1. Chromosomes and Gene Expression
  2. Evolutionary Biology
Download icon

Chromatinization of Escherichia coli with archaeal histones

  1. Maria Rojec
  2. Antoine Hocher
  3. Kathryn M Stevens
  4. Matthias Merkenschlager
  5. Tobias Warnecke  Is a corresponding author
  1. Medical Research Council London Institute of Medical Sciences, United Kingdom
  2. Faculty of Medicine, Imperial College London, United Kingdom
Research Article
  • Cited 3
  • Views 2,746
  • Annotations
Cite this article as: eLife 2019;8:e49038 doi: 10.7554/eLife.49038

Abstract

Nucleosomes restrict DNA accessibility throughout eukaryotic genomes, with repercussions for replication, transcription, and other DNA-templated processes. How this globally restrictive organization emerged during evolution remains poorly understood. Here, to better understand the challenges associated with establishing globally restrictive chromatin, we express histones in a naive system that has not evolved to deal with nucleosomal structures: Escherichia coli. We find that histone proteins from the archaeon Methanothermus fervidus assemble on the E. coli chromosome in vivo and protect DNA from micrococcal nuclease digestion, allowing us to map binding footprints genome-wide. We show that higher nucleosome occupancy at promoters is associated with lower transcript levels, consistent with local repressive effects. Surprisingly, however, this sudden enforced chromatinization has only mild repercussions for growth unless cells experience topological stress. Our results suggest that histones can become established as ubiquitous chromatin proteins without interfering critically with key DNA-templated processes.

https://doi.org/10.7554/eLife.49038.001

Introduction

All cellular systems face the dual challenge of protecting and compacting their resident genomes while making the underlying genetic information dynamically accessible. In eukaryotes, this challenge is solved, at a fundamental level, by nucleosomes,~147 bp of DNA wrapped around an octameric histone complex. Nucleosomes can act as platforms for the recruitment of transcriptional silencing factors such as heterochromatin protein 1 (HP1) in animals (Danzer and Wallrath, 2004; Zhao et al., 2000) and Sir proteins in yeast (Gartenberg and Smith, 2016), but can also directly render binding sites inaccessible to transcription factors (Beato and Eisfeld, 1997; Zhu et al., 2018). As a consequence, gene expression in eukaryotes is often dependent on the recruitment of chromatin remodelers. By controlling access to DNA, histones play a key role in lowering the basal rate of transcription in eukaryotic cells and have therefore been described as the principal building blocks of a restrictive transcriptional ground state (Struhl, 1999).

Histones are not confined to eukaryotes, but are also common in archaea (Adam et al., 2017; Henneman et al., 2018). They share the same core histone fold but typically lack N-terminal tails, which are the prime targets for post-translational modifications in eukaryotes (Henneman et al., 2018). As tetrameric complexes, they wrap ~60 bp instead of ~147 bp of DNA (Reeve et al., 2004). At least in some archaea, these tetrameric complexes can be extended, in dimer steps, to form longer oligomers that wrap correspondingly more DNA (~90 bp,~120 bp, etc.) and assemble without the need for dedicated histone chaperones (Xie and Reeve, 2004; Mattiroli et al., 2017; Maruyama et al., 2013). Archaeal and eukaryotic nucleosomes preferentially assemble on DNA that is more bendable, a property associated with elevated GC content and the presence of certain periodically spaced dinucleotides, notably including AA/TT (Ammar et al., 2011; Nalabothula et al., 2013; Pereira et al., 1997; Bailey et al., 2000; Ioshikhes et al., 2011). They also exhibit similar positioning around transcriptional start sites (Ammar et al., 2011; Nalabothula et al., 2013), which are typically depleted of nucleosomes and therefore remain accessible to the core transcription machinery. Whether archaeal histones play a global restrictive role akin to their eukaryotic counterparts, however, remains poorly understood, as does their involvement in transcription regulation more generally (Gehring et al., 2016).

Thinking about the evolution of restrictive chromatin and its molecular underpinnings, we wondered how the presence of histones would affect a system that is normally devoid of nucleosomal structures. How would a cell that has neither dedicated nucleosome remodelers nor co-evolved sequence context cope with chromatinization? Could global chromatinization occur without fundamentally interfering with DNA-templated processes? How easy or hard is it to transition from a system without histones to one where histones are abundant? What are the key adaptations required, if any, to accommodate histones?

Motivated by these questions, we built Escherichia coli strains expressing histones from the hyperthermophilic archaeon Methanothermus fervidus (HMfA or HMfB), on which, thanks to the pioneering work of Reeve and co-workers, much of our foundational knowledge about archaeal histones is based. HMfA and HMfB are 85% identical at the amino acid level but differ with regard to their DNA binding affinity and expression across the M. fervidus growth cycle, with HMfB more prominent toward the latter stages of growth and able to provide greater DNA compaction in vitro (Sandman et al., 1994; Marc et al., 2002). We find that HMfA and HMfB, heterologously expressed in E. coli, bind to the E. coli genome and protect it from micrococcal nuclease (MNase) digestion, allowing us to map nucleosomes in E. coli in vivo. We present evidence for sequence-dependent nucleosome positioning and occupancy and consider how the presence of histones affects transcription on a genome-wide scale. Importantly, we find evidence for local repressive effects associated with histone occupancy yet only mild repercussions for growth and cell morphology, unless cells are forced to deal with excess levels of DNA damage or topological stress. Under favourable conditions, E. coli copes remarkably well with enforced chromatinization, despite evidence that histones disrupt the binding of native nucleoid-associated proteins (NAPs). Our findings have implications for how histones became established as global repressive regulators during the evolution of eukaryotes and for the evolvability of transcriptional ground states.

Results

Archaeal histones bind the E. coli genome in vivo, assemble into oligomers, and confer protection from MNase digestion

We transformed an E. coli K-12 MG1655 strain with plasmids carrying either hmfA or hmfB, codon-optimised for expression in E. coli and under the control of a rhamnose-inducible promoter (see Materials and methods, Figure 1—figure supplement 1). Below, we will refer to these strains as Ec-hmfA and Ec-hmfB, respectively, with Ec-EV being the empty vector control strain (Supplementary file 1). Following induction, both histones are expressed at detectable levels and predominantly found in the soluble fraction of the lysate in both exponential and stationary phase (Figure 1—figure supplement 2). We did not observe increased formation of inclusion bodies. Based on dilution series with purified histones (see Materials and methods, Figure 1—figure supplement 2), we estimate HMfA:DNA mass ratios of up to ~0.6:1 in exponential (~0.7:1 in stationary phase), which corresponds to one histone tetramer for every 76 bp (64 bp) in the E. coli genome. Given that a tetramer wraps ~60 bp of DNA, this implies a supply of histones that is, in principle, sufficient to cover most of the E. coli genome. However, it is important to note that, at any given time, not all histones need to be associated with DNA.

We carried out MNase digestion experiments using samples from late exponential and stationary phase, corresponding to 2 hr and 16–17 hr after induction, respectively (see Materials and methods). In response to a wide range of enzyme concentrations, MNase digestion of chromatin from Ec-hmfA/B (see Materials and methods) yields a ladder-like pattern of protection that is not observed in Ec-EV (Figure 1A–B). Across many replicates, we could usually discriminate the first four rungs of the ladder, with the largest rung at 150 bp. On occasion, we observe multiple larger bands (e.g. for Ec-hmfA in Figure 1A). Sequencing digestion fragments < 160 bp using single-end Illumina technology recapitulates the read length distribution seen on gels, with peaks around 60 bp, 90 bp, 120 bp, and 150 bp (Figure 1C), consistent with oligomerization dynamics described for archaeal histones in their native context (Maruyama et al., 2013; Mattiroli et al., 2017). Indeed, we obtained remarkably similar digestion profiles when we applied the same protocol, modified to account for altered lysis requirements (see Materials and methods), to M. fervidus cultures (Figure 1A,C). Modal fragment sizes of ~60 bp and ~90 bp in exponential and stationary phase (Figure 1C), respectively, suggest that larger oligomers become more prevalent later in the growth cycle, which might reflect elevated histone:DNA ratios but also reduced perturbation from replication and transcription, as further discussed below. In exponential phase only, an additional peak is evident at ~30 bp. Fragments of this size were previously observed during in vitro reconstitution experiments with HMfA/B and, at the time, attributed to the binding of histone dimers (Grayling et al., 1997). However, in our digestion regime, this peak is also present in Ec-EV, and we cannot therefore rule out the possibility that it is caused by specifics of the digestion protocol, library construction or native E. coli proteins found exclusively in exponential phase. Below, we therefore focus on larger peaks (60 bp, 90 bp, etc.) that are absent from Ec-EV, but present in M. fervidus and our histone-bearing E. coli strains.

Figure 1 with 2 supplements see all
MNase digestion of M. fervidus and E. coli strains expressing M. fervidus histones.

(A) Agarose gel showing profiles of DNA fragments that remain protected at different MNase (MN) concentrations. (B) Ladder-like protection profiles are only observed when hmfA/B expression is induced. (C) Length distribution profiles of sequenced fragments show peaks of protection at multiples of 30 bp in histone-expressing strains. Structural views below highlight how these 30 bp steps would correspond to the addition or removal of histone dimers, starting from the crystal structure of a hexameric HMfB complex (PDB: 5t5k), which wraps ~90 bp of DNA.

https://doi.org/10.7554/eLife.49038.002

Intrinsic sequence preferences govern nucleosome formation along the E. coli genome

Mapping digestion fragments to the E. coli genome, we find that binding is ubiquitous. On a coarse scale, coverage across the chromosome appears relatively even (Figure 2A). On a more local scale, however, protected fragments group into defined binding footprints (Figure 2C). Local occupancy (measured for 60 bp windows, overlapping by 30 bp) is highly correlated across replicates (Figure 2D), consistent with non-random binding. Ec-hmfA and Ec-hmfB are also highly correlated (Figure 2E); minor differences may reflect subtly different binding preferences, as previously reported (Bailey et al., 2000). Areas of apparent histone depletion often coincide with AT-rich domains (Figure 2C,F): nucleosomes are depleted from AT-rich transcriptional start sites (TSSs), mimicking a key aspect of nucleosome architecture in eukaryotes and archaea (Figure 2G), and extension into longer oligomers is less likely when tetramer binding footprints are flanked by AT-rich sequence (Figure 2—figure supplement 1), as is the case in M. fervidus (Hocher et al., 2019).

Figure 2 with 1 supplement see all
Distribution of MNase-protected fragments across the E. coli genome.

