Membrane protein expression triggers chromosomal locus repositioning in bacteria

Abstract

It has long been hypothesized that subcellular positioning of chromosomal loci in bacteria may be influenced by gene function and expression state. Here we provide direct evidence that membrane protein expression affects the position of chromosomal loci in Escherichia coli. For two different membrane proteins, we observed a dramatic shift of their genetic loci toward the membrane upon induction. In related systems in which a cytoplasmic protein was produced, or translation was eliminated by mutating the start codon, a shift was not observed. Antibiotics that block transcription and translation similarly prevented locus repositioning toward the membrane. We also found that repositioning is relatively rapid and can be detected at positions that are a considerable distance on the chromosome from the gene encoding the membrane protein (>90 kb). Given that membrane protein-encoding genes are distributed throughout the chromosome, their expression may be an important mechanism for maintaining the bacterial chromosome in an expanded and dynamic state.

Studies of several different bacteria have revealed that their chromosomes are organized structures, with genetic loci occupying relatively defined positions along the long axis of the cell (1–7). However, the potential impact of gene function and expression state on chromosome organization and on subcellular positioning of specific loci remains relatively unexplored. Such modulation of the spatial organization of the chromosome may affect numerous cellular processes, including the regulation of gene expression, gross transcriptional control, assembly of macromolecular complexes and domains, and the generation of cellular asymmetry (8–14).

One long-standing hypothesis posits that genes actively expressing membrane proteins are localized to regions proximal to the membrane (8, 11, 15). To date, however, there is no direct evidence for membrane protein expression affecting the subcellular positions of chromosomal loci. We therefore sought to test directly whether expressing a membrane protein affects the localization of its encoding gene in Escherichia coli. We tested two loci and found that induction of membrane protein expression rapidly results in a dramatic repositioning toward the membrane. We also show that the positions of chromosomal loci as far away as 90 kb from the induced gene are affected. This shift in position is a significant perturbation on the scale of the cell and may therefore be a major determinant of chromosome conformation.

Results

We first tested the effect of inducing the lac operon on the intracellular position of the chromosomal lac locus. The operon lacZYA encodes the membrane protein lactose permease (LacY), in addition to two cytoplasmic proteins, beta-galactosidase (LacZ) and galactoside acetyltransferase (LacA). To follow the intracellular position of lacZYA, we inserted an array of Tet repressor binding sites (tetO) in the gene cynX, which is adjacent to lacA and ∼2 kb from lacY. TetR–YFP binding to this array produces foci that are easily identified by fluorescence microscopy (16, 17). Cell membranes were labeled with the fluorescent dye FM4-64, and the distances from chromosomal loci to the membrane (peak FM4-64 fluorescence) were determined (Fig. 1A). Because each 2D fluorescence image is a projection of a 3D cell, we refer to the measured distances as projected distances. For the approximately cylindrical geometry of the E. coli cell, chromosomal loci that fall on average closer to the membrane will have distributions of projected distances, measured from fluorescence images of a cell population, that are relatively enhanced at the membrane. Therefore, a decrease in the distance between a chromosomal locus and the membrane across a population of cells will appear as a shift of the distribution of projected distances toward the membrane.

Fig. 1.

LacY expression triggers chromosomal repositioning at the native locus. (A) Measurement of distances between chromosomal loci and the membrane. (Left) Sample false color composite image of an E. coli cell with the lac locus labeled with a tetO array (inserted in cynX) growing in the presence of inducer (2 mM IPTG). The cell was labeled with FM4-64 (red) and TetR–YFP (green). Fluorescence images were increased in size (resampled) by a factor of four using a cubic spline interpolation. (Right) The intensity profile along the dashed line for the two fluorescence channels. Distances were measured between the peak YFP and nearest peak FM4-64 fluorescence in the resampled images as described in SI Methods. All such distances reflect projections onto the plane of focus and are therefore referred to as projected distances. (Scale bar: 1 μm.) One pixel in the original image (before resampling) is ∼80 nm. (B) Distribution of projected distances of the lac locus to the membrane across a population of cells growing in the absence or presence of inducer (2 mM IPTG) for wild-type lac. (Upper) Distribution showing distances measured in nanometers. (Lower) Distribution representing the same data as in Upper but with each distance normalized by the cell’s radius (cell width/2; SI Methods). Each distribution is comprised of measurements of at least 300 loci. (C) Sample images of induced cells from B. Fluorescence profiles from line scans and the corresponding projected distance to the membrane are also indicated. For each image, the upper and lower foci correspond to the left and right graphs, respectively. Images were resampled as in A. Note that for the growth conditions used here, cells had on average two copies of the chromosomal region containing the lac locus. (D) Distribution of projected distances of the lac locus to the membrane across a population of cells growing in the absence or presence of inducer (2 mM IPTG) for a mutated lac operon with the coding sequence of lacY replaced with the corresponding sequence of aadA, which encodes a cytoplasmic protein. For comparison, D also includes the data from B for lacY⁺ cells without inducer. Each distribution reports the mean values from two independent experiments; bars report the ranges. Distributions from each experiment comprise measurements of at least 187 loci.

