNA for both lon and bluR mutants grown at 25 versus 37°C is similar to that of wild type AG100 (Figure 2B). However, a lon bluR mutant carrying both mutations fails to increase ompF expression when grown at 25°C.

To further characterize the role of Lon and BluR in regulating OmpF expression, we constructed a transcriptional reporter fusion of ompF promoter with lacZ (PompF-lacZ). The resulting fusion did not carry ompF 5′-UTR affected by MicF (Andersen and Delihas, 1990), but instead carried lacZ 5′-UTR. In this case, we prevented post-transcriptional effects on ompF 5′-UTR. A single copy of the reporter fusion was then integrated at the att(λ) site of wild type E. coli BW25113, a lac-λ- strain, and of derivative mutants lon, bluR, and lon bluR. LacZ activity was then assayed in cells grown at 37 and 25°C. Our results from the β-galactosidose assays show that transcription of PompF-lacZ increases by approximately threefold in the wild type strain grown under low temperature conditions (Figure 3). In contrast, the lon bluR double mutant fails to increase PompF-lacZ expression at 25°C and displays similar PompF-lacZ activities at 25 and 37°C (Figure 3). Plasmid mediated expression of bluR in the lon bluR mutant restores PompF-lacZ levels similar to that of wild type (Figure 3). Our data also show that a single mutation in either lon or bluR has no effect on PompF-lacZ transcription (Figure 3). We suspect that the decreased OmpF mRNA level observed for the lon mutant (see above) is likely due to a lower stability of the messenger in the absence of the Lon protease since no difference is detected at the transcriptional level.

The simplest model illustrating the effect of both lon and bluR mutations on ompF transcription under the growth conditions used in our experiments implies that BluR represses a locus coding for an intermediate protein that is a substrate of the Lon protease. This intermediate protein would act as a repressor of the ompF promoter. In this model, the inactivation of bluR leads to an increased transcription of the intermediate locus, and only in the absence or with a reduced activity of Lon, can the intermediate protein accumulate and considerably repress ompF at 25°C, a condition that normally permits increased expression of the porin in wild type E. coli. In an otherwise direct activation of ompF promoter by BluR, deletion of bluR alone would lead to decreased ompF transcription.

We subsequently aimed to identify the intermediate locus controlled by BluR and involved in the regulation of ompF expression. BluR has been previously described to directly repress the transcription of the ycgZ-ymgABC operon (ZABC; Tschowri et al., 2012). This operon encodes small proteins of 78–90 amino acid residues involved in biofilm formation through a mechanism that is yet to be determined (Tschowri et al., 2009). Using RT-qPCR, we measured the expression of the ZABC operon in AG100 (wt), AGEZ3 (bluR), and M113REZ3 (lon bluR) derivative strains. When comparing YcgZ mRNA levels at 37 and 25°C for the wild type strain, our results shows a ∼46-fold increased expression of ycgZ when the cells were grown at ambient temperature (Figures 4A,B). Our data also determine that a bluR mutant increases expression of the ZABC operon by ∼100- and ∼4-fold in cells grown at 37 and 25°C, respectively (Figure 4A), indicating a strong repression of the ycgZ promoter by BluR at 37°C. It was previously shown by Tschowri et al. (2009) that BluR–DNA interaction is released in the presence of BluF, a direct antagonist of BluR whose activity is induced by low temperature. We believe that the weaker repression of the ycgZ promoter observed at 25°C likely comes from the BluR inactivation by BluF under these growth conditions. At 37°C, BluR repressor is fully active and its deletion leads to a large de-repression of the ycgZ promoter. Our results in Figure 4A also show that the addition of a lon mutation to a bluR mutation (lon bluR) leads to a slight increased in ycgZ mRNA levels when compared to that of bluR. For AGEZ3 (bluR) and M113REZ3 (lon bluR), the low temperature mediated induction is only approximately twofold, overall indicating that bluR is mainly responsible for the temperature-dependent expression of ycgZ.

