Histone-like nucleoid structuring (H-NS) protein silences the beta-glucoside (bgl) utilization operon in Escherichia coli by forming a DNA loop

Graphical abstract


Introduction
The b-glucoside (bgl) utilization operon in Escherichia coli (E. coli) encodes necessary proteins for the uptake and utilization of aromatic b-glucosides such as salicin, esculin and arbutin [1][2]. The operon is silent in wild-type E. coli strains such that cells are not able to grow on these b-glucosides as sole carbon sources. In spite of this, the bgl operon can be activated by mutations, and it then becomes inducible in the presence of b-glucosides, so that E. coli cells can utilize b-glucosides for growth, leading to the Bgl positive (Bgl + ) phenotype. One of the most common types of mutations giving rise to the Bgl + phenotype is the transposition of IS elements like IS1 and IS5 into the bgl regulatory region [2][3]. In addition, the operon can be activated by non-insertional mutations in the bgl regulatory region or outside the bgl operon, which counteract the silencing of the operon [4][5][6][7][8][9][10][11].
The bgl operon includes three genes, bglG, bglF and bglB, with two rho-independent terminators flanking bglG [2,12]. In the absence of b-glucosides, transcription is terminated at the two terminators due to the formation of hairpin-like secondary structures. However, in the presence of b-glucosides, the product of the first gene, bglG, acts as a transcriptional antiterminator which binds to an upstream region that partially overlaps each terminator. This prevents the formation of the hairpin-like structures of the two terminators, thus relieving the termination and allowing full length transcription of the operon [2,[13][14][15]. The second gene, bglF, encodes an enzyme II (EII) of the phosphoenolpyruvate:sugar phosphotransferase system (PTS), which is responsible for the transport and phosphorylation of extracellular b-glucosides [16][17][18]. The BglF protein can also phosphorylate and dephosphorylate BglG, regulating BglG antitermination [12,18]. In the presence of bglucosides, BglG is dephosphorylated by EII and phosphorylated at a different site by a phosphocarrier protein (HPr) of the PTS, which leads to full activity of BglG-mediated antitermination [12,[18][19][20][21] The third gene, bglB, encodes a phospho-b-glucosidase which hydrolyzes the aglycon-sugar linkage in the phospho-bglucosides that are transported and phosphorylated by BglF [2,22]. Even though the two terminators terminate transcription, and the activated BglG prevents termination, it is still unclear if the terminators contribute to and play a significant role in the silencing of the bgl operon.
There are several transcriptional regulators reported to be involved in upregulating expression of the bgl operon, such as the cyclic-AMP receptor protein (Crp), LeuO and BglJ [6][7][8][10][11]. Crp is a global regulator, which once bound to cyclic-AMP (cAMP), activates the bgl operon promoter by binding to a site in the regulatory region [23][24]. A single mutation in the Crp operator, creating a stronger Crp-cAMP binding site, leads to transcriptional activation of the operon [6]. In addition, LeuO and BglJ are transcriptional regulators which when present at high (overexpressed) levels, also bind to the bgl regulatory region and increase the expression of the bgl operon [7][8]10]. BglJ forms heterodimers with another transcriptional regulatory protein RcsB, which binds to the bgl regulatory region and activates the operon [11]. Both LeuO and BglJ, when overexpressed, relieve repression of the bgl operon when bound to the bgl regulatory region, but these transcription factors have no effect on the region downstream of the promoter [10].
As mentioned above, the bgl operon is silent in wild-type E. coli cells. The major repressor of the bgl operon is the histone-like nucleoid structuring protein (H-NS), which is involved in chromosomal organization and acts as a global negative regulator [4,9,23,[25][26]. H-NS preferentially binds to AT-rich and curved DNA regions [27]. It consists of a dimerization/oligomerization domain at the N-terminus and a DNA-binding domain at the Cterminus [28][29]. Once bound to DNA, H-NS can form higher order oligomers [30]. H-NS-mediated nucleoprotein complexes are believed to be important for effective gene repression and silencing. This protein acts by bridging DNA duplexes through H-NS oligomerization, thus allowing the formation of DNA loops [31][32]. The formation of such nucleoprotein complexes prevents transcription by either trapping RNA polymerase or blocking RNA polymerase from binding to the promoter region [33]. However, the change of leucine to proline at the 30th codon (L30P) position within the N-terminal domain of the protein abolishes H-NS's ability to form higher order oligomers while still retaining its ability to form dimers and bind to DNA [34].
In relation to the bgl operon, H-NS binds close to the Crp-cAMP binding site within the regulatory region, as well as a downstream region within the bglG gene [6,9,35]. However, the exact H-NS binding sites in the bgl operon are still not defined. In addition, it is still unclear what the exact mechanism is by which H-NS silences the bgl operon, although it is known that the binding of H-NS to both locations is required for full repression, and the repression involving H-NS at these two sites is synergistic [36]. A missense H-NS mutant carrying an amino acid substitution at position 30 (Leu30 to Pro or L30P), when expressed in a plasmid, is thought to be deficient in dimerization, thereby losing the ability to function as a transcriptional repressor [29]. In another report, this same H-NS mutant was reported to maintain its dimerization and DNA-binding property [34]. Also, when H-NS is truncated at the N-terminus with loss of its DNA binding ability, another repressor of the bgl operon, StpA can function as a DNA-binding adapter of H-NS to repress the operon [37][38]. However, StpA is negatively regulated by H-NS and can only regulate the bgl operon when H-NS is absent [39][40]. Therefore H-NS is the major repressor of the bgl operon in wild-type E. coli cells.
DNA looping, mediated by transcription factors, is a mechanism sometimes used for strong and effective regulation of gene transcription [41]. For the gal operon, the binding of the galactose repressor (GalR) to the two operators in the gal operon alone is insufficient to repress the operon. Instead, the formation of a DNA loop, resulting from the protein-protein interaction between GalR proteins, is required for strong repression of the gal operon [42]. Also, a similar DNA looping mechanism has been shown in the E. coli lac operon, with the binding of the lac repressor, LacI, to two of the three lac operators within the operon [43]. The tetrameric repressor CI of Lambda phage binds to its two operators O R and O L (separated by 2.3 kb) and loops them together by octamerization when the phage is incorporated into the E. coli chromosome [44][45]. The length of the intervening DNA in the DNA loop can be as long as 5 kb [46]. Please note that all these repressors have their specific operators (DNA binding sites). Each DNA loop consists of the intervening DNA between two operators and a fixed oligomer (a dimer for GalR, a tetramer for LacI, and an octamer for CI) [42][43][44][45].
As a common repressor, H-NS can exert its repressive effect on transcription by first binding to two or more DNA targets and then looping these target sites, trapping RNA polymerase at the promoter or excluding binding of RNA polymerase from its operator [33,47]. Such a repression loop has been visualized and reported in previous studies on the transcriptional regulation of the hdeAB and rrnB operons in E. coli [33,48]. Once an H-NS dimer binds to a target site, it undergoes self-oligomerization, forming a stiffened filament along the DNA strand [49]. The formation of such an H-NS/DNA nucleoprotein complex (that is, a stiffened filament) is required for efficient repression [50]. Likely, a similar mechanism applies to the repression of the bgl operon by H-NS, but this has not been demonstrated directly. Our prior studies have suggested that H-NS binding to its binding site(s) in the bgl promoter region is not sufficient to repress the operon, so we proposed that H-NS represses the bgl operon by the formation of a DNA loop between its two binding sites [39]. This proposed mechanism of bgl operon repression was investigated in-depth in this study.
Here, we first show both terminators flanking the bglG gene are not required for bgl operon silencing as the operon remains uninducible when these terminator sequences are deleted. We then examine the effect of preventing the formation of the H-NSmediated DNA loop on bgl operon transcriptional activity, as well as the ability of E. coli cells to grow on aromatic b-glucosides. In order to do so, the formation of the DNA loop was altered using the following approaches: (1) preventing H-NS selfoligomerization, (2) preventing or reducing H-NS binding in the bgl operon regulatory region, (3) preventing or reducing H-NS binding within the bglG gene, (4) changing the phase of the DNA between the two H-NS binding sites, and (5) changing the distance between the two H-NS binding sites. We demonstrate that H-NS binding to both upstream and downstream sites simultaneously is insufficient to abolish transcription. However, a DNA loop between these binding sites, formed by H-NS oligomerization, is conceivably required for full repression. The proposed H-NSmediated DNA loop is robust as it is independent of DNA phasing and the length of intervening DNAs. The presence of such a strong DNA loop would explain why other transcription factors (reported to be involved in bgl regulation), when expressed at wild type levels, have negligible effects on bgl operon expression. These results provide a new perspective on the mechanism by which the bgl operon is repressed by H-NS in wild-type E. coli cells.
tors play a role in silencing the bgl operon, we first removed both terminators from our bgl operon reporter, Pbgl-bglG-lacZ strain [51], creating a new operon reporter strain, G50-Z. As shown in Fig. 1A, the new operon reporter (without terminators) showed a slight increase in b-galactosidase (LacZ) activity compared to BW25113_Z.
The two terminators were also removed from the native bgl operon, yielding strain G50. As seen in Fig. 1B, the b-glucosidase (BglB) activity of strain G50 (with no terminators) was similar to that of the wild-type, although with a slight increase. The G50 strain also showed a Bgl À phenotype, as its colonies did not show a visible red color after 1 day of incubation on MacConkey indicator plates with 0.5% salicin at 37°C, indicating its inability to ferment salicin. Also, no visible colonies were seen on M9 minimal agar plates with salicin after 2 days of incubation at 37°C, which further confirmed its Bgl À phenotype. All of these results showed that the bgl operon remains silent (and non-inducible by salicin) in the absence of both terminators flanking the bglG gene. This terminator-less strain was used in most of the experiments in this study.

