Strains and genetic constructs used in this study are listed in Table 1. Strain AP18 was chosen as the base strain for most of the genetic modifications described here. The strain descends from the M. buryatense 5G(B1) strain, a lab-adapted variant of wild type M. buryatense 5G. The strain is resistant to rifampicin, has improved transformation efficiency, and lacks the 76 kB plasmid, the glycogen synthase genes glgA1 and glgA2, and the sucrose-phosphate synthase (sps) gene (Puri et al., 2015). These modifications resulted in a very modest increase in the FA pool as well as an inability to produce glycogen and/or sucrose. So far no significant impacts on cell growth and methane consumption have been observed. The strain AP18 and plasmid constructs used to create acetate kinase-deleted and acetyl-CoA carboxylase-overexpressing mutants were kindly provided by the Lidstrom Laboratory (University of Washington).

M. buryatense strains were grown on methane as described (Ojala et al., 2011), with some modification of the growth medium (Table S1). Escherichia coli strains were grown on Luria-Bertani media supplemented with kanamycin (100 μg/ml) and ampicillin (100 μg/ml). Genetic manipulations with M. buryatense were done as described (Puri et al., 2015). For unmarked gene deletions, the pCM433kanT plasmid carrying ~600-bp of sequences flanking the to-be-deleted genes was introduced. After conjugation, single-crossover kanamycin-resistant M. buryatense clones were plated on rifampicin to counter-select against E. coli. Then, to select for Kan-sensitive double crossover clones with deleted genes of interest, single-crossover clones were passaged on plates with 2.5% sucrose and the resulting colonies were PCR-genotyped for the absence of the gene of interest followed by sequencing. For overexpression of genes, the pAWP78 plasmid was used as a vector. All genes were cloned under the tac-promoter, which provides strong constitutive expression in M. buryatense (Puri et al., 2015). The nucleotide positions of the sequences cloned into plasmids are outlined in Table S2.

The acetyl-CoA carboxylase expression system (pAWP78::accABCD) was obtained from Lidstrom and Puri. The construct includes the native acetyl-CoA carboxylase genes (subunits A, B, C, and D) cloned into pAWP78 under the tetracycline-inducible pTet promoter (Puri, unpublished data). The plasmid was transformed into E. coli S17–1 and transferred into the 5GB1C, AP18, AP18 Δack, AP18ΔfadABEΔack, abd AP18ΔfadABE backgrounds via conjugation. Cells were grown until mid-exponential phase (at OD ~0.5). Each cell line was represented by biological (3) and technical (2 per biological replicate) replicates. Expression of acetyl-CoA carboxylase was induced by addition of anhydrotetracycline (1 μM final) followed by a 6-h incubation before collection of cells by centrifugation.

Fifty milliliters of cell cultures grown at 30°C with 50 ml of methane to OD₆₀₀ ~1 were collected by centrifugation, washed twice with 10 mM Tris-Cl (pH8.0) and 50 mM of NaCl, and lyophilized. Biological replicates (n = 2 ÷ 10) were submitted to Matrix Genetics (Seattle, WA, http://matrixgenetics.com) for FAME derivatization and GC-MS analysis.

Strains (i) AP18 (ii) AP18ΔfabR, and (iii) AP18::pAWP78 (fabR) were grown in duplicate in a 50 ml volume at 30°C with shaking with methane. RNA from pelleted cells was extracted according to (Griffiths et al., 2000) and treated with PureLink DNaseI (ThermoFisher Scientific) according to the manufacturer's instructions. Samples were sequenced on Illumina HiSeq4000 with ~50 million/sample SR50 reads by IGM Genomics Center, University of California, San Diego.

To convert the flux balance model for M. buryatense 5G(B1) into the COBRA Toolbox format (Schellenberger et al., 2011) we used the published genome-scale metabolic network of this strain (De la Torre et al., 2015) developed in Pathway Tools (Karp et al., 2010). Subsequently, all zero-flux reactions were removed from the model. The information about zero-flux reactions was extracted from the solution file of the Pathway Tools. We manually extended the COBRA model by adding exchange rates for both nutrients and secreted metabolites as well as by the biomass equation. Moreover, we used the Paint4Net tool (Kostromins and Stalidzans, 2012) installed with the COBRA Toolbox to define dead-ends in the model network and to add corresponding exchange rates for them. To standardize reaction and metabolite IDs, we used abbreviations according to BiGG Models ID Specification and Guidelines (King et al., 2016) and KEGG abbreviations for enzymatic reactions for which we could not find corresponding BIGG reactions. To link genes with certain enzymatic reactions we turned to account NCBI Reference Sequence annotation: NZ_KB455575. Moreover, we added missing formulas for metabolites and checked out the model in COBRA Toolbox to ensure all reactions were mass balanced. The model was also extended by the addition of enzymatic reactions (EC 3.1.2.2; 1.3.8.-; 4.2.1.17; 1.1.1.35; 2.3.1.16) for FA degradation via the β-oxidation pathway and a reaction catalyzed by phosphoketolase (EC 4.1.2.22). Flux balance analysis (FBA) was performed by solving the Linear Programming problem for biomass optimization on the basis of GNU Linear Programming Kit (GLPK) (http://www.gnu.org/software/glpk/) solver in MATLAB. The converted model for M. buryatense 5G(B1) in COBRA format is available on the web-site: http://sci.sdsu.edu/kalyuzhlab/.

