Originally discovered in 1983 [1], heliobacteria have been a source of great interest, as they are the only known phototrophic group within the Firmicutes [2]. Heliobacteria also possess the simplest known photosynthetic apparatus. Comprised of a type I homodimeric reaction center (RC) in the cytoplasmic membrane, this photosynthetic apparatus uses a unique pigment, bacteriochlorophyll g (BChl g) [3]. Unlike other phototrophs, heliobacteria lack a peripheral antenna system. They are not autotrophic and must obtain carbon from organic molecules [4,5], but are capable of robust nitrogen fixation [6]. Like many other bacteria in the order Clostridiales, heliobacteria can grow fermentatively in the dark using simple organic acids, such as pyruvate [7]. In the light, they use the same carbon sources and grow at a significantly faster rate (“photoheterotrophic growth”). These characteristics identify heliobacteria as distinct among the seven photosynthetic bacterial lineages. Several heliobacterial species have been discovered, but only the genome sequence of H. modesticaldum has been reported [8]. H. modesticaldum is an obligatory anaerobe and is moderately thermophilic with an optimal growth temperature of 50- 52°C [9]. The genome of H. modesticaldum is a single 3.1-Mb circular chromosome possessing 3,138 putative open reading frames (ORFs) [8]. The photosynthetic gene cluster is contained in a single operon, encoding the HbRC core polypeptide (PshA) and pigment biosynthetic enzymes (e.g. BchBNL). The order of these genes is the same as in the cluster in Heliobacillus mobilis [10]. While many enzymes involved in the central pathway of carbohydrate metabolism were predicted, it was recently reported that H. modesticaldum has an incomplete reductive TCA cycle lacking the key enzyme ATP-citrate lyase [11], as was also observed through ¹³C-labeling [7]. Like all heliobacteria, H. modesticaldum is capable of fixing nitrogen, even at the elevated temperatures under which it grows [9]. Nitrogen fixation is catalyzed by the nitrogenase enzyme complex, and results in the reduction of atmospheric dinitrogen (N₂) to ammonium (NH₄⁺) and the production of molecular hydrogen [12]. This process requires large amounts of chemical energy (16 ATP) and reducing power (8 Fd^red) to convert one N₂ to two molecules of NH₄⁺ [13], which is then assimilated into many biomolecules. Sequence similarity predicts the use of a Mo-Fe group I nitrogenase consisting of a homodimer of NifD/K polypeptides [14]. The primary pathway for NH₄⁺ assimilation in heliobacteria is the glutamine synthetase/glutamate synthase pathway [15]. This pathway is essential for growth, because glutamine is the primary intracellular nitrogen donor for purine and pyrimidine synthesis. Both ATP and reducing power are required in carbon metabolism, nitrogen assimilation, and hydrogen production, inextricably linking these pathways. In H. modesticaldum, the high-energy demand required for nitrogen fixation during diazotrophic growth has resulted in strict regulation of the nif genes encoding for nitrogenase. Thus, the addition of NH₄⁺ to cultures results in repression of nitrogenase activity [15,16]. However, the effects of N2-fixing versus non-fixing conditions on genome-wide expression and metabolism in Heliobacterium remained to be determined. The sequencing of the genome of H. modesticaldum, along with recent proteomic studies, have provided insights into the energy metabolism of this phototroph. However, metabolic studies have been correlated to only a few genes related to energy and carbon metabolism at the transcriptomic level [17]. To enhance our understanding of the energy metabolism of H. modesticaldum, it is necessary to explore the entire mRNA link between genome and proteome. We report here a singlenucleotide resolution map of the H. modesticaldum Ice1 transcriptome under N₂-fixing and non-fixing conditions. In general, we observed low-level repression of transcription genome-wide upon a shift to N₂-fixing conditions. On the contrary, a few genes, such as the genes involved in N₂-fixation and ammonium scavenging, were upregulated. There were also several cases of even more drastic down-regulation, including the core genes of cyclic photophosphorylation.

Growth of H. modesticaldum

Isolated colonies of H. modesticaldum strain Ice1 were cultured in gel-rite media modified by Lin and Casida (1984) for thermophilic application [18] inside an anaerobic Coy glove box at 52°C under infrared lights at 780 nm. Cells were inoculated in Pyruvate-Yeast Extract (PYE) growth media [16], which contains 1 g/L of NH₄SO₄ as the source of nitrogen and “vitamin levels” of yeast extract (0.02%). The PYE-NH₄⁺ medium was made by elimination of NH4SO4 and increasing the amount of Na₂S₂O₃•5H₂O from 0.2 g to 0.4 g. The pH was adjusted to 6.8 with H₂SO₄ prior to autoclaving. Growth was monitored spectroscopically at an Optical Density (OD) at 625-nm, as minimal photosynthetic pigments absorb at this wavelength in heliobacteria [3,17]. Cells were grown under anaerobic conditions at 52°C. From PYE or PYE-NH₄⁺ conditions, 3 ml of cells in late exponential growth phase were inoculated into 300 ml of similar media. This ensured no traces of ammonia were carried over to the next generation of cells. Extraction was performed when cells were in mid-log phase of growth at an OD of 0.303 (PYE-NH₄⁺) and 0.410 (PYE). A biological replicate for each condition was also prepared.

Isolation and purification of mRNA

Total RNA was isolated using the Purelink total RNA isolation kit (Invitrogen, USA). Cells were anaerobically extracted and lysed using the needle homogenization method offered in the kit protocol. Once total RNA was extracted the solutions were treated with DNaseI via the Ambion RiboPure Bacteria procedure to remove all genomic DNA. Depletion of the 16S and 23S rRNAs was performed via the Ambion MICROBE Express kit through subtractive hybridization with capture oligonucleotides. The resulting solution contained tRNAs, 5S rRNAs, and enriched mRNA.

Library preparation and Ion Torrent sequencing

Library preparation was performed using the Ion Total RNASeq Kit (Ambion, USA). Both sample and WT control RNAs (1 μg/ μL HeLa total RNA) were fragmented using RNase III at 37°C for 10 minutes. The fragmented RNA was purified via Ambion’s RiboMinus^TM Concentration Module. The resulting mRNA yield was measured on the Bioanalyzer (Agilent, USA). These enriched mRNA samples were hybridized and ligated with Ion Torrent adaptors. Strand specificity was retained using the Ion Adaptor Mix containing oligonucleotides with a single-stranded degenerate sequence at the 3’ end and a defined sequence at the 5’ end. This effectively constrains the RNA orientation, with sequencing only performed from the 5’ end of the sense strand. Reverse transcription was performed and the cDNA was purified and size selected using Agencourt’s AMPure^® XP reagent for an optimal fragment size between 30-200 bp. The cDNA samples were amplified using the provided Ion 5’ and 3’ PCR Primers for 16 cycles. Fragment size was verified on the Bioanalyzer (Figure S1). The prepared cDNA libraries were loaded on the chip and sequenced using the Ion PGM sequencer per the manufacturer’s instructions (Life Technologies, USA).

