Media & growth conditions

All Har. hispanica strains were derived from Haloarcula hispanica ATCC33960 type strain. Har. hispanica was routinely grown on rich medium for Haloferax volcanii modified to contain 23% basal salts (YPC23) supplemented with 0.1% glucose (w/v) [49]. During glucose limitation experiments, strains were grown in casamino acid media modified to 23% basal salt concentration (Hh-CA). Briefly, 23% basal salts contain, per liter, 184 g of NaCl, 34.5 g of MgSO₄ · 7H₂O, 23 g of MgCl · 6 H₂O, 5.4 g of KCl, and 15.3 mM Tris HCl (pH 7.5). If glucose was supplemented, it was added to a final concentration of 0.1% (w / v) unless otherwise noted. All media were supplemented with uracil (50 μg/ml) to complement the biosynthetic auxotrophy of the Δ pyrF parent strain unless specified. Other ingredients and media preparation are in line with Allers et. al. [49]. All plates were incubated at 37°C for 8–10 days for single colonies and liquid cultures were cultivated aerobically at 37°C with 250 rpm orbital agitation.

Strains, plasmids, & primers

Deletion and integration plasmids were constructed using isothermal assembly [50]. For growth complementation assays, a heterologous expression vector for Har. hispanica was constructed from pWL502, a pyrF-based expression vector for Haloferax mediterranei [51]. Briefly, a mevinolin resistance cassette was amplified from pNBKO7 and replaced the Hfx. mediterranei pyrF at the SmaI and BamHI sites to yield pAKS83. All plasmid sequences were confirmed via Sanger sequencing and propagated in Escherichia coli NEB5α. Strains, plasmids, and primers used are presented in S1, S2 and S3 Tables, respectively. Gene deletions and chromosomal integrations were performed using two-stage selection and counterselection as described previously [45]. Strains were generated using the spheroplasting transformation method and plated on Hh-CA without supplemental uracil [52]. Resulting colonies were inoculated in 5 ml of YPC23 + glucose and grown for 48 hours and then plated onto Hh-CA + 150 μg/ml 5-Fluoroorotic acid (5-FOA) + glucose for counterselection.

To verify the complete deletion of all copies of trmB in the genome and to check for secondary mutations, genomic DNA was extracted using phenol:chloroform:isoamyl alcohol (25:24:1) followed by ethanol precipitation. TruSeq libraries were prepared and sequenced on Illumina MiSeq by the Center for Genomic and Computational Biology at Duke University (Durham, NC). Reads were aligned to the Har. hispanica ATCC33960 genome (GCF_000223905.1, assembly ID ASM22390v1, accessed 2022-10-19) using breseq with default options [53].

Sequence analysis of Har. hispanica TrmB homologs

To identify sugar-sensing TrmB homologs, reference proteomes were searched for protein sequences similar to TrmB_Hbt and results were filtered to include hits with identical domain architecture. Domain identities and confidence values were confirmed using hmmscan on the Pfam database [54, 55]. Har. hispanica TrmB protein and gene sequences were accessed from NCBI and globally aligned to Halobacterium salinarum NRC-1 VNG1451C using EMBOSS Needle [56].

Quantitative phenotyping

Strains were streaked from freezer stocks for each experiment and incubated for 10 days. Single colonies were inoculated in 5 mL YPC23 + glucose and grown to stationary phase (OD₆₀₀ ∼ 4). Stationary precultures were then collected, washed twice in Hh-CA and diluted into 200 μl fresh Hh-CA with or without 25 mM of additional carbon sources to an initial OD₆₀₀ of 0.05. Growth was measured every 30 minutes at 37°C with continuous shaking for 72–90 hours in Bioscreen C analysis system (Growth Curves USA, Piscataway, NJ). For quantitative analysis, growth curves were blank-adjusted within independent experiments, fitted, and holistic differences across growth phases were summarized using the area under the curve (AUC) [57]. AUC values were averaged across technical replicates, the standard deviation was calculated across biological replicates, and significant differences were evaluated by a two-tailed paired Student’s t-test.