H-NS oligomerization is necessary for bgl operon silencing
In an earlier paper, we proposed that H-NS represses the bgl operon by forming a DNA loop between its two binding sites [39]. Here, we investigated this proposed DNA-looping mechanism in-depth. To form a DNA loop, H-NS proteins have to bind to the upstream site and the downstream site simultaneously and then perform self-oligomerization to bridge the two sites together. We examined the effects of altering each of these three factors (up-stream binding, downstream binding and self-oligomerization) on formation of such a loop by measuring bgl operon activity.
We first determined the effect by preventing the oligomerization of H-NS using the hnsL30P mutant (a missense mutation in the hns gene leading to a change of leucine to proline at the 30th codon) ( Fig. 2A) [29,39]. This derivative of H-NS is still capable of dimerizing, thereby binding to the target DNA, but it has lost the ability to oligomerize [34]. Therefore, possibly, this modified H-NS is unable to form nucleoprotein complexes involving DNA bridges and DNA looping. Fig. 2B reveals that the operon activity (44 units) of hnsL30P increased 550 times compared to strain BW25113 (0.08 units) which expresses the wild type H-NS, indicating this H-NSL30P mutant protein lost most of the repressive effect on the operon. However, the operon activity of hnsL30P was lower (by 53 units) than that of the Dhns mutant (Fig. 2B). Most likely this is attributable to the binding of H-NSL30P to both the Pbgl and bglG binding sites for H-NS, leading to some repression (see below). Both Dhns and hnsL30P showed Bgl + phenotypes on both MacConkey + salicin plates and M9 + salicin plates.
Our previous results have shown that H-NS and H-NSL30P are able to bind to the bgl promoter region in the absence of bglG, while H-NSL30P even showed a higher binding affinity [39]. Here, we further demonstrated H-NS binding to its downstream site within bglG, using the lacIQ promoter driving expression of the bglG-lacZ transcriptional fusion at the lac locus (Iq-G_Z). As seen in Fig. 2C, using the Iq-G_Z reporter, the presence of H-NS (strain BW25113) or H-NSL30P (strain hnsL30P) led to lower LacZ activity than the absence (Dhns) of either protein. On the other hand, both H-NS and H-NSL30P have similar repressive effects on bglG expression. These results confirmed that H-NSL30P retains the ability to bind to bglG like the wild-type H-NS, despite its inability to oligomerize. Together with our previous observations [39], this H-NS derivative, H-NSL30P, is still able to bind to its two sites although it has lost its oligomerization property. These data also indicate that H-NS binding to bglG alone (without DNA looping) only partially represses bgl operon expression.
A stronger promoter, Ptet, driving bglG-lacZ expression (Ptet-G-Z) was also tested for possible inhibitory effects of H-NS and H-NSL30P. However, no obvious operon repression was observed when H-NS or H-NSL30P was present (data not shown). This might be due to the fact that H-NS repression, due to binding to the bglG gene, is less efficient when a stronger promoter is used [36]. With a strong promoter like Ptet, the RNA polymerase might be able to transcribe through the H-NS binding site when H-NS or H-NSL30P is bound at this site. This explains our results, suggesting that H-NS repression is seen in the background of wild-type hns and mutant hnsL30P when a moderate promoter (Iq-G_Z) is used, but not when a stronger promoter (Ptet-G-Z) is used.

