To determine the essentiality of the six tbp and seven tfb genes in Halobacterium sp. NRC-1 and study their involvement in gene expression, we attempted to individually knock out all 13 genes using the Δura3 system 26. For this work, ~500 bp of the 5' and 3' flanking sequence and the first four and last four codons of each of the 13 tbp and tfb genes were cloned into the pBB400 or pMPK408 suicide vector (see additional files 3 and 4), and transformed into the Δura3 mutant of wild-type Halobacterium sp. NRC-1. Integrant strains were selected on uracil dropout media and excisants were selected on agar plates containing 5-fluoroorotic acid (5-FOA) 313842. PCR amplification was conducted on DNA template from 20 to 70 Foa^R colonies using flanking primers to screen for each gene deletion. Knockouts were obtained for six genes: tbpC, tbpD, tbpF, tfbA, tfbC, and tfbG, indicating that these genes are non-essential under standard laboratory growth conditions (Figures 1 and 2, Table 1).

For the remaining seven genes, for which deletants were not obtained, we assayed for the presence of knockout alleles using populations of 1,000 Foa^R colonies batchwise. Control experiments showed that our detection limit for deletants was one knockout allele for every 100–500 wild-type alleles (data not shown). For all of the tbp and tfb genes for which we did not obtain knockouts by the initial screens (tbpABE and tfbBDEF), bands corresponding only to the wild-type alleles, but not the deletion alleles were observed. This indicated that these genes were not deleted, even at a frequency of 0.1%, which is consistent with the requirement of these genes for cell viability under standard laboratory growth conditions (Figures 1 and 2).

To test the hypothesis of a novel regulatory system requiring the recognition of different promoters by specific TBP-TFB combinations operating in Halobacterium sp. NRC-1, we compared the transcriptomes of two mutants, ΔtbpD and ΔtfbA (see additional files 1 and 2). We used custom DNA microarrays fabricated using Agilent inkjet technology with oligonucleotides designed with the program OligoPicker. The arrays contained 10,807 60-mer nucleotide features representing 2,677 open reading frames (ORFs) 4142. Each knockout mutant was compared to the parent Δura3 strain and both biological and technical replicates were performed. A significant change in transcript abundance was defined as at least 1.5-fold change (Figures 3 and 4).

Of the 2,677 ORFs on the array, representing 99.9% of the Halobacterium sp. NRC-1 genes, 272 (10%) showed lower transcript levels and 141 (5.3%) showed higher transcript levels in the ΔtbpD strain while 304 (11.4%) showed lower transcript levels and 178 (6.6%) showed higher transcript levels in the ΔtfbA strain (see additional files 1 and 2). Strikingly, a large fraction of the transcriptionally affected genes, 262 genes with less transcript and 101 with more transcript levels were changed by 1.5-fold or greater magnitude in both knockout strains. A total of 96.3 and 86.2% of the genes found to have lower transcript levels in either the ΔtbpD and ΔtfbA strains, respectively, corresponded to lower transcript levels in the other. This similar depression in transcript level of a large number of genes suggests that the great majority of genes requiring the TbpD factor for transcription also requires the TfbA factor, and vice versa (Figures 3, 4, 5, and Table 2).

Genes where expression was changed by 1.5-fold or greater in either the ΔtbpD or ΔtfbA mutants were grouped into one of 13 functional categories (Table 2). In all cases, the largest groups were the unknowns, accounting for 30 to 45% of all regulated genes. Interestingly, the percentage of genes co-regulated on each replicon is roughly the same for the ΔtbpD and ΔtfbA microarray. For the genes whose transcripts are decreased in both mutant strains, 85% are on the chromosome, 3% are on pNRC100, and 11% are on pNRC200. For genes whose transcripts are increased in the mutant strains, 87% are on the chromosome, 6% are on pNRC100, 6% are on pNRC200. This suggests that TbpD, whose gene is on pNRC100, is able to pair with TfbA, whose gene in on the chromosome, to coordinate regulation of genes all over the genome but predominately in the chromosome. The location of the regulated genes was checked and found to be distributed essentially randomly over the entire genome.

As expected, the expression of the ΔtbpD or ΔtfbA mutants was significantly reduced compared to the parent strain, 12.4-fold and 10.8-fold respectively (Table 3). Further, tfbA was significantly down regulated (1.5-fold) in the ΔtbpD strain. Transcription of the other tbp and tfb genes was not significantly changed, except tfbF in the ΔtbpD and ΔtfbA strains and tbpF in the ΔtfbA strain. Further, the transcripts of several important enzymes such as triosephosphate isomerase (tpiA), chitinases (chi and chi2), chemotaxis proteins (cheC2), gas vesicle gene cluster (gvpA-L), small heat shock protein (hsp1), and the thermosome (cctA) are all down regulated in both mutants (see additional files 1, 2 and 5).