Isolation of uracil prototrophic AKS133 strains

Colonies used for AKS133 RNA-seq experiment that showed pyrF expression were subjected to a quantitative growth assay in Hh-CA media ± uracil ± glucose to test for growth in the absence of uracil. All 10 wells containing AKS133 grew in the absence of uracil. These cultures were inoculated into Hh-CA—uracil + glucose for 24 hours and then plated in medium lacking uracil for individual colonies. Of the ten cultures, single colonies were isolated from 7, from which genomic DNA was extracted. Amplification of pyrF from genomic DNA confirmed that the endogenous deletion was intact in all prototrophic isolates, ruling out gene conversion, and revealed pyrF sequence elsewhere in the genome (primer sequences are listed in S3 Table). We were unable to locate pyrF with multiple attempts at arbitrary PCR [58], but note that there is substantial sequence homology between the vector and the Har. hispanica genome, particularly near the origins of replication.

AKS319 strain construction & verification

To construct AKS319, transformations were carried out as described above, except that all media and plates were supplemented with glucose to alleviate selection. Colonies harboring the deletion vector were passaged twice in YPC23 + glucose (for a total of 96 hours), and then exposed to a higher concentration (250 μg/ml) of 5-FOA to select against pyrF and vector retention. After initial confirmation of genotype by PCR, prior to storage and WGS, 5-FOA-resistant colonies were screened for uracil prototrophic growth for 90 hours in Hh-CA media ± uracil ± glucose. Genomic DNA was extracted and WGS was carried out to confirm no reads mapped to the pyrF or trmB loci.

Chromatin immunoprecipitation

Strains DF60 and AKS155 were streaked from freezer stocks onto YPC23 plates and incubated for 8 days. Independent colonies were used to start two cultures of DF60 and four cultures of AKS155 in 10 mL YPC23. Precultures were grown for 29 hours (OD₆₀₀ ∼ 2) then collected at 6000 rpm for 2 minutes and washed twice with Hh-CA. AKS155 pellets were resuspended in Hh-CA or Hh-CA + glucose and grown to midlog phase (average OD₆₀₀ ∼ 0.34) before crosslinking (S1 File). Samples were processed as described by Wilbanks et al. using anti-HA polyclonal antibody (Abcam catalog #ab9110) to immunoprecipitate cross-linked fragments [59]. Libraries were constructed by the Center for Genomic and Computational Biology at Duke University (Durham, NC) using KAPA Hyper Prep kit and Illumina TruSeq adapters, and the 50 base pair, single-ended libraries were sequenced on an Illumina HiSeq 4000.

ChIP-seq read processing & peak calling

Raw FastQ files were trimmed of adapter sequences with Trim_galore! 0.4.3 and Cutadapt 2.3 and read quality was checked with FastQC 0.11.7. Reads were aligned to the Har. hispanica genome with Bowtie2 2.3.4.3 [60]. Aligned sequence files were then sorted, indexed, and converted to binary format with samtools 1.9 [61]. Before calling the peaks, the fragment length was optimized for each IP and input control sample using ChIPQC 1.30 [62]. Binding peaks were called from sorted bam files using Mosaics 2.32 in R 4.1.2 with the calculated fragment lengths and FDR < 0.01. DiffBind 3.4.11 was used to merge peaks that were shared in at least 3 biological replicates for samples grown without glucose (N = 4) and all samples grown in the presence of glucose (N = 2) [63]. Briefly, binding peaks were merged if they overlapped by at least one base pair, and consensus peaks were then trimmed to 300 base pair width centered around the consensus peak maximum. Samples were RLE normalized. Separately, we used DiffBind to identify peaks shared between glucose-replete and depleted samples that exhibited differential binding. Peaks were visually verified using the trackvieweR package to determine an enrichment cutoff [64]. Then, consensus peaks were annotated using IRanges and GenomicFeatures to identify genomic features overlapping and adjacent to TrmB binding peaks [65]. For the per base enrichment plot, bam files were extended to the estimated average fragment size by ChIPQC (150 base pairs), converted to BEDGRAPH format, and scaled according to sequencing depth before calculating the enrichment ratio and averaging across replicates as described by Grünberger et al. [66]. The fifteen regions identified via peak-calling exactly correspond with regions surpassing log2 fold enrichment of 4.