We then investigated whether the ZABC operon, which is de-repressed at 25°C, was involved in the control of ompF expression. In our model, if one of the proteins encoded by the ZABC operon accumulates in a lon bluR strain and represses ompF, we expect to see an increased ompF expression when the ZABC operon is deleted, i.e., in strain lon bluR ZABC. Using RT-qPCR, we compared OmpF mRNA levels in wild type, lon bluR and lon bluR ZABC strains grown at 37 and 25°C. When grown at 37°C, the three strains show similar OmpF mRNA levels (Figure 5A). Under low temperature growth conditions, the lon bluR ZABC strain expressed OmpF mRNA level similar to that of the wild type strain (Figure 5A). We further confirmed that the ZABC operon was involved in the control of ompF transcription using our PompF-lacZ reporter fusion; Figure 5B shows that deletion of the ZABC operon in a lon bluR background restores LacZ activity similar to that of the wild type. Taken together, our results suggest that the ZABC operon encodes a protein able to repress ompF expression. To further identify this repressor, we used a low copy pBAD vector allowing the expression of a target gene from the L-arabinose-dependent promoter araBAD. We independently expressed ycgZ, ymgA, ymgB, and ymgC in the lon bluR ZABC strain and subsequently measured the LacZ activity in the transformants grown at 25°C in the presence of L-arabinose. Our results show that the activity of PompF-lacZ decreases in response to increasing amount of YcgZ, while YmgA, YmgB, and YmgC have no significant effect on PompF-lacZ activity (Table 2 and see also Supplementary Figure S2). While our results identifies YcgZ as a repressor of the ompF promoter and our data show that ycgZ expression is high in a bluR mutant (Figure 4), a bluR mutant does not decrease ompF transcription (Figures 2, 3). However, a lon bluR strain grown at 25°C expresses less ompF transcript than the bluR mutant. These results strongly suggest that YcgZ is unstable in the presence of Lon under the growth conditions used in our experiments. That a lon bluR strain grown at 37°C does not decrease transcription of ompF is probably due to growth conditions that inherently lead to a high repression of ompF expression. In this case, YcgZ amount may be too low to further repress the ompF promoter.

To evaluate whether YcgZ was a substrate of the Lon protease, we compared YcgZ protein amount in E. coli wild type and lon mutant. For completion purpose, we also evaluated the stability of YmgA, YmgB, and YmgC. Lacking antibodies that can specifically interact with the proteins, we decided to evaluate the steady-state level of each native protein when overexpressed from an araBAD promoter using pBAD derivative plasmids (see Table 1). Protein expression was compared between BW25113 (wild type) and JW0419-1 (lon) carrying the plasmids and grown in the presence of L-arabinose. Our experiments indicated that the amount of YcgZ was significantly lower when expressed in the wild type strain, while a lon mutant accumulated a large amount of YcgZ (Figure 6A). Of note, we observed this phenomenon when the cells were grown at both 25 and 37°C. We verified that the lower amount of YcgZ in the wild type was not due to a lesser expression of ycgZ by quantifying ycgZ messenger levels. RT-qPCR experiments showed that ycgZ was similarly expressed in both strains (see Supplementary Table S8), suggesting that the higher level of YcgZ protein observed in the lon mutant is likely due to a higher stability of the protein in the absence of Lon. Our data illustrated in Figure 6A also shows similar amounts of YmgA and YmgC proteins when overexpressed in the wild type and in the lon strains. We could not detect significant amount of YmgB in any of the strains tested even though high level of ymgB messenger was measured by RT-qPCR in both strains (data not shown). Using an XPress-tagged YcgZ (XPYcgZ), we confirmed an L-arabinose-dependent accumulation of YcgZ in the lon strains, while the protein was barely detectable in the wild type (Figure 6B). Additionally, we examined the stability of native YcgZ in the wild type and lon strains after stopping the protein synthesis with chloramphenicol and determining the amount of YcgZ in multiple samples taken over a certain period of time (Figure 6C). We observed a decline in YcgZ quantity in the wild type strain, while a high amount of the protein was maintained over time in the lon mutant, demonstrating a significant stability of YcgZ in the absence of Lon.

We further aimed to confirm the lon-dependent repression of ompF promoter by YcgZ. YcgZ was expressed in the ZABC and lon bluR ZABC mutant strains, both carrying a chromosomal PompF-lacZ fusion and providing both lon+/lon- backgrounds. In the absence of lon, expression of YcgZ significantly decreased LacZ activity with addition of 0.0025% of L-arabinose (Figure 7). In a lon+ background, similar repression of ompF promoter was reached by increasing the L-arabinose concentration by at least threefold (0.0075%). Ultimately, when a higher concentration of L-arabinose was used (>0.01%), both strains expressed a similar level of LacZ indicating that under those conditions, an abundant level of YcgZ is sufficient to repress ompF promoter even in the presence of the Lon protease (Figure 7). Of note, the level of ycgZ transcript measured when ycgZ is expressed from its own promoter is much lower than that expressed from the araBAD promoter with concentration of L-arabinose above 0.01% (compared Supplementary Tables S4, S5 with Table S8).