Bioinformatic analysis of RNA-seq data

The sequenced reads captured on the Ion Torrent platform were analyzed by following procedure described here. Initially, the sequencing reads were subject to filtering out of those with the poor quality and more than 50% of Ns in reads using PRINSEQ [19]. Bowtie2 [20] with default parameters was used to align filtered RNA-seq reads against H. modesticaldum chromosome as the reference genome. The resulting sequence alignment files were imported into Partek Genomics Suite (Partek Inc., St. Louis, MO) to compute raw and fragments per kilobase of exon model per million mapped (RPKM) reads for the normalized expression values of each transcript. A stringent filtering criterion with RPKM value of 1.0 [21] was used to obtain expressed transcripts. The RPKM values of filtered transcripts were log-transformed using log2 (RPKM + offset) with an offset value of 1.0, and fold changes in transcript expression, differential expression, and p-values were generated from these using the Partek software with default settings. Metabolic pathways that were significantly enriched between +/-NH₄⁺ were identified using Cytoscape [22] with the reference pathways of H. modesticaldum from the Kyoto Encyclopedia of Genes and Genomes (KEGG) database [23].

Real-time qPCR

A 1 μg concentration of non-fragmented mRNA was added to an iScript cDNA Synthesis Kit (Bio-Rad, USA) and reverse transcription was performed. Primers were designed and synthesized for qPCR using the SsoFast EvaGreen Supermix kit (Bio-Rad, USA). Enzyme activation was performed at 95°C for 1 minute, followed by 40 cycles of denaturation at 95C for 15 seconds and annealing/extension at 57°C for 30 seconds. The Ct values were determined based on the PCR cycle number that crossed an arbitrarily defined threshold. Gene expression was determined using the 2^-ΔΔCt method [24]. Three technical replicates were performed and the mean values were reported (Table S1). Primers used are listed in Table S2.

Reaction center measurements

Cells were grown in PYE and PYE-NH₄⁺ to mid-log phase of growth. Aliquots from both cultures were diluted to OD₆₅₀ = 0.30 to equalize cell density between the two conditions. Cells were transferred into a 1 cm anaerobic cuvette and brought to room temperature before analysis with a JTS-10 LED spectrometer (Bio-Logic; Claix, France). Absorption from P₈₀₀ was monitored using 10 μs pulses of light from a 810 nm LED; 780 nm high-pass filters were placed in front of the detectors to block actinic light provided by a saturating flash (31 mJ) from a frequency-doubled Nd/YAG laser [25]. The bleaching values were measured six times per sample with a 90-s dark adaptation and cuvette inversion between measurements. After the photobleaching measurements, an acetone extraction was performed on both samples anaerobically; the supernatant was measured for BChl g concentration and the pellet was resuspended in 50 μl of 50 mM MOPS (pH 7). A bicinchoninic acid assay was performed on the pellet and the resulting protein concentration values were used to normalize both the photobleaching kinetic data and BChl g concentrations (Thermo Scientific, USA).

The baseline transcriptome

In order to be able to correlate the transcriptomic data with previous proteomic studies, it was important to use comparable growth conditions. Therefore, we chose phototrophic growth in Pyruvate Yeast Extract (PYE) medium as the baseline condition, as it is the typical growth mode used in most studies on this organism. The presence of yeast extract in this media was shown to have no significant effect on the expression levels of genes involved in metabolism and other cellular functions [17]. Following cellular extraction during mid-log phase growth, depletion of 16S and 23S rRNA was performed. This procedure reduces the sequencing bias towards highly expressed genes, as it was shown that 98% of the transcriptome consisted of rRNA in three different bacterial species when total RNA was used for transcriptome analysis [26]. These highly expressed transcripts reduce the coverage of mRNA reads and undermine transcriptome analysis accuracy. Electrophoretic analysis of the RNA pools indicated that the depletion was largely successful (Figure S1 of Supplemental Material). After rRNA depletion, cDNA libraries of three biological replicates were generated and sequenced on the Ion Torrent. The sequencing reads of the PYE culture were pooled together, resulting in 3.75 million reads and 256 Mbp (> 83- fold sequencing depth of the genome). Following the same procedure as described above, the sequenced reads of the PYE-NH₄⁺ culture resulted in 3.71 million reads and 345 Mbp (>112-fold genome coverage). Further, we filtered the sequencing reads with poor quality or more than 50% ambiguous bases, and the remaining reads were subject to the alignment. Reads totaling 74 and 69 Mbps from the ammonium-replete and ammonium-deplete cultures, respectively, were mapped onto the genome of H. modesticaldum. Both data sets resulted in a single-nucleotide resolution transcriptome with an average coverage of 24- and 22-fold per nucleotide, respectively. The inherent strand specificity of the Ion Torrent methodology revealed several interesting features regarding H. modesticaldum’s transcriptome. First, one of the genome strands encodes two thirds of the protein coding sequences while the opposite strand encodes the remaining coding sequences (Figure S2A). This observation is consistent with the location of open reading frames on the sequenced genome [8].

Additionally, the strand-specific sequencing revealed non-coding RNAs, which appear to include cis-encoded antisense RNAs. These are RNAs that both overlap and are complementary to the sense strand. Examples of antisense RNAs in H. modesticaldum’s transcriptome can be seen throughout the NiFe hydrogenase cluster as both cis (red) and complementary to the sense strand (green) (Figure S3). The function of antisense RNAs in bacteria is increasingly being recognized as an important regulator of metabolic and physiological processes [27]. In the cyanobacterium Synechocystis sp. PCC 6803, there are antisense RNAs for approximately 10% of the total protein-coding genes, indicating that regulation by antisense RNA may be a significant form of gene regulation in phototrophic bacteria [28]. We have annotated 775 potential antisense transcripts using an arbitrary cut-off of ≥ 10 in coverage (Data Set S1), indicating that the use of antisense RNA could be as high (or even higher) in heliobacteria.

Another feature of the baseline transcriptome is several regions of high intergenic transcription, which may indicate non-annotated genes. For instance, the region between 1352.5-1358.5 Kbps contains a high level of intergenic transcription, and nBLAST attributes a 99 percent identity to 16S rRNA for both the non-annotated transcripts and the intervening “hypothetical gene” HM1_3154 (Figure S2B). Table S1 lists 53 transcripts that map to intergenic regions (i.e. not overlapping with genes on either strand) and have coverage ≥ 10 and minimal length of 50 bases. BLAST searches were performed on all peptides potentially encoded by this region, but the vast majority did not have significant matches. Three of them are potential pseudogenes, with significant matches to peptides 50-55 residues long. One transcript, however, has 58% identity (73% similarity) to the ParA chromosome partitioning protein of Paenibacillus polymyxa over the entire length of 282 residues, and thus very likely represents a gene that was missed in the primary genome annotation (Table S1).