RNA-isolation & sequencing

Strains DF60 and ASK133 or AKS319 were streaked from freezer stocks onto YPC23 + 0.1% glucose plates and incubated for 10 days. Four single colonies of DF60 and ASK133 or AKS319 were inoculated in 5 mL YPC23 medium with 0.1% glucose and grown aerobically to stationary phase (OD₆₀₀ ∼ 4). Cells were washed twice with Hh-CA medium without glucose and transferred into fresh 50 mL Hh-CA with or without 0.1% glucose at an initial density of OD₆₀₀ ∼ 0.4. After 24 hours, cultures were harvested and RNA extracted using Absolutely RNA Miniprep kit (Agilent Technologies, Santa Clara, CA) followed by additional DNAse treatment with Turbo DNAse (Invitrogen, Waltham, MA). Total RNA was quantified using Agilent Bioanalyzer RNA Nano 6000 chip (Agilent Technologies, Santa Clara, CA). The absence of DNA contamination was determined on 200 ng of RNA in 25 PCR cycles. Ribosomal RNA was removed with the PanArchaea riboPOOL kit according to the manufacturer protocol (siTOOLs Biotech, Germany), and sequencing libraries were constructed with NEBNext UltraII Directional RNA Library Preparation Kit (Illumina, #E7760) as described previously [67]. The fragment size of the libraries was measured using the Agilent Bioanalyzer DNA 1000 chip and then pooled and sequenced on NovaSeq6000 at the Center for Genomic and Computational Biology at Duke University (Durham, NC).

Differential expression analysis & clustering

FastQ files were processed as described for ChIP-seq, except alignment by Bowtie2 was done using pair-end mode. After alignment, counts were calculated using featureCounts requiring complete and concordant alignment of paired reads [68]. Samples were considered outliers and removed prior to differential expression analysis if they had a pairwise correlation lower than R = 0.6 with all other replicates. Differential expression analysis was carried out in DeSeq2 with log2 fold-change and FDR cutoffs of 1 and 0.05, respectively, using an experimental design informed by TrmB model of regulation in Hbt. salinarum [10, 69]. To cluster differentially expressed genes based on expression patterns across strain and condition, normalized counts were mean and variance standardized and subjected to k-means clustering and visualized as described in [70].

Functional enrichment

Annotated and predicted gene functions were accessed from EggNOG 5.1 [71]. Genes adjacent to TrmB binding sites were then tested for functional enrichment relative to the whole genome using a hypergeometric test with Benjamini and Hochberg multiple testing correction.

Motif discovery & scanning

TrmB binding motif was identified using the command line version of MEME Suite 5.5.1. Specifically, sequences under consensus peaks (300 base pairs) were extracted and analyzed with XSTREME for motif discovery, motif enrichment, and motif comparison [72]. Default settings were used except that the maximum motif width was set to 19 base pairs and a first-order Markov background model was generated using non-coding regions of the Har. hispanica genome to control for dinucleotide frequencies. The recognition motif reported for Hbt. salinarum was included for downstream motif enrichment and similarity assessment (see reference [10] supplemental table 5). The motif identified was robust to zero and second-order background models. To identify motif occurrences genome-wide, the Haloarcula hispanica reference genome was scanned using FIMO with a p-value cutoff of 1 × 10⁻⁴.