H-NS binding to the bgl operon regulatory region is necessary for strong bgl repression
Next, we sought to prevent H-NS-mediated DNA looping by decreasing or abolishing H-NS binding to the upstream site of the bgl operon regulatory region (Fig. 3A). The approaches included: (1) deletion of an upstream region carrying the proposed H-NS binding site; (2) insertion of IS elements upstream of Pbgl where the upstream H-NS binding site is thought to be, and (3) overexpression of bgl operon positive regulators. Fig. 3B shows that strain G50DO HNS -Pbgl (deleted for the upstream region, where the H-NS binding site is thought to be, in front of the Crp operator) exhibited a 90-fold greater operon activity than strain G50, in which H-NS can bind to both the upstream and downstream sites. This fact clearly indicates that with no binding (or a significantly decreased binding) to the upstream site, H-NS almost completely loses its repressive effect on bgl operon expression. As expected, strain G50DO HNS -Pbgl is Bgl + . Fig. 3C shows the bgl operon activity of an IS5 insertional mutant, IS5r, in which an IS5 element (in the reverse orientation) was inserted upstream of the Crp operator in the bgl regulatory region of BW25113. Compared to the wild-type strain, IS5 insertion dramatically elevated the operon activity (see columns 1 and 2). Furthermore, the IS5 insertional mutant, IS5r, shows an almost 2-fold increase in operon activity compared to the hns deletion mutant (Dhns). This may be because without H-NS, the other repressor, StpA is expressed at a greater level and therefore exerts a repressive effect on bgl operon expression [39]. This was confirmed using our b-glucosidase (BglB) assay. As shown in Figure S1, the stpA mutation had no effect on bgl operon activity, probably due to the strong H-NS repression of stpA expression. However, bgl operon transcription of the hns/stpA double mutant (DhnsDstpA) increased 2-fold compared to the hns single mutant, confirming that StpA exhibits an inhibitory effect on bgl operon expression in the absence of H-NS. On the other hand, the hns and IS5 insertional double mutant (IS5rDhns) showed a slightly higher operon activity than that of the IS5 insertional mutant alone (see columns 2 and 4 in Fig. 3C). It is possible that IS insertions not only prevent H-NS binding to the bgl regulatory region, but also generate new promoters, or attract new regulators to the bgl operon. As expected, the IS5r mutant showed a Bgl + phenotype on both MacConkey + salicin plates and M9 + salicin plates.
We further examined the effects of IS varieties, locations and orientations on alteration of H-NS binding to the bgl regulatory region. Fig. S2A shows that IS1 elements tended to be inserted further from the bglG transcriptional start site, followed by IS5 elements, with IS2 elements being closest to the bglG transcriptional start site. However, more data are needed to support this preliminary observation. Also, the same IS elements inserted in different orientations (direct or reverse) at the same location showed different operon activities (Fig. S2B). These results indicate that although insertions of IS1, IS2 and IS5 all activated the bgl operon, leading to a Bgl + phenotype, the activation levels by IS insertions are not closely correlated with the type of IS elements, the insertional positions, or the IS element orientations. Our previous work showed that IS elements insert preferentially into the SIDD site (AT-rich and easy-to-open to form single strand DNA) that most likely overlaps with the upstream H-NS binding site [52]. H-NS preferentially binds to AT-rich and curved DNA regions. Thus, it is conceivable that insertions of IS elements completely or partially abolish H-NS binding, leading to various levels of bgl operon activation.
Positive regulators (suggested from previous studies) of the bgl operon, including Crp, BglJ, LeuO and RcsB, all of which bind to the bgl regulatory region, were used to reduce H-NS binding to the bgl regulatory region. First, strain G51 is essentially the same strain as G50 except for the presence of a point mutation at the Crp binding site, creating stronger Crp-cAMP binding to the bgl regulatory region (Fig. 3D) [6]. The operon activity of G51 showed a moderate increase compared to G50, although it is still Bgl À . To see if the increased operon activity was due to stronger Crp-cAMP binding to the bgl regulatory region, the crp gene was removed from the operon reporter G51-Z (G50-Z with a point mutation in the Crp operator), creating operon reporter DcrpG51-Z. In the absence of crp, the b-galactosidase (LacZ) activity dropped to the level observed for G50-Z without the point mutation (Fig. S3). This confirmed that the actual stronger binding of Crp to the bgl regulatory region led to a substantial increase in the operon activity of strain G51-Z.
In addition, we overexpressed bglJ, leuO and rcsB in the background of G50, so that these positive regulators could (partially) outcompete H-NS binding to the bgl regulatory region (Fig. 3E). The operon activities increased compared to G50 when either leuO or bglJ was overexpressed individually (Fig. 3F). The operon activity doubled when bglJ and rcsB were overexpressed, but it dropped slightly when rcsB alone was overexpressed, as well as when bglJ was overexpressed in DrcsB cells. A similar trend was observed in the activity of operon reporter G50-Z when these positive regulators were overexpressed, indicating the validity of our results (Fig. S4). All of the strains showed Bgl À phenotypes and were incapable of growing on salicin minimal agar plates. However, colonies of leuO, and bglJ overexpression strains appeared to be slightly red on MacConkey + salicin plates although these strains cannot utilize salicin as the sole carbon source. These growth phenotypes are consistent with low levels of the operon activity. These results indicate that over-production of LeuO, BglJ, RcsB or the BglJ/RcsB complex only partially decreased H-NS binding to the upstream site within the regulatory region.

H-NS binding within the bglG gene is necessary for bgl repression
Previous studies had shown that the downstream H-NS binding site is approximately within positions +468 to +607, relative to the bglG translation initiation site [9]. We used several approaches to impede H-NS binding to this region, thus preventing the H-NSmediated DNA looping (Fig. 4A). These approaches include: (1) deletion of the entire bglG50 region; (2) deletion of the 3 0 region, thought to contain the H-NS binding site; and (3) substitutions of the H-NS binding region with one or two TetR operators.
The entire bglG gene was removed from G50 (DbglG50) to ensure that no binding of H-NS to bglG could occur. As shown in Fig. 4B, the operon activity of the resultant strain DbglG50 increased dramatically (200-fold) in comparison to G50, while DbglG50 shows a Bgl + phenotype. These facts indicate that the H-NS binding site in the bglG gene is essential for bgl operon silencing by H-NS.
When a short version of bglG50, deleted for the 3 0 region, from +466 to the end of the gene, was used for the operon lacZ reporter expression (that is, Pbgl-bglG50 0 -lacZ in strain G50 0 -Z), an over 5-fold increase of operon activity was observed compared to G50-Z which carries the complete bglG50 gene in front of the reporter (Fig. 4C). This fact suggests that the deleted 3 0 region is important for H-NS repression of the bgl operon. Meanwhile, the smaller increase (5-fold increase for the truncated bglG50 versus 200-fold increase for the deletion of the entire bglG50 gene) may be due to the less effective binding of H-NS to the remaining region of the truncated bglG gene.
Strains G50.1R and G50.2R were created by replacing part of the putative H-NS binding site within the bglG gene with either one or two TetR operators, respectively (Fig. 4D). Both strains showed similar operon activities as for G50 when the TetR protein was absent, suggesting that such small region changes do not appreciably affect H-NS binding within the bglG gene. However, these strains showed a 2-fold increase in operon activities when the tet repressor TetR was expressed (Fig. 4E). These two strains showed Bgl À phenotypes. Similar results were seen using operon reporters G50.1R-Z and G50.2R-Z (G50-Z with the H-NS binding site within bglG replaced by one or two TetR operators, respectively) (Fig. S5).