We found that induction of LacY expression produced such a shift. In a population of cells growing in the absence of inducer, the position of the native lac locus showed a distribution that was biased toward midcell, away from the cell membrane (Fig. 1 B and C). Induction of the lac operon shifted the distribution to smaller projected distances, indicating a shift of the lac locus toward the membrane. In contrast, for a strain in which the lacY coding sequence was replaced with a sequence encoding a cytoplasmic protein (aadA, encoding spectinomycin adenylyltransferase), there was no significant difference in the distribution of lac-membrane projected distances for cells growing in the presence or absence of inducer (Fig. S1).

To compensate for cell-to-cell variation in cell widths in each sample population, we normalized the measured distances by cell radius (one half of the peak to peak FM4-64 distance; SI Methods). Distributions with this normalized distance similarly demonstrate a LacY expression-dependent shift of the locus toward the membrane (Fig. 1B), whereas the three cases in which a membrane protein is not produced—or produced only at very low levels—showed remarkably similar spatial distributions (Fig. 1D).

IPTG induction of the Δ(lacY)::aadA⁺ strain resulted in levels of beta-galactosidase that were approximately half of those produced by wild-type lac (Fig. S2A). To rule out the possibility that decreased transcription could account for the lack of chromosomal repositioning upon induction, we replaced the lac promoter in the Δ(lacY)::aadA⁺ strain with the stronger trc promoter (18). For this strain, induced beta-galactosidase levels were ∼50% higher than those measured for the wild-type lac operon (Fig. S2A). However, we again did not observe any difference in the distribution of projected differences for this Δ(lacY)::aadA⁺ strain when comparing inducing and noninducing conditions (Fig. S2B).

To test a second membrane protein, we used the tetracycline efflux pump, TetA, derived from the transposon Tn10 (19). A DNA segment encoding the tetracycline-inducible repressor TetR and a functional fusion of TetA to the fluorescent protein mCherry (tetR tetA–mcherry) was inserted in the chromosome at the phage lambda attachment site (attB_λ). An array of LacI binding sites (lacO) was integrated just downstream of tetA–mcherry and labeled with LacI–YFP (17, 20). Full induction of TetA–mCherry expression with anhydrotetracycline (aTc) produced a significant shift in the distribution of tet-membrane projected distances, indicating repositioning of the tet locus toward the membrane (Fig. 2A and Fig. S3A). In contrast, there was no such shift in a strain that had tetA–mcherry replaced with mcherry, which encodes a cytoplasmic protein (Fig. 2B and Fig. S3B). Fluorescence measurements of the tetA–mcherry⁺ and mcherry⁺ strains indicated that mCherry expression was lower than that of TetA–mCherry (Fig. S4). We therefore constructed a strain with a modified 5′ untranslated region (5′UTR) for mcherry to increase expression. Comparison of this strain with the tetA–mcherry⁺ strain expressing comparable levels of mCherry fluorescence indicated that the absence of a shift toward the membrane for the cytoplasmic protein mCherry is not due to decreased expression. (Fig. S4).

Fig. 2.

Chromosomal repositioning from TetA expression. (A and B) Distributions of projected distances of the tet locus to the membrane across a population of cells growing in the presence or absence of inducer (100 ng/mL aTc) for tetA–mcherry (A) and mcherry (B). The plot in B also includes data from A, tetA–mcherry without inducer, for comparison. (C) Effect of tetA–mcherry induction on the distribution of projected distances of the lac locus (∼440 kb away from tet) to the membrane. For all measurements, fluorescence from TetA–mCherry and mCherry was negligible compared with FM4-64 fluorescence. Each distribution reports the mean values from two independent experiments; bars report the ranges. Distributions from each experiment comprise measurements of at least 160 loci.

We also found that induction of TetA–mCherry did not have a significant effect on the localization of the lac locus, which is ∼440 kb from attB_λ (Fig. 2C). This result indicates that TetA–mCherry induction does not cause repositioning throughout the entire chromosome.