Four TrmB homologs are encoded in the Har. hispanica genome

Using bioinformatic analysis, we investigated the conservation of TrmB. We detected 619 proteins with identical domain architecture as Hbt. salinarum TrmB across archaeal and bacterial Pfam reference proteomes (S1 File). Of these, almost half were encoded in haloarchaeal genomes (Fig 2A). On average, Haloarchaea genomes (n = 97) contain three TrmB homologs per genome, compared to an average of one homolog outside of haloarchaea (n = 238), indicating lineage-specific expansion of this class of regulators (p-value < 2.2 × 10⁻¹⁶, one-sided Wilcoxon test). Some regulators containing the TrmB DNA-binding domain have been recently shown to function during oxidative stress in haloarchaea [22], but proteins with both a TrmB DNA-binding domain and carbohydrate-binding domain have not been functionally characterized in haloarchaea aside from VNG_1451C in Hbt. salinarum (hereafter referred to as TrmB_Hbt).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Sugar-sensing TrmB homologs.

A) Phylogenetic distribution of TrmB proteins that also contain a carbohydrate-binding domain. B) Multiple sequence alignment of the sugar-binding domain of Har. hispanica TrmB homologs. Red, underlined residues are essential for sugar binding in Tcc. litoralis, conserved residues are in bold underline [73, 74]. Asterisks below the alignment denote identical residues and dots represent similar residues. Organism abbreviations and locus tags are as follows: TLI_TRMB, Tcc. litoralis OCC_03542; PFU_TRMBL1, Pyr. furiosus PF0124; TKO_TGR, Tcc. kodakaraensis TK1769; HBT_TRMB, Hbt. salinarum VNG_1451C. Har. hispanica TrmB homologs are named according to their identity to characterized proteins.

https://doi.org/10.1371/journal.pgen.1011115.g002

To determine the most likely functional ortholog of Hbt. salinarum TrmB (VNG_1451C) in the Har. hispanica genome, we searched the 148 proteins with predicted DNA-binding capabilities. Ten of these putative transcription factors have a TrmB DNA-binding domain and four contain both the TrmB DNA-binding domain and sugar-binding domain (PF01978 and PF11495, respectively): HAH_0923, HAH_1548, HAH_1557, and HAH_2795 (Fig 2B). Multiple sequence alignment revealed that HAH_0923, HAH_1548, and HAH_2795 maintain two of the six active site residues required for sugar-binding in Thermococcus litoralis, while HAH_1557 preserved four of the six residues [73]. Both G320 and E326 residues, which when mutated in Tcc. litoralis drastically reduce the binding affinity for sucrose and maltose, are conserved across the haloarchaeal TrmB proteins analyzed [73]. In contrast, N305, which is specific for maltose but not sucrose binding, is not conserved outside of hyperthermophiles [74]. Of the TrmB homologs in Har. hispanica, HAH_1548 exhibits the highest amino acid sequence identity to TrmB_Hbt, and best agreement with the sugar-binding domain consensus (e-value = 3.3 × 10⁻²⁰, Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Har. hispanica TrmB homologs.

https://doi.org/10.1371/journal.pgen.1011115.t001

In Tcc. litoralis and Tcc. kodakarensis, the trmB homolog is located near genes that encode carbohydrate ABC transporters. However, in Hbt. salinarum and Pyr. furiosus, these genes are absent from the genomic region surrounding trmB [10, 75]. Genomic context analysis revealed that HAH_1548 is upstream of a putative carbohydrate ABC transporter and glucose-1-dehydrogenase. HAH_1557 flanks the same ABC transport system but has a substantially lower identity to TrmB_Hbt (Fig 2B, Table 1). Thus, HAH_1548 was distinguished via sequence alignment and genomic context as the likely functional homolog of TrmB_Hbt (for clarity, hereafter we refer to the HAH_1548 gene as trmB_Har and the translated protein as TrmB_Har).