(A) Genome-wide coverage (and normalized coverage) tracks of MNase-protected fragments along the E. coli K-12 MG1655 and (B) the M. fervidus genome. (C) Fragments of defined size cluster into footprints in E. coli and M. fervidus, as illustrated for two example regions. (D) Correlation in coverage measured for two biological replicates of Ec-hmfA. Coverage here is expressed as a proportion of total reads in a given replicate. (E) Correlation in normalized coverage between Ec-hmfA and Ec-hmfB. Reads were pooled across replicates for each strain. (F) Two examples from Ec-hmfA highlighting that drops in coverag frequently correspond to regions of low GC content. (G) Coverage as a function of both distance from experimentally defined transcriptional start sites (see Materials and methods) and fragment size.

https://doi.org/10.7554/eLife.49038.005

The above observations point to a role for sequence composition in determining nucleosome positioning and/or occupancy but likely also reflect known MNase preferences for AT-rich DNA (see Ec-EV in Figure 2G in particular). To discriminate between these two factors, we first analysed read-internal nucleotide enrichment patterns, which should be unaffected by MNase bias. Considering fragments of exact size 60 bp (90 bp, etc. see Materials and methods), we find dyad-symmetric nucleotide enrichment patterns that are absent from size-matched Ec-EV fragments but mirror what is seen in fragments from native M. fervidus digests (Figure 3A), despite large differences in overall genomic GC content. Next, to disentangle conflated signals of MNase bias and nucleosomal sequence preferences directly, and to assess their relative impact on inferred occupancy across the genome, we normalized coverage in Ec-hmfA/B by coverage in Ec-EV (see Materials and methods). We then trained LASSO models for different fragment size classes (60 bp, 90 bp, 120 bp) to predict normalized occupancy across the genome from the underlying sequence, considering all mono-, di-, tri-, and tetra-nucleotides as potential predictive features (see Materials and methods, Supplementary file 2). We find that sequence is a good predictor of normalized occupancy in stationary phase (Figure 3B–C), particularly for larger fragments (e.g. 120 bp footprints in Ec-hmfA: ρ = 0.72, p<2.2×10−16; 120 bp footprints in Ec-hmfB: ρ = 0.76, p=<2.2×10−16, Figure 3C). GC content as a simple metric captures much of the variability in occupancy (Figure 3B,D).

Sequence and other predictors of histone occupancy in E. coli.

(A) Read-internal nucleotide enrichment profiles for reads of exact length 60/90/120 bp. Symmetric enrichments are evident for Ec-hmfA and M. fervidus native fragments but not Ec-EV. (B) Left panel: top and bottom 20 individually most informative k-mers to predict fragment size-specific normalized histone occupancy in different strains. Red and blue hues indicate positive and negative correlations between k-mer abundance and normalized occupancy, respectively. Right panel: performance of the full LASSO model on training and test data (see Materials and methods). expo: exponential phase; stat: stationary phase. (C) Correlations between predicted and observed coverage of 120 ± 5 bp fragments predicted at single-nucletoide resolution across the genome. All p<0.001. (D) GC content and normalized coverage are positively correlated in stationary but not exponential phase. All p<0.001. Coverage and GC content are measured by gene. (E) The correlation between GC content and occupancy is stronger in genomic regions where transcriptional output is lower. Regional transcriptional output is computed as median transcript abundance in a 200-gene window. To assess potential interactions between replication and transcription, windows are computed separately for genes where the directions of transcription and replication coincide and those where they differ. (F) There is no negative correlation between mRNA abundance in Ec-EV and normalized histone occupancy in Ec-hmfA, suggesting that low levels of transcription do not facilitate higher occupancy. (G) The strength of the correlation between GC content and occupancy varies along the E. coli chromosome. Correlations are computed for 500 neighbouring genes using a 20-gene moving window.

https://doi.org/10.7554/eLife.49038.007

Interestingly, however, the predictive power of sequence is dramatically reduced in exponential phase (Figure 3B,D). Why would this be? We suspect that stationary phase represents a comparatively more settled state, characterized by reduced replication, transcription, and other DNA-templated activity, that is more conducive to the establishment or survival of larger oligomers and where nucleosome formation is better able to track intrinsic sequence preferences. In support of this hypothesis, we find that transcriptional activity modulates the relationship between GC content and occupancy: the relationship is stronger where transcriptional activity is weaker (ρ = −0.46, p=0.039; Figure 3E). Importantly, this does not imply that higher transcription leads to reduced histone occupancy. In fact, there is no negative correlation between transcript levels in Ec-EV and histone occupancy (Figure 3F, ρ >0.1 for all growth phase/histone combinations). Rather, these results are consistent with transcription increasing the fuzziness of nucleosome positioning. We also find a better correlation between sequence composition and occupancy further away from the origin of replication, suggestive of replication-associated perturbation (Figure 3G).

Evidence that nucleosome formation locally represses transcription

Next, we asked whether the presence of histones in E. coli affects transcription. We first consider whether histones exert direct repressive effects in cis. Further below, we look at genome-wide transcriptional responses to histone expression more broadly to understand how E. coli is challenged by and adapts to the presence of histone proteins.

To address the first question, we generated two additional strains, Ec-hmfAnb and Ec-hmfBnb, where hmfA and hmfB, respectively, were recoded to carry three amino acid changes (K13T-R19S-T54K) previously shown to abolish DNA binding of HMfB (Soares et al., 2000). MNase treatment of these strains resulted in digestion profiles similar to Ec-EV, consistent with compromised ability to form protective nucleosomal structures (Figure 4—figure supplement 1). Using RNA-Seq, we quantified differential transcript abundance in Ec-hmfA versus Ec-EV and Ec-hmfAnb versus Ec-EV (see Materials and methods) and then excluded genes from further analysis that were significantly up-regulated (or down-regulated) in both comparisons, reasoning that coincident patterns of change are not uniquely attributable to binding and might instead derive from systemic responses to heterologous expression. We then considered differential expression in Ec-hmfA/B versus Ec-EV for the remaining genes as a function of nucleosome occupancy.

Looking at normalized coverage across gene bodies, annotated promoters and experimentally mapped transcriptional start sites, we find evidence for nucleosome-mediated dampening of transcriptional output. Notably, genes that are significantly (Padj <0.05) down-regulated in histone-bearing strains display significantly higher nucleosome occupancy at TSSs than upregulated genes (Figure 4A). This is true regardless of whether we consider occupancy at a single base assigned as the TSS, occupancy in a ± 25 bp window around that site, or occupancy across annotated promoters (see Materials and methods). This signal is lost almost entirely when considering a promoter-proximal 51 bp control window centred on the start codon (Figure 4—figure supplement 2). This finding argues against a model where histone occupancy increases as a consequence of downregulation. Under such a model, we would have predicted histone occupancy to increase not only at the promoter but also downstream of it. The relationship between transcriptional changes and average histone occupancy across the gene body is more complex; weaker effects in the expected direction are evident for Ec-hmfA but not Ec-hmfB (Figure 4—figure supplement 2).

Figure 4 with 3 supplements see all
The impact of archaeal histones on transcription in E. coli.

(A) Reduced transcript abundance in histone-expressing strains is associated with higher average histone occupancy at the TSS. Top panels: Ec-hmfA. Bottom panels: Ec-hmfB (B) Genes that are significantly downregulated in histone-expressing strains exhibit higher coverage of large (90+bp) but not small (60 bp) fragments. Top panels: Ec-hmfA. Bottom panels: Ec-hmfB. ****p<0.001; ***p<0.005; **p<0.01; *p<0.05.

https://doi.org/10.7554/eLife.49038.008

Interestingly, repressive effects at TSSs in particular appear to be driven by larger oligomeric nucleosomes (90 bp, 120 bp, 150 bp, Figure 4B, Figure 4—figure supplement 2). This might be because larger oligomeric complexes are intrinsically more stable (Figure 4—figure supplement 3), harder to bypass/displace, and therefore more significant barriers to transcription initiation and elongation. In analogy to H-NS, larger oligomers might also, from an initial point of nucleation, extend to cover sequences that disfavour nucleation – a property that might facilitate promoter occlusion (Henneman et al., 2018; Hocher et al., 2019).

Histone binding is associated with mild phenotypic effects under favourable conditions

Despite evidence for repressive effects, gross cell morphology and growth rate appear surprisingly normal. Histone-expressing cells are longer than Ec-EV cells, particularly in stationary phase, but they do not exhibit an altered nucleoid/cytoplasm ratio and, following a transient reduction in growth rate after induction, appear to divide normally (Figure 5, Figure 5—figure supplement 1). Under favourable conditions, growth of histone-expressing E. coli appears remarkably unremarkable. But how do these strains respond to stress? To find out, we monitored growth in response to transcriptional stress (rifampicin), oxidative stress (H2O2), DNA damage (UV), and supercoiling stress (novobiocin). To capture effects of histone occupancy during lag phase and ensure that stress responses are measured in cells where histones are established, we inoculated new cultures with cells that had already been expressing histone genes for 2 hr (see Materials and methods). When these pre-induced cells are re-inoculated, we observe a slightly prolonged lag phase (Figure 6A). However, histone-expressing strains recover quickly to catch up with non-binding/EV control strains. Lag phase is extended further in strains treated with rifampicin or H2O2 (Figure 6A). Again, histone-expressing strains recover well. Under these conditions, histones have a mild bacteriostatic but no bactericidal effect. In contrast, the presence of histones clearly affects the ability of cells to respond to UV and novobiocin treatment: colony formation and growth, respectively, are severely affected (Figure 6A–B). In novobiocin-treated histone-expressing cells, we also observe marked morphological changes, as cells become conspicuously elongated (Figure 6C).

Figure 5 with 1 supplement see all
The impact of archaeal histones on E. coli growth.

(A) Morphological changes triggered by HMfA and HMfB expression. Compared to the empty vector control, Ec-hmfA and Ec-hmfB become significantly longer, particularly toward the final stage of the cell cycle. DAPI staining suggests that the increase in cell length is not due to impaired cell division. Magnification 100x. (B) Quantification of cell length and area in histone-expressing and control strains. Some unexpectedly low values are likely attributable to debris being misidentified as cells. *p<0.0001. (C) Growth curves for induced and uninduced histone-expressing and control strains. Rhamnose was added for induction at 200 min.

https://doi.org/10.7554/eLife.49038.012
Growth responses of histone-expressing E. coli strains under stress.

(A) Growth curves for pre-induced histone-expressing E. coli strains and controls in LB medium and LB medium with added rifampicin, H2O2, or novobiocin. See Materials and methods for growth/induction protocol and drug/chemical concentrations. (B) Colony counts for E. coli strains exposed to UV radiation or left untreated (all p<0.05). (C) Novobiocin treatment of Ec-hmfA results in a strong filamentation phenotype.

https://doi.org/10.7554/eLife.49038.014

Systemic transcriptional responses to histone expression in E. coli

The results above suggest that histones do not compromise dynamic responses to stress in general but that their presence is problematic when sensing or dealing with altered DNA topology or damage. To better understand the molecular basis of altered growth, we compared the transcriptome-wide signature of differential expression in Ec-hmfA (versus Ec-hmfAnb, exponential phase) to >950 previously published differential expression profiles from a broad range of perturbations (see Materials and methods).

Calculating dot products as a measure of similarity between two differential expression vectors (see Materials and methods), we find that correlations between expression profiles is modest (maximum ρ = 0.34), indicating that the transcriptional response to histone expression has a strong unique component. Histone-expressing strains are most similar to perturbations that are marked by transient growth arrest and induction of the stringent response (amino acid starvation, cadmium shock, heat stress, Figure 7A, source data file 1) and to growth under metabolically challenging conditions, that is conditions where carbon sources are either scarce (stationary phase, minimal media) or suddenly altered (glucose-to-lactose shift, Figure 7A). Specific similarities include the downregulation of flagellar genes – a hallmark of the stringent response – and upregulation of the general stress response (RpoS regulon, Figure 7B). These transcriptional signatures are very much in line with the mild bacteriostatic growth phenotype (extended lag phase) we observed (Figure 6). Cells delay division until they have had sufficient time to adjust and even though stress responses are induced, these are not necessarily required for survival (Figure 7—figure supplement 1).

Figure 7 with 1 supplement see all
Global transcriptional responses in histone-expressing strains highlight effects on E. coli physiology and native chromatin organization.