The ammonium-deplete transcriptome

With the general features of H. modesticaldum’s transcriptome established, it becomes possible to quantify the expression of the genome at the RNA level and how this changes in response to an environmental change. We repeated the RNA-seq experiment with cells deprived of ammonium, thus requiring them to fix N₂ present in the headspace gas (~97% N₂). Cells were diluted 100-fold into PYE medium lacking ammonium (PYE-NH₄⁺) and grown to saturation. This culture was then used to inoculate a new PYE-NH₄⁺ culture. The growth curves of the cultures grown in ammonium replete (PYE) and ammonium deplete (PYE-NH₄⁺) media are shown in Figure 1, and cells were harvested for RNA extraction at an optical density (OD) of 625 nm at 0.410 (PYE) and 0.303 (PYENH₄⁺). In both cases, cells were grown photoheterotrophically using pyruvate as the carbon source. We determined that the cells were dependent upon the N₂ in the headspace for their nitrogen needs by also inoculating a PYE-NH₄⁺ culture in a bottle that had been purged with argon to remove all N₂. This culture stopped growing very soon after inoculation and did not attain a density more than 25% that of the PYE-NH₄⁺ culture that had a headspace of 97% N₂ (data not shown).

Figure 1: Growth of H. modesticaldum with and without ammonium. Cells grown in pyruvate yeast extract with NH₄⁺ (black) and without NH₄⁺ (red). The Optical Density (OD) at 625 nm was monitored and RNA was extracted from cells once the cells reached an OD of 0.4 (PYE) or 0.3 (PYE-NH₄⁺), as indicated by the arrows.

Figure 2: Proposed connections between fundamental carbon metabolism and nitrogen assimilation. The major pathways of carbon metabolism when cells are grown with pyruvate as the sole carbon and electron source are shown, along with the major nitrogen incorporations reactions. Ammonia is transported into the cell when available, or is made by nitrogenase by reduction of N₂ when unavailable. The ammonia is first incorporated into an organic molecule by Gln synthetase. Gln is used to make Glu, which can be used to make Asp and the rest of the amino acids via transaminases. Gln, Glu, and Asp are the nitrogen donors for almost every other N-containing biomolecule, including amino acids (aa) and nucleobases (nb). For simplicity, the consumption/generation of reducing equivalents and ATP is largely left off this figure. Since the direction of the TCA cycle between succinate and 2-OG is unclear at this time, the arrows of the SCS and 2OGS reactions are dotted. Abbreviations: 2OGS, 2-oxoglutarate synthase; Acon, aconitase; Amt, ammonium transporter; CS, citrate synthase; FR, fumarate reductase; GlnS, glutamine synthetase; GluS, glutamate synthase; IDH, isocitrate dehydrogenase; MDH, malate dehydrogenase; N₂ase, nitrogenase; PEPCK, phosphoenolpyruvate carboxykinase; PEP, phosphoenolpyruvate; PFOR, pyruvate:ferredoxin oxidoreductase; PPDK, pyruvate phosphate dikinase; SCS, succinyl-CoA synthetase.

With the high energetic requirements of N2 fixation, we predicted a significant change in metabolism from expression levels of genes in the nif operon, which encodes the nitrogenase enzyme and ancillary functions, which was observed. Although the genes encoding nitrogenase (nifDHK) are hardly expressed in ammoniumreplete medium, they are strongly expressed in the medium without ammonium (Data Set S3, hmo00910b). We also observed a decreased rate of growth (Figure 1), consistent with previous observations [17] and indicative of a global change in metabolism. Our primary findings are detailed below, and all genes are shown in Data Set S2 along with the corresponding expression values.

If one removes all genes for which no reads were found in either condition (and for which a change in expression level is therefore impossible to calculate), one is left with 2430 genes, representing 73% of the 3268 genes annotated in the genome. This tends to remove genes that are either very poorly expressed in general and/ or for which the change in expression is very large. The average fold change in expression upon removal of ammonium for these 2430 genes is -3.93± 8.05; i.e. a 4-fold decrease in expression, with a large spread in values from a 12-fold decrease to a 4-fold increase (within one standard deviation). The overall change in expression across the genome appears to be significant (P-value = 0.0124, based on a two-tailed, paired t-test of the data; Data Set S2). The number of genes displaying different fold change values is tabulated in Table 1. There is a strong bias towards down-regulation. For example, there are over 20 times more genes with a >10-fold decrease in expression than there are genes with a >10-fold increase in expression (224 vs. 10). It is important to note that some of the most highly expressed genes (e.g. ribosomal RNA and protein) exhibited almost no change in message level, which explains why most genes could exhibit a decrease in expression. It is thus likely that the total cellular RNA level hardly decreased at all, as one might expect.

Table 1: Description: Summary of effect of ammonium depletion upon gene expression.

In order to put into useful context, the changes in expression of the 2757 genes for which we possess data, we have made use of the annotated metabolic pathways in the Kyoto Encyclopedia of Genes and Genomes (KEGG). All of the pathways in the KEGG for the H. modesticaldum genome were analyzed and a table was constructed for each pathway by incorporating the expression data for all genes encoding products involved in the pathway. These tables are named according to the KEGG numbering system (e.g. hmo00010 is the glycolysis/gluconeogenesis pathway). Each can be found as a pathway in one of the Data Sets described in the Supplemental Materials. Each Data Set contains a set of pathways that are similar in nature, and the organization of the Data Sets mirrors the discussion that follows. The results from all of these pathways are summarized in Table 2. The statistical significance of the overall expression changes of the genes in each pathway were assessed by paired Student’s t-test (two-tailed) of the RPKM values for each gene in the two conditions (plus and minus ammonium). Overall expression changes with P-values lower than 0.05 were deemed insignificant. By this measure, expression changes in 39 of the 62 analyzable KEGG pathways were significant. To correct for the multiple comparisons problem, we applied the Bonferroni correction (i.e. α=0.05/62). After this procedure, only 7 of the KEGG pathways were deemed statistically significant (Table 2).

Table 2: For each pathway in the KEGG for H. modesticaldum, a table was created (present in the Supplemental Information named after the KEGG pathway; e.g. Table hmo00010 is for pathway hmo00010, which is glycolysis). Each row in Table 2 summarizes one of these tables, providing the KEGG designation, name and brief description of the pathway, as well as a summary of the RNA-seq data.

Nitrogen fixation, uptake, and incorporation

Providing nitrogen in the form of ammonium eliminates the need to fix N₂ via the energy-expensive nitrogenase system (encoded by the nif operon), which requires 16 ATP and 8 Fd^red for each N₂ molecule fixed [13]. Not only is nitrogenase activity absent in cells grown in medium with ammonium, but the activity is lost very quickly after addition of ammonium to N₂-fixing cultures, in this species [9] and other members of the family [6,29]. The RNA-sequencing data for all genes involved in nitrogen metabolism can be found in Data Set S3. However, due to the very different regulation seen for different sets of genes, we pulled out the genes involved in nitrogen fixation and put them into a separate table (pathway hmo00910b), leaving the rest of the genes involved in basic nitrogen metabolism in the first table (pathway hmo00910a).