TrmB_Har is essential for growth in gluconeogenic conditions

To assess whether TrmB_Har plays a physiological role in carbohydrate metabolism and gluconeogenesis, we deleted trmB_Har. Because haloarchaea are polyploid [76], it is possible for copies of the wild-type locus to persist at levels too low to detect via Sanger sequencing [77]. Therefore, the genotypes of all strains in this study were confirmed with short-read whole genome sequencing (WGS). WGS of the trmB_Har deletion strain confirmed it to be free of second-site mutations and chromosomal copies of trmB_Har (S1 Fig, SRA accession PRJNA947196). Notably, DF60, the Δ pyrF parent strain, harbors a previously unreported 2.6 kb deletion disrupting HAH_2675–79, encoding flgA1 and cheW1. This region is absent in all strains derived from the Δ pyrF strain in this study.

We subjected both Δ trmB_Hbt and Δ trmB_Har to a panel of nine ecologically and physiologically relevant carbon sources at an equimolar concentration to facilitate comparisons between species (Fig 3). As expected from previous reports, the Hbt. salinarum Δ trmB strain exhibits a severe growth defect under gluconeogenic conditions [10]. Only glucose and glycerol stimulate growth in the deletion strain (Fig 3A). The slight differences between these results and those reported by Schmid et al. are due to differences in supplemental carbon concentration (389mM and 760mM for glucose and glycerol, respectively, [10]).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. TrmB_Har is essential for gluconeogenic growth.

Proportion of parent growth achieved by (A) Δ trmB_Hbt and (B) Δ trmB_Har with various carbon supplements, as measured by area under the growth curve (AUC). Error bars indicate the standard deviation over a minimum of 3 biological replicates, each in technical triplicate. Insets show the fitted, log-transformed growth curves of Δ trmB strains in each condition. Asterisks indicate the significance of growth difference relative to the parental strain in each respective condition. C) Effect of increasing glucose concentrations on the growth of Har. hispanica Δ trmB and Δ pyrF strains. Solid lines represent the mean of the log-transformed growth curve, and shaded regions depict 95% confidence intervals. For all panels, asterisks indicate significance: * p-value < 0.05; ** < 0.01; **** < 0.0001.

https://doi.org/10.1371/journal.pgen.1011115.g003

Growth assays also revealed a severe growth defect of the Δ trmB_Har strain in Hh-CA (Fig 3B). Without additional carbon sources, amino acids are the sole nitrogen and carbon source in Hh-CA, so cells rely on gluconeogenesis to synthesize all required glucose. Normal growth could be restored with heterologous expression of TrmB_Har (S2 Fig). In contrast to Hbt. salinarum, Δ trmB_Har growth could be restored by the addition of glucose, fructose, sucrose, or glycerol. Xylose stimulated growth to 57% of the parent strain, as measured by the area under the growth curve (AUC), but growth remains depressed relative to the parent strain in the presence of ribose, pyruvate, acetate, and galactose. Accelerated growth of the parent strain in media supplemented with pyruvate and ribose contributes to the lower growth ratios reported for those carbon sources (S2 Fig).

We next titrated the concentration of glucose to identify the minimum amount that restored normal growth (Fig 3C). Relative to parent strain growth in the absence of carbon, Δ trmB_Har was stimulated 0.31, 0.81, 1.42, 1.73, 1.74-fold with no supplemental carbon source, 0.01%, 0.05%, 0.1%, and 0.5% glucose (w/v), respectively. Glucose concentrations above 0.1% did not further stimulate growth, identifying 0.1% glucose, or 5.55 mM, as the minimal glucose concentration needed to rescue growth. These data demonstrate that TrmB is indispensable for growth under gluconeogenic conditions in haloarchaea with various metabolic capabilities, and implicate glucose in the function of TrmB in both Hbt. salinarum and Har. hispanica.

TrmB_Har binds promoters of genes involved in central carbon metabolism

To determine whether TrmB_Har functions as a transcriptional regulator, we located TrmB_Har-DNA interactions with chromatin immunoprecipitation coupled to sequencing (ChIP-Seq) in the presence and absence of glucose. A trmB_Har::hemagglutinin (HA) epitope fusion at the endogenous locus was constructed for these experiments and confirmed via WGS (S1 Fig). There was no difference in the growth of the tagged strain from the parent strain in the absence of glucose, suggesting the C-terminal fusion retained full functionality (S2 Fig).