(A) Comparative analysis of global transcriptional responses, comparing up- or down-regulated genes in Ec-hmfA (versus Ec-hmfAnb) to other perturbations (underlying data provided as Figure 7A – source data). Perturbations with high similarity to Ec-hmfA versus Ec-hmfAnb along at least one dimension are highlighted and coloured according to the nature of the perturbation. Values < 0 indicate overall dissimilarity, equivalent to a negative correlation coefficient between the transcriptional responses. Note that the absolute similarity values here have no intrinsic meaning; only the relative distance from the maximum, hmfA/hmfAnb (expo), is meaningful. Note also that similarity should only be interpreted in reference hmfA/hmfAnb (expo). Points labelled ‘exponential phase’ constitute rare cases where, in the original study, differential expression was assessed as expo/stat rather than the more common stat/expo. When flipped, these fall into or close to the pink cluster of stationary phase datasets. (B) Genes controlled by RpoS (identified by comparing the response to isoleucine starvation in WT and ΔrpoS cells, upper panel) are upregulated upon isoleucin starvation but also in histone-expressing strains (illustrated for Ec-hmfA in the lower panel). Based on GSE11087 as provided in GenexpDB. (C) Correspondence between transcriptomic changes in Ec-hmfB versus Ec-hmfBnb and a Δh-ns strain (GSE123554). Direct H-NS targets, as inferred by Gawade et al. (2019), are highlighted in green. (D) Histone occupancy in regions previously found to be bound or unbound by a particular nucleoid-associated protein in E. coli. ∆ histone occupancy is defined as the difference in histone occupancy in a region bound by a given NAP and the nearest unbound region downstream. Negative ∆(histone occupancy) values therefore indicate greater histone occupancy in areas not bound by the focal NAP, suggestive of competition for binding or divergent binding preferences. *p<0.005 **p<0.001.

https://doi.org/10.7554/eLife.49038.015

Downregulation of gyrases (gyrA/B, Figure 7—figure supplement 1), which introduce negative (or relax positive) supercoils, might be part of such an adaptive readjustment. Histones wrap DNA in negatively constrained supercoils so reducing gyrase expression might counteract histone-associated build-up of negative supercoiling. This might provide a quick fix, but at the cost of rendering cells more susceptible to novobiocin. In line with this idea, histone-expressing strains share transcriptional similarities to cells expressing CcdB, a gyrase poison (Figure 7A).

Evidence that histones interfere with the binding of native nucleoid-associated proteins

We were further intrigued to see that, specifically with regard to upregulated genes, the effect of histones is similar to deleting h-ns (ρ = 0.19, p<2.2×10−16). Most notably, genes previously identified as direct H-NS targets (green icons in Figure 7C) are amongst the most upregulated genes not only when h-ns is deleted (as one would expect), but also upon HMf expression. This might indicate that histones displace H-NS, but fail to provide similar silencing, leading to de-repression of H-NS target genes. In line with this hypothesis, we find that histone occupancy is not significantly reduced at known binding footprints of H-NS (Kahramanoglou et al., 2011), indicating that histones successfully compete for binding at those sites (Figure 7D). In addition to de-repression of its usual target genes, the release of H-NS might also cause gain-of-function effects, for example through the binding of AT-rich promoters that would normally not be silenced. It is interesting to note in this context that strong (>40 fold) overproduction of H-NS has previously been reported to trigger a transient (several-hour) growth arrest after which cells resume growth (McGovern et al., 1994). This situation, which the authors dubbed ‘artificial stationary phase’, is qualitatively reminiscent of the prolonged lag phase we observe upon HMf expression.

We also find little, if any, evidence for competitive exclusion at known binding sites of other endogenous NAPs (Figure 7D). In contrast to Δh-ns, however, transcriptional responses in ΔhupA/hupB, Δdps, and Δfis strains are uncorrelated to those in Ec-hmfA/B (all ρ<|0.04|).

The above results suggest that histones readily invade genomic real estate normally occupied by endogenous NAPs. Might histones therefore, in some instances, complement NAP deletions? To address this question, we examined the effects of HMfA expression on growth in a small collection of NAP deletion strains, using the larger YFP protein as a conservative control for the burden of gratuitous protein expression. Note first that NAP deletions in E. coli are not associated with a strong growth phenotype, with the notable exception of the hupA/hupB double deletion (ΔΔHU) strain, which grows notably more slowly compared to its C600 wild-type progenitor (Figure 8B). HMfA expression generally leads to an increase in lag phase duration, operationally defined as the time to maximum growth rate (Figure 8A). This is particularly pronounced when fis is deleted and – for unknown and hard to interpret reasons – in M182, the wild-type progenitor strain of Δh-ns. HMfA expression is also associated with a small but consistent increase in doubling time. However, in most cases, this effect is not compounded by deleting the focal NAP. The exception, again, is HU. Growth retardation associated with hupA/hupB deletion and HMfA expression are not additive, suggesting that histone expression might partially alleviate defects associated with the absence of HU, perhaps because both proteins constrain negative supercoils.

Effects of histone expression in NAP deletion strains.

(A) Duration of lag phase, operationally defined as the time between inoculation and the point of maximum growth rate, in cultures of YFP- and HmfA-expressing cells from different genetic backgrounds. (B) Doubling time at the point of maximum growth rate as determined by curve fitting. All pairwise comparisons significant at p<0.05 unless indicated. Different genetic backgrounds are separated by dotted vertical lines.

https://doi.org/10.7554/eLife.49038.018

Discussion

Our experiments demonstrate that archaeal histones are surprisingly well tolerated when expressed in E. coli, a system that has not evolved to deal with nucleosomal structures. Despite binding ubiquitously to the E. coli genome, they do not fundamentally compromise critical DNA-templated processes under favourable growth conditions. In particular, while we find some evidence that nucleosome occupancy locally restricts the output of the transcription machinery and that histones displace endogenous NAPs, gene expression is insufficiently perturbed to affect growth beyond a mild extension of lag phase. Transcriptomic analysis revealed induction of several stress responses as well as downregulation of DNA gyrases, which likely help the cells to adapt to the unique challenge of nucleosome formation. With the system already stretched, histones constitute a more severe problem when cells are forced to deal with double strand breaks or topological stress.

E. coli has not evolved to specifically deal with nucleosomal structures. Why then, did histone expression not cause much more drastic effects? We suggest that, both in E. coli and during evolution, global wrapping of DNA into nucleosomes was facilitated by two factors in particular: first, by virtue of their AT-rich nature, promoters remain comparatively accessible to the transcription machinery, even in a naïve prokaryote whose sequence and functional repertoire did not co-evolve to accommodate histones. Nucleosome-free regions at the TSS, a key features of nucleosome architecture in eukaryotes, might therefore have emerged, in the first instance, as a simple consequence of promoter composition. Once established, nucleosomes bordering the TSS were uniquely positioned to be co-opted into gene regulatory roles in eukaryotes and perhaps along different archaeal lineages, with nucleosome positioning later refined by evolution at specific loci to provide more nuanced control over transcriptional processes. Second, compared to their eukaryotic counterparts, archaeal nucleosomes appear to be more surmountable barriers to transcription elongation. Even at high histone concentrations, transcription through a HMf-chromatinized template in vitro is slowed but not aborted (Xie and Reeve, 2004), in line with the absence of recognizable histone remodelers from archaeal genomes. Thus, near-global coating of the genome with archaeal-type histone proteins might have evolved without severe repercussions for basic genome function before a more restrictive arrangement, perhaps coincident with the advent of octameric histone architecture, took hold during eukaryogenesis. From an evolutionary point of view, one might therefore call the ground state mediated by archaeal histones proto-restrictive.

To what extent restrictive, proto-restrictive, or permissive ground states exist in different archaea in vivo remains unclear. Experiments with histones from M. fervidus, Methanococcus jannaschii, and Pyrococcus furiosus have shown that archaeal nucleosomes can interfere with transcription initiation and elongation in vitro (Wilkinson et al., 2010; Soares et al., 1998; Xie and Reeve, 2004; Sanders et al., 2019). However, significant inhibitory effects were only observed at high histone:DNA ratios (close to or above 1:1). Ratios of that magnitude, while regularly found in eukaryotes, need not be prevalent in archaea. Direct measurements of histone:DNA ratios are scarce and variable, with prior estimates in M. fervidus reporting stoichiometries as high as 1:1 (Pereira et al., 1997) and as low as 0.2–0.3:1 (Stroup and Reeve, 1992). Considering transcript levels as a (really rather imperfect) proxy, histones appear very abundant in Thermococcus kodakarensis and Methanobrevibacter smithii (Figure 9), strengthening the case for histones as global packaging agents in these species. In contrast, histone mRNAs are much less plentiful in Haloferax volcanii and Halobacterium salinarum (Figure 9), where histones likely have a limited role in DNA compaction (Dulmage et al., 2015) and less than 40% of the chromosome is resistant to MNase digestion (Takayanagi et al., 1992). In these species, non-histone proteins might be more important mediators of chromatin architecture and packaging. Thus, histone:DNA stoichiometry likely varies substantially across taxa as well as along the growth cycle (Takayanagi et al., 1992; Dinger et al., 2000; Sandman et al., 1994).

Relative transcript levels of histone genes across different archaeal species.

Histones were assigned a percentile rank based on their relative expression in a given species and transcriptomic dataset (0 = least abundant mRNA in the dataset; 100 = most abundant mRNA in the dataset). For species with more than one histone gene, transcript levels were summed across histone genes. Because of significant variability between studies, two separate estimates are given for H. volcanii. Data sources: H. salinarum (Gene Expression Omnibus accession GSE99730), M. barkeri (GSE70370), T. onnurineus (GSE85760), M. acetivorans (GSE64349), M. smithii (GSE25408), H. volcanii (Blombach et al., 2018 Nucl Acid Res 46:2308–2320; Gelsinger and DiRuggiero, 2018 J Bacteriol 200:e00779-17), T. kodakarensis (Jäger et al., 2014 BMC Genomics 15:684).

https://doi.org/10.7554/eLife.49038.019

Attempts to delete histone genes have also revealed considerable diversity across archaea. Histones are required for viability in T. kodakarensis and Methanococcus voltae (Čuboňováa et al., 2012; Heinicke et al., 2004), but can be removed with surprisingly muted effects on transcription in Methanosarcina mazei (Weidenbach et al., 2008) and H. salinarum (Dulmage et al., 2015). In both species, a comparatively small number of transcription units were affected by histone deletion, the majority of which was down- rather than upregulated.

Taken together, these observations suggest that histones likely play a more variable, species- and context-dependent role in archaea, may only sometimes act as global repressive agents and, more generally, that care should be taken in projecting properties of eukaryotic histones onto those of archaea. In many instances, archaeal histones might be better understood with reference to bacterial NAPs, especially when considering how concentration drives opportunities for oligomerization, cooperativity, and bridging interactions with DNA. In this context, we note that our results are reminiscent of a recent study by Janissen and colleagues, who found that dps deletion in E. coli results in nucleoid decompaction but does not greatly impact transcription (Janissen et al., 2018). This provides some generality to the notion that architectural DNA-binding proteins, even if they bind to most of the genome and alter its compaction and gross structure, need not unduly interfere with transcription. The same study also highlights that, while polymerases may continue to access DNA and operate as usual, the same need not be true for other DNA-binding proteins: Dps substantially reduced the ability of several restriction enzymes to recognize and cut their target sites. Whether archaeal histones have similar effects in E. coli (beyond their ability to protect from MNase treatment), remains to be established. However, access regulation outside of a transcriptional context might well have provided the original evolutionary impetus for histones to spread across the genome, as genomes evolved to defend themselves against selfish elements that target the host genome for integration (Talbert et al., 2019). We note in this regard that our chromatinized E. coli strains might be of use for future synthetic biology applications. As more complex, combinatorial control of gene expression becomes a desirable genome engineering objective, limiting access to desired target sites will become an increasingly important design consideration (Cardinale and Arkin, 2012), as will chassis integrity in the face of potential invaders. As we find interference with transcription and replication to be limited, it will be interesting to experiment with expressing archaeal histones to restrict global access to the genome for specific DNA-binding factors or protect the genome against selfish element invasion (Sultana et al., 2019; Aslankoohi et al., 2012).

