Archaeal nucleosome positioning in vivo and in vitro is directed by primary sequence motifs

Nucleosome positioning motifs in vivo

The discovery of archaeal histones [14] and most subsequent studies [9, 15], have investigated histones from methanogens for which genetic procedures are not available. We therefore first determined, and confirmed, that archaeal nucleosome positioning in vivo was directed by same sequence motifs in M. thermautotrophicus and T. kodakarensis and then focused on T. kodakarensis, as the experimental system, for which genetic procedures and strains with histone genes deleted were available [13]. As previously documented for M. thermautotrophicus[16], MN digestion of chromatin from T. kodakarensis generated a ladder of discrete-length DNA molecules, with a predominant population of ~90 bp molecule first accumulating, consistent with the length of a DNA molecule required to completely encircle an archaeal histone tetramer core [17, 18]. With further MN digestion, the number of ~90 bp molecules decreased and ~60 bp molecules accumulated (Figure 1a), the length of DNA directly bound by the histone folds of a histone tetramer. Deep-sequencing of the ~60 bp fragments resulted in 7 and 9 million unique reads, respectively, of molecules with lengths ranging from 55 to 65 nucleotides from M. thermautotrophicus and T. kodakarensis (Figure 1b). As established for DNA molecules preferentially packaged into eukaryotic nucleosomes [1, 3, 5–8, 19–21], these sequences were dominated by helical repeats (10 bp) of AA/AT/TA/TT dinucleotides offset by 5 bp from repeats of GG/GC/CG/CC dinucleotides with the centers preferentially filled by GC-rich sequences (Figure 2a and 2b).

In contrast to A/T-rich dinucleotides that, in a 10 bp periodicity, offer flexibility and so facilitate DNA incorporation, poly (dA:dT) tracts are relatively rigid. This deters their incorporation into nucleosomes [1–3, 6, 8, 22, 23] and poly (dA:dT) tracts are significantly underrepresented in DNA incorporated into eukaryotic nucleosomes [1, 6–8, 23–33]. Analyses of the archaeal ~60 bp nucleosomal DNA fragments revealed that A/T-rich oligonucleotides were also excluded from incorporation into archaeal nucleosomes assembled in M. thermautotrophicus and T. kodakarensis. For example, as illustrated in Figures 2c and 2d, when the frequencies of occurrence of each of the 1064 pentamers was determined, all of the 32 pentamers that contain only A and/or T were underrepresented relative to the presence in the genome sequences with ATATA (=TATAT) being the most disfavored pentamer in both Archaea. Oligo A/T-rich sequences are consequently located preferentially in nucleosome-depleted regions. In contrast, all of the 32 pentamers that contain only G and/or C were enriched in the ~60 bp MN-protected fragments generated from both Archaea relative to their abundances in the genome sequences (Figure 2c and 2d).

Exclusion of nucleosomes from intergenic regions

In T. kodakarensis, at least 92% of the genome is coding sequence [34], and most intergenic regions are <100 bp, with many <50 bp. Within these regions, both transcription initiation (TATA-box sequences) and termination (oligo A/T-rich sequences) are directed by A/T-rich sequences [35, 36]. Given compliance with the nucleosome positioning code [5–8, 19, 37, 38], archaeal nucleosome assembly should avoid intergenic regions and this was confirmed by aligning the ~60 bp nucleosomal sequences with the genome sequence. A transcriptome map has not been established for T. kodakarensis but, based on bioinformatic predictions [39], non-transcribed intergenic regions are preferentially depleted of nucleosomes. Genome-wide, there was a substantial underrepresentation of nucleosomes immediately upstream of translation initiation codons (Figure 3a) and, in multigene operons, this was predominantly upstream of the promoter proximal gene. For example, TK1761-TK1762-TK1763 constitutes an operon transcribed divergently from TK1760 [40]. The intergenic region separating TK1760 and TK1761 contains several oligo A/T-rich sequences and had minimal nucleosome occupancy in vivo (Figure 3b). There was no similar region of nucleosome exclusion downstream from the promoter within the TK1761-1763 operon.