Consistent with the model of regulation reported in Hbt. salinarum and homologs in hyperthermophiles [10, 17, 78], TrmB_Har primarily binds DNA when glucose is absent (Fig 4A). Under this condition, fifteen reproducible TrmB_Har binding sites, or peaks, were enriched relative to the input control (Table 2, Fig 4B, Methods). Of the 15 peaks, seven exhibited significantly increased binding in the absence of glucose relative to glucose-replete samples (FDR ≤ 0.1). Seven other peaks were exclusively detected in the absence of glucose, and one was observed in all samples regardless of glucose availability (S1 File, S3 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Genome-wide TrmB_Har binding sites.

A) Proposed TrmB mechanism of regulation from Hbt. salinarum and Thermococcus kodakaraensis [10, 19]. B) Per base enrichment of reads across the genome. Samples were normalized to paired input samples and averaged across replicates. Averages were normalized relative to untagged controls and log-transformed. C) Genomic context of TrmB_Har binding sites. Sites were trimmed to 300 base pairs centered around peak summits. Promoter regions are shown in purple, intergenic regions in green, coding regions in yellow, and motif locations in black. Regions are sorted by enrichment, or peak size. D) XSTREME motif analysis of selected regions. The 19-base pair motif identified is similar to the binding motif reported for TrmB_Hbt (p-value = 0.017, TomTom [79]).

https://doi.org/10.1371/journal.pgen.1011115.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Har. hispanica TrmB binding-sites.

https://doi.org/10.1371/journal.pgen.1011115.t002

TrmB_Har binding sites were predominantly located in non-coding regions of the genome (Fig 4C): 41% of the bases in enriched regions were non-coding (p-value < 1 × 10⁻¹⁵, binomial test), despite non-coding sequences comprising only 13% of the Har. hispanica genome. In the absence of experimentally characterized promoters for Har. hispanica up to 250 base pairs of non-coding sequences upstream of translational start sites were considered promoter regions. TrmB_Har binding sites overlap the promoter regions of 20 genes (Table 2). Of those, five are homologous to TrmB_Hbt targets by reciprocal protein blast, all in the canonical EMP gluconeogenic pathway: pyruvate oxidoreductase (encoded by porA/B), phosphoenolpyruvate synthase (ppsA), GAPDHII (gapII), and fructose-bisphosphatase (fbp). Overall, the predicted function of genes near binding sites were significantly enriched for carbohydrate transport and metabolism (adjusted p-value < 7.3 × 10⁻⁵, hypergeometric test). Notably, the second largest peak is located in the promoter of HAH_5129, a sugar major facilitator superfamily (MFS) transporter (PF00083, e-value < 3.0 × 10⁻¹²³), highlighting this gene as a candidate for the primary glucose transporter in Har. hispanica.

Computational de novo motif detection using 300 base pairs around binding summits revealed a 19 base pair partially palindromic binding motif (Fig 4D). The motif is similar to the TrmB recognition sequence reported for Hbt. salinarum (p-value = 0.017, calculated with TomTom [79]), and robust to various background models (S4 Fig). The motif occurs 235 times throughout the genome, and of those, 67 instances occur in the promoter regions of 58 genes (motif occurrences on opposite strands were considered distinct, S1 File). Like genes near peaks, genes with motifs located in promoter regions are enriched for functions in the transport and metabolism of carbohydrates and amino acids (hypergeometric test adjusted p-value < 5.2 × 10⁻⁵ and 0.003, respectively). Together, these data support the model that in the absence of glucose TrmB_Har recognizes a conserved cis-regulatory sequence to bind DNA near genes primarily involved in carbon metabolism.

