The structure of the bacterial chromosome and the molecular mechanisms underlying its organization are poorly understood, in part due to the lack of appropriate tools for visualizing the chromosome in vivo. It has been shown by fluorescence microscopy that DNA only occupies the central part of the bacterial cell, referred to as the nucleoid (1
), but the diffraction-limited optical resolution prevents a detailed characterization. Ultrastructural characterization of the nucleoid by electron microscopy has provided varying results depending on the procedures used to fix, dehydrate, and embed the cells (1-2
). Recently, labeling of specific gene loci using fluorescence in situ hybridization and fluorescent repressor-operator systems has allowed imaging of individual gene positions, and their relationship to DNA replication and segregation, in fixed and live bacterial cells (3-5
). However, these studies probe only a set of specific loci at a time, and the global chromosome organization remains unclear.
In bacteria, major nucleoid-associated proteins (NAPs) are the most abundant factors that associate with the chromosome (6-7
). In E. coli
, major NAPs include H-NS, HU, Fis, IHF, and StpA (6
). Each of these NAPs binds up to hundreds of specific sites per chromosome (6, 8-9
). Moreover, due to their substantial nonspecific DNA-binding affinity, the majority of cellular NAPs are bound to the chromosomal DNA with a coverage of about one NAP per 100 base pairs of DNA (10
). NAPs have two major functions: gene regulation and chromosome organization (7
). In particular, H-NS preferentially binds to AT-rich sequences (8-9, 11-13
), functions as a global transcriptional silencer of genes with high AT content (14-15
), and is thought to reside at the center of the nucleoid (16
). The oligomerization of H-NS can promote higher-order DNA structures in vitro (17
), potentially through DNA looping, bridging, and/or stiffening (18-20
). It has been hypothesized based on these biophysical properties of NAPs and their numerous binding sites on DNA that NAPs potentially act as chromosome organizing centers (21
). However, it remains unknown whether the implicated higher-order DNA structures induced by NAPs exist in vivo, and how the chromosome is globally organized by the NAPs.
In a live bacterial cell, a single protein, upon binding to the less mobile structures, such as the cell membrane or chromosome, can be detected and localized against a strong cellular autofluorescence background (22-24
). However, the diffraction-limited optical resolution limits this imaging approach to proteins with low copy numbers in the cell (24
). To obtain the subcellular distribution and organization of the abundant bacterial NAPs, sub-diffraction-limit image resolution is required.
Here, we used localization-based super-resolution imaging (stochastic optical reconstruction microscopy (STORM) or photoactivated localization microscopy) (25-29
) to survey the subcellular distributions of major NAPs, H-NS, HU, Fis, IHF, and StpA. We tagged the target of interest with a monomeric photoactivatable fluorescent protein, mEos2 (30
), unless otherwise specified. We then created E. coli
strains in which the fusion proteins were expressed from their native promoters at the endogenous loci, allowing the targets to be fully labeled and expressed approximately at the wildtype level (Table S1
). All of these mEos2 fusion strains exhibited the same growth rates (cell doubling times) as the wildtype (31
). Cells were imaged in a M9 minimal medium supplemented with glucose at room temperature shortly after taken out of the 37°C culture at the early log phase (31
To acquire a super-resolution image, the mEos2 molecules were activated by a weak 405 nm light, such that only an optically resolvable subset of molecules were activated at any given instant; the activated molecules were imaged using a 561 nm light and their centroid positions were determined in three dimensions (3D) using astigmatism imaging, as previously shown in 3D STORM (32
). The molecular localizations accumulated over time allowed a sub-diffraction-limit image to be constructed. A continuous activation and imaging mode (33
) was used, allowing ~1,000 molecules per cell to be imaged every minute. We note that only a subset of the mEos2 label could mature and become fluorescent due to the long maturation time of mEos2 compared to the E. coli
doubling time (30
). Among those that matured, only a subset could be activated by the 405 nm light. The laser illumination used for imaging did not exert appreciable effect on cell viability, as evident from the nearly identical (within 10%) cell doubling times observed with or without illuminations.
Notably, H-NS formed a few compact clusters within each cell (, Movie S1
). The majority of H-NS molecules resided in these clusters, whose fluorescence accounted for 60 ± 25% of the total activated mEos2 signal (31
). To test the functional integrity of the mEos2-tagged H-NS, we measured the expression levels of hdeA
, two genes repressed by H-NS (15
). Indeed, the strain expressing the fluorescent fusion protein retained wildtype activity in repressing these two genes (Fig. S1
). As a control, the expression levels of lacZ
, a gene not regulated by H-NS, were similar in the wildtype, fluorescent fusion, as well as H-NS deletion strains (Fig. S1
Fig. 1 Super-resolution imaging of major nucleoid-associated proteins in living E. coli cells. (A) Compact H-NS clusters in the nucleoid. The E. coli cells shown in the bright-field image (left) expressed photoactivatable fluorescent protein mEos2 fused to H-NS, (more ...)