Materials and methods

Plasmid design

Request a detailed protocol

The coding sequences of hmfA and hmfB were codon-optimised for E. coli and synthesised as part of a rhamnose-inducible pD861 plasmid (Figure 1—figure supplement 1) by ATUM (Newark, CA). Originally, both plasmids also encoded a chromogenic protein to enable visual screening for induction. However, as the chromogenic protein was expressed at very high levels (Figure 1—figure supplement 2) and since we did not want to unduly increase cellular burden we removed the corresponding gene to yield pD861-hmfA. To generate non-binding histone mutants, hmfA/hmfB sequences were re-coded to carry three changes (K13T-R19S-T54K), previously shown to jointly abolish DNA binding of HMfB (Soares et al., 2000). These sequences were codon-optimized, synthesised and integrated onto a pD861 plasmid as above, without the chromogenic proteins, as was hmfB, for which cloning had failed. Plasmids pD861-hmfA, pD861-hmfB, pD861-hmfAnb, and pD861-hmfBnb are identical expect for the sequences of the respective histone genes. hmfA was removed from pD681-hmfA to obtain Ec-EV.

Bacterial transformation and growth

Request a detailed protocol

E. coli K-12 MG1655 cells were transformed via heat-shock with either pD861-hmfA, pD861-hmfB or pD861-EV, or the non-binding histone mutants pD861-hmfAnb or pD861-hmfBnb to generate strains Ec-EV, Ec-hmfA, Ec-hmfB, Ec-hmfAnb and Ec-hmfBnb, respectively. All strains were grown in LB medium plus kanamycin (50 μg/ml) at 37°C with agitation (170 rpm). Histone expression was induced by adding L-Rhamnose monohydrate to a final concentration of 15 mM at OD600 ~ 0.6. Cells were harvested after 2 hr or 16–17 hr following induction.

Protein purification

Request a detailed protocol

HMf protein purification was performed as in Starich et al. (1996).

Coomassie staining

Request a detailed protocol

Bacteria were harvested by centrifugation (4000 rpm for 15 min at 4°C), the supernatant discarded, and the pellet resuspended in a small volume of Histone Wash Buffer (50 mM Tris-HCl pH 7.5, 100 mM NaCl, 1 mM EDTA). Cell envelopes were disrupted using a Bioruptor Plus sonication system (Diagenode s.a., Belgium) for 10 cycles, 30 s on/off with power set to high. The soluble protein fraction was separated from cellular debris by centrifugation at 15,000 x g for 15 min at 4°C, while the insoluble fraction was obtained by re-suspending the pelleted debris in Histone Wash Buffer. The protein concentration in the cell lysate was quantified with a Pierce BCA Protein Assay Kit (ThermoFisher Scientific, UK) using the provided albumin as standard. Protein fractions were separated by means of 16.5% Tris-tricine precast gels (Bio-Rad Laboratories, California) and bands were revealed by colloidal Coomassie (InstantBlue, Sigma-Aldrich) staining. Histone-expressing strains showed a band close to the size expected for HMfA/B (Figure 1—figure supplement 2). This band was excised and protein identity confirmed as HMfA/B via mass spectrometry.

Growth assays

Request a detailed protocol

Overnight pre-cultures were diluted 1:500 into LB medium plus kanamycin (50 μg/ml). Samples were plated in replicate into a flat bottom Nunc 96-well plate (ThermoFisher Scientific, UK) and incubated at 37°C at 100 rpm for 30 min. OD measurement were performed using a high-throughput microplate reader (FLUOstar Omega, BMG LABTECH GmbH, Ortenberg, Germany) in which bacteria were grown at 37°C under continuous shaking (~500 rpm, double orbital). Optical density was measured at 600 nm every 30 min for 12.5 hr. For induction, the microplate reader was paused at cycle 6 and L-Rhamnose monohydrate added manually to the relevant wells to a final concentration of 15 mM. Results presented are from three biological replicates per strain, each averaged across six technical replicates.

MNase digestion – E. coli

Request a detailed protocol

Bacterial cultures were harvested by centrifugation (4000 rpm for 15 min at 4°C), the supernatant discarded and the pelleted cells re-suspended in chilled 1x PBS (Gibco, ThermoFisher Scientific, UK). Cells were then fixed by adding a fixation solution (100 mM NaCl, 50mMTris-HCl pH 8.0, 10% formaldehyde) for 10 min at room temperature under slow rotation, after which fixation was quenched by adding 140 mM glycine. Following a further round of centrifugation (4000 rpm for 5 min at 4°C), bacteria were washed twice with 10 ml chilled 1x PBS and incubated in a lysozyme buffer (120 mM Tris-HCl pH 8.0, 50 mM EDTA, 4 mg/ml Lysozyme) for 10 min at 37°C to generate protoplasts. Cells were pelleted (15000 rpm for 3 min at room temperature) and re-suspended in 500 μl of lysis buffer (10 mM NaCl, 10 mM Tris-HCl pH 7.4, 3 mM MgCl2, 0.5% NP-40, 1x Pi, 0.15 mM Spermine, 0.5 mM Spermidine), transferred to a new microcentrifuge tube and incubated on ice for 20 min. Subsequently, the lysate was spun down and the pellet washed with 500 μl of -CA buffer (15 mM NaCl, 10 mM Tris-HCl pH 7.4, 60 mM KCl, 1x Pi, 0.15 mM Spermine, 0.5 mM Spermidine) without re-suspending. The washed pellet was finally re-suspended in 500 μl of +CA buffer (15 mM NaCl, 10 mM Tris-HCl pH 7.4, 60 mM KCl, 1 mM CaCl2, 0.15 mM Spermine, 0.5 mM Spermidine) to a uniform suspension. 50 μl of this suspension were digested with micrococcal nuclease (LS004798, Worthington Biochemical Corporation, NJ; 500 U/ml for Ec-hmfA and Ec-hmfB, 50 U/ml for Ec-EV) for 10 min (20 min for cells in stationary phase) at room temperature and finally blocked with a STOP solution containing calcium-chelating agents (100 mM EDTA, 10 mM EGTA). Each sample was further diluted with -CA buffer and treated with 10% SDS and 150 ng/ml proteinase K overnight at 65°C with shaking at 500 rpm. Undigested DNA fragments were purified by two rounds of phenol:chloroform extraction separated by an RNase A digestion step (100 μg/ml, 2 hr at 37°C with shaking at 500 rpm). Finally, DNA fragments were precipitated in ethanol and re-suspended in 40 μl distilled water. The quality of the digest and the size of the retrieved fragments were assessed by agarose DNA electrophoresis (2.5% agarose gel in 1x TBE run at 150V for 30 min).

MNase digestion – M. fervidus

Request a detailed protocol

Frozen pellets of M. fervidus harvested in late exponential and stationary phase were purchased from the Archaeenzentrum in Regensburg, Germany. We then followed the MNase protocol outlined above with the following modifications: first,~0.5 g of frozen pellet were thawed and re-suspended in 9 ml of 1x PBS before fixation. Second, due to differences in cell wall composition between M. fervidus and E. coli, the lysozyme digestion step was replaced by mechanical disruption with a French press: after the wash that follows fixation, the pellet was re-suspended in 20 ml of chilled 1x PBS, the cell suspension passaged twice through a TS Series French press (Constant Systems) at 15kpsi and then spun down at 4000 rpm for 15 min at 4°C before proceeding with cell lysis. Finally, the extracted chromatin was re-suspended in 250 μl of +CA buffer (instead of 500 μl). Digestion, fragment purification, sequencing and analysis were performed as for E.coli but with a micrococcal nuclease concentration of 100 U/ml.

MNase digest sequencing

Request a detailed protocol

Size distributions of the DNA fragments retrieved by MNase digestion of strains Ec-EV, Ec-hmfA, Ec-hmfB and M. fervidus were analysed with an Agilent Bioanalyser DNA1000 chip. For each of these strains, three biological replicates were selected for sequencing. Twenty nanograms per sample were used for library construction with the NEBNext Ultra II DNA Library Prep Kit for Illumina and NEBNext Multiplex Oligos for Illumina. The output was then taken to 10 PCR cycles and purified using a 1.8x Ampure XP bead clean-up kit. Libraries were quantified via Qubit and quality assessment carried out on an Agilent Bioanalyser DNA 1000 chip. Libraries were then sequenced on an Illumina MiSeq sequencer using single-end 160 bp reads.

Read processing

Request a detailed protocol

Reads were trimmed using Trimmomatic-0.35 (single-end mode, ILLUMINACLIP:2:30:10) to remove adapter sequences. This did not remove short remnant adapter sequences so that we submitted reads to a further round of trimming using Trimgalore v0.4.1 with default parameters. Trimmed reads were aligned, as appropriate, to either the Escherichia coli K-12 MG1655 genome (NC_000913.3) or the M. fervidus DSM2088 genome (NC_014658.1) using Bowtie2 (Langmead and Salzberg, 2012). Only uniquely mapping reads were retained for further analysis. Per-base coverage statistics were computed using the genomeCoverageBed function in the bedtools2 suite (Quinlan and Hall, 2010).

Peak calling

Request a detailed protocol

Nucleosome peaks were called using the NucleR package in R as described previously (Hocher et al., 2019). See Supplementary file 2 for the relevant Fourier parameters.

LASSO modeling

Request a detailed protocol

LASSO modeling was carried out for different footprint size classes (60 ± 5 bp, 90 ± 5 bp, 120 ± 5 bp) using empty vector-normalized coverage. Empty vector coverage was computed across fragment sizes and coverage across the genome uniformly increased by one to enable analysis of zero-coverage regions. K-mer counts (k={1,2,3,4}) were computed using the R seqTools package over windows of three different sizes (61 bp, 91 bp, 121 bp). Subsequent LASSO modeling was then carried out as described previously (Hocher et al., 2019), with models trained on one sixth of the E.coli genome (genomic positions 0–773608) and tested on the remainder of the genome.

Transcriptional start sites

Request a detailed protocol

Experimentally defined transcriptional start sites were obtained from RegulonDB (Salgado et al., 2013) (http://regulondb.ccg.unam.mx/menu/download/datasets/files/High_throughput_transcription_initiation_mapping_with_5_tri_or_monophosphate_enrichment_v3.0.txt). The position inside each broad TSS associated with the most reads (column three in the file above) was defined as the TSS for downstream analysis. Promoter annotations were obtained from the same source (http://regulondb.ccg.unam.mx/menu/download/datasets/files/PromoterSet.txt).

Comparison with other transcriptomes

Request a detailed protocol

All available transcriptomic data corresponding to E.coli K-12 strains were downloaded from the E. coli Gene Expression Database (GenExpdb, https://genexpdb.okstate.edu), which aggregates differential transcriptional responses (increased/decreased mRNA expression computed from pairwise comparisons in different individual studies). Similarity between differential expression in Ec-hmfA versus Ec-hmfAnb and other pairwise comparisons (Figure 7A) was calculated as the dot product of the two differential expression vectors.