H-NS-mediated DNA looping within the bgl operon is DNA Phaseindependent
To investigate the characteristics of the putative H-NS loop, formed in the bgl operon, we first examined the impact of DNA phasing between the two H-NS binding sites on DNA loop formation (Fig. 5A). The DNA double helix is around 10.5 bp per turn. Two small DNA fragments of 5 bp and 10 bp were individually inserted within the bglG gene between the two H-NS binding sites, creating strains G50.P5 and G50.P10, respectively. As compared to strain G50, if DNA phasing between two binding sites were important, in contrast to strain G50.P10, repression would be diminished in strain G50.P5. Fig. 5B shows that these two strains had similar operon activities compared to G50, and they were still Bgl À . These results reveal that H-NS silencing of the bgl operon is DNA phase independent, and the DNA loop is stable.

H-NS-mediated DNA looping in the bgl operon is distanceindependent
To further characterize H-NS-mediated DNA loop formation, we changed the distances between the two H-NS binding sites on the bgl operon (Fig. 6A). The distances were increased by individually inserting a 1017 bp DNA fragment and an 85 bp DNA fragment within bglG in the wild-type strain BW25113. These insertions were essentially without effect on bgl operon expression (Fig. 6B), suggesting that H-NS-mediated repression via the DNA loop is not affected by increased distances between two binding sites. However, these experiments could not rule out the possibility that the DNA fragment insertions inactivated the anti-terminator BglG, and that bgl operon transcription was terminated at each of the two terminators flanking bglG.
To eliminate a possible effect of one or both of the two terminators on DNA loop formation, we used strain G50 deleted for both terminators. As shown in Fig. 5B, the insertions of 5 bp and 10 bp DNA fragments between the upstream and the downstream binding sites within the bglG50 gene did not affect DNA looping. To further increase the distance, a 1017 bp fragment was inserted into bglG50 as well (1017-G50). As shown in Fig. 6B, this large insertion had a negligible effect on operon activity compared to G50. Similar results were observed when a 1305 bp fragment encoding Km resistance was inserted into the same location within the bglG50 gene. As expected, all of these modified strains with various sizes of DNA insertions within the bglG gene still showed Bgl À phenotypes using our standard assays involving MacConkey + salicin plates and M9 + salicin plates. To further confirm the absence of a terminator sequence within those two DNA fragments (1017 bp and 1305 bp) inserted at the bglG50 gene, which may otherwise abolish transcription of downstream bglF and bglB, the hns mutation was individually transferred to G50, 1017-G50 and 1305-G50 (DhnsG50, Dhns-1017-G50 and Dhns-1305-G50). These new strains were examined for their BglB activities using our standard approach. As expected, a dramatically increased operon activity was seen for each strain, and there is little difference (10% or less) among these three new strains (Fig. S6), indicating that there is no terminator within each of these two DNA fragments inserted into bglG50. All of these results demonstrated that increasing distances between the two H-NS binding sites does not appreciably affect strong repression of the bgl operon by H-NS. This therefore indicates that H-NS-mediated DNA looping is robust.