In contrast to the clustered distribution of H-NS, HU was largely scattered throughout in the nucleoid (), consistent with recent data from another bacterial species, C. crescentus
). Similar distributions were observed for Fis and IHF, albeit at lower expression levels (Fig. S2
). StpA also displayed a scattered distribution in the nucleoid (Fig. S2
), despite being a paralogue of H-NS (14
). Interestingly, Rok, a B. subtilis
protein functionally analogous to H-NS but lacking any sequence homology, is distributed non-uniformly in the nucleoid (35
). In addition to the NAPs, we also imaged the ribosomes, which should be excluded from the nucleoid (1
). As expected, the ribosomes were enriched in the cell periphery (Fig. S3
Next, we examined the molecular mechanisms responsible for the formation of H-NS clusters in vivo. H-NS has two structural domains: an N-terminal domain that promotes dimerization and oligomerization and a C-terminal domain that binds to DNA (17-19
). We tested the effects of these functions on the cluster formation by introducing an N-terminal point mutation L30P that inhibits H-NS dimerization (36
) or a C-terminal point mutation P116S that inhibits DNA binding (37-38
) into the chromosomally expressed H-NS-mEos2 fusion protein. The expression levels of the two mutants were comparable to that of the wildtype (Table S1
). Both mutations abolished the silencing effect of H-NS on hdeA
, but had little influence on lacZ
expression (Fig. S1
). In contrast to the wildtype H-NS, H-NSL30P
did not form clusters, but was scattered throughout the nucleoid (), indicating that cluster formation was induced by the N-terminal-domain driven oligomerization of the protein. In the cells expressing H-NSP116S
, the number of observed localizations was reduced by ~20 fold compared to the H-NS expressing cells (), indicating that the localizations of H-NS were primarily due to molecules bound to DNA.
To quantify the effect of H-NS clustering on chromosome organization, we first characterized the number of H-NS clusters per chromosome, and the physical location and size of these clusters. Given the cylindrical symmetry of the cells and that the H-NS clusters were rarely observed to line up with each other in the z
-direction, we used 2D projection images for the following quantitative characterizations to take advantage of the superior resolution in the xy
plane (measured to be ~35 nm in full width at half maximum (FWHM)) compared to that along the z
direction (~75 nm, FWHM) (31
Most newly divided cells had approximately two clusters, and the number increased with the cell length (). For the longest cells prior to division, the cluster number reached four on average (). These data suggest an increase in the number of H-NS clusters with the chromosome copy number. To test whether the specific fluorescent protein tag, mEos2, had influenced the cluster formation, we fused H-NS with a different monomeric photoactivatable fluorescent protein, PAmCherry1 (39
). Similar clustering was observed for PAmCherry1-labeled H-NS (Fig. S4
). Subsequent analyses of H-NS clusters were performed on mEos2-labeled samples.
Fig. 2 Quantitative characterizations of the H-NS clusters. (A-C) The number of clusters per cell. (A) Overlay of the phase contrast images showing the cell contours (segmentation shown in green) and the super-resolution images of H-NS (magenta) for three cells (more ...)
To determine the number of clusters per chromosome, we reduced the cell growth rate by using a glycerol-supplemented minimal medium to ensure that each newly divided cell had exactly one copy of the chromosome (31
). Under this condition, we observed two H-NS clusters in the shortest cells (), suggesting that there are ~2 H-NS clusters associated with each copy of the chromosome. These two clusters were preferentially located near the one-quarter and three-quarter positions along the long axis of the cell (). In cells that had three clusters, the additional cluster tended to appear in the middle (Fig. S5
To characterize the cluster size, we determined the localization distributions within the clusters. The widths of the distributions were on average ~160 nm measured at half maximum density and ~360 nm at 3% of the maximum (). The background localization density outside the clusters was only ~1% of the peak densities in the clusters. The cluster size was substantially larger than both our localization precision (~35 nm) and the cluster movement during the observation time (~40 - 50 nm over 0.5 - 2 min). Moreover, the measured cluster size did not change appreciably when we changed the imaging time from 0.5 to 1 min. These results indicate that neither localization precision nor motion blurring had substantial effects on our measurements on the H-NS clusters.
The above quantifications indicate that the volume occupied by the H-NS clusters in each cell represents only a small fraction of the total nucleoid volume that was estimated to be ~0.2 μm3
from the volume occupied by HU molecules as well as from previous experiments (40
). Given that H-NS binding sites are broadly distributed throughout the E. coli
), collapsing of the DNA-bound H-NS into two compact clusters must therefore lead to substantial folding of DNA and reorganization of the chromosome at the global scale.
To probe whether the cluster organization of H-NS correlates with its regulatory role, we studied the spatial relationship between H-NS clusters and H-NS regulated genes. The positions of the gene loci were determined by imaging eYFP-labeled Tet repressor (TetR-eYFP) bound to tet
) sequences inserted upstream of the genes of interest (). Unlike previously used fluorescent repressor-operator systems, which typically contain tens to hundreds of tandem repeats of repressor binding sites, we inserted only six tetO
repeats (219 bp) immediately upstream of the target genes to more precisely mark their positions. Using a negative feedback loop regulated by MalI () (41
), we achieved a low expression level of TetR-eYFP that allowed the clear detection of the tetO
-bound TetR-eYFP above the background (), though not all target loci were necessarily bound by TetR-eYFP due to the small number of tetO
sites and the low expression level of TetR-eYFP. These strains had the same growth rates as the wildtype (31
), whereas the strains with a large number of inserted repressor binding sites tend to exhibit growth defects.