RNA extraction and sequencing

Request a detailed protocol

250 μl of culture were harvested from late exponential and stationary phase by centrifugation (15,000 x g at 4°C for 15 min). The supernatant was discarded and the pellet re-suspended in 100 μl of Y1 Buffer (1M Sorbitol, 0.1M EDTA, 1 mg/ml lysozyme, 0.1% β-mercaptoethanol) and incubated at 37°C for 1 hr at 500 rpm. The cell suspension was added to 350 μl of RLT buffer, 250 μl 100% ethanol and loaded onto an RNeasy column from the RNeasy Kit (Qiagen, Germany). RNA was then washed and eluted following the manufacturer’s protocol. Eluted samples were incubated with DNase I (New England Biolabs, MA) for 10 min at 37°C and then cleaned up with a second passage through the RNeasy column (loading, washes and elution according to manufacturer’s instructions). Samples were finally eluted in 30 μl of RNase-free water and RNA quantified with Nanodrop. Quality assessment of the extracted RNA was carried out with an Agilent Bioanalyser RNAnano chip and five replicates per strain/condition were chosen for sequencing.

RNA sequencing

Request a detailed protocol

For each replicate/strain/condition, 1.5 μg of total RNA were depleted of rRNA using the Ribo-Zero rRNA depletion kit (Illumina) and libraries constructed using a TrueSeq Stranded RNA LT Kit (Illumina). After 12 PCR cycles, library quality was assessed with an Agilent Bioanalyser HS-DNA chip and quantified by Qubit. No size selection was carried out and the samples were sequenced on a HiSeq 2500 machine using paired-end 100 bp reads.

Transcriptome analysis

Request a detailed protocol

Using Bowtie2, reads were first aligned to all annotated non-coding RNA genes (rRNA, tRNA, etc.). Reads that mapped to any of these genes were discarded, even if they mapped to more than one location in the genome. We then used Trim Galore v0.4.1 with default parameters to trim adapters and low-quality terminal sequences. Trimmed reads were aligned to the E. coli K-12 MG1655 genome (NC_000913.3) with Bowtie2 (--no-discordant --no-mixed). As a technical aside, we note that, despite the above filtering steps, some of the samples had an unusually low alignment rate (<30%). We found that most of the unaligned reads were perfect matches to rRNA sequences from Bacillus subtilis but not E. coli and had therefore eluded the above filter. As contamination at this scale is unlikely (no bacteria other than E. coli are grown or sequenced in the lab and a plain LB control was added to check for contamination when growing the samples for RNA extraction), we suspect these reads are the result of carrying over RiboZero oligos. The addition of a further round of filtering to discard reads that match non-coding RNA sequences from Bacillus subtilis increased the alignment rate to E. coli index up to ~90%.

By-gene read counts were computed from read alignments using the summarizeOverlaps function (mode=‘Union’, singleEnd = FALSE, ignore.strand = FALSE, fragments = TRUE) from the GenomicAlignments package in BioConductor. Differential gene expression analysis was carried out using DESeq2 (Love et al., 2014). Replicates found to be outliers in principal component analysis and that were subsequently excluded from differential expression analysis are listed in Supplementary file 3.

Microscopy

Request a detailed protocol

Overnight pre-cultures of Ec-EV, Ec-hmfA, and Ec-hmfB were diluted in fresh LB medium plus antibiotic and grown as described above. ~300 μl of culture were harvested by centrifugation (15,000 rpm for 15 min). Pellets were resuspended in 1% FA in PBS and fixed for 10 min at room temperature. Fixating agent was removed by spinning (15,000 rpm for 15 min) and pellets were resuspended in 1 ml PBS. 5 μl of cellular suspension was spread onto an agarose pad, covered in VectaShield containing DAPI (Vector Laboratories) and the excess liquid removed. Slides were imaged using a Manual Leica DMRB with phase contrast and DIC for transmitted light illumination. For quantification, images from three independent experiments were analysed with MicrobeJ (Ducret et al., 2016) to perform automatic cell detection and size measurements. MicrobeJ image profiles were manually curated to remove background and wrongly detected debris. For each sample/condition, measurements of cell length and area are derived from averages across ~10 independent pictures. Average DAPI profiles and associated cell contours were determined and visualized using Coli-Inspector (https://sils.fnwi.uva.nl/bcb/objectj/examples/Coli-Inspector/Coli-Inspector-MD/coli-inspector.html).

NAPs binding regions

Request a detailed protocol

Genomic regions bound by Fis and H-NS were obtained from Kahramanoglou et al. (2011), regions bound by IHF from Prieto et al. (2012), and regions bound by Dps from Antipov et al. (2017). Differential histone occupancy was computed between regions bound by a given NAP and the unbound region immediately downstream.

Modeling of HmfB oligomer stability and DNA affinity

Request a detailed protocol

The HMfB dodecamer unit cell was reconstructed from the crystal structure (PDB ID: 5t5k). Dimers were removed sequentially from the structure to build different sized HMfB polymers. We used the AMBER ff14SB forcefield (Maier et al., 2015) and solvated the unit cell with 14 Å of explicit TIP3P water and NaCl. We relaxed the system with 10,000 steps of steepest descent and conjugate gradient energy minimisation, heated the system to 300 K and performed 100 ns of NPT classical molecular dynamics using pmemd with a two fs timestep (Salomon-Ferrer et al., 2013). Binding affinity and stability calculations were performed using MMPBSA (Miller et al., 2012).

Stress assays and NAP deletion mutants

Request a detailed protocol

To minimise the risk of suppressor mutations all cultures involving NAP deletions were inoculated straight from −80°C stocks. 5 mL LB cultures containing the antibiotic necessary for plasmid selection were inoculated in 50 mL tubes at OD = 0.1 from overnight LB cultures that contained antibiotics for both plasmid and mutant selection. After 2 hr, rhamnose was added to a final concentration of 15 mM to induce histone or YFP expression. Two hours after induction, 10 µL of cultures were used to inoculate 200 µL wells in Nunc 96-well microplates (at least three biological replicates, each with three technical replicates). Growth was monitored as described above. For chemical stress assays, novobiocin, rifampicin and H2O2 were used at respective concentrations of 10 µg/mL, 0.31 ng/mL, and 1.25 mM. For UV stress, cells used as inoculum were grown as for chemical stress assay and 106 and 103 dilutions (in 1x PBS) were plated onto plates containing 15 mM rhamnose and kanamycin. Plates were irradiated with 10000µJ using Stratalinker 2400.

Data availability

Request a detailed protocol

Datasets generated for this study have been deposited in the NCBI Gene Expression Omnibus under accession number GSE127680 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE127680).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56

Decision letter

  1. David M Truong
    Reviewing Editor; NYU Langone Health, United States
  2. Jessica K Tyler
    Senior Editor; Weill Cornell Medicine, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

We commend the authors for the detailed responses to the initial decision. The manuscript has been notably improved by these efforts. While the major findings of this paper have largely stayed the same, the authors have discovered new subtleties that strengthen this study. By growing the archaeal histone expressing E. coli strains in additional growth conditions, the authors show that the histones inhibit growth under topological stress. Furthermore, they find that archaeal histones induce expression changes similar to growth arrest and H-NS deletion strains. Finally, the authors have taken pains to carefully address other concerns such as global transcriptional effects, new findings related to archaeal histones, and volumetric analyses of the cells. Thank you for submitting this fascinating study.

Decision letter after peer review:

Thank you for submitting your article "Chromatinization of Escherichia coli with archaeal histones" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Jessica Tyler as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

Rojec and colleagues present a simple but rather provocative experiment: would expressing archaeal histones from Methanothermus fervidus in the histone-less bacterium Escherichia coli – a truly blank-slate for histones – recapitulate key features of chromatin-based regulation of transcription? The answer is both yes and no. Archaeal histone proteins have received renewed interest in recent years and important advances have been made in terms of understanding nucleosomal structures both biochemically and within cells. By expressing the archaeal histones hmfA and hmfB from the well-studied M. fervidus, many key characteristics about nucleosome binding were evident. The key findings are: (1) the histone proteins assemble genome-wide into multimeric structures of varying sizes according to known intrinsic sequence preferences; (2) nucleosome displacement during transcription and replication; (3) local transcriptional repression due to nucleosome binding at promoters and gene bodies; (4) modest to inconclusive global effects on transcription; and (5) surprisingly, no major repercussions to cell growth.

The manuscript is well-written and a pleasure to read. It reports a large body of convincing data on the neochromatinization of the E. coli genome with M. fervidus histones. The experiments have been planned with great care. A nice and novel observation is that the size of nucleosomes bound at transcription start sites correlates with repression efficiency. The paper will be intriguing for different subfields because it suggests that early histone-fold like proteins may not have evolved solely to regulate transcription, although the other posited roles remain hypothetical. Nevertheless, the amount of novel information generated in relation to archaeal histone function is limited: the reported results largely confirm earlier observations. We ask that the authors add more discussion as to whether there are additional biological insights about hmfA/B. Furthermore, the manuscript poses questions in the Introduction, which are perhaps, eukaryotic-centric. We ask that the authors reframe these questions as current work on archaeal histones suggests many similarities between bacteria and archaeal DNA organizing proteins. Finally, there was consensus that the authors will need to further justify their assertion that the effect to E. coli "…is remarkably unremarkable" and provide more analyses on the effects to transcription.

Essential revisions:

1) All reviewers agreed that the exponential growth experiments in rich medium (LB) were inadequate to support the statement that "growth of E. coli [with archaeal histones].… is remarkably unremarkable". Rich medium obviates the requirement for many E. coli genes for survival. Further, exponential growth has a relatively "fixed" transcriptome. In fact, the lagging phase (where transcriptional changes occur most) are likely where the nucleosomes have the largest effect. We would like to see a sampling of different types of stress: transcription (e.g., rifampicin), DNA damage (e.g., UV and/or chemical), metabolic (e.g., minimal media, oxidative stress, etc.). This can be done in a reasonable length of time. A more detailed analysis of gene expression, including the data they excluded, may offer clues as to the type of defect, if any, the histone expressing cells may be experiencing. Perhaps they could even include a cell cycle analysis.

2) It is rather striking that the hmfA/B expressing strains have both larger cell size and greater size heterogeneity. This was only briefly mentioned in the Results but is rather striking. The fact that the SOS response is activated suggests that the histones are affecting more processes than suggested. Would growing the cells in conditions that cause DNA stress (UV for example) exacerbate the morphology? Are the cells overall bigger or just narrower and longer (the cell area analysis seems to favor the former)? Could this be due to increased volume of chromatinized DNA? We would be satisfied with a more thorough discussion on this topic but would also like to see whether HfmA/B/nb expressing cells display additional morphological changes. It would also be useful if the authors could present additional information, if available, on cell-to-cell variability in terms of expression of the histone proteins. This might aid in making their point that hmfA/B have little impact on cell physiology.

3) We generally found that the "locally repressive effects" was not robust and needed further exploration. The authors could partially address this by explicitly showing, examining, and categorizing by function the genes that are down- or up-regulated in hmfA/B/nb vs. EV, and determine the corresponding level of nucleosome signal in hmfA/B expressing cells. They reasonably exclude these as potentially generic responses to ectopic protein expression. But is that the case? Do these genes change expression in response to any ectopically expressed proteins or are there residual hmfAnb effects? The idea is to deconvolute whether histones incorporate more readily with already downregulated genes or cause their repression. Could these include genes that enable E. coli to adapt to histones regardless of whether they are bound to DNA? Finally, the correlation between GC content vs. histone occupancy is not formally quantified (only browser tracks are shown). What is the relationship genome-wide? Is it significant? Here we are only asking for additional analyses of the existing data.