When the bacterium was placed in N2-fixing conditions, expression of genes for nitrogen fixation was up-regulated (Figure 3A and Data Set S3, pathway hmo00910b). In particular, the nifD and nifK transcripts, encoding the core subunits of nitrogenase, increased in abundance by a factor of 25-50. A similar increase was seen for nifH, which encodes the Fe protein that provides electrons to the nitrogenase enzyme, but it was difficult to estimate, as no reads were detected aligning to this gene in the baseline transcriptome. In cases such as this, a value of 1 RPKM in was used for the baseline value in order to estimate a minimum expression difference. If one excludes genes without reads in the baseline transcriptome, the average up-regulation of nitrogen fixation genes was ~25-fold (n=9). If one includes such genes (for nifH, nifN, fdxB, HM1_0868 and HM1_0867), using an assumed baseline value of 1 RPKM, then an average increase of 23-fold (n=14) was estimated, indicating that this is a fairly robust value. This up-regulation in structural genes, maturation factors, and regulatory protein gene transcription was expected, since all of these components are needed for the reduction of molecular N2 and are in general not required for other functions.

Figure 3: Ammonium regulation of nitrogenase operon and ammonium transporter. (A) The absence of ammonium induces the nif operon during N₂-fixing conditions (top) and the presence of ammonium represses expression of the nitrogenase operon, in which nifD and nifK encode the a/ß subunits of the nitrogenase complex. (B) The high affinity ammonium transporter (amt) and the nitrogen regulatory protein PII (HM1_0742) are highly expressed in ammonium-deplete medium (top) and repressed in the presence of ammonium (bottom).

The exception to this may be nifS and HM1_1865, which encode a cysteine desulfurase and FeS cluster assembly protein, respectively, and are not found in the nif operon. Thus, the corresponding gene products likely play a role in functions required during normal growth (e.g. FeS cluster assembly) as well as assembly of the FeMoCo cofactor in dinitrogenase. Expression of these two genes is quite high in the baseline transcriptome, and only increases about 3-fold under N2-fixing conditions (Data Set S3, pathway hmo00910b). The other exception is nifX, which is found in the nif operon, and is only up-regulated by ~3-fold as well. Its expression is quite low in both transcriptomes (3.4 and 10 RPKM), so this change may not be as reliable as for the other members of the operon.

The nif gene cluster had been previously cloned and sequenced from Heliobacterium chlorum, and the gene order is completely conserved in H. modesticaldum. In both species, there are three genes at the 5’ end of the nif operon that appear to be involved in regulation of N2 fixation. Immediately upstream of nifH are the nifI1 and nifI2 genes (HM1_0868 and HM1_0867, respectively), which are members of the GlnB family of PII proteins. These PII-like proteins are typically regulated by glutamine (as a signal for nitrogen), 2-oxoglutarate (as a signal for carbon), and ATP (as a signal for energy) in a wide variety of Bacteria and Achaea [30-32]. The nifI1/nifI2 gene pair diverges from this family and is found in the nif operons of several anaerobic diazotrophs, including Clostridium acetobutylicum and Methanococcus maripaldus [33]. Just upstream of this gene pair is a conserved protein that appears to be unique to heliobacteria, called “orf1” [34]; it was proposed to alter nitrogenase expression by controlling termination of transcription after the orf1 sequence, thus attenuating expression of the downstream nif genes [33]. Under N2-fixing conditions, expression of all 3 genes increases (Figure 3A and Data Set S3, pathway hmo00910b), as was seen previously in Heliobacterium chlorum [34].

Growth in ammonium-deplete medium also resulted in a large upregulation of the amt gene (~59-fold; Figure 3B), a gene similar to the bacterial amtB gene encoding a high-affinity ammonium transporter [35]. In fact, this represents the largest increase in expression that can be reliably measured (i.e. where the expression in the baseline transcriptome was significant). It is likely that the cells in lowammonium conditions up-regulate amt in an attempt to scavenge any available NH₄⁺, which would represent a lower energetic cost than fixing N₂. Conversely, NH4+ is relatively abundant in the baseline condition, which implies that there is adequate NH4+ influx via low-affinity transporters present during mid-log phase growth. Therefore, amt expression is likely down-regulated through a feedback mechanism in that situation. The AmtB protein has also been implicated in regulation of nitrogen metabolism in general and of nitrogenase in particular in multiple bacterial species [36,37]. It is interesting that only 13 genes display an increase in expression of 15-fold or higher in the absence of ammonium, and all of them are either in the nif operon or in a threegene cluster containing the amt gene. In the latter, amt is preceded by HM1_0742, a PII homolog, which exhibits a 28-fold increase in expression. It is followed by HM1_0740, which displays similarity to the sporulation-specific YabG peptidase of Bacillus subtilis and whose expression increases 48-fold; it is unclear at this time what the role of this gene product would be in heliobacteria. All other genes in the genome exhibit fold changes of 11 or less in the absence of ammonium.

In contrast to the genes involved in N₂ fixation, the rest of the genes involved in basic nitrogen metabolism were not significantly affected by the absence of ammonium (Data Set S3, pathway hmo00910a). Once NH4 + is either synthesized or transported into the cell, it is incorporated into glutamine via glutamine synthetase (glnA). Glutamate is synthesized by amination of 2-OG using the amide N of glutamine and NADPH as a reductant, catalyzed by the enzyme glutamate synthase. Glutamate is used as a nitrogen donor for synthesis of many amino acids and other biomolecules. As this organism appears to lack a glutamate dehydrogenase, glutamine synthetase is thus the only entry point for ammonium into biomolecules, and glutamate synthase is the only way to make glutamate. The glnA gene is highly expressed under both conditions (306 vs. 102 RPKM), being reduced by only 3-fold upon transfer to ammonium-deplete conditions. (This level of decrease of mRNA abundance is typical of most genes, as discussed below.) This is expected, given that glutamine synthetase will be the major entry point for ammonium under both conditions. The genes encoding glutamate synthase (gltB and gltD) were expressed at a lower level. Most of the genes for incorporation of ammonia were down-regulated in ammonium-deplete media by a factor of 4-5 on average. As will be seen, this behavior is typical for most metabolic pathways. Thus, expression of genes for fundamental nitrogen metabolism behaved much as expected upon a shift from ammonium-replete to N2-fixing conditions.

Bioenergetics and electron transfer

Photosynthesis is an efficient mechanism for ATP production that does not compete for the carbon of pyruvate, much of which is wasted as CO2 and small molecules (e.g. acetate, ethanol) during fermentation [7]. The heliobacterial reaction center (HbRC) uses absorbed photons to oxidize periplasmic cyt c553 and reduce cytoplasmic ferredoxin (Fd). The electrons from Fdred are sent back to cyt c553 via complex I and the cyt b6c complex in a cyclic electron transport (cET) process that produces a proton gradient that can be used by ATP synthase to phosphorylate ATP [38-40]. Because the KEGG does not have the heliobacterial cET pathway, we made use of the oxidative phosphorylation pathway (hmo00190) as the baseline pathway, as it contains complexes I (NADH dehydrogenase), II (succinate dehydrogenase), III (cyt b6c), and V (ATP synthase), despite the fact that fact that heliobacteria lack complex IV (cytochrome oxidase), being strict anaerobes. Other genes important in bioenergetics not represented in the KEGG pathways were added to separate tabs; hmo00190b contains the genes encoding additional proteins involved in the cET pathway (e.g. the HbRC, its electron donors and acceptors), while hmo00190c lists the genes for hydrogenases.