Transcriptome profiling suggests cryptic integration of deletion vector, resulting in uracil prototrophy in Δ trmB_Har strains

We conducted transcriptome profiling to determine the direction of TrmB-dependent regulation of genes near binding sites and to elucidate TrmB_Har targets in instances where peaks were located over the promoters of divergent genes. Unexpectedly, we observed high pyrF expression even though the pyrF gene (the ura5 homolog) is deleted in the parent strain to enable genetic manipulation [45]. Transcripts mapping to the pyrF locus were not present in any parent samples but were detected in all Δ trmB_Har (AKS133) samples regardless of glucose availability. Subsequent phenotyping and genotyping confirmed stably inherited uracil prototrophy in 7 of 10 independent AKS133 isolates and ruled out stock contamination, vector propagation, and gene conversion at the endogenous locus (S5 Fig; Methods). This points to homologous recombination of the deletion vector, which harbors pyrF, elsewhere in the genome.

We carefully generated a new Δ trmB_Har strain, AKS319, and repeated the experiment. Despite additional safeguards during strain generation and genotype confirmation by WGS, 4 of 8 Δ trmB_Har samples exhibited high pyrF expression (S6 Fig; Methods). Due to this variation, however, we were able to investigate the fitness and transcriptomic consequences of pyrF expression in the AKS319 strain and across Δ trmB_Har strains (S6 Fig). AKS319 exhibits a more extreme growth defect relative to AKS133, which can be mostly, but not completely, ameliorated by glucose supplementation. However, expression of pyrF has little effect on the overall transcriptome: average counts per gene are highly correlated across all Δ trmB_Har samples regardless of pyrF expression both before (S6 Fig) and after batch correction(Fig 5A, S1 File). The repeated and apparently independent instances of vector integration resulting in uracil prototrophy suggests that both deletion strains are likely mixed populations with a small proportion of cells retaining vector sequence. TrmB_Har may be conditionally essential in the DF60 background, explaining this phenomenon.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. TrmB_Har-dependent expression.

A) Principal component analysis of AKS133 and AKS319 samples after batch correction. Samples with transcripts mapping to pyrF are outlined in black. B) Correlation of 32 differentially expressed genes from analyses including all Δ trmB_Har samples (y-axis) and only samples with no detected pyrF expression (x-axis). C) Normalized expression pattern of 32 genes in (B). Yellow indicates elevated expression and purple indicates reduced expression. Genes near peaks are bolded and genes predicted to be co-transcribed are indicated by a black bar. If available, the predicted function of the operon was provided. D) Three and 29 genes have patterns consistent with TrmB-dependent repression and induction, respectively. Light grey lines connect normalized expression values for each transcript across strain and condition.

https://doi.org/10.1371/journal.pgen.1011115.g005

TrmB_Har activates genes involved in gluconeogenesis and tryptophan biosynthesis and represses those encoding glucose uptake

Based on the model of TrmB regulation described in Hbt. salinarum, we reasoned that direct targets of TrmB_Har would exhibit differential expression only when TrmB is present and active, i.e., in the parent strain grown in the absence of glucose (Fig 4A). This assumption was made explicit in the DeSeq2 model used to identify differentially expressed genes [69]. Using this framework, we detected 32 genes that were significantly differentially expressed (FDR < 0.05; LFC > 1) in the parent vs mutant strain across two independent analyses, one including and the other excluding samples with counts mapping to pyrF (Fig 5B, S1 File). Because pyrF expression, when present, is not glucose-dependent, it was not considered significant using this explicit model.

Of the 32 genes, three are significantly up-regulated in Δ trmB_Har relative to the parent strain in the absence of glucose: glutamate synthase (encoded by HAH_0919), a universal stress protein (HAH_5129), and the MFS family sugar transporter (HAH_5130) (Fig 5C and 5D). The remaining 29 genes exhibit the opposite pattern: they are significantly down-regulated in the deletion strain, exhibiting a pattern across strains and conditions consistent with transcriptional activation by TrmB. The predicted products of these 29 genes are enriched for functions in amino acid transport and metabolism (adjusted p-value = 1.085 × 10⁻¹⁰, hypergeometric test) including two tryptophan biosynthesis operons, cysteine synthase, a putative amino acid transporter, and a TRAP transporter (Fig 5D). Genes whose expression is induced also include those predicted to function in gluconeogenesis (gapII and class I fructose-bisphosphate encoded by fbp) and two small hypothetical proteins less than 50 amino acids in length (HAH_0188 and HAH_2805).