Discussion
In this work, we have provided evidence that the nucleoid structuring protein, H-NS, silences the bglGFB operon using a DNA looping mechanism. As the major repressor, dimerized H-NS proteins have long been thought to bind to an upstream site within the regulatory region and a downstream site within the bglG gene. However, these bindings alone are not sufficient to abolish operon expression, although they do contribute to partial repression. Instead, by individually disrupting H-NS self-oligomerization or by preventing DNA binding to either of the two sites, we proposed that H-NS, when bound to both sites, bridges by loop formation the regulatory region and the bglG gene via oligomerization, forming the DNA loop and thus blocking transcription.
H-NS is characterized by self-dimerization and oligomerization via its N-terminal domain, leading to formation of higher-order oligomers [29][30][53][54]. H-NS proteins exert their repressive effects on gene transcription by first binding to target DNA sites (usually AT-rich and curved DNA regions) through their C-terminal DNAbinding domains, forming nucleoprotein complexes. For stronger repression, as is responsible for gene silencing, H-NS oligomerizes along the DNA strand which either stiffens the DNA strand (forming a stiffened filament) or further forms a bridged nucleoprotein complex (a DNA loop) [33,48,55]. In relation to the bgl operon, H-NS is a strong dominant repressor which simultaneously binds to the bgl regulatory region and the 3 0 region of the bglG gene [9]. H-NS binding to both locations is believed to be synergistic in repressing bgl operon expression, as the repression is much greater when H-NS binds to both locations than when it binds to either one [36].
In this work, we found that H-NS lost the vast majority (85%) of its repressive ability when it is deficient in self-oligomerization, even though it still maintains its DNA binding property. It is generally recognized that self-oligomerization is vital for H-NS to loop DNA [31,33,[56][57]. Furthermore, when binding to either of these two sites is disrupted or partially disrupted, H-NS repression is abolished or partially abolished, although it can oligomerize from one binding site. It is worthwhile to note that H-NSL30P fails to oligomerize due to perturbation of the dimerization domain (site 1) instead of its multimerization domain (site 2) [54]. Our results and previously published reports show that H-NSL30P is still capable of binding to DNA, thus repressing transcription [39]. Most likely, H-NSL30P proteins dimerize through the site 2, implying the formation of ''non-canonical dimers". Protein-protein interaction via site 1 is believed to be stronger than that via site 2. Based on these results, we propose a model by which H-NS silences the bgl operon (Fig. 7). Briefly, dimerized H-NS proteins first bind to both sites simultaneously or to either one site. Then DNA-bound H-NS dimers recruit more dimerized H-NS proteins extending along with the DNA strand(s), forming H-NS-DNA stiffened filaments. Subsequently, two binding sites are brought together to form a DNA loop. This may occur through binding of the Cterminal domain of each dimer within the filament to a nearby region of the second binding site (in the case of simultaneous binding to two locations) or directly to the region containing the second binding site (in the case of binding to one location). The entire bgl promoter and the 5' region of bglG are embedded within the loop. Such an H-NS-mediated DNA loop efficiently abolishes the transcription of the bgl operon by (1) blocking the binding of RNA polymerase to the bgl promoter, and/or (2) trapping RNA polymerase when bound to the bgl promoter to prevent readthrough.
Instead of usually repressing canonical transcription, Singh et al. reported that H-NS primarily functions to repress spurious transcription of non-coding RNAs within the genes previously described as being H-NS repressed [58]. In the case of bgl operon regulation, in addition to the downstream binding site within bglG, there may be some other unidentified intragenic or intergenic promoters (for sigma70 to bind) overlapped with H-NS binding sites in both sense and antisense orientations. Transcription (leading to non-coding RNAs) from these promoters could be (strongly) repressed by H-NS as well. Although the possible presence of such new promoters and their repression by H-NS contribute to our understanding of the complex regulation of the bgl operon, these facts should not replace the main DNA looping mechanism by which H-NS silences the operon.
Similar to the bgl operon, the proU operon, encoding an osmoprotectant uptake system, is subject to strong H-NS repression as well [59][60]. The repression requires H-NS binding to the regulatory region and a downstream region within the first gene proV of the operon [61]. The repression via these two sites is synergistic [36]. The distance between these two binding sites and their relative orientations on the face of the DNA helix had no major effect on proU repression by H-NS [62]. Combining all these reported findings, a looping mechanism similar to that in the bgl system would most likely be applicable to the proU operon. Alterations of DNA phasing and the distances between two binding sites have negligible effects on H-NS-mediated looping, indicating that the formed DNA loop is exceptionally stable. This may be correlated to the large size (about 1 kb nucleotides) of the DNA loop, which confers rotational flexibility. The presence of such a robust DNA loop would definitely contribute to our understanding of the bgl operon as to its silencing state and its non-inducibility in wild-type cells. It is also conceivable that with such a stable DNA loop, involving the entire promoter region, other transcription factors, such as BglJ, RcsB, LeuO and Crp, when expressed at the wild-type levels, either fail to access or are trapped in the promoter region, thus being essentially without effects on bgl operon expression. When these proteins are produced in large amounts or have better binding sites, the binding of H-NS to the upstream site may be partially prevented, thereby elevating rates of transcription.
Indicator agar plates such as those containing MacConkey agar with salicin or bromothymol blue with salicin have been used to determine the phenotypes of growth/fermentation on/of bglucosides, with a dark color (red or blue) developed within individual colonies that are Bgl + (able to utilize b-glucoside for growth) [7,11]. However, the color changes of colonies on these indicator plates do not always represent the actual capability of the cells to grow on b-glucosides as the sole carbon source, as all cells are able to grow on the indicator plates. In fact, some mutant strains that were claimed to be Bgl + in previous studies, were not able to grow on minimal agar plates with salicin as the sole carbon source. In this report, we show that overproduction of BglJ, RcsB, the BglJ/ RcsB complex or LeuO did not give rise to appreciable growth on minimal salicin agar media although visible red color could be seen at the centers of some of their colonies on MacConkey agar plates containing salicin. Therefore, phenotypic results obtained from MacConkey indicator plates with salicin were sometimes misleading. As a result, strains considered to have a Bgl + phenotype in this study must fulfill two requirements: (1) appearance of colonies with visible red color on MacConkey agar plates containing 0.5% salicin after 1 day of incubation at 37°C, and (2) appearance of colonies of visible sizes on M9 agar plates with salicin as the sole source of carbon after 2 days of incubation at 37°C.
Flanking the bglG gene are two terminators that effectively block transcription due to the formation of hairpin-like secondary structures in the absence of the activated antiterminator, BglG [12]. It was of interest to know if these transcriptional terminators contribute to bgl operon silencing. Using a mutant strain (G50) only deleted for these two terminators with all other regions unchanged, we showed that its operon activity increased only slightly compared to the wild-type, and it showed a Bgl À phenotype (Fig. 1B). This indicated that the two terminators flanking (ii) Alternatively, RNAP might still be able to bind to the bgl promoter (considering the relatively large DNA loop), but the polymerase is trapped inside the loop, therefore failing to transcribe mature bglG mRNA. This model is applicable to both the wild-type bgl operon (carrying both terminators) and the terminator-less bgl operon. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) the two sides of bglG do not exert an appreciable effect on the silencing of the bgl operon in wild-type E. coli cells. We thus distinguished the roles of H-NS and both terminators with respect to their regulatory effects on bgl operon expression. Strain G50 (without the two bgl terminators) was used throughout this study, showing that H-NS is still able to form a DNA loop, thus silencing bgl operon expression, when the terminator structures are not present.
Overexpression of positive regulators, BglJ and LeuO, increased bgl operon transcriptional activities under H-NS repression (Fig. 3F). This is consistent with the notion that reducing H-NS binding to the promoter region can (partially) eliminate DNA looping, thereby enhancing operon transcription. However, this is not the case for another positive regulator, RcsB, as its overproduction did not appreciably affect operon expression activity (columns 3 and 4 in Fig. 3F). As BglJ and RcsB form heterodimers when regulating bgl operon expression [11], BglJ might be the limiting factor with RcsB present in normal (wild type) amounts. The expression of bglJ is strongly repressed by H-NS, leading to very low levels of BglJ in wild-type E. coli cells [11,63]. This explains our results showing that bgl operon expression was increased when bglJ was overexpressed but not when rcsB was overexpressed. Interestingly, even though the operon activity increased when both bglJ and rcsB were overexpressed simultaneously, the increased level was not as great as that when bglJ was overexpressed alone (columns 3 and 5 in Fig. 3F). Similar trends were obtained when using either bglucosidase (BglB) or b-galactosidase (LacZ) assays, substantiating the validity of the data. At the same time, a growth defect has been observed for the cells when rcsB was overexpressed (data not shown). As RcsB is a global regulator, overexpression of rcsB might have caused a pleiotropic effect on transcription of other genes and cell growth in addition to affecting bgl operon transcription [64].
A moderate increase of bgl operon expression was seen when stronger Crp-cAMP binding to the bgl regulatory region occurred. However, all strains, either producing higher levels of positive regulators or carrying a better binding site, are still Bgl À due to the low levels of operon expression (Fig. 3D and F). Most likely, these regulators cannot efficiently bind to their operators within the regulatory region due either to an insufficient affinity, or to the presence of a stable DNA loop formed by H-NS. It would be interesting to know how these regulators at elevated levels or with better operators affect bgl operon transcription in the absence of H-NS although there was no effect on bgl operon transcription when the gene for any one of these factors was deleted in the absence of H-NS [39].
Previous studies had reported that H-NS binds to a region approximately 600 to 700 base pairs downstream of the bgl promoter, which is around positions +468 to +607 relative to the bglG translational initiation site [9]. In this study, we replaced part of this region, presumably the H-NS binding site, within the bglG gene by one or two TetR operators, resulting in small increases in bgl operon activity when compared to the control (Fig. 4E). However, the operon activities increased even more when TetR was present. These results suggest that: (1) the binding of TetR to the bglG gene only partially or occasionally decreased H-NS binding to its target site with the bglG gene; (2) the binding of TetR to two operators within the bglG gene further impeded H-NS binding to its target site, reducing the H-NS looping ability and increasing transcription, even though the bound TetR might act as a roadblock, blocking transcription; and (3) the region replaced by the TetR binding sites might not be the actual H-NS binding site within the bglG gene, or H-NS might be able to bind to (an)other location(s) in the gene as deletions of the entire bglG gene or a larger portion of the gene either eliminated or substantially disrupted DNA looping.
Our recent work revealed that the frequency of Bgl + mutations mediated by small IS transposons are positively regulated by the level of the anti-terminator protein, BglG [51]. It will be most interesting to investigate: (1) whether and how H-NS or H-NS-mediated DNA looping is involved in affecting IS insertional mutations; and (2) whether and how H-NS binding affects the SIDD (Superhelical stress-Induced DNA Duplex Destabilization) structure [52] in the regulatory region. Further studies should be also carried out to examine interactions between the two H-NS bound regions within the bgl operon using the 3C technology [65] and to directly visualize (and measure) H-NS-mediated DNA looping within the bgl operon in the various types of live E. coli cells used in this study [66]. Such investigations should contribute to our understanding of the looping structure and its dynamics.