Because TetA expression can be continuously tuned by varying the amount of tetracycline in the growth medium (21), we were able to use this system to characterize the extent of tet localization near the membrane for different levels of TetA–mCherry expression. To quantify membrane localization, we measured the fraction of tet loci that were within a distance of 0.3R from the FM4-64–labeled membrane in fluorescence images, where R is the cell radius. We observed that membrane localization increased in a graded fashion with increasing TetA–mCherry expression, reaching a maximum of ∼0.3 at full protein induction (100 ng/mL aTc) (Fig. 3A). We note that a locus localized to the membrane of a cylindrical cell would appear within 0.3R of the membrane in ∼50% of the 2D fluorescence images (Fig. S5). The maximal value for this measure of membrane localization (for a cylindrical shape) is therefore 0.5.

Fig. 3.

Dose dependence and kinetics of chromosomal repositioning. The fraction of loci proximal to the membrane is defined to be the fraction of loci that are within 0.3R of the membrane (R is the cell radius). (A) Steady-state membrane localization as a function of TetA–mCherry expression, at various levels of induction (from left to right: 0, 0.05, 0.1, 0.45, 0.5, 1 μg/mL tetracycline and 100 ng/mL aTc). mCherry fluorescence is normalized by the 100 ng/mL aTc value. (B) Kinetics of locus repositioning following addition of aTc to 100 ng/mL (Inset) Detail of early time points. Points and vertical bars denote the means and ranges of two measurements from at least 150 loci each. The horizontal bars in A also denote the ranges of two measurements. The horizontal bars in B denote the time interval of each measurement.

We also characterized the time scale over which the tet locus repositions toward the membrane following maximal induction. The process is remarkably rapid; a significant change was detectable at the first measurement interval, spanning 1–3 min following the addition of aTc (Fig. 3B). This behavior is consistent with the rapid onset of TetA–mCherry protein expression, which by 2 min postinduction showed detectable protein expression in ∼30% of the population and had reached a maximal rate of expression by 4 min (Fig. S6).

To determine whether translation is required for repositioning, we mutated the tetA–mcherry start codon to TAA. This mutation reduced TetA–mCherry fluorescence to background levels under inducing conditions (Fig. S7) and had a similarly strong effect on membrane localization, with no significant difference for cells grown in the presence or absence of inducer (Fig. 4). We note that elimination of translation may affect mRNA stability. However, the results nevertheless indicate that translation is required for the observed membrane localization. We also tested the effects of blocking transcription or translation with antibiotics. Treatment with the transcription inhibitor rifampicin abrogated repositioning of the tetA locus in response to induction with aTc (Fig. S8A). Treatment with the translation inhibitor kasugamycin similarly prevented localization to the membrane (Fig. S8B). Although these inhibitors affect many aspects of cell physiology, including nucleoid structure, the results are consistent with the conclusion that the chromosomal repositioning described here requires both transcription and translation.

Fig. 4.

Effect of mutating the tetA–mcherry start codon on chromosomal repositioning. Fraction of loci proximal to the membrane (within 0.3R of the membrane) for the tetA–mcherry⁺ strain (WT) and a strain in which the start codon of tetA–mcherry was mutated (ATG→TAA). The data represent the means and ranges of two measurements, consisting of at least 138 loci each.

As discussed above, we found that induction of TetA–mCherry expression did not perturb the subcellular position of the lac locus (Fig. 2C), a site far from tet in the chromosome. To determine how sites closer to tet are affected, we constructed a sequence of strains with lacO arrays inserted at various distances from tetA–mcherry (6, 44, 90, and 170 kb from tet in the counterclockwise direction on the standard E. coli K-12 genetic map). Chromosomal repositioning to the membrane in response to inducing tetA–mcherry expression showed a monotonic decrease with distance and was detectable from as far away as 90 kb from the tet locus (Fig. 5).

Fig. 5.

Repositioning of chromosomal loci at various distances from tetA–mcherry following induction with aTc. Fraction of loci proximal to the membrane (within 0.3R of the membrane) are shown for strains containing insertions of lacO arrays at the indicated distances from tet. The loci (ybhJ, mngA, kdpE, dsbG) are all on the same side of tet, corresponding to decreasing map coordinates. Distances were taken to be the distance from the end of tetA–mcherry to the right end of the gene that was the insertion site for the lacO array. This distance includes 2,600 bp of sequence upstream of tetA integrated at attB_λ. The points and bars represent the means and ranges of two measurements consisting of at least 129 loci each.

Discussion

It has been hypothesized that for membrane proteins, transcription, translation, and insertion into the membrane are concurrent—a process termed transertion—and therefore lead to membrane localization of the encoding genes (8, 11, 15). Available evidence indicates that translation initiates during transcription (22) and that most membrane proteins are inserted in the membrane cotranslationally (23). However, these two observations do not imply that transertion must occur; it is possible that, for some or all membrane protein-encoding transcripts, cotranslational insertion in the membrane begins after transcription has terminated and the mRNA has disengaged from the chromosome. Indeed, recent studies suggest that the bacterial signal recognition particle may slow or arrest the translation of membrane proteins (24, 25), which could provide time not only for the ribosome-nascent chain complex to engage the secretion machinery but also for transcription to terminate before membrane insertion commences. To our knowledge, transertion has never been demonstrated, although observations of plasmid supercoiling (26), nucleoid collapse from translation inhibitors (10), and DNA localization in large cells (27) can be explained by such a mechanism. The direct observation of locus repositioning to the membrane reported here is consistent with a transertion mechanism as well.