The genome sequence and annotation files of M. buryatense 5G(B1) were obtained from NCBI (RefSeq accession number GCF_000341735.1). Reads were aligned to the genome sequence using the Rockhopper 2 software system (Tjaden, 2015) with default settings. It is worth noting that the tool reports integral values for analyzed transcriptomic characteristics, even if the underlying value is a decimal. If a read aligns partially to a transcript, it will contribute a fractional amount to the raw count. Rockhopper 2 normalizes each RNA-seq data set using upper quartile normalization, because the normalization procedure of the mapped read counts is an essential preprocessing step for differential expression detection (Bullard et al., 2010). To quantify transcript abundance levels, Rockhopper 2 employs a modified RPKM value based on more robust normalizer of upper quartile transcript expression. The normalized data were used to compare gene expression in the wild type AP18 strain vs. the overexpressed farE and ΔfarE strains using the built-in algorithm of DESeq (Anders and Huber, 2010). To determine whether a transcript shows differential expression in data from two any conditions we used a threshold (q < 0.05) of q-values that are corrected p-values to control the false discovery rate using the Benjamini-Hochberg procedure (Benjamini and Hochberg, 1995). Appropriateness of the q-value threshold <0.05 was verified with a set of housekeeping genes and genes encoding key C1-enzymes, and those showed no differential expression [for example: 3-hexulose-6-phosphate synthase, q(AP18 vs. AP18ΔfarE) = 0.28, q(AP18 +pAWP78 vs. AP18::farE) = 0.53; pyruvate carboxylase, q(AP18 vs. AP18ΔfarE) = 0.41, q(AP18 +pAWP78 vs. AP18::farE) = 0.64; arginyl-tRNA-protein transferase, q(AP18 vs. AP18ΔfarE) = 1, q(AP18 +pAWP78 vs. AP18::farE) = 1].

Methane metabolism in the M. buryatense 5G(B1) strain is outlined in Figure 1A. The automatic annotation of the genome predicts a typical Type II FA biosynthesis pathway. The FA-biosynthesis genes form several clusters: (1) a plsXfabHDGacpfabF-cluster, which includes genes for a fatty acid/phospholipid synthesis protein, 3-oxoacyl-ACP synthase III, malonyl-CoA-ACP transacylase, 3-oxoacyl-ACP reductase, acyl carrier protein (ACP), and 3-oxoacyl-ACP synthase II. The cluster also includes a hypothetical protein and a 50S ribosomal protein L32 gene upstream of plsX. The organization of this cluster is strongly conserved among all methanotrophic bacteria; (2) a fabAB-cluster encoding 3-oxoacyl-ACP synthase I and beta-hydroxydecanoyl thioester dehydrase; (3) a fabGFZ/A cluster, encoding 3-oxoacyl-ACP synthase II, 3-oxoacyl-ACP reductase, and beta-hydroxyacyl-ACP dehydratase; and (4) an accCB-gene cluster encoding biotin carboxylase and biotin carboxyl carrier protein (BCCP). Genes encoding the acetyl-coenzyme A carboxylase carboxyltransferase alpha subunit (accA) and beta subunit (accD), a second copy of the acetyl-CoA carboxylase biotin carboxylase subunit (accC), an additional (3R)-hydroxymyristol ACP dehydratase (fabZ), and two genes for ACP are located at sites distant from each other and other FA-biosynthesis genes.

Synthesis of phospholipids in M. buryatense 5G(B1) appears to follow the E. coli paradigm where a phosphatidic acid precursor is synthesized from glycerol 3-phosphate and FA precursors (Yao and Rock, 2012). However, no plsB (Glycerol-3-phosphate acyltransferase) homologs were identified within the M. buryatense 5G(B1) genome. Instead, homologs of plsX and plsY were found. PlsXY has been shown to preferentially utilize acyl-acyl carrier protein and not acyl-CoA as an acyl donor (Lu et al., 2006). A set of phospholipid-modifying enzymes whose homologs have been shown to act on phospholipids post-synthetically were found in the genome. The first of these enzymes is a membrane-bound desaturase, DesC, which shows about 60% identity to other well-characterized desaturases that catalyze the oxygen-dependent introduction of a double bond at the Δ9 position of saturated FAs, particularly stearic acid (Zhu et al., 2006; Parsons and Rock, 2013). Additionally, M. buryatense 5G(B1) contains two copies of cfa, a cyclopropane FA synthase, which uses S-adenosyl methionine to generate a cyclopropyl group at a double bond position of a FA (Grogan and Cronan, 1997). Cyclopropane FAs have been previously observed in other methanotrophic bacteria (Methylococcus capsulatus, Methylobacterium organophilum) during growth on methane and during oxygen limitation (Patt and Hanson, 1978; Jahnke and Nicholst, 1986). An intriguing aspect of these lipid-modifying enzymes in M. buryatense 5G(B1) is that both copies of cfa and desC are clustered together as part of a larger operon. Additionally the operon appears to encode a histidine sensor kinase, a lipase/esterase, and a sterol reductase gene. The physiological significance of this is unclear; however, it is tempting to speculate that these various activities are genetically linked as an operon for a coordinated membrane remodeling response in M. buryatense 5G(B1). The M. buryatense 5G(B1) genome lacks free fatty transport gene, fadL, but contains an acyl-CoA synthetase gene, acsA, which implies that M. buryatense 5G(B1) cannot transport fatty acids. However, the rest of genes, fadA, fadB, and fadE, responsible for the degradation and recycling of FAs via β-oxidation, were identified. The presence