Given the importance of the HbRC to cellular metabolism, we expected the pshA gene encoding the HbRC core polypeptide to be highly expressed during photoheterotrophic growth. In fact, pshA is the twelfth highest protein-encoding gene expressed (1446 RPKM; Data Set S4, pathway hmo00190b, Fig. 4B). During the elucidation of the HbRC structure to 2.2 Å, we discovered a previously unknown subunit called PshX [41]. This small polypeptide consists of a single transmembrane alpha-helix with a few residues on either side. The resolution of the structure was sufficient to allow identification of sidechains and subsequent identification of the gene, which was confirmed by mass spectrometry. The gene encoding this subunit (HM1_0821), which had not been previously annotated, is not in the photosynthetic gene cluster nor does it appear to be in an operon. However, its normalized expression level (1685 RPKM; Data Set S4, pathway hmo00190b, Fig. 4C) is almost exactly the same as pshA. This makes sense as the stoichiometry of the subunits is 2:2 in the HbRC.

The electron donor to the HbRC is a membrane-anchored cyt c₅₅₃ encoded by the cyhA gene; it is highly expressed (490 RPKM; Data Set S4, pathway hmo00190b). A high expression level was also found for genes pshB2 and pshB1 (Figure S4), which encode the [4Fe- 4S] dicluster ferredoxins, now thought to be the primary electron acceptors of the HbRC [42]. The pshB1 and pshB2 genes are adjacent in the genome and are likely co-transcribed, but for unknown reasons the second mRNA (pshB1) is about five-fold more abundant than the first one (pshB2), and is in fact the fifth most highly expressed proteinencoding gene (2507 RPKM; Data Set S4, Figure 4B and Fig. S4). The ferredoxins are expected to interact with Fd-NADP⁺ oxidoreductase (FNR) to reduce NAD(P)+ to NAD(P)H and generate reducing power for metabolic reactions as well as for cyclic electron flow. While the FNR enzyme was not originally annotated in the genome, the enzyme and its gene (HM1_0289) were subsequently identified and found to be well expressed [17]; it has, however, nowhere near as highly expressed as the other genes in this pathway (29 RPKM; Data Set S4, pathway hmo00190b).

The genes involved directly in photosynthetic electron transfer are strongly regulated by ammonium abundance. In the absence of ammonium, there is a 44-fold decrease in the pshA transcript, a 78- fold decrease in the cyhA transcript, a 57-fold decrease in the pshB1 transcript, and a ~140-fold decrease in the pshB2 transcript (Figures 4B and S4, Data Set S4, pathway hmo0019b). This result is surprising given the distinct advantage provided by the HbRC for growth, and one would not expect the abundance of the HbRC, or its electron donors and acceptors, to be drastically lowered in N₂-fixing conditions. In contrast, the mRNA levels of the other proteins in the proposed cyclic electron transfer pathway (complex I and III) as well as ATP synthase (complex V) do not follow the same pattern (Data Set S4, pathway hmo00190). They are not as highly expressed, but they are also not nearly as strongly repressed by the absence of ammonium (average repression of 2- to 3-fold). The same is true of the genes encoding the cytochrome bd menaquinol oxidase (cydA/B), whose primary role is likely to scavenge O₂ rather than contribute to energy production.

Such a large drop in transcript abundance after a shift to ammonium-deplete medium is not a common feature of the other abundant mRNAs. For example, rpoE (σ24 factor) only decreases by 1.6-fold and cstA (carbon starvation protein) increases by 1.4-fold. In addition, the four most highly expressed genes in ammoniumreplete conditions (HM1_3146, HM1_3147, HM1_3149, HM1_3151; 6200-19000 RPKM), all of which are conserved hypothetical proteins of unknown function located next to each other in the genome, exhibit very little repression in the absence of ammonium (at most 2-fold; Data Set S2). In order to make the comparison more meaningful, we compiled a list of the most highly expressed protein-encoding genes in ammonium-replete conditions (i.e., with normalized transcript levels >300 RPKM). When the log2 of the expression change (-NH4+ compared to +NH4+) is plotted against the log2 of the expression level in +NH4+ medium of these 42 genes, something striking emerges: the structural genes for the HbRC (pshA, pshX), its electron donor (cyhA) and acceptor (pshB1, pshB2) form a discrete group of genes that are highly repressed by the lack of ammonium (red dots in Figure 4B).

Figure 4: Ammonium regulation of the photosynthetic reaction center. (A) The absence of ammonium repressed the pshA gene encoding the HbRC core polypeptide. (B) Log-log plot of expression change vs. expression for the 42 most highly expressed protein-coding genes. This list includes all protein-encoding genes with normalized transcript levels of >300 RPKM (in the presence of NH₄ ⁺) that do not overlap with rRNA or tRNA genes (potentially raising their transcript counts). The x-axis is the log2 of the normalized transcript level (in RPKM). The y-axis is the log2 of the ratio of the normalized transcript levels in ammonium-deplete to ammoniumreplete conditions. The names of the six genes with repression levels >16-fold are marked. The five red dots correspond to structural genes for the HbRC (pshA, pshX), ferredoxins (pshB1, pshB2), and cyt c₅₅₃ (cyhA). (C) Photobleaching of P₈₀₀ induced by a 6-ns laser flash (at time 0) was monitored in cells using 10-µs pulses of light from an 803-nm LED. The ammonium-replete and ammonium-deplete cultures had A625s of 0.42 and 0.30, respectively, at the time of the measurement. Absorption changes at 803 nm are reported as mAU normalized to the A625 values of the culture to provide an estimate of cellular HbRC content. The inset is the post-flash data plotted on a log₁₀ time scale.

To test the effects of ammonium depletion on expression of the HbRC, we grew cells in PYE or PYE-NH4 + media, took samples at a few points along the growth curve, and directly assayed functional HbRC levels by a quantitative P₈₀₀ photobleaching experiment. Briefly, cells were excited with a saturating laser flash and the absorbance at 800 nm was monitored (Figure 4C). The primary electron donor of the HbRC (P₈₀₀) absorbs maximally at 800 nm and is bleached upon lightdriven electron transfer to the ferredoxin acceptors. The bleaching recovers within a few milliseconds due to reduction of oxidized P800 by the membrane-attached cyt c553 donor [43,44]. We observed only a modest drop (~30%) in cellular P800 levels in PYE-NH4+ cultures relative to the PYE cultures (Table S4), and in both cases there was a small drop (~20%) as the cultures aged. Moreover, the kinetics of P₈₀₀ recovery were about the same in both (t1/2 ≈ 2.5 ms; Figure 4C), which is inconsistent with a large drop in cyt c₅₅₃; the reduction of P₈₀₀ + by cyt c₅₅₃ is a second-order reaction and is thus sensitive to the concentration of the electron donor [44]. Thus, despite the large drop in pshA and cyhA mRNA levels, we find no evidence for correspondingly large drops in the levels of the encoded proteins. Thus, the transcript levels of the structural genes for the RC and its donors and acceptors (pshA, pshX, cyhA, and pshB1/2) may be much higher in the presence of ammonium than actually necessary, although it is unclear why this would be the case.