Analysis of the microarray data showed transcriptional changes in several gene clusters, indicating that they are responsive to TbpD and TfbA: vng1–3, vng80–84 (moe genes), vng161–162 (glutamate dehydrogenase and medium-chain acyl-CoA ligase), vng811–813 (spermidine/putrescine-binding protein), vng814–817 (quinine oxidoreductase and chitinases), vng1238–1240 (amino acid transporter yhdG), vng1314–1315, vng1340–1342 (3-oxoacyl-reductase and N5, N10-methylenetretrahydromethanopterin reductases), vng1550–1566 (cobalamin biosynthesis), vng1632-1365 (cobalt transport), vng1801–1802 (hsp1), vng1972–1973 (chains of glycerol-3-phosphate dehydrogenase), vng2217–2220 (pyruvate dehydrogenase, dihydrolipoamide S-acetyltransferase, dihydrolipoamide dehydrogenase, prephenate dehydratase), vng2377–2378 (nitrite/nitrate reduction and copper transport), vng5028–5033 (gas vesicle cluster A), vng6229–6247 (gas vesicle cluster B).

It was evident from the microarray data that the gene for one small heat shock protein (hsp1) and the thermosome (cctA) had reduced transcript levels in both the ΔtbpD and ΔtfbA mutant strains (-10.3 and -3.1-fold respectively for ΔtbpD and -15.8 and -2.2-fold respectively for ΔtfbA). Since a reduction in small heat shock proteins has been previously shown to increase the amount of aggregated cytosolic protein 44, we hypothesized that both strains (ΔtbpD and ΔtfbA) would have slower growth rates at elevated temperatures (i.e. above 42°C). Doubling times were measured for the parent, ΔtbpD, and ΔtfbA strains to gauge the growth effects at elevated temperatures. Growth curves at 37°C, calculated from the line of least squares for the growth of each strain, indicated that all three strains had doubling times of 13 to 13.5 hours (data not shown). However, growth curves at 49°C showed that the ΔtbpD and ΔtfbA strains grew more slowly than the Δura3 strain. Doubling times were 14.4 ± 0.5 for Δura3, 16.5 ± 0.7 for ΔtbpD and 17.3 ± 1.1 hours for ΔtfbA (data not shown).

To further confirm the heat sensitive phenotype of both strains, we tested the ability of the ΔtbpD and ΔtfbA strains to survive at elevated temperatures. Incubating the Δura3 strain for one hour at a sublethal temperature, 49°C, before exposure to a lethal temperature, 56°C, resulted in protection by a classic heat shock response that could be observed after about two hours, with greater survival (89% survival) compared to direct exposure to 56°C (62% survival) 45. In contrast, the ΔtbpD and ΔtfbA strains showed little or no protection from preincubation at 49°C versus direct exposure to 56°C (46 versus 51% for ΔtbpD and 58 versus 67% for ΔtfbA). These experiments corroborated the results from analysis of growth rates, confirming that the ΔtbpD and ΔtfbA strains have a reduced ability to grow and survive at higher temperatures (Table 4). The observed reduction in growth rate and decrease in survivability at increased temperatures in the mutants is consistent with a role for the TbpD and TfbA transcription factors in protection of cells to elevated temperature.

In order to examine the evolutionary relationships of the haloarchaeal TBP and TFB proteins, sequences from Halobacterium sp. NRC-1, Haloarcula marismortui, Haloferax volcanii, Haloquadratum walsbyi, Natronomonas pharaonis, and Pyrococcus woesei, as an outgroup, were used to generate alignments using the ClustalX program and produce phylogenetic trees with the core regions using the PAUP* program. The resulting trees are shown in Figure 6. For the TBP tree, the largest family consists of TbpE from Halobacterium sp. NRC-1 and a single member from each of the other four haloarchaea, including the sole TBP from H. marismortui and N. pharaonis. The other families contained up to three homologs of the canonical TBP protein, with representatives from NRC-1 in two clades, with at least one member being dispensable by gene deletion (asterisks in Figure 6A).

The situation was significantly more complex for TFB proteins, which were present in greater copy numbers (7–9) in all five haloarchaea (Figure 6B). The largest family of TFBs contained three members from Halobacterium sp. NRC-1, which did not show knockouts (TfbB, TfbD, and TfbF), and between two and five for each of the other haloarchaea. Another large family contained two members from Halobacterium sp. NRC-1 (TfbC and TfbG), both of which were found to be dispensable by knockouts, but each of the other haloarchaea also contained one or more members. A third family contained one member from each of the five haloarchaea, including TfbA from Halobacterium sp. NRC-1 and a fourth family contained TfbE from Halobacterium sp. NRC-1, for which we were unable to isolate a knockout, and H. volcanii (four members) and H. marismortui (one member) homologs.

Escherichia coli strain DH5α was used for construction and propagation of plasmids. Plasmid pBB400 or pMPK408, containing the ura3 and bla genes, was used as the vector for construction of gene knockouts 26. Knockout constructs were created by first cloning, into the SmaI site, a PCR amplified fragment containing the coding region of the target gene and ~500 bp of sequence upstream and downstream (see additional file 3). For construction of gene deletion plasmids, inverse primer pairs hybridizing near the start and stop codons of the target genes, including several codons, were used for PCR amplification and the linear product was ligated to circularize (see additional file 4).