E. coli strains
E. coli K12 strain BW25113 [67] was used as the parental strain that is referred to as ''wild-type" strain in this study. All other strains are derived from this strain. BW-RI is another wild-type strain used in this study, which constitutively produces the TetR protein, the repressor of the tet promoter, Ptet [68].

Construction of an intermediate strain, DPbglG
Using wild type strain BW25113, the chromosomal region containing the bgl promoter (Pbgl) and the bglG gene plus the bglGflanking terminators was replaced by a cat gene amplified from pKD3 [67] by oligos Pbglcat-P1 and bglFcat-P2 (Table S2). The cat gene was flipped out by pCP20, yielding strain DPbglG, in which both Pbgl and bglG with the two terminators were replaced by an 85-bp FRT scar. This strain was then used as a beginning strain to make several other strains.

Removal of two terminators flanking the bglG gene
The first gene in the bgl operon, bglG, is flanked by two transcriptional terminators. To determine if these terminators are important for H-NS mediated bgl operon silencing, they were deleted from the chromosome. To do this, the strain DPbglG was used as the starting strain. Present in plasmid pKES50 [69], a Pbgl-bglG50-lacZ cassette consisted of the lacZ gene (with its own Shine-Dalgarno sequence) fused to a region downstream of the bgl promoter (Pbgl), a modified bglG gene, bglG50, (note that this modified gene has no flanking terminators, and it is not translated because the first two ATG codons were changed to GCG). The Pbgl-bglG50 DNA region together with the downstream lacZ's RBS (between bglG50 and lacZ) was amplified from plasmid pKES50 using oligos Pbgl-Xh-Fbg and bglG-Bm-Rm (Table S2). The PCR products were gel purified, digested with XhoI and BamHI, and then ligated into the same sites of pKDT [70], where ''T" refers to the rrnB terminator (rrnBT). This yielded pKDT_Pbgl-G50, in which the native bgl promoter drives the terminator-less bglG gene (bglG50). The cassette containing the km r gene, rrnBT and Pbgl-G50 (km r :rrnBT:Pbgl-G50) was amplified from pKDT_Pbgl-G50 using oligos Pbgl.G50-P1 and bglF.G50-P2 (Table S2). The PCR products were gel purified and electroporated into DPbglG cells expressing the Lambda Red proteins to replace the 85-bp FRT scar. This yielded strain G50, in which the native bgl promoter drives a terminator-less bglG gene (bglG50) followed by bglF and bglB. The intergenic region (aatagcttcacaggaaacagct) between bglG and bglF is the same as one between bglG and lacZ in pKES50 [69], and it carries the lacZ's RBS for bglF.
The strain G50 was used as the recipient strain for P1 transduction in order to make other strains. First, the transcription unit including a constitutively expressed tetR gene and a spectinomycin resistance (sp r ) marker was transferred to G50 by P1 transduction as mentioned above, yielding strain RIG50. Also, the deletion of hns was transferred to G50 by P1 transduction, yielding DhnsG50.

Removal of the presumptive H-NS binding site within the bgl regulatory region
H-NS binds to two locations in the bgl operon, one within the bgl regulatory region and the other within the bglG gene, although exact binding sites have not been defined [9][10]. To abolish or reduce H-NS binding to the bgl regulatory region, the upstream region (located between positions À585 to À207 relative to the bglG translational initiation site) carrying the possible H-NS binding site in G50, was first replaced by a km r gene amplified from pKD4 using oligos Pbglcat-P1 and Pbgl2-P2 (Table S2). The km r gene was then flipped out by pCP20, yielding strain G50DO HNS -Pbgl.

Screening IS insertional Bgl + mutants
Insertion of IS elements is believed to abolish or reduce H-NS binding to the bgl regulatory region. To obtain IS insertional mutants, wild-type BW25113 cells were grown in LB medium at 30°C for at least 6 hrs. The cells were washed twice with M9 salts with no carbon source, and around 10 8 cells were inoculated onto M9 or M63 minimal agar plates containing 0.5% salicin as indicated below. The plates were incubated at 30°C and were examined daily for the appearance of Bgl + colonies [51]. The colonies were subject to PCR analyses using Pbgl-F2 and Pbgl-R2 (Table S2) to verify the presence of any insertional mutations at the bgl regulatory region. The type, location, and orientation of IS elements inserted within the bgl regulatory region were identified by DNA sequencing.
The IS insertional mutant, IS5r with IS5 inserted in reverse orientation at À207.5 relative to the bglG translational initiation site, was used as the positive control. The deletion of hns was transferred to IS5r by P1 transduction, yielding strain IS5rDhns.