Conservation of archaeal nucleosome positioning in vivo and in vitro

Most archaeal histones are single histone folds that form homodimers in solution [9]. However, when mixed with other archaeal histone homodimers, there is rapid and spontaneous reassembly that generates an equilibrium mixture of the homodimers and all possible heterodimers [18]. M. thermautotrophicus has three histones [41] and T. kodakarensis has two histones [34] and, practically, it is impossible to know their homo- versus heterodimer configurations in vivo, a concern that limits the reproducibility in vitro of results obtained in vivo. The two histones in T. kodakarensis, HTkA and HTkB are encoded by TK1413 and TK2289, respectively, and although constructing a strain with both genes deleted proved impossible, T. kodakarensis LC124 (ΔTK1413) and LC125 (ΔTK2289) were constructed [13]. These strains therefore contain only one archaeal histone and, by heterologous expression of TK1413 and TK2289, preparations of recombinant HTkA and HTkB homodimers were obtained. Direct comparisons could therefore be made of the locations of archaeal nucleosomes assembled by only HTkA or HTkB, in vivo and in vitro, on T. kodakarensis genomic DNA. The archaeal nucleosome profiles were very similar in vivo and in vitro throughout the length of the T. kodakarensis genome (Figure 4a). The pattern of nucleosome depletion immediately upstream of translation codons was conserved (Figure 4b) and there was only minimal archaeal nucleosome assembly on the rDNA operon both in vivo and in vitro (Figure 5).

As established for the parental strain (Figure 1a), the ~60 bp fragments of the T. kodakarensis genome protected from MN digestion by only HTkA or HTkB assembly in vivo and in vitro also contained 10 bp helical-periodicity repeats of AA/AT/TA/TT and GG/GC/CG/CC dinucleotides, offset by 5 bp, and pentamers containing only A and/or T were under-represented, and those containing only G and/or C were over-represented, relative to their occurrences in the T. kodakarensis genome (Additional file 1: Figure S1). Together these results confirm that the positions at which HTkA and HTkB assemble to form archaeal nucleosomes are predominantly determined by the T. kodakarensis genome sequence and, as concluded from eukaryotic nucleosome studies [7, 33, 38, 43–46], from an archaeal genome sequence [39], it should be possible to predict where archaeal nucleosomes will preferentially assemble in vivo.

Evolutionary conservation of the nucleosome positioning code

source (eukaryotic versus archaeal), we assembled nucleosomes using eukaryotic histones with archaeal DNA and archaeal histones with eukaryotic DNA. As expected, based on many previous studies, chicken histone octamers bound M. thermoautotrophicus and T. kodakarensis genomic DNAs into nucleosomes that protected ~147 bp DNA fragments from MN digestion and, as in Figure 1a, ~60 bp fragments of yeast genomic DNA were protected from MN digestion by archaeal histone assembly. Sequencing these fragments generated 2 to 5 million unique reads, equating to ~60- to 270-fold coverage per bp, per 60 bp or 147 bp fragment. Analyses of the sequences confirmed that the presence of oscillating dinucleotide repeat patterns and the exclusion of A/T-rich pentamers in all the nucleosome-incorporated DNAs (Figure 6). Aligning the nucleosome profiles confirmed that most sites at which archaeal nucleosomes assembled in vivo were sites at which eukaryotic nucleosomes also assembled preferentially, although less abundantly, in vitro. Chicken histone octamers often formed nucleosomes that encompassed two adjacent sites of preferred archaeal nucleosome assembly (Additional file 2: Figure S2).

Archaeal histone deletion changes transcription and nucleosome positioning

T. kodakarensis LC124 (ΔTK1413) and LC125 (ΔTK2289) exhibit no detectable growth defects but microarray hybridizations revealed that transcripts of 3% to 4% of genes increased or decreased in abundance when compared with their abundances in the parental strain T. kodakarensis TS517 [13]. Comparing the nucleosome profiles in the intergenic regions immediately upstream of these genes in T. kodakarensis TS517 versus LC124 or LC125 revealed clear differences, but no consistent pattern that correlated directly with an increase or decrease in transcript abundance. For example, in T. kodakarensis LC124 that lacks HTkA, transcripts of TK2196 and TK1927 increased in abundance 3.1- and 3.4-fold, respectively, but although there was decreased nucleosome assembly upstream of TK2196, there was increased assembly upstream of TK1927 (Figure 7a). Similarly, in the absence of HTkA, transcripts of TK0166 and TK0982 decreased 5.3- and 3.5-fold, respectively, and there was increased nucleosome assembly upstream of TK0166, but there were both regions with decreased and increased nucleosome assembly upstream of TK0982 (Figure 7b).