4) We ask for additional analyses on bacterial organizational structure in the context of AT/GC compared to endogenous bacterial NAPs. What is conceptually the difference between over-expressing a bacterial DNA organizing protein, one from another bacterium, or an archaeal histone? Why do we not see a phenotype similar to that of E. coli lacking one or more of its own DNA organizing proteins? We would like to see some straightforward estimates (using RT-qPCR) of a few stress-regulated genes under direct control of bacterial NAPs when grown under stress. We also ask whether there is phenotypic complementation of H-NS and HU knockout strains. In the case of HU this would make clear whether the similar structural properties of the two proteins are enough to have complementation. In the case of H-NS it could suggest that invasion of hmfA/B into large H-NS regulated regions indeed has only mild effects on global transcription. Such growth experiments carried out under a few conditions (see 1) should be fairly straightforward. What is the difference between AT-rich non-histone binding regions and AT-rich H-NS binding regions? What is the effect of histones in gene bodies; is some of the repression of spurious intragenic transcription by H-NS taken over by bound archaeal histones?

5) One of the most interesting questions put forward by the authors is 'how easy or hard is it to transition from a system without histones to one where histones are abundant?', but the answers to this question leaves something to be desired. Although the authors use an inducible system (2 h and 16 h), perhaps only subpopulations of the cells accommodate histones? We think it is important to rule out potential adaptive mechanisms especially by looking more deeply into the gene expression changes. It is quite possible that this occurs at the level of single-cells, which is beyond the scope of this study, but perhaps explains some of the variability. We again suggest simple experiments such as determining whether the SOS response is required to tolerate expression of histones or their binding to DNA.

https://doi.org/10.7554/eLife.49038.039

Author response

[…] The amount of novel information generated in relation to archaeal histone function is limited: the reported results largely confirm earlier observations. We ask that the authors add more discussion as to whether there are additional biological insights about hmfA/B.

We share the reviewers’ desire to learn more about the native functions of archaeal histones. Given that we express archaeal histones in a heterologous system, we have been cautious in drawing firm conclusions about native functionality. That said, we do think that our experiments have revealed interesting leads that might help us better understand how the histones of M. fervidus and other archaea operate in their native cellular context. Even though not the focal point of the manuscript, it is worth highlighting two such leads:

First, as highlighted by the editor, we provide some evidence – yet to be reported for any archaeon – that longer oligomers elicit stronger repression. We now also provide evidence from molecular dynamics simulations that longer oligomeric complexes are more stable (Figure 4—figure supplement 3). It is certainly conceivable that archaea make use of this property to dynamically regulate genome accessibility, either by changing absolute histone concentrations (which should lead to longer oligomers on average – see exponential-stationary transition in E. coli in Figure 1C) or by varying the relative levels of different histone paralogs that are more or less likely to form stable longer oligomers.

Second, while our focus in this manuscript is on the question how well a naïve system tolerates histone proteins, we recently used some of the MNase data reported here to understand sequence-dependent oligomer formation in archaea themselves (Hocher et al.,2019 bioRxiv 564930). There, we suggest that oligomer formation is subject to sequence-dependent nucleation-extension dynamics. We show that initial tetramer formation relies on a relatively high affinity site but that, from this nucleation point, oligomers can reach into less conducive sequence space. Importantly, this might enable repression of promoters that – by virtue of their sequence composition – disfavour nucleation. We anticipate that our heterologous E. coli system, which displays similar nucleation-extension dynamics (see Author response image 1, Figure 2 —figure supplement 1 in the revised manuscript), will help to elucidate this further, especially since targeted manipulations (e.g. of promoter sequences) remain impossible or difficult for the majority of archaea. We now highlight these insights in the subsections “Intrinsic sequence preferences govern nucleosome formation along the E. coli genome and “Evidence that nucleosome formation locally represses transcription”.

Author response image 1
The relationship between oligomer formation and the underlying sequence.

(A) The upper half of this figure corresponds to Figure S4b in Hocher et al.2019, bioRxiv: 564930. The data are for M. fervidus native chromatin digests. The figure compares different types of narrow peaks, i.e. peaks principally formed by tetramers. Narrow peaks where longer fragments are common (i.e. extension is likely to occur, central panel) are compared with narrow peaks where longer fragments are rare (i.e. extension is uncommon, right panel). Narrow peaks where extension rarely happens are associated with significantly elevated flanking AT content (see Hocher et al. for quantification, W=A or T, S=G or C). (B) Patterns very similar to those observed in (A) are also evident in E. coli expressing HMfA, supporting the notion that sequence-dependent oligomerization behaviour is an intrinsic property of archaeal histones.

https://doi.org/10.7554/eLife.49038.025

Regarding insights into HMfA versus HmfB in particular (rather than archaeal histones in general) we note that differences between the two paralogs clearly do exist, as evident from MNase profiles (Figure 1C, Figure 2E), the transcriptomic response (Figure 4, 7) and prior work by John Reeve’s group. However, these differences are comparatively subtle. As this issue is not central to the manuscript, we decided not to focus on these differences further. However, our data will allow a more focused exploration of this in the future.

Furthermore, the manuscript poses questions in the Introduction, which are perhaps, eukaryotic-centric. We ask that the authors reframe these questions as current work on archaeal histones suggests many similarities between bacteria and archaeal DNA organizing proteins.

We agree that there are important similarities between bacterial NAPs and archaeal histones. In fact, we stated this forcefully in the Discussion.

“… care should be taken in projecting properties of eukaryotic histones onto those of archaea. In many instances, archaeal histones might be better understood with reference to bacterial NAPs, especially when considering how concentration drives opportunities for oligomerization, cooperativity, and bridging interactions with DNA.”

It is therefore not because we are unaware of these similarities that we pose eukaryotic-centric questions in the Introduction. Rather, the framing is a considered choice. We believe that most readers will approach this manuscript from a eukaryote-based knowledge of histones and histone function. We aim to pick them up from their place of relative comfort and journey towards a better understanding of archaeal histones, ending up with a more nuanced understanding of the nature of archaeal histones – functionally somewhere between eukaryotic histones and bacterial nucleoid-associated proteins.

Finally, there was consensus that the authors will need to further justify their assertion that the effect to E. coli "…is remarkably unremarkable" and provide more analyses on the effects to transcription.

We appreciate that what is and is not remarkable is very much in the eye of the beholder. We have therefore characterized growth in a number of different stress conditions, as detailed below (point #5). We have also added a more rigorous comparative analysis of global and local transcriptional effects. To reflect that our astonishment is necessarily subjective we have rephrased the relevant sentence (subsection “Histone binding is associated with mild phenotypic effects under favourable conditions”). Whether the reviewers and readers share our astonishment is, of course, entirely up to them.

Essential revisions:

1) All reviewers agreed that the exponential growth experiments in rich medium (LB) were inadequate to support the statement that "growth of E. coli [with archaeal histones].… is remarkably unremarkable". Rich medium obviates the requirement for many E. coli genes for survival. Further, exponential growth has a relatively "fixed" transcriptome. In fact, the lagging phase (where transcriptional changes occur most) are likely where the nucleosomes have the largest effect. We would like to see a sampling of different types of stress: transcription (e.g., rifampicin), DNA damage (e.g., UV and/or chemical), metabolic (e.g., minimal media, oxidative stress, etc.). This can be done in a reasonable length of time.

We share the reviewers’ curiosity to see how our histone-expressing strains respond to various stresses. In the revised manuscript we now present data on growth in response to rifampicin (transcriptional stress), H2O2 (oxidative stress), UV (DNA damage), and novobiocin treatment (supercoiling stress). We added the latter condition to explore whether histones, known to constrain negative supercoils, exacerbate the effects of gyrase inhibition. To stress-test the cells, we slightly modified the growth protocol, inducing histone expression in exponentially growing cells with rhamnose, growing these cells for a further two hours in rhamnose, and then using them to inoculate a new culture that contains rhamnose as well as the drug of interest (rifampicin etc). In this way, we ensure that stress responses are measured in cells where histones are already established, and that we can adequately capture effects of histone occupancy during lag phase. We then tracked growth/survival of these strains by measuring optical density or, in the case of UV irradiation, colony formation. The results of these experiments are displayed in Figure 6.

Briefly, the salient insights are as follows:

- When pre-induced cells are transferred into new LB medium, there is a slightly prolonged lag phase. However, histone-expressing strains recover quickly and eventually catch up with non-binding/EV control strains.

- Lag phase is extended further in strains treated with rifampicin or H2O2. Again, histone-expressing strains recover well. Under these conditions, histones can therefore be said to have a mild bacteriostatic but no bacteriocidal effect.

- In contrast, the presence of histones matters in the face of UV and novobiocin treatment, where colony formation and growth, respectively, are severely affected. In novobiocin-treated cells, we observe strong morphological changes, as cells become conspicuously elongated (Figure 6). The presence of histones therefore seems to interfere with the cells’ ability to effectively deal with double strand breaks (UV) and altered supercoiling levels.

Based on these results, which are complemented by new insights from transcriptomic analysis (see below), it would be reasonable to surmise that histones do not compromise dynamic responses to stress in general. However, their presence is problematic where the stress specifically relates to sensing or dealing with altered DNA topology or damage. Given what we know about histones, this is not unexpected. We present and discuss these results in the subsection “Systemic transcriptional responses to histone expression in E. coli”.

A more detailed analysis of gene expression, including the data they excluded, may offer clues as to the type of defect, if any, the histone expressing cells may be experiencing. Perhaps they could even include a cell cycle analysis.

We have carried out substantial further analysis of our gene expression data, including uncensored data, and now provide a more in-depth comparison of our strains to other perturbations that had previously been assessed via global transcriptome profiling (https://genexpdb.okstate.edu/). This allows us to place the effect of histone expression into a comparative framework and, by virtue of similarity to other conditions, suggest molecular pathways through which histones affect E. coli physiology.

Figure 7A summarizes the key results of this analysis. It displays the similarity in transcriptional responses between histone expression (Ec-hmfA versus Ec-hmfAnb) and previous experiments that involved differential gene expression analysis.

Comparing differential expression in exponential phase Ec-hmfA (versus Ec-hmfAnb) to other experiments, several interesting patterns emerge:

- The most similar differential expression patterns are observed for other histone-expressing strains, regardless of whether non-binding mutants or the empty vector control is used as a comparator. This provides further reassurance that both EV and the non-binding histone constitute reasonable controls.

- In terms of genes that are down-regulated, the histone-expressing strains exhibit global transcriptional similarities to amino acid starvation, cadmium shock, and heat stress. What unites these conditions is a transient growth arrest phenotype and induction of the stringent response as well as the general stress response (rpoS regulon). Genes in the rpoS regulon are also upregulated in histone-expressing strains (Figure 7B) while flagellar genes are downregulated, a response our strains share with the stringent response.

- In terms of genes that are up-regulated, histone-expressing strains are similar to conditions associated with environments where carbon sources are scarce/exhausted (stationary phase, minimal media) or altered (glucose-to-lactose shift).