A uracil auxotroph Halobacterium strain (Δura3) was used as the host for construction of the tbp and tfb knockouts 31. The Δura3 strain was grown in CM⁺ media supplemented with 5-FOA and competent cells were obtained through the addition of EDTA and PEG, as previously described 24. After transformation with knockout plasmids, integrants were selected on agar plates containing uracil drop-out media. DNA was purified from individual colonies and integrant strains confirmed by PCR amplification of the bla gene. Excisants were selected on CM⁺ media with 0.25 mg/ml 5-FOA. DNA was purified from either individual colonies or by collecting ~1,000 colonies from a CM⁺-5-FOA plate and screened by PCR amplification using upstream and downstream primers 3142 (see additional files 3 and 4).

The Δura3, Δura3ΔtbpD, and Δura3ΔtfbA Halobacterium strains were aerobically grown in batch cultures, shaken in a New Brunswick G25 rotary shaker at 220 rpm, to exponential phase (OD₆₀₀ = 1) at 42°C. Before harvesting the cells, cultures were swirled in an ice-water bath to rapidly cool the cultures and "freeze" the RNA profile. Total RNA was isolated using the Agilent total RNA isolation mini kit (Agilent Technologies, Palo Alto, CA) and then treated with DNase. cDNA was prepared with equal amounts of RNA from control and experimental samples and then fluorescently labeled with Cy3-dCTP and Cy5-d-CTP. Concentrations of RNA and cDNA were measured using a Nanodrop (ND-1000 spectrophotometer). Incorporation of the Cy-3 and Cy-5 labels was checked via gel electrophoresis and scanning on a GE Typhoon fluorescence scanner. Both biological replicates (independent cultures) and technical replicates (dye swap) were performed for each paring (i.e. knockout and Δura3 strain) 42. The same Δura3 culture was used as the control in all arrays. Washing and hybridization of the arrays was performed as recommended by Agilent and previously described 41. Slides were scanned for Cy-3 and Cy-5 signals with an Agilent DNA-microarray scanner.

Oligonucleotide arrays, in situ synthesized using ink-jet technology, were used for transcriptome analysis of the strains 56. Oligomer (60-mer) probes have been designed for 2,677 (99.9%) ORFs utilizing OligoPicker 57. These microarrays were thoroughly tested for linearity of response and statistical significance in related studies of Halobacterium sp. NRC-1 4142. Signal intensities with a dynamic range in excess of three orders of magnitude were found allowing simultaneous analysis of low- and high-intensity features. Probe signals were extracted and initial analysis was done with the Agilent Feature Extraction Software, where signal from each channel was normalized using the LOWESS algorithm 58 to remove intensity-dependent effects within the calculated values. The data was also log₂ transformed and parsed using an Excel script 43. Genes showing greater than 1.5 fold change in transcript abundance in at least two of the replicates with an illuminant intensity greater than 7 and less than 14 were selected for further analysis 59.

Growth curves were measured for 25 ml cultures grown in CM⁺ in Klett flasks, starting with an initial Klett reading of 20, grown at either 37 or 49°C in an Innova 4230 rotary shaker shaken at 220 rpm. Measurements were taken for four cultures at 37°C and eight cultures at 49°C per knockout until late log phase on a Klett-Summerson Photoelectric Colorimeter and doubling times derived by calculating the line of least squares from 20–45 hours of growth post inoculation (exponential phase). The effects of heat killing were also measured for each strain. For this, cultures were grown at 42°C and then incubated at 49°C for 1 hour then shifted to 56°C for 2 hours or 56°C for 3 hours. Aliquots were taken from cultures at all temperatures at one hour intervals and plated onto CM⁺ media and allowed to grow at 37°C.

The amino acid sequences of the TBPs and TFBs from Halobacterium sp. NRC-1 (AAC82815, AAC82955, AAC82884, AAC82898, NP_280885, NP_395899, NP_280839, NP_279732, NP_395840, NP_279833, NP_395867, NP_279414, NP_279370), Haloarcula marismortui (YP_135388, YP_135994, YP_138046, YP_136462, YP_134372, YP_136614, YP_134329, YP_135787, YP_137236), Haloferax volcanii (http://archaea.ucsc.edu [UCSC archaeal genome browser]), Haloquadratum walsbyi (CAJ53507, CAJ52639, CAJ53741, CAJ53505, CAJ52683, CAJ52676, CAJ51995, CAJ51817, CAJ51753, CAJ51356, CAJ51283), Natronomonas pharaonis (YP_326192, YP_326499, YP_326761, YP_326774, YP_327502, YP_330892, YP_331136, YP_331271, YP_331316), and Pyrococcus woesei (U10285, P61999) were downloaded from NCBI and sequences were aligned with ClustalX (Version 1.81) for Windows XP. The core regions of the molecules were used in distance analysis with the neighbor-joining algorithm, with 10,000 bootstrap replicates, performed using PAUP* in the GCG Wisconsin Package (version 10) running on a Silicon Graphics workstation. Core regions were defined as those residues aligning with amino acids 3–293 of the TBP and 101–300 of the TFB from P. woesei.