Integrating across experiments, nine genes are adjacent to TrmB_Har binding sites containing a palindromic cis-regulatory motif, and are differentially expressed (Table 2, Fig 4D). Taking the ChIP-seq, RNA-seq, and motif evidence together, we conclude that these genes comprise the high-confidence regulon under the direct transcriptional control of TrmB_Har (Fig 5C bolded, Fig 6A). For these targets, we also determined whether the relative TrmB motif distance from the putative TATA box or start codon was predictive of the direction of regulation (S4 Table). The archaeal promoter architecture resembles a simplified version of that found in eukaryotes, including a TATA box located around -26 bp and a BRE element around -33 bp [80, 81]. Generally, motifs were upstream of predicted transcription initiation features, consistent with our observation that TrmB acts primarily as an activator. In contrast, genes repressed by TrmB binding had motifs that overlapped the predicted B recognition elements, though HAH_0188 is a notable exception (S4 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Comparison of TrmB targets in gluconeogenesis in haloarchaea.

Summary of experimental evidence of TrmB regulon in (A) Har. hispanica and (B) Hbt. salinarum. Data for Hbt. salinarum summarized from figures 4 and 6 of Schmid et. al. [10]. C, D: Summary of TrmB targets that encode enzymes in central carbon metabolism for each species. Targets in the high-confidence regulon are indicated with solid, colored arrows. Genes differentially expressed but lacking a nearby binding site are indicated with dashed arrows. Targets that were near binding sites but not differentially expressed are indicated in light blue. Solid gray arrows indicate that specific carbohydrate transport systems are unknown or that genes are present but no enzymatic activity has been detected. Genes predicted to encode the necessary enzymes are labeled (see S8 Fig for normalized expression values). Additional putative pyruvate oxidation systems that are not regulated by TrmB are listed in gray. G6P, glucose-6-phosphate; F6P, fructose-6-phosphate; F1,6P, fructose-1,6-bisphosphate; DHAP, dihydroxyacetone phosphate; DKFP, 6-deoxy-5-ketofructose-1-phosphate; GAP, glyceraldehyde-3-phosphate; 1,3BPG, 1,3-bisphosphoglycerate; PEP, phosphoenolpyruvate; A-CoA, acetyl coenzyme A; KDG, 2-keto-3-deoxygluconate; KDPG, 2-keto-3-deoxy-6-phosphogluconate.

https://doi.org/10.1371/journal.pgen.1011115.g006

Given the small number of TrmB binding sites across the genome, it was surprising that only half of the binding sites were near genes that exhibited differential expression. We wondered if there might be other regulatory mechanisms at play, namely TrmB_Har-dependent regulation of small or antisense RNAs, which have recently been described in other haloarchaeal species [82–85]. Since we required correct strand orientation when generating transcript counts, reads mapping to the non-coding strand would not have been considered in the downstream differential expression analysis. We visualized strand-specific reads near peaks and identified an unannotated transcript approximately 200 base pairs in length near the sole intergenic peak in our data (peak 3, Table 2). This transcript is induced when TrmB is active and contains a predicted promoter sequence upstream (S7 Fig). We also saw evidence for a transcript anti-sense to HAH_0887 that appears to be induced when TrmB is active, and that both genes encoding putative small proteins (HAH_0188 and HAH_2805) appear to have extended 3’ UTRs (S7 Fig). We did not find any evidence for unannotated transcripts for peaks near HAH_1010, HAH_1264, HAH_3039, or HAH_4332. Unannotated transcripts require validation and further characterization to be considered direct targets of TrmB_Har, but support the hypothesis that TFs may regulate the expression of regulatory RNAs in haloarchaea, as has been reported for methanogens [86].