Hydrogenase enzymes can catalyze the two-electron reduction of protons to molecular hydrogenase (H₂) and/or the two-electron oxidation of H2 [45]. The genome of H. modesticaldum contains two copies of the genes (hupSLC) for a membrane-bound [NiFe] uptake hydrogenase; one of these operons also contains the genes encoding the maturation factors (hupD hypABCDEF) [11]. This type of hydrogenase is often used by N₂-fixing organisms to recover the H₂ produced by nitrogenase [46]. However, we found that expression of the hup and hyp genes was not induced by a shift to ammonium-deplete conditions, but was instead slightly repressed by a factor of 7 ± 5 (Data Set S4, pathway hmo00190c). The genome also appears to encode two copies of a [FeFe] hydrogenase, which are often used by Clostridia to get rid of excess electrons during fermentation. The predicted products of HM1_1617 and HM1_2613 resemble each other (59% identical residues) and the catalytic subunits of clostridial [FeFe] hydrogenases, but genes for maturation factors were not found. It is thus unclear if this organism assembles an active [FeFe] hydrogenase. In any case, these genes are also not induced by the shift to N₂-fixing conditions.

Regulation of fundamental carbon metabolism by ammonium

As mentioned, H. modesticaldum cannot grow as a photoautotroph, as it lacks at least one key enzyme for each carbon fixation pathway. For example, heliobacteria seem to possess every enzyme for the reverse TCA (rTCA) cycle except for ATP-citrate lyase, which prevents them from using this cycle to fix CO₂ in an autotrophic manner [11,17]. Thus, H. modesticaldum is an obligate heterotroph; it uses pyruvate, lactate, or acetate as carbon sources [16]. Figure 2 summarizes central carbon metabolism and connections to nitrogen metabolism in heliobacteria.

Pyruvate was the sole carbon source during the growth of heliobacteria in these experiments. Expression of genes encoding enzymes in pyruvate metabolism was reduced by approximately 6-fold in the absence of ammonia (Data Set S5, pathway hmo00620). Pyruvate can be oxidatively decarboxylated to produce acetyl-CoA with concomitant reduction of 2 ferredoxins by the pyruvate:ferredoxin oxidoreductase (PFOR), as this organism lacks an NAD(P)-dependent pyruvate dehydrogenase. During fermentative growth, acetyl-CoA synthetase converts acetyl-CoA to acetyl-phosphate, which can be used as a phosphate donor to ADP to make ATP via acetate kinase. These latter 2 enzymes are moderately expressed (7-16 RPKM) under both conditions, while PFOR is highly expressed (30-168 RPKM). This may be because the cultures used were grown in the light and could use photosynthetic cET to produce ATP, allowing pyruvate to be used primarily as a carbon source.

This organism appears to lack pyruvate carboxylase to convert pyruvate to oxaloacetate (OAA). However, it can convert pyruvate to phosphoenolpyruvate (PEP) by pyruvate phosphate dikinase (PPDK). Incorporation of carbon into biomolecules can then be achieved through PEP carboxykinase (PEPCK) [17], which carboxylates PEP to OAA. OAA can then be converted to malate, fumarate, and succinate by the “left arm” of the TCA cycle; these enzymes are highly expressed in H. modesticaldum (Data Set S5, pathway hmo00020). The transcript level of pckA encoding PEPCK was reduced by ~16-fold during N₂- fixing conditions, which is the largest fold change for any gene in the TCA cycle, which exhibited an average ~3.5-fold down-regulation in transcript level. Malate dehydrogenase, which would reduce OAA to malate in the reverse direction, exhibited an ~11-fold reduction in transcript level, the second highest change. It may be that the organism slows the flow of carbon into the left arm of the TCA cycle when ammonium is low.

We note that there are two genes annotated to be similar to the alpha subunit of OAA decarboxylase (oadA), which are moderately expressed under both conditions (30-120 RPKM). The OAA decarboxylase complex usually consists of 3 subunits – the beta and gamma subunits are integral membrane proteins required for coupling Na+ efflux to decarboxylation of OAA by the alpha subunit [39]. These latter two subunits are not annotated in the genome, so the existence of this enzyme is unclear at this time. We note, however, that the first oadA gene exhibits similarity to pyruvate carboxylase, and is preceded by HM1_0379, which is annotated as a biotin carboxyl carrier protein; both genes are moderately expressed (80-120 RPKM in +NH₄⁺) and exhibit little repression in the absence of ammonium. There is a similar story with the second oadA gene, which is followed by a gene similar to the biotin carboxylase subunit of acetyl-CoA carboxylase (accC). Thus, it is possible that at least one of these putative oadA genes is part of a functional pyruvate carboxylase, which could allow synthesis of OAA in low ammonium.

The key enzyme of the TCA cycle – citrate synthase – appears to be missing in the H. modesticaldum genome. There is no gene encoding the typical (Si)-citrate synthase, but it has been suggested that this organism uses an (Re)-citrate synthase, an enzyme evolutionarily related to homocitrate synthases [7]. There are two genes in the H. modesticaldum genome that show high similarity (>40% identity) to the putative (Re)-citrate synthase of Clostridium kluyveri [46]. The most similar gene (50% identity) is aksA (HM1_2993), which is well expressed under both conditions (21-28 RPKM). The other is nifV (42% identity) and is part of the nif operon. Expression of this gene was increased by a factor of ~33-fold in the absence of ammonium. We thus propose that this gene is a true homocitrate synthase, and this increase in expression is expected, as homocitrate is an essential part of the FeMoCo cofactor in dinitrogenase, serving as one of the ligands to the Mo metal [47,48]. Heliobacteria do not have the lysine biosynthetic pathway via homocitrate – they use the pathway via aspartate 4-semialdehyde (Data Set S5, pathway hmo00300) – and there is no other known use for homocitrate. Thus, it is likely that the aksA gene product is actually a (Re)-citrate synthase, as proposed by Tang et al. [7]. If so, then OAA could be converted to citrate by this enzyme, and thence to 2-oxoglutarate via isocitrate by the “right arm” of the TCA cycle (Figure 2). This would allow the two arms of the TCA cycle to work in the opposite directions and OAA would be a major branch point compound.