Conservation and ancestral origin of the nucleosome positioning code

Whether a sequence will favor or disfavor assemble into a nucleosome can be predicted from the below- or above-average energy requirements needed to distort that sequence into the circular configuration of DNA wrapped around the nucleosome core [19, 47, 48]. To a large extent, these predictions have been confirmed experimentally by the sequences of DNA isolated from eukaryotic nucleosomes and so in the definition of a nucleosome positioning code [5–8, 19, 21, 33, 38, 43]. The results reported here confirm that two basic features of this code, namely a 10 bp periodicity of AA/AT/TA/TT dinucleotides offset by 5 bp from GG/GC/CG/CC dinucleotides, and the exclusion of A/T-rich oligonucleotides also direct archaeal nucleosome assembly in vivo and in vitro. In the structures established for the eukaryotic nucleosome [49, 50], the AA/AT/TA/TT dinucleotide repeats facilitate DNA wrapping as the dinucleotides that most readily accept the distortion needed [48], where the minor groove faces inwards towards the histone octamer. The GG/GC/CG/CC dinucleotide repeats, offset by half a helical turn, facilitate the distortion needed at each site where the minor groove faces outwards and so away from the nucleosome core [3, 49, 50]. To date, despite a significant effort, there is no high resolution structure of an archaeal nucleosome available but, given that the archaeal and eukaryotic histone folds are virtually identical [9, 51] and that DNA-histone interactions are almost entirely mediated by histone fold residues [52], it seems highly likely that the same wrapping mechanism is employed and so DNA distortion is required to assemble archaeal nucleosomes. Given this conservation, the nucleosome positioning code almost certainly predates the Eukaryotes, and co-evolved with histones and the histone-fold based mechanism of DNA compaction early in the archaeal lineage, before the divergence of Euryarchaea, Nanoarchaea and Thaumarchaea[9–11].

Archaeal nucleosome positioning and gene expression

As in Bacteria, many archaeal genes are organized into operons and so are co-transcribed from a single promoter and, in the absence of a nuclear membrane; transcription and translation are coupled [53]. However, archaeal promoters and the transcription machinery are substantially more similar to their eukaryotic than bacterial counterparts. Archaeal transcription initiation requires that both transcription factors and RNA polymerase be recruited to the promoter [35] and archaeal histone binding to promoter regions has been shown, alternatively, to inhibit or to stimulate transcription in vitro [54, 55]. Without a transcriptome, we chose to determine the locations at which HTkA and HTkB assembled into nucleosomes on T. kodakarensis genomic DNA, both in vivo and in vitro, relative to translation start codons. This is a reasonable proxy for transcription initiation sites as most archaeal promoters are located within short intergenic regions and so are close to a start codon. As the results revealed, on a genome wide basis, archaeal nucleosomes are excluded from the DNA immediately upstream of open reading frames, and these nucleosome depleted regions (NDRs) are flanked by regions with above average nucleosome occupancy (Figure 4b). This is very similar to the pattern of nucleosome occupancy established in promoter regions and around the downstream transcription start sites in the yeast genome [1, 3, 26, 56]. The conservation of this nucleosome organization argues strongly for a participatory role in gene expression. Possibly, avoiding nucleosome assembly in the promoter region coupled with the presence of a nucleosome at the transcription start site evolved as a generic system to facilitate pre-transcription complex assembly while preventing immediate transcription initiation. If so, a mechanism would then also be needed to remove the inhibitory nucleosome when transcription should occur. In Eukaryotes, many complexes have evolved that catalyze histone modifications, nucleosome remodeling, repositioning and/or eviction that relieve transcription inhibition [3] but, to date, there is no evidence for archaeal histone modifications nor for archaeal nucleosome remodeling complexes.