Recently, it was shown that mRNA encoding the membrane protein BglF localizes to the membrane in E. coli independently of translation, indicating that the membrane targeting information is encoded directly in the bglF mRNA (28). The observation that repositioning of the tet locus requires translation suggests that a mechanism based on mRNA alone cannot account for the chromosomal repositioning reported here, although it could play a role. We also note that our results do not necessarily imply a direct physical association between the chromosomal loci and the membrane. For example, it is possible that loci reposition to the surface of the bacterial nucleoid, which would have the effect of moving these loci closer to the membrane. However, based on our results, the mechanism for this repositioning would have to be specific to loci encoding membrane proteins.

Although it has been argued that most mRNA species in the bacterial cell are likely to be mature transcripts that are detached from RNA polymerase (29), a single nascent transcript that is physically associated with its encoding gene could be sufficient to reposition a local region of the chromosome. Our observation that the extent of localization near the membrane increases continuously with increasing protein expression level does not contradict this view, especially if one takes into account that transcription occurs in bursts (30, 31). Transcriptional bursting could explain the behavior in Fig. 3A if the frequency of bursts increases with protein expression level, and if the chromosomal locus is most likely to occupy positions proximal to the membrane only during transcription. This model suggests the membrane association is transient, with the frequency and duration of association dependent on expression level. The rapid increase in localization near the membrane following addition of saturating concentrations of inducer (Fig. 3B) is consistent with this model as well. We also note that repositioning toward the membrane does not require an active process and could instead be mediated by diffusion. For example, the requirement that a transcribing DNA locus move a distance of order the radius of the cell (∼400 nm) in 2 min is within the range of apparent diffusion constants measured for mRNA and DNA in E. coli (29, 32–34).

We found that inducing expression of tetA–mcherry, which is integrated in the lambda phage attachment site, perturbed loci as far away as 90 kb (kdpE). This result suggests that in the intervening region, there are no chromosomal positions that are tightly tethered to the membrane for the growth conditions used in this study, because tight tethering would insulate distal sites such as kdpE from the effects of tetA–mcherry movement toward the membrane. There are ∼20 predicted membrane protein-encoding genes between tetA–mcherry and kdpE. The above results suggest that the expression levels of these genes are sufficiently low to prevent a strong membrane association under the growth conditions used for this work.

A simple polymer scaling estimate for the DNA in the cell suggests the long-distance effect in Fig. 5 reflects a condensed structure of the chromosome (SI Methods, Polymer Scaling Estimate). The data in Fig. 5 cannot be explained by the random fluctuations of a DNA polymer constrained only by excluded volume effects associated with a short-range repulsion between polymer segments; such a description gives an unrealistically small polymer segment length for DNA (<0.5 nm). However, the data are consistent with a chromosome composed of highly condensed DNA. This description gives a segment length of order 40 nm, which is more in line with studies of DNA bound to nucleoid-associated proteins (35–38). A recent study of the positioning of chromosomal loci in E. coli similarly concluded that the chromosome contains a higher-order structure that can be modeled as a condensed DNA fiber (5). It has also been argued that a strong lengthwise condensation of DNA in the bacterial chromosome is important for segregating duplicated chromosomes (39).

E. coli K-12 has ∼1,000 genes spread throughout the chromosome that are predicted to encode inner membrane proteins, and it has been argued that a significant fraction are expressed under standard laboratory culture conditions (40). Based on the above results, and as previously hypothesized (6, 8, 11, 41), membrane protein expression across the entire genome is likely to play a key role in shaping chromosome conformation. Our results further suggest that repositioning at any given locus is likely to be transient, occurring concomitantly with bursts of transcription. The resulting movement toward and away from the membrane at points distributed around the chromosome may be an important mechanism for maintaining the nucleoid in a sufficiently dynamic state to ensure accessibility to regulatory proteins, ribosomes, and RNA polymerase. Many other effects of chromosome–membrane associations have been proposed (8–12, 15, 26, 42, 43), but they will require further experiments to determine whether membrane protein expression plays a direct role.

Methods

See SI Methods for a description of growth conditions, analysis methods, and strain constructions. Table S1 lists strains and plasmids used in this study. Table S2 lists strains used for the data in specific figures.