Fig. 3 Colocalization of H-NS clusters and specific gene loci. (A) Two-color imaging scheme of mEos2-labeled H-NS and eYFP-labeled gene locus as described in the text. (B) Two-color live-cell images of H-NS (magenta) and the hdeA, hchA, or lacZ loci (green), (more ...)
The two-color super-resolution images of mEos2-labeled H-NS and eYFP-labeled gene loci were taken using a sequential imaging approach to avoid the spectral crosstalk between eYFP and the pre-activation form of mEos2: The mEos2 molecules were first activated using a 405 nm laser and imaged with a 561 nm laser; after all mEos2 molecules were photobleached, the eYFP molecules were imaged using a 514 nm laser. The negligible displacement of the H-NS clusters (~20 nm) during the time taken for eYFP imaging allowed the colocalization between the gene loci and H-NS clusters to be probed in live cells.
Using this approach, we imaged H-NS together with the loci of hdeA, hchA
, and lacZ
genes, the former two of which are regulated by H-NS (Fig. S1
). As shown in , hdeA
colocalized with H-NS clusters to a substantially larger extent than lacZ
, indicating that the H-NS-regulated operons are preferentially sequestered into the H-NS clusters. Interestingly, while the H-NS clusters themselves appeared largely static, both hdeA
loci were mobile and did not always colocalize with the clusters, suggesting that the nucleoprotein complex is a heterogeneous and dynamic entity.
By sequestering the regulated genes, the H-NS clusters likely cause a significant reorganization of the chromosome. To test this effect, we probed the positions of hdeA, hchA and lacZ in the wildtype versus hns-null strains. The position of the labeled gene locus was determined relative to the cell's long and short axes, and shows the probability density maps of these gene loci obtained from many cells. Notably, the positions of hdeA and hchA loci in hns-null cells were both shifted by ~300 nm compared to the wildtype cells, a distance comparable to the radius of the nucleoid. In contrast, the position of the lacZ gene remained largely unchanged (shifted by less than 60 nm) upon hns deletion.
To further test the long-range chromosome interactions induced by H-NS clustering, we performed a chromosome conformation capture (3C) assay (42
) to probe the spatial proximity among various H-NS regulated genes. In this assay, DNA segments brought into proximity by protein-mediated interactions were captured by formaldehyde crosslinking, followed by restriction enzyme digestion. The crosslinked DNA segments were then ligated and probed by quantitative PCR (qPCR) with specifically designed primer pairs. The amount of PCR products relative to those obtained from non-crosslinked cells should scale with the crosslinking frequency, which in turn reflects the relative proximity between the DNA segments.
Using this approach, we tested the pair-wise proximity among nine H-NS regulated genes broadly distributed along the E. coli genome (), which gave a total of 36 possible pairs. In addition, we selected three random loci on the genome plus lacZ as four negative control sites (). The six pairs among these four sites and the nine pairs between lacZ and each of the nine H-NS regulated genes constituted a total of 15 control pairs. Out of the 36 pairs of H-NS-regulated gene loci, ten showed PCR priming errors, while the remaining 26 pairs gave quantifiable results. Likewise, 14 out of 15 control pairs gave quantifiable results. Because all control pairs showed crosslinking frequencies < 2 (, right), we designated 2 as the background value. The vast majority of the H-NS regulated locus pairs (25 out of 26 pairs) yielded crosslinking frequency values larger than the background level, all of which showed reduced crosslinking frequency values upon hns deletion, suggesting protein-induced juxtaposition of these loci (, left). Taken together, these results indicate that the H-NS clusters bring many gene loci into proximity and thereby mediate long-range interactions in the chromosome.
Fig. 4 Proximity between gene locus pairs probed by chromosome conformation capture (3C). (A) Nine H-NS regulated gene loci (labeled as A-I, black circles) and four negative control loci (labeled as a-d, red crosses) on the circular E. coli chromosome map. The (more ...)
Overall, our results demonstrate that H-NS forms a few compact clusters in the chromosome, with cluster formation driven by oligomerization of H-NS bound to DNA. The genes regulated by H-NS are specifically sequestered into these clusters. Given that H-NS is a global transcriptional silencer that regulates ~5% of all E. coli
) and binds to many DNA sites broadly distributed along the E. coli
genome (8-9, 11-13
), the cluster formation of H-NS and, consequently, the juxtaposition of DNA segments interacting with H-NS must cause substantial folding and condensation of the bacterial DNA. The H-NS clusters could thus serve as anchoring points for numerous DNA loci distributed throughout the genome, potentially creating DNA loops connecting the anchored loci. These anchor points (or organizing centers) can act in concert with the previously described chromosome domains (4-5
) to shape the 3D architecture of the E. coli