Consistent with the presence of an archaeal nucleosome impeding transcription [54], presumably to maximize rRNA synthesis, archaeal nucleosome assembly was strongly deterred by the rDNA operon sequences, in vivo and in vitro, in both T. kodakarensis (Figure 6) and in M. thermautotrophicus (Additional file 3: Figure S3). To test the prediction that the presence of an archaeal nucleosome in a promoter region was sufficient to prevent downstream transcription, we compared the archaeal nucleosomes profiles upstream of genes whose transcription was known to increase or decrease in response to the absence of HTkA or HTkB [13]. There were substantial differences in these profiles, in both the abundance and positioning of archaeal nucleosomes, when compared with the parental strain with both histones (Figure 7) but there was no consistent correlation. An increase or decrease in transcription did not simply result from the absence or presence of an archaeal nucleosome in a promoter region.

Supporting report

The results of a similar and complementary investigation were published [57] while this report was being finalized. Ammar et al. [57] determined the locations of archaeal nucleosomes assembled in vivo in Haloferax volcanii, also a euryarchaeon, but a species with a relatively high 65% G + C content, and one with a single atypical archaeal histone that has two histone folds fused into one polypeptide [9, 15, 58]. Both positioning investigations established that G/C-rich sequences predominate at the center of an archaeal nucleosome but a 10 bp periodicity of AA/AT/TA/TT dinucleotides was not detected in the DNA incorporated into nucleosomes in Hlf. volcanii. In both studies, a NDR flanked by increased nucleosome assembly was documented in intergenic regions but, with a transcriptome available, Ammar et al. [57] were able to define the location of the NDR relative to the sites of transcription initiation rather than, as here, to translation start codons. They did not report confirmatory in vitro positioning studies, but the nucleosome positioning in vivo in Hlf. volcanii was, as established here both in vivo and in vitro, almost certainly directed primarily by the Hlf. volcanii genome sequence.

Isolation of genomic DNA and archaeal nucleosomes assembled in vivo

Cells from exponentially growing cultures of M. thermautotrophicus, T. kodakarensis TS517 (ΔpyrF; Δ trpE::pyrF; Δ TK0664), LC124 (Δ pyrF; ΔtrpE::pyrF; Δ TK0664; Δ TK1413) and LC125 (Δ pyrF; Δ trpE::pyrF; ΔTK0664; ΔTK2289) [13] were harvested by centrifugation, flash frozen and genomic DNA preparations isolated from aliquots of these cells as previously described [36, 40]. The remainder were ruptured by grinding in frozen micrococcal nuclease (MN) buffer [50 mM Tris (pH 8), 1 mM CaCl₂, 100 mM NaCl], and the lysates allowed to thaw at 4°C. Aliquots were incubated with MN (1 U/μl) at 37°C, and the nuclease digestion then terminated, after increasing periods of digestion, by addition of 250 mM EDTA, 1% SDS, 200 mM NaCl. Following incubation with RNase A (10 mg/ml) for 60 min at 42°C, the DNA molecules that remained were purified by phenol:chloroform extraction, concentrated by ethanol precipitation, and separated by electrophoresis through 3.5% NuSieve agarose gels (Fisher Molecular Biology, Trevose, PA) or 6% polyacrylamide gels. Gel fragments that contained DNA molecules with ~60 bp lengths were excised, crushed and the DNA molecules eluted by incubation overnight at 37°C in 300 mM sodium acetate, 1 mM EDTA (pH 8), 0.1% SDS. The DNA molecules were concentrated by ethanol precipitation, and prepared for sequencing (see below).