Construction of strain G51 with a stronger Crp-cAMP binding site at Pbgl
To make strain G51, plasmid pKES51 [69] was used as the template to construct plasmid pKDT_Pbgl-G51, which is essentially the same as pKDT_Pbgl-G50 except that the presence of a stronger Crp-cAMP operator (O* Crp ) in Pbgl resulted from a C to T substitution at position À134 relative to the bglG translational initiation site [6]. Using similar methods as described above for G50, the cassette of Pbgl_O* Crp , driving bglG expression, was integrated upstream of bglF in DPbglG cells, yielding strain G51. This strain is the same as G50 except that it carries a stronger Crp-cAMP binding site in the bgl promoter region.
To make a bglJ/rcsB double overexpression strain, the km r gene was first flipped out from G50Ptet-bglJ by pCP20, yielding G50Ptet-bglJ-km s , which is sensitive to kanamycin. The rcsB gene, driven by the strong tet promoter with a km r marker, was then transferred to G50Ptet-bglJ-km s by P1 transduction, yielding strain G50Ptet-bglJ/ rcsB. The rcsB mutation was transferred to G50Ptet-bglJ-km s as well, yielding strain G50Ptet-bglJDrcsB, in which the strong tet promoter drives bglJ in the absence of rcsB.

Removal of the entire bglG gene
To completely abolish H-NS binding in the bglG gene, the entire bglG gene and its two flanked terminators were deleted. First, the region from À78 to +946 relative to the bglG translational initiation site, containing the first terminator, the bglG gene and the second terminator, was replaced by a km r gene amplified from pKD4 [67] using oligos G50-P1 and G50-P2 (Table S2). The km r gene was then flipped out using pCP20, yielding strain DbglG50, in which the bglG gene with the two flanking terminators was replaced by an 85 bp fragment.

Replacing the proposed H-NS binding site within the bglG gene with one or two TetR Operator(s)
In vitro protein/DNA binding assays showed that H-NS proteins are capable of binding to a region within the bglG gene, which is located at +479 to +606 relative to the bglG translational initiation site [9]. To reduce H-NS binding to this region (the presumptive H-NS binding site), part of this region in strain G50 was replaced by one or two operators of TetR using the TetA-SacB positive and negative selection system [71]. Briefly, to add one TetR operator, the 19 nucleotides with sequence ''aaatgctgcaattaataaa", between +544 and +564 relative to the bglG translational initiation site, was replaced by one TetR operator with sequence ''tccctat cagtgatagaga" [72], yielding G50.1R. Similarly, a region of 40 nucleotides with sequence ''gtgtcacgcagttaatgcgcgaaatgctgcaatt aataaa" (between +523 and +564) was replaced by two identical TetR operators with sequences ''tccctatcagtgatagagaCGtccctatcagtg atagaga" (two TetR operators separated by two nucleotides CG, in the center), which yielded G50.2R. The tetR-sp r cassette (constitutively expressing tetR) was transferred from BW-RI to G50.1R and G50.2R, yielding strains RIG50.1R and RIG50.2R, respectively.

Insertions of 5 bp and 10 bp fragments between the two H-NS binding sites
DNA is a double-stranded helix with 10-10.5 bps per turn. To determine if DNA looping within the bgl operon, formed via H-NS protein oligomerization, is DNA phase-dependent, a 5-bp fragment with nucleotides ''ATAGA" and a 10-bp fragment with nucleotides ''ATAGAATAGA" were inserted individually into the bglG gene between +108 and +109 relative to the bglG translational initiation site in strain G50. Briefly, these two small fragments (5 bp and 10 bp) were first integrated into the designated locus (between +108 and +109) within the bglG50 gene in plasmid pKDT_Pbgl-G50 using fusion PCR, yielding plasmids pKDT_Pbgl-G50.5 bp and pKDT_Pbgl-G50.10 bp, respectively. Using these plasmids as templates, the ''km r :rrnBT:Pbgl-G50.P5" cassette and the ''km r :rrnBT:Pbgl-G50.P10" cassette were amplified using oligos Pbgl.G50-P1 and bglF.G50-P2 (Table S2). The PCR products were electroporated into DPbglG cells to substitute for the 85-bp scar that replaced the bgl regulatory region and the bglG gene together with the two flanking terminators. Several Km resistant colonies were confirmed by colony PCR and DNA sequencing to carry a 5bp insertion or a 10-bp insertion at the expected location within the bglG gene, yielding strains G50.P5 and G50.P10. Compared to the original G50 strain, the phasing of those two H-NS binding sites is almost unchanged in strain G50.P10. However, the phasing should be essentially opposite in strain G50.P5.

Changing the distance between the two H-NS binding sites
To determine if the stability of H-NS-mediated looping in the bgl operon is size dependent, the distance between the two H-NS binding sites was increased. A 1017-bp fragment carrying a cat gene from pKD3 was inserted into the beginning of the bglG gene of BW25113, between +209 and +210, relative to the bglG translational initiation site. In the resulting strain, the 1017-bp DNA fragment carrying the cat gene with its own promoter and two FRT sites were inserted into the bglG gene between the two H-NS binding sites, oriented in an opposite direction as the bglG gene, yielding strain 1017-G. The cat gene was then flipped out using pCP20, yielding strain 85-G, in which an 85-bp fragment is left between the two H-NS binding sites. Similarly, a cat gene was inserted into the same location of the bglG50 gene in the G50 background, yielding 1017-G50, with a 1017 bp distance increase between the two H-NS binding sites. Further, a km r gene (1305 bp), encoding kanamycin resistance, was amplified from pKD13 and subsequently inserted into the same location as for 1017-G50, yielding 1305-G50 that carries a 1305 bp fragment between the two H-NS binding sites.