These transcriptional signatures of transient growth arrest and/or slowed growth are in line with the growth phenotypes we observe, which are characterized by an extended lag period (see Figure 7B). This suggests that histone expression slows initial growth until cells have had time to adjust. Such adaptive adjustments almost certainly involve the downregulation of gyrases, as previously noted. Histones wrap DNA in negatively constrained supercoils. Gyrases introduce negative (or relax positive) supercoils so counteracting the histone-associated built-up of negative supercoiling by downregulating gyrases makes sense. Downregulation of gyrases then likely renders cells more susceptible to novobiocin (see Figure 7A). This model is further consistent with the observation that expressing CcdB, a gyrase poison, leads to gene expression changes that have some similarity to histone expression (Figure 7A).

We were further intrigued to see that, with regard to upregulated genes, the effect of histone expression is similar to the effects of deleting h-ns.

Finally, and importantly, we note that “similarity in expression” here is strictly relative. The strongest correlation to a dataset outside this study is modest (ρ=0.34). Thus, the transcriptional response to histone expression by no means precisely mimics any of the perturbations discussed above but has a strong unique component.

2) It is rather striking that the hmfA/B expressing strains have both larger cell size and greater size heterogeneity. This was only briefly mentioned in the Results but is rather striking. The fact that the SOS response is activated suggests that the histones are affecting more processes than suggested. Would growing the cells in conditions that cause DNA stress (UV for example) exacerbate the morphology?

We have considered how stress might exacerbate the impact of histone expression on cell morphology in the context of novobiocin treatment. Gyrase inhibition does indeed exacerbate morphology in histone-expressing cells, which become extremely elongated (see Figure 6).

Are the cells overall bigger or just narrower and longer (the cell area analysis seems to favor the former)? Could this be due to increased volume of chromatinized DNA? We would be satisfied with a more thorough discussion on this topic but would also like to see whether HfmA/B/nb expressing cells display additional morphological changes.

We agree that it is interesting to explore changes to nucleoid morphology and nucleoid/cytoplasm ratio in histone-expressing E. coli cells. As a potential starting point for future more in-depth analysis, we therefore decided to carry out a quantitative analysis of the relationship between nucleoid and cell volume based on DAPI-stained cells. As evident from Figure 5—figure supplement 1, we find no obvious changes to the nucleoid/cytoplasm ratio in histone-expressing cells and therefore no evidence for either compaction or decompaction. Ec-hmfA/B have a somewhat larger relative nucleoid volume than Ec-EV, but are indistinguishable from WT, suggesting that this is an Ec-EV-specific effect.

It would also be useful if the authors could present additional information, if available, on cell-to-cell variability in terms of expression of the histone proteins. This might aid in making their point that hmfA/B have little impact on cell physiology.

We initially considered tagging the histones with a fluorescent reporter, which would have allowed direct quantification. However, prior attempts at tagging by others were abandoned as even small tags were found to affect DNA binding (Sandman, pers. comm.). Tagging is much easier in eukaryotic histones, where the tag can be attached to the long flexible tails that are absent from archaeal histones. As an imperfect alternative, we sought to approximate heterogeneity by monitoring cell-to-cell variability of YFP expressed from the same promoter, either with or without HMfA simultaneously expressed (from a different plasmid but the same promoter) in the same cell. What we can conclude from Author response image 2 is somewhat limited but, to us, two features stand out: first, YFP is well-expressed in the majority of cells that also express HMfA. These estimates should be conservative given that, when HMfA is in play, both YFP and HMfA compete for transcriptional/translational resources. Second, there is an enrichment of cells with low YFP expression when HmfA is overexpressed. This is consistent with translational arrest in a subset of cells, in line with the transient growth arrest/stringent response phenotype we observe.

Author response image 2
YFP expression measured in cells with or without concurrent expression of HmfA.
https://doi.org/10.7554/eLife.49038.026

3) We generally found that the "locally repressive effects" was not robust and needed further exploration. The authors could partially address this by explicitly showing, examining, and categorizing by function the genes that are down- or up-regulated in hmfA/B/nb vs. EV, and determine the corresponding level of nucleosome signal in hmfA/B expressing cells. They reasonably exclude these as potentially generic responses to ectopic protein expression. But is that the case? Do these genes change expression in response to any ectopically expressed proteins or are there residual hmfAnb effects? The idea is to deconvolute whether histones incorporate more readily with already downregulated genes or cause their repression.

We think the reviewers make several distinct points here. Below, we have tried to disaggregate these points as we understand them.

a) The reviewers would like to see additional information to support the finding of locally repressive effects. They reasonably suggest that an alternative explanation might be true, namely that cause and effect are reversed, i.e. increased histone occupancy follows (rather than elicits) downregulation. There are two key pieces of evidence that strongly argue against this interpretation. First, as now highlighted in Figure 3F, unlike sometimes observed in eukaryotes, there is no negative global relationship between mRNA levels (as measured in Ec-EV to approximate the pre-histone state) and histone occupancy in our system. In fact, the relationship is slightly positive (ρ≈0.13-0.18, depending on histone type and growth phase). Thus, lower expression per se is not associated with higher histone occupancy. Second, we showed previously (Figure 4—figure supplement 2) that the relationship between histone occupancy and expression, while present at the promoter, breaks down when considering a control window just downstream of the promoter. This is inconsistent with the alternative model where downregulation precedes histone deposition. If histones simply accumulated on the DNA because transcription was reduced or switched off, the same relief from transcriptional perturbation should be observed immediately downstream of the promoter. It is not.

b) The reviewers also wonder, if we understand correctly, whether finding downregulation in the binding strain and the non-binding strain necessarily means that downregulation in the binding strain is really unrelated to DNA-binding. This is a good question. To begin with, we cannot and do not claim that coincident downregulation in both the binding and non-binding strain implies that downregulation in the binding strain is necessarily unrelated to binding. It is perfectly possible, for example, that histones exert a mild repressive effect at gene X and the same gene X in the non-binding histone strain also experiences downregulation, but for different reasons. We chose to focus on genes that are up- or down-regulated only in strains with a DNA-binding histones to minimise type I errors. Our overriding concern here is to isolate the effects of DNA-binding on gene expression. Naturally, in doing so we almost inevitably decrease sensitivity and miss genes at which histones exert a genuine effect via DNA binding and nucleosome formation. In the context of obtaining robust results on the direct effects of nucleosome formation, we believe this is a price worth paying.

c) Finally, it is possible, that non-binding mutant histones still associate with DNA. The unexpectedly poor growth upon novobiocin treatment also hints at that possibility for HmfAnb. However, we have shown experimentally that the non-binding mutants do not form nucleosomes and therefore would have to associate with DNA in a qualitatively distinct manner. Does that mean that overexpression of non-binding histones has impacts that are distinct from expressing other proteins? Yes. Is this a good thing? We would argue that it is, since it provides a more tailored control for other features of histones (their aggregation potential, propensity to interact with other peptides, etc.) that would not be controlled for by expressing a more inert gratuitous protein, and it is this well-matched control that allows us to home in on the specific effects of nucleosome formation along the E. coli chromosome.

We struggled to follow the reviewers’ comment that categorizing genes by function would shed light on the issues above. There is no a priori reason to suspect that direct, locally mediated repressive effects of histone occupancy are divided along functional lines (rather than, say, driven by sequence composition at the promoter). One would expect functional structure for transcriptional responses launched to cope with the stress induced by histones (in terms of both heterologous expression and stress associated with binding), but not in terms of the initial targets of local repression. NB: we did carry out a classic gene set enrichment analysis (GSEA), which showed depletion of genes in COG M (caused by flagellar genes) and COG J (translation), very much in line with the findings from comparative transcriptomic analysis. We did not present those results for two reasons: first, we think the comparative transcriptomic analysis is more insightful. Second, we think that toolkit to carry out GSEA in prokaryotes is fundamentally incomplete. Tools inherited from eukaryotes do not take into account operons. This is problematic because genes cluster into operons by function. As a consequence, changes to the regulation of a single operon get amplified by the number of genes of the same function that operon contains. In light of this problem, it is typically unclear what statistical enrichments in prokaryotic GSEA actually mean.

Could these include genes that enable E. coli to adapt to histones regardless of whether they are bound to DNA?

Yes, absolutely, although it is arguably harder to pinpoint specific genes that react in this manner (i.e. respond specifically to histones rather than any heterologous protein, but respond irrespective of binding). Our experiments were not designed to address this specifically.

Finally, the correlation between GC content vs. histone occupancy is not formally quantified (only browser tracks are shown). What is the relationship genome-wide? Is it significant? Here we are only asking for additional analyses of the existing data.

We already provided a quantitative assessment of the relationship between GC content and histone occupancy in Figure 3, in particular 3D. There are strong correlations in stationary phase between GC content and histone occupancy (normalized by coverage in the empty vector strain to control for MNase-associated biases; ρ=0.64 for hmfA, P<2.2x10-16).

4) We ask for additional analyses on bacterial organizational structure in the context of AT/GC compared to endogenous bacterial NAPs. What is conceptually the difference between over-expressing a bacterial DNA organizing protein, one from another bacterium, or an archaeal histone? Why do we not see a phenotype similar to that of E. coli lacking one or more of its own DNA organizing proteins? We would like to see some straightforward estimates (using RT-qPCR) of a few stress-regulated genes under direct control of bacterial NAPs when grown under stress. We also ask whether there is phenotypic complementation of H-NS and HU knockout strains. In the case of HU this would make clear whether the similar structural properties of the two proteins are enough to have complementation. In the case of H-NS it could suggest that invasion of hmfA/B into large H-NS regulated regions indeed has only mild effects on global transcription. Such growth experiments carried out under a few conditions (see 1) should be fairly straightforward. What is the difference between AT-rich non-histone binding regions and AT-rich H-NS binding regions? What is the effect of histones in gene bodies; is some of the repression of spurious intragenic transcription by H-NS taken over by bound archaeal histones?

We agree with the reviewers that it would be interesting to understand in greater molecular detail how histone expression impacts nucleoid organization, how histones interact with endogenous NAPs, and whether histone expression poses a challenge distinct from overexpressing one of these NAPs.

We now present several new lines of investigation that shed light on these questions. We identified intriguing similarities in the global transcriptional response between hmfA/B-expressing strains and an h-ns deletion strain (ρ=0.19, P<2.2x10-16). As direct targets are arguably better characterized for H-NS than any other NAP, we focused our investigation on H-NS. We began by comparing gene expression changes in our strain to those in a previously published h-ns deletion strain (same K-12 background, Gawade et al., 2019). Perhaps most notably, genes previously identified as direct H-NS targets (green rectangles in Figure 7C), are amongst the most upregulated genes not only when h-ns is deleted (as one would expect), but also upon HMf expression. This might indicate, as suggested by the reviewers, that histones displace H-NS, but fail to provide similar silencing, leading to de-repression of H-NS target genes. This hypothesis is consistent with our previous analysis (now Figure 7D) showing that histones are not significantly excluded from previously determined H-NS binding sites, implying that they successfully compete for binding at those sites.

We also attempted to look at spurious (intragenic and intergenic) transcription in this dataset. However, we found that the proportions of reads mapping to intergenes or antisense were effectively indistinguishable in the h-ns deletion and the corresponding wildtype strain. In other words, we cannot detect what are considered typical effects of H-NS removal in this dataset. This might be because spurious transcripts are short and/or unstable and therefore not well represented in data from a typical RNA-Seq workflow (compared to TSS mapping as done, for example, by David Grainger’s lab). Irrespective of the underlying reason, this precludes an informed comparison of spurious transcription between this and our data. As the focus of this manuscript is not on H-NS, we did not pursue this further experimentally.