Archaeal histone gene cloning, expression and purification of recombinant HTkA and HTkB

The genes, TK1413 and TK2289, that encode HTkA and HTkB respectively in T. kodakarensis TS517 [34], were PCR-amplified and cloned into plasmid pQE-80 (Qiagen, Valencia, CA) generating plasmids pTS600 (TK1413) and pTS601 (TK2289) that were transformed into Escherichia coli Rosetta 2 (EMD-Millipore, Billerica, MA). Cultures of the transformants were grown to the late exponential phase in LB medium that contained 100 μg ampicillin/ml and 30 μg chloramphenicol/ml at 37°C, and recombinant HTkA or HTkB synthesis was then induced by adding isopropyl β-D-1-thiogalacto-pyranoside (500 μM final concentration) and continued incubation for 3 h at 37°C. The cells were harvested by centrifugation, resuspended (0.33 g wet cell pellet/ml) in 25 mM Tris–HCl (pH 7), 0.1 mM EDTA, 50 mM NaCl, lysozyme (100 μg/ml) added and the mixtures held ice for 30 min. Phenylmethanesulfonyl fluoride (Sigma, St. Louis, MO) was added (100 μg/ml) and cells were ruptured by repeated passage through a French press. The lysates were clarified by centrifugation at 4°C (60,000 g, 20 min), MgCl₂ (5 mM) and DNase I (40 μg/ml) added, the mixtures incubated for 1 h at 37°C and then at 85°C for 20 min. Following further centrifugation (60,000 g, 30 min, 4°C), the supernatants generated were loaded onto 5 ml Hi-Trap heparin columns (GE Healthcare; Pataskala, OH). Recombinant HTkA and HTkB were eluted by passage of 10 column volumes of linear 50 to 500 mM, and 200 to 700 mM gradients of NaCl, respectively, dissolved in 25 mM Tris–HCl (pH 7). The eluate fractions that contained HTkA or HTkB were identified by Commassie-staining of the proteins in samples of the fractions separated by electrophoresis through 22% (w/v) denaturing polyacrylamide gels. These fractions were combined and the protein solution concentrated (final volume of ~0.5 ml) by centrifugation through a pre-rinsed Vivaspin 6 centrifugal concentrator (5 K molecular weight cut off; Sartorious AG, Bohemia, NY). The solutions were adjusted to contain 600 mM NaCl in 25 mM Tris–HCl (pH 7) and then passaged through Sephacryl S-100 HR 16/40 column (GE Healthcare) at a flow rate of 0.5 ml/min. Fractions that contained HTkA or HTkB, identified by Commassie-blue staining after electrophoresis of aliquots through 22% denaturing polyacrylamide gels, were pooled and concentrated (final volume of ~2 ml) by centrifugation again through pre-rinsed Vivaspin 6 centrifugal concentrators (5 K molecular weight cut off). These proteins solution, >99% purified archaeal histone, were dialyzed against in 25 mM Tris–HCl (pH 7), 500 mM NaCl, 50% (v/v) glycerol, and stored at −20°C.

Purification of eukaryotic histones

Chicken histone octamers were purified from erythrocytes by salt extraction and by hydroxyapatite column chromatography as previously described [63].

Archaeal and eukaryotic nucleosome assembly in vitro

Eukaryotic nucleosomes were assembled in vitro by salt dialysis in 200 μl reaction mixtures that contained 50 μg of genomic DNA and 30 μg of chicken histone octamers [6]. Archaeal nucleosomes were reconstituted by mixing 50 μg of genomic DNA with 30 μg archaeal histone tetramers. The complexes formed were dialyzed into MN digestion buffer, and aliquots containing ~2.5 μg of DNA were incubated with 0.1 U MN/ μl for 5 min at 37°C. The MN digestions were stopped by addition of 125 mM EDTA, 200 mM NaCl, and the DNA molecules remaining were isolated by phenol:chloroform extraction, concentrated by ethanol precipitation and separated by electrophoresis through 6% polyacrylamide or 3.5% NuSieve agarose gels. Gel fragments that contained the ~60 bp, or ~147 bp, DNA molecules protected from MN digestion by incorporation into archaeal or eukaryotic nucleosomes, respectively, were excised and the DNA molecules extracted, purified and prepared for DNA sequencing as described above.