Construction of the bgl operon LacZ reporter
Strain BW25113_Z [51] carries a bgl operon LacZ reporter, which is composed of the bglGFB promoter, the first terminator, bglG, the bglG/bglF intergenic region (with the second terminator) and the first 9 codons of bglF, plus a stop codon at the 3 0 end, immediately upstream of the lacZ's RBS at the lac locus. A km r gene followed by a rrnB terminator is located upstream of this reporter. Using the same protocol for strain BW25113_Z construction [51], the cassettes of ''km r :rrnBT:Pbgl-bglG50" on plasmid pKDT_Pbgl-G50 and ''km r :rrnBT:Pbgl-bglG51" on plasmid pKDT_Pbgl-G51 were individually integrated into the chromosome of the default strain MGIBKY [70] to replace the lacI gene and the lacI/lacZ intergenic region, yielding strains G50-Z and G51-Z. These two reporters were transferred into a Dcrp Glp + mutant (able to grow on glycerol) [73], yielding DcrpG50-Z and DcrpG51-Z, respectively.
As described above, H-NS proteins are capable of binding to a region within the bglG gene, which is located between +479 and +606 relative to the translational initiation site of bglG. To examine the effect of this region on b-galactosidase (LacZ) activity, part of this region of the bglG50 gene in plasmid pKDT_Pbgl-G50 was replaced by one or two operators of TetR using fusion PCR, yielding pKDT_Pbgl-G50.1R and pKDT_Pbgl-G50.2R, respectively. The cassette of ''km r :rrnBT:Pbgl-bglG50.1R" on pKDT_Pbgl-G50.1R and the cassette of ''km r :rrnBT:Pbgl-bglG50.2R" on pKDT_Pbgl-G50.2R were individually transferred to the same location on the chromosome as for G50-Z, yielding strains G50.1R-Z and G50.2R-Z, respectively. The two operon reporters are the same as G50-Z, except that part of the H-NS binding site within bglG was replaced by one or two TetR operators. Using G50.1R-Z and G50.2R-Z as recipient strains, two new operon reporters were made by transferring the tetR expression unit with a sp r marker using P1 transduction, yielding RIG50.1R-Z and RIG50.2R-Z, respectively.
To further prevent H-NS binding to bglG50, a new operon reporter similar to G50-Z, carrying the bgl promoter and only the 5 0 region of bglG50 with no 3 0 region, was made. The cassette of ''km r :rrnBT:Pbgl-bglG50'" (bglG50' refers to the region from +1 to +465 relative to the bglG translational initiation site), amplified from pKDT_Pbgl-G50 by oligos P1n-P1n and G50tn-Z-P2 (Table S2), was fused to the lacZ gene as for other operon reporters, yielding operon reporter G50'-Z. There is a stop codon (TAA) added to the end of bglG50', and the RBS for lacZ remains the same as for G50-Z.
To determine if H-NS or H-NSL30P has an inhibitory effect on expression of bglG when bound to it, another new lacZ transcriptional reporter was made in which the bglG-lacZ fusion is driven by a constitutive promoter. The cassette, ''km r :rrnBT:Iq-bglG" (with lacIq driving bglG expression), was amplified from strain Iq-G [51] and subsequently fused upstream of lacZ as for other operon lacZ reporters at the lac locus. This reporter, carrying a lacIq constitutive promoter driving the terminator-less bglG gene that is transcriptionally fused to a lacZ gene, was transferred into strains Dhns and hnsL30p, yielding Dhns_Iq-G-Z and hnsL30P_Iq-G-Z, respectively. Similarly, the cassette ''km r :rrnBT:Ptet" from strain Ptet-G [51] was fused upstream of lacZ and subsequently transferred to Dhns and hnsL30P, yielding Dhns_Ptet-G-Z and hnsL30P_Ptet-G-Z, respectively.
4.13. Determining the Bgl + /Bgl À phenotypes of strains E. coli cells were grown on MacConkey indicator plates with 0.5% salicin at 37°C overnight. Visible red color was shown for colonies with a Bgl + phenotype, indicating their ability to ferment salicin. Otherwise, colonies with a Bgl À phenotype remained white in color. In addition, cells were grown on M9 minimal agar plates with 0.5% salicin as the sole carbon source at 37°C for 2 days to further confirm their phenotype. To have a Bgl + phenotype, cells should form visible colonies during the 2-day incubation. Otherwise, the cells are considered to be Bgl À . In this study, we define a Bgl + phenotype as the appearance of red colonies on MacConkey + salicin plates during a 24-hr incubation plus the formation of visible colonies on M9 + salicin plates during a 48-hr incubation.

b-Glucosidase (BglB) assay
The protocols for b-glucosidase assay as described in previous studies were followed with minor changes [12,74]. Briefly, the cells were grown in LB medium at 37°C with shaking. After at least 6 hrs, 20 lL of the culture were transferred to 1x M63 minimal medium with 0.66% casamino acids; then it was left overnight for growth at 37°C with shaking. An appropriate amount of the overnight culture was inoculated into 6 mL of M63 minimal medium with 0.66% casamino acids and 0.5% salicin, at a starting OD 600 of 0.025. The cells were grown at 37°C with shaking. Five samples with 0.8 mL each were collected during the late exponential growth phase when the bgl operon was fully induced with an OD 600 of above 1.5. The samples were centrifuged at a speed of 5,500 rpm for 2.5 mins. The supernatant was removed, and the cells were suspended with 1 mL of Z-buffer with 50 lg/mL chloramphenicol.
The samples were warmed in a 37°C water bath. To start the reaction, 200 lL of p-nitrophenyl-b-D-glucoside (PNPG, 8 mg/mL) were added to the cell suspension in Z-buffer. After a visible yellow color appeared, 0.5 mL of 1 M sodium carbonate was added and vortexed to stop the reaction. The reaction mixture was diluted, if appropriate, and centrifuged at a speed of 15,000 rpm for 2.5 mins. The absorbance of the reaction mixture was measured at 420 nm and 550 nm. The BglB activity of each sample was calculated using the equation: b-glucosidase (BglB) activity = [10 00 Â (OD 420nm -1.75 Â OD 550nm ) Â Dilution factor]/[OD 600nm Â Time of reaction (min) Â Volume of sample (mL)]. The activity of the strain was determined by averaging the BglB activities of the samples measured.

b-Galactosidase (LacZ) assay
Reporter strains were grown in LB medium at 37°C with shaking. After at least 6 hrs, 20 lL of the culture were transferred to 3 mL of M63 minimal media with 0.5% glycerol. Then it was left overnight for growth at 37°C with shaking. The overnight culture was inoculated into 5 mL of M63 minimal medium with 0.5% glycerol, with a starting OD 600 of 0.03. The cells were grown at 37°C with shaking. Four samples were collected during the exponential growth phase, when the OD 600 reaches around 0.2, 0.4, 0.6 and 0.8. Collected samples could be kept frozen at À20°C prior to the measurements.
200 lL of samples and 800 lL of Z-buffer were mixed with 25 lL of chloroform by vortexing for 10 s twice. The samples were warmed in a 37°C water bath. To start the reaction, 200 lL of onitrophenyl-b-D-galactopyranoside (b-ONPG) was added to each sample. After a visible yellow color appeared in some tubes, 0.5 mL of 1 M sodium carbonate was added and vortexed to stop the reaction. The reaction mixtures were diluted, if appropriate, and centrifuged at a speed of 15,000 rpm for 2.5 mins. The absorbance of the reaction mixture was measured at 420 nm and 550 nm. The LacZ activity of each sample was calculated using the equation [75]: b-galactosidase (LacZ) activity (Miller units