PEP would also be a major branch point, as it can be converted to OAA by PEPCK to feed into the TCA cycle, or converted to carbohydrate via the enzymes of gluconeogenesis. As H. modesticaldum does not utilize carbohydrates as a carbon or energy source, it is most likely that the enzymes of glycolysis are all used in the anabolic direction (i.e. gluconeogenesis) for making fructose-6-phosphate, which then can be converted into whichever hexoses are needed. The genes involved in the gluconeogenesis pathway were all down-regulated in the absence of ammonium by an average of ~8-fold (Data Set S5, pathway hmo00010). The fructose-6-phosphate generated by this pathway can also be converted to pentose-phosphates via the pentose-phosphate pathway. It should be noted that the oxidative enzymes of this pathway (glucose-6-phosphate dehydrogenase and 6-phosphogluconate dehydrogenase) are absent in this organism, so it would have to use primarily transketolase and transaldolase to convert the hexosephosphates and triose-phosphates from gluconeogenesis into pentosephosphates. The genes involved in the pentose-phosphate metabolism were all down-regulated in the absence of ammonia by a factor of ~7.5 (Data Set S5, pathway hmo00030). Gluconeogenesis products also feed into the fructose/mannose and galactose metabolic pathways, allowing production of any needed hexoses. The genes encoding enzymes in these pathways were not significantly affected by the absence of NH₄⁺ (Data Set S5, pathways hmo00051 and hmo00052).

There has so far been no report of carbon storage polymers in H. modesticaldum, such as polyhydroxybutyrate or glycogen/starch. While the genome lacks enzymes for synthesis of the former, there are genes encoding a putative glycogen synthase (glgA) and glycogen branching enzyme (HM1_2990) in the genome. There are two genes for glycogen phosphorylase (glgP); one is highly expressed (108 -135 RPKM) and the other is barely expressed (0.3 – 0.7 RPKM). There is also a debranching enzyme (malQ). Thus, the minimum complement of enzymes is present for glycogen/starch synthesis and breakdown; they are down-regulated only ~3-fold in the absence of ammonium (Data Set S5, pathway hmo00500). If the rate of carbon incorporation exceeds that of ammonia synthesis during N2-fixing conditions, it is possible that the cells store some of the excess carbon as glycogen.

Expression of enzymes in fatty acid biosynthesis was unaffected by the absence of ammonia (Data Set S5, pathway hmo00061). H. modesticaldum does not appear to use fatty acids as a carbon or energy source [5,9]. The only genes that showed significant expression in this pathway were HM1_0073 and aas, which are involved in other pathways, such as fatty acid biosynthesis. The metabolic pathways of lipopolysaccharide and glycerolipid synthesis were not significantly affected by the absence of ammonium, while glycerophospholipid synthesis genes collectively showed a ~6-fold decrease in gene expression (Data Set S5, pathways hmo00540, hmo00561 and hmo00564). A deviation from this trend was seen with glpK, which was up-regulated by a factor of ~3.5-fold. This gene encodes glycerol kinase, which can be involved in other pathways.

Regulation of intermediary metabolism and cofactor synthesis

The transcripts of genes involved in the synthesis of alanine, aspartate, glutamate, lysine, arginine, proline, glycine, serine, threonine, cysteine, methionine, and D-alanine, as well as the sulfur relay system, were all reduced by 2- to 3-fold on average in the absence of ammonium (Data Set S6). The genes involved in the nitrogen metabolism pathway—which utilizes the products formed during nitrogen fixation in order to synthesize glutamine, arginine and proline—displayed a 5-fold decrease in expression. The same was seen for genes involved in selenocompound metabolism. However, in both the cysteine/methionine and selenocompound metabolism pathways, there was an exception to both trends. The gene metC (cystathionine β-lyase) deviated significantly with an increase in expression of 4.5- fold (Data Set S6, pathway hmo00270). Expression of genes in basal sulfur metabolism was unaffected by the absence of ammonium, except for one of the cysD genes, which was repressed ~9-fold (Data Set S6, pathway hmo00920). Metabolism of the branched aliphatic amino acids (valine, leucine and isoleucine) demonstrated a 7-fold reduction in gene expression, while biosynthesis of the aromatic amino acids (histidine, phenylalanine, tyrosine and tryptophan) displayed a higher level of repression in the absence of ammonium (~12-fold). An interesting exception is the two aroF genes, which encode the enzyme catalyzing the first committed step of aromatic amino acid biosynthesis (3-deoxy-7-phosphoheptulonate synthase); the more highly expressed one is strongly repressed (~60-fold; Data Set S6, pathway hmo00400; see Figure 4B), while the less highly expressed one is unaffected by the absence of ammonium. Thus, the repression of genes involved in biosynthesis of most amino acids was not very different from the average, with the exception of the aromatic and (to a lesser extent) the branched aliphatic amino acids.

The purine and pyrimidine pathways of H. modesticaldum synthesize the ribo- and deoxyribonucleotides that are polymerized to make DNA and RNA, while the amino/sugar nucleotide pathway involves the incorporation of sugars into nucleotides. Peptidoglycan, which forms the cell wall that encases the cell membrane and provides structural support and protection, is synthesized as amino-sugar (rather than nucleotide-sugar) polymers. In the absence of ammonium, the overall expression of genes in the nucleotide and peptidoglycan synthetic pathways decreased by an average of 4-fold. All individual genes were down-regulated by a factor of less than 17-fold or upregulated by a factor less than 3-fold. Overall, nucleotide metabolism and peptidoglycan synthesis do not seem to be greatly affected by the absence of ammonia, as one might expect (Data Set S6, pathways hmo00230 and hmo00240).

There are several pathways for the biosynthesis of molecules used as cofactors or prosthetic groups. The flux through these pathways is thus smaller than the pathways for intermediary metabolism. For most of these pathways, changes in transcript levels of the enzymes was either not significant or not very large (~6-fold, on average; Data Set S6, pathways hmo00670 to hmo00790). Synthesis of porphyrin is initiated by the conversion of glutamyl-tRNA to 5-aminolevulinate; it then diverges into two pathways: one leads to the synthesis of Vitamin B12, while the other path leads to synthesis of heme and chlorophyll. The porphyrin and chlorophyll biosynthesis pathway was one of the few pathways for which the change in transcript abundance (~9-fold) was statistically significant after the Bonferroni correction (Table 2, Data Set S6, pathway hmo00860). The thiD gene, which encodes phosphomethylpyrimidine kinase involved in thiamine and Vitamin B6 synthesis, was an outlier, being repressed ~23-fold in the absence of ammonium (Data Set S6, pathways hmo00730 and hmo00750). Other outliers were bchG (encoding chlorophyll synthase, 28-fold), as well as HM1_2396 and cbiB (encoding cobalamin biosynthesis protein, CbiB/ CobD; 20-fold and 24-fold, respectively).