In contrast to Δh-ns, we found no similarities of note when comparing our data to other NAP deletion mutants (ΔhupA/hupB, Δdps, Δfis, all ρ<|0.04|). As far as we are aware, transcriptome-wide data on the effects of NAP overexpression (vis-à-vis deletion) in a comparable genetic background are not available. They would also be intrinsically harder to compare and interpret given that a) effects will inexorably depend on the level to which the protein is overexpressed and b) phenotypes reflect expression burden as well as protein-specific effects unless suitable controls are in place. However, it is interesting to note that strong (>40-fold) overproduction of H-NS has previously been reported to lead to a transient (several-hour) growth arrest after which cells resume growth (McGovern et al.,1994). This situation, which the authors dubbed “artificial stationary phase”, is qualitatively reminiscent of the prolonged lag phase we observe in the presence of HMf (see point #5) and further underlines the transcriptional similarity between histone-expressing strains and E. coli growing in stationary-phase(-like) conditions. Why would histone expression resemble both h-ns deletion and h-ns overexpression? We suggest, based on the chain of events proposed above, that histone displacement of H-NS not only leads to dysregulation of target genes, but – like overexpression – also generates a pool of free H-NS that might go on to bind off-targets (including perhaps AT-rich promoters) or have toxic effects unrelated to DNA binding.

In contrast to H-NS, HU overexpression from the same promoter led to 1.8-fold slower growth but no growth arrest, again underlining the outlier status of H-NS.

Our results, both old and new, indicate that histones likely invade genomic real estate normally occupied by NAPs. Histones also share some key properties with NAPs. Notably, they resemble HU in their ability to constrain negative supercoils. We therefore agree that it is interesting to ask if histones complement NAP deletion mutants. To address this question, we have assayed effects of HMfA expression on growth in a collection of NAP deletion strains. The results are summarized in Figure 8, considering differences in the time to maximum growth rate (an operational proxy for the duration of lag phase) and doubling time at the time of maximum growth as determined by curve fitting (see Materials and methods). We note the following:

- NAP deletions in E. coli are not associated with a strong growth phenotypes, with two exceptions:

fis deletion (Δfis) is associated with a reduced lag phase;

the hupA/hupB deletion (ΔΔHU) strain grows more slowly (increased doubling time) compared to its C600 WT progenitor strain.

- HMfA expression generally leads to an increase in lag phase duration, which is particularly pronounced in Δfis (and – for unknown reasons – in M182 WT)

- HMfA expression is associated with a relatively small but consistent increase in doubling time, but this effect does not appear to be compounded in NAP deletion strains. Interestingly, the growth retardation associated with hupA/hupB deletion and HMfA expression are not additive, suggesting that histone expression might partially alleviate defects associated with the absence of HU.

Note that we use YFP here for historical reasons – we had already carried out some of these assays early on in the project, prior to constructing the non-binding histone control strains. YFP serves as a control for the burden imposed by gratuitous expression.

Finally, we want to comment more generally on whether there are (what the reviewers call) “conceptual differences” between over-expressing a histone or an endogenous or heterologous bacterial NAP. We believe there are differences, but we would not call them conceptual. Rather, differences are grounded in the specific properties of histones vis-à-vis DNA binding proteins found in bacteria as well as in evolutionary context. What are these differences? To begin with, histones are arguably unusual in the way they wrap DNA and in the fact that they bind DNA strongly enough to provide protection from MNase. One might therefore suspect that they provide a stronger barrier than many other NAPs to the molecular machines that need to access DNA. In addition, many well-characterized bacterial NAPs, certainly those in E. coli, prefer binding to AT-rich DNA. This might be because they have an ancestral or extant function in the silencing of genomic invaders, which – for reasons that remain poorly understood – tend to have a GC content lower than the host genome. Histones, in contrast, prefer the GC-richer parts of the genome. This has some interesting implications. For example, as we suggest in the manuscript, histones might naturally have a lesser impact on sequences like promoters that are naturally AT-rich, reducing their impact on transcription. These divergent sequence preferences might also imply that, at low dosage, histones and NAPs such as H-NS might come to occupy different parts of the genome. If dosage is sufficiently high, however, histones, having saturated their preferred binding sites, will start to compete with NAPs for their preferred targets. Of course, these differences are not dogmatic. For example, we recently showed that a HU ortholog in Thermoplasma acidophilum prefers GC-rich sequences and behaves in many respects more like an archaeal histone than a “classic” bacterial HU (Hocher et al.,2019). Arguably the biggest difference, however, is evolutionary context: the absence of molecular machinery that has co-evolved with histone proteins.

5) One of the most interesting questions put forward by the authors is 'how easy or hard is it to transition from a system without histones to one where histones are abundant?', but the answers to this question leaves something to be desired. Although the authors use an inducible system (2 h and 16 h), perhaps only subpopulations of the cells accommodate histones?

The observation that, under standard conditions, histone expression has a mild bacteriostatic but no bacteriocidal effect argues strongly against the possibility that a significant fraction of cells drop out of the population because they cannot accommodate histones. That does not mean, of course, that there is no variability in histone expression and that some cells have an easier time than others, but note that across the population we observe histone:DNA ratios of up to ~0.7:1, so cells with a reduced burden are offset against those where the ratio exceeds 0.7:1.

We think it is important to rule out potential adaptive mechanisms especially by looking more deeply into the gene expression changes. It is quite possible that this occurs at the level of single-cells, which is beyond the scope of this study, but perhaps explains some of the variability. We again suggest simple experiments such as determining whether the SOS response is required to tolerate expression of histones or their binding to DNA.

To understand the role of the SOS response in allowing E. coli cells to adapt to histone expression, we assayed growth in a recA deletion strain expressing histones. Histone expression leads to a significant reduction in growth compared to the control but is by no means lethal (see Figure 7—figure supplement 1). This suggests that the SOS response facilitates adaptation to some aspect of the presence of histones, but is not critical for survival. Consistent with the SOS response being triggered following histone expression, we also show concerted upregulation of the rpoS regulon in our strains (see above).

On a more general note, we did not and do not argue that E. coli does not respond to histone expression in an adaptive manner, i.e. by activating or repressing genes that help it cope with the presence of histones. In fact, we previously highlighted downregulation of gyrases as a likely adaptive response. The key, from an evolutionary point of view, is that E. coli copes through activation of generic stress responses rather than through a mechanism specifically evolved to allow the accommodation of histones. The cells survive without machinery dedicated to dealing with histones or adaptations to the transcriptional or replicative apparatus.

https://doi.org/10.7554/eLife.49038.040

Article and author information

Author details

  1. Maria Rojec

    1. Medical Research Council London Institute of Medical Sciences, London, United Kingdom
    2. Institute of Clinical Sciences, Faculty of Medicine, Imperial College London, London, United Kingdom
    Contribution
    Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—review and editing
    Contributed equally with
    Antoine Hocher
    Competing interests
    No competing interests declared
  2. Antoine Hocher

    1. Medical Research Council London Institute of Medical Sciences, London, United Kingdom
    2. Institute of Clinical Sciences, Faculty of Medicine, Imperial College London, London, United Kingdom
    Contribution
    Formal analysis, Investigation, Visualization, Methodology, Writing—review and editing
    Contributed equally with
    Maria Rojec
    Competing interests
    No competing interests declared
  3. Kathryn M Stevens

    1. Medical Research Council London Institute of Medical Sciences, London, United Kingdom
    2. Institute of Clinical Sciences, Faculty of Medicine, Imperial College London, London, United Kingdom
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
  4. Matthias Merkenschlager

    1. Medical Research Council London Institute of Medical Sciences, London, United Kingdom
    2. Institute of Clinical Sciences, Faculty of Medicine, Imperial College London, London, United Kingdom
    Contribution
    Resources, Supervision, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2889-3288
  5. Tobias Warnecke

    1. Medical Research Council London Institute of Medical Sciences, London, United Kingdom
    2. Institute of Clinical Sciences, Faculty of Medicine, Imperial College London, London, United Kingdom
    Contribution
    Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing
    For correspondence
    tobias.warnecke@imperial.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4936-5428

Funding

Medical Research Council (MC_A658_5TY40)

  • Tobias Warnecke

Engineering and Physical Sciences Research Council (EP/R029407/1)

  • Tobias Warnecke

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Ziwei Liang, Till Bartke, Ben Foster, and Kathleen Sandman for experimental advice, training, and sharing protocols; Finn Werner for his continued support and mentorship; Madan Babu, Ben Lehner, Peter Sarkies and members of the LMS Quantitative Biology section for discussions; Jacob Swadling for help with structure visualizations and modelling, the MRC LMS Genomics and Proteomics facilities for sequencing and mass spectrometry, and David Grainger and Jacques Oberto for NAP deletion strains. This work was supported by Medical Research Council core funding to TW. This project made use of time on UK Tier 2 JADE granted via the UK High-End Computing Consortium for Biomolecular Simulation, HECBioSim, supported by the EPSRC (grant no. EP/R029407/1).

Senior Editor

  1. Jessica K Tyler, Weill Cornell Medicine, United States

Reviewing Editor

  1. David M Truong, NYU Langone Health, United States

Publication history

  1. Received: June 4, 2019
  2. Accepted: November 5, 2019
  3. Accepted Manuscript published: November 6, 2019 (version 1)
  4. Version of Record published: November 20, 2019 (version 2)

Copyright

© 2019, Rojec et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,746
    Page views
  • 472
    Downloads
  • 3
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Cell Biology
    2. Chromosomes and Gene Expression
    Leah Bury et al.
    Research Article Updated

    Although originally thought to be silent chromosomal regions, centromeres are instead actively transcribed. However, the behavior and contributions of centromere-derived RNAs have remained unclear. Here, we used single-molecule fluorescence in-situ hybridization (smFISH) to detect alpha-satellite RNA transcripts in intact human cells. We find that alpha-satellite RNA-smFISH foci levels vary across cell lines and over the cell cycle, but do not remain associated with centromeres, displaying localization consistent with other long non-coding RNAs. Alpha-satellite expression occurs through RNA polymerase II-dependent transcription, but does not require established centromere or cell division components. Instead, our work implicates centromere–nucleolar interactions as repressing alpha-satellite expression. The fraction of nucleolar-localized centromeres inversely correlates with alpha-satellite transcripts levels across cell lines and transcript levels increase substantially when the nucleolus is disrupted. The control of alpha-satellite transcripts by centromere-nucleolar contacts provides a mechanism to modulate centromere transcription and chromatin dynamics across diverse cell states and conditions.

    1. Chromosomes and Gene Expression
    2. Genetics and Genomics
    Yicheng Luo et al.
    Research Article Updated

    Ribosomal RNAs (rRNAs) are essential components of the ribosome and are among the most abundant macromolecules in the cell. To ensure high rRNA level, eukaryotic genomes contain dozens to hundreds of rDNA genes, however, only a fraction of the rRNA genes seems to be active, while others are transcriptionally silent. We found that individual rDNA genes have high level of cell-to-cell heterogeneity in their expression in Drosophila melanogaster. Insertion of heterologous sequences into rDNA leads to repression associated with reduced expression in individual cells and decreased number of cells expressing rDNA with insertions. We found that SUMO (Small Ubiquitin-like Modifier) and SUMO ligase Ubc9 are required for efficient repression of interrupted rDNA units and variable expression of intact rDNA. Disruption of the SUMO pathway abolishes discrimination of interrupted and intact rDNAs and removes cell-to-cell heterogeneity leading to uniformly high expression of individual rDNA in single cells. Our results suggest that the SUMO pathway is responsible for both repression of interrupted units and control of intact rDNA expression.