ABI SOLiD sequencing of DNA fragments

The ends of the ~60 bp and ~147 bp DNA fragments were repaired and 5’-phosphorylated by incubation in DNATerminator end-repair kits, as recommended by the manufacturer (Lucigen Corp., Middleton, WI). SOLiD adapters were ligated and the DNA molecules PCR amplified (very low cycle number) and sequenced by using the Applied Biosystems protocol for SOLiD fragment paired-end sequencing [64]. Sequencing generated from 2 to 12 million unique reads which, depending on the experiment, equated to 60- to 800-fold coverage per 60 bp or 147 bp nucleosome footprint.

Analysis of DNA reads generated by pair-end sequencing

We first selected reads of length 55–65 bp (nucleosome of 60 bp lengths) to construct the center-weighted nucleosome occupancy scores. If a read length was odd, a Gaussian weight of exp(−0.5 * (d/10)²) was assigned to a position d bp away from the center of the read for d ≤ 25. If a read length was even, then positions i − 1 and i were treated as the possible nucleosome centers. For example, for a 60 nucleotide sequence i = 31, and so the two potential centers were at positions 30 and 31. Each center in an even read was, in turn, assigned a weight of 0.5 * exp(−0.5 * (d/10)²) for a position d bp away from the center and the values for both positions were then divided by 2. The center-weighted occupancy score for any given position was defined as the aggregation of the weighted scores from all reads. We identified well-defined peaks on the reads occupancy-curve as putative nucleosome centers by controlling the peak height and steepness simultaneously. To generate AA/AT/TA/TT frequency plots, after defining the nucleosome center positions based on the peaks of center-weighted occupancy score, dinucleotide frequency scores were computed as described by Segal et al.[5]. We searched for a sequencing tag of length 60 bp nearest to the peak position in the +/−5 bp region. If no such read existed, we further searched for reads of lengths 61, 59, 62 and 58 bp sequentially within +/− 5 bp region of the peak until the first read was identified. The center of the identified read was treated as the nucleosome center to generate the AA/TT/TA/AT frequency plot. If no such read was identified in the +/−5 bp region, the peak position was treated as the true nucleosome center to generate the alignment. For paired-end MNase sequencing data for 147 bp long nucleosomes, read lengths of 137–157 bp were used. We followed a similar approach as described above and also employed by Brogaard et al. [21, 65] to identify the nucleosome centers.

Analyses of the DNA reads generated by single-end sequencing

For the single-end reads with known start position on the Watson strand, their end positions are unknown. However, since the DNA inserts are mapping nucleosomes, their length must be subject to the constraint of being around one nucleosome repeat length. Thus, if we observe a single-end read on the Watson strand at position i, we could practically assume that its end position should be within a region, say [i + a, i + b], and follows some distribution. For practical purpose, we let a = 51, b = 68. We further assumed that the start and end positions of the DNA inserts are independently distributed around the two edges of the nucleosome they map. Let c_i + 51, …, c_i + 68 be the Crick strand tag numbers in this region. Then the relative frequency defined as $c_{i+k}/{\displaystyle \sum _{j=51}^{68}{c}_{i+j}}$ can be used to estimate the probability of a DNA insert ending at position i + k for k = 51, …, 68. Thus, if we observe w _i single-end tags at position i from the Watson strand, then we could regard that we had observed w _i paired-end tags ending at i + k for k = 51, …, 68 with respective frequency $w_i{c}_{i+k}/{\displaystyle \sum _{j=51}^{68}{c}_{i+j}}$. Likewise, if we observe c _i single-end tags at position i from the Crick strand, we would regard that there were c _i paired-end tags ending at i − k for k = 51, …, 68 with respective frequency $c_i{w}_{i-k}/{\displaystyle \sum _{j=51}^{68}{w}_{i-j}}$. By this calculation the observed data with ${\displaystyle \sum _i(w_i+c_i})$ single-end tags are converted approximately to a pseudo data set consisting of ${\displaystyle \sum _i(w_i+c_i})$ paired-end tags. The approach defined above for paired-end data was then used to define the center-weighted reads occupancy score and the nucleosome centers.

Primary data

The sequences obtained and detailed descriptions of the computational analyses are available [66]. The M. thermoautotrophicus and T. kodakarensis genome coordinates and RefSeq transcript annotations used were from the methTher1[67] and therkoda1[42] genome assemblies available on the Archaeal Genome Browser web site [39].