DNA, RNA, and protein metabolism

The genes involved in replication and repair of genomic DNA, including base excision repair, nucleotide excision repair, mismatch repair, and homologous recombination, were reduced by an average of ~5-fold in the absence of ammonium (Data Set S7). The expression of genes encoding subunits of RNA polymerase and those involved in RNA degradation were unaffected by the absence of ammonium (Data Set S7, pathways hmo03020 and hmo03018).

The ribosome is a major cellular component – ribosomal RNA (rRNA) is by far the most abundant form of RNA and ribosomal proteins are also abundant proteins. On average, expression of rRNA and ribosomal proteins was not affected by the absence of ammonium (Data Set S7, pathway hmo03010). Expression of tRNAs and aminoacyltRNA synthetases exhibited a ~3.4-fold reduction in message level (Data Set S7, pathway hmo00970).

Like all Gram-positive bacteria, heliobacteria lack an outer membrane and periplasm. Only one membrane must be crossed to direct a protein outside the cell. Heliobacteria possess both the Sec-SRP pathway and the Twin-arginine targeting (Tat) pathways for protein export. The expression of genes involved in the bacterial secretion systems was not significantly affected by ammonium depletion (Data Set S7, pathway hmo03060).

Regulation of active transport, flagella, and chemotaxis

ABC (ATP binding cassette) transporters couple ATP hydrolysis to active transport of a wide variety of molecules. H. modesticaldum has 11 different ABC transporter systems, which import phosphate, sulfate/thiosulfate, mineral ions (zinc, nickel/cobalt, tungstate, molybdate), monosaccharides (ribose/xylose), amino acids (glutamine, spermidine/putrescine, branch chain amino acids), and vitamins (biotin). There are also two ABC-2 type systems for export of teichoic acid (for lipopolysaccharide synthesis) and bacitracin (presumably for resistance to this antibiotic), as well as the FtsEX complex involved in cell division. Expression of genes encoding ABC transporters decreased by an average of 2.6-fold in the absence of ammonia (Data Set S8, pathway hmo02010).

The phosphotransferase system (PTS) is responsible for importing carbohydrates, such as hexoses and disaccharides. In order to prevent the acquired substrate from prematurely exiting the cell, phosphotransferase chemically modifies the substrate via phosphorylation using PEP as the phosphoryl donor. Only one gene in this pathway was identified in H. modesticaldum, ptsI (encoding phosphoenolpyruvate-protein phosphotransferase), and it was downregulated by 7-fold in the absence of ammonia (Data Set S8, pathway hmo02060). It is unclear if the PTS system exists in this organism.

Two-component signal transduction systems consisting of a sensor and regulator allow cells to sense changes in environment, such as carbon sources, pH, quorum signals, and osmolarity. Expression of the genes involved in these systems decreased 3.5-fold on average in the absence of ammonium (Data Set S8, pathway hmo02020). A notable exception to this trend is csrA, which underwent a 7-fold increase in expression. This gene encodes the carbon storage regulator protein, an RNA-binding protein that is a pleiotropic regulator of many genes. The CsrA protein facilitates the destruction of many mRNAs, presumably by binding to their Shine-Dalgarno sequences and inhibiting translation, thus removing the protection from nucleases provided by the ribosome [49-53]. It is possible that increased expression of this protein is at least partially responsible for the decrease in the steady-state level of many other mRNAs.

Bacterial chemotaxis allows H. modesticaldum to regulate the speed of rotation and direction of its flagella in response to perceived chemical gradients. Transcripts of genes in this pathway were reduced an average of ~6-fold in the absence of ammonium (Data Set S8, pathway hmo02030). Similarly, the expression of genes involved in flagellar assembly (i.e. flagellar components or assembly proteins) was reduced by ~7-fold (Data Set S8, pathway hmo02040). There were 6 genes in this group that exhibited a moderate increase in expression; exclusion of these 6 genes was required to calculate a change in expression that was statistically significant. The gene that demonstrated the largest increase in expression, flgM, encodes for a negative regulator for flagellin synthesis. It is unclear why the transcripts of some genes involved in flagellar synthesis are increasing, while those of most of the genes are decreasing in abundance after the shift to ammoniumdeplete medium.

The Heliobacteria appear to be the result of a lateral gene transfer of photosynthetic genes into an ancestral bacterium likely resembling a Clostridial cell, and indeed they can grow in the dark fermentatively like their non-photosynthetic cousins. The transcriptome reported here provides insights into how such a hybrid organism operates under normal conditions, and gives us the first view at a large scale of how the genetic information in its genome can be used under different conditions, such as when it is forced to use atmospheric N₂ as a source of nitrogen. These results should prove to be foundational for future studies on this particular organism (H. modesticaldum), which is rapidly becoming a model organism for the Heliobacteria.

This study reports the first transcriptome of H. modesticaldum. We also report the genome-wide effect that the absence of NH₄⁺ has on expression. These findings show that the shift from ammoniumreplete to N₂-fixing conditions is accompanied by a general decrease of ~4-fold in 74% of the genome. There was also increased expression of nitrogenase and the high-affinity ammonium transporter, along with a few other proteins, including some regulators (e.g. PII proteins and csrA). We also observed a significant decrease in the level of the mRNAs encoding the photosynthesis RC core polypeptide and electron donors/ acceptors (cyt c₅₅₃ and ferredoxins), which are among the most highly expressed protein-coding genes in ammonium-replete conditions. However, functional assays of the RC revealed only a modest decrease of the protein in vivo, indicating a significant role for regulation of translation or post-translational processes. Proteomic studies will be required in the future to unravel these contradictory responses. Future applications of the two H. modesticaldum transcriptomes reported will likely benefit from additional RNA-seq studies using a variety of growth conditions and comparative RNA-seq studies of other heliobacterial species.

This work was partially supported by the Division of Chemical Sciences, Geosciences, and Biosciences, Office of Basic Energy Sciences, of the U.S. Department of Energy through grant DE-SC0010575 to Kevin Redding. Daniel Sheehy was supported by funds from the Howard Hughes Medical Institute through the Undergraduate Science Education Program and from the School of Life Sciences Undergraduate Research Program (SOLUR) at Arizona State University. We thank Scott Bingham at the Genomics Core at Arizona State University for all sequencing performed on the Ion Torrent PGM. The following ASU Honors students from Kevin Redding’s biochemistry course constructed all of the data sets and most of the tables: Denysia Allen, Michelle Arnold, Jami Bonifacio, David Belohlavek, Matthew Calhoun, Claire Cambron, Taylor Camiliere, McKinsey Chartier, Justin Cheung, Devin Corrigan, Paige Davis, Fredrick Frost, Melynn Grebe, Amritpal Khela, Keifer McKenna, Savanah McMahon, Alexander Michael, Ryan Muller, Anoosha Murella, Daniel Murphy, Miranda Ngan, Geneva Nguyen, Kiri Olsen, Dixit Patel, Justin Patel, Anne Richards, Tirth Shah, Alexander Tomlinson, Kelli Trimble, Cameron Turro, Andrew Vo, Dustin Weigele, Sean Wilson, Dana Woell, Justin Zeien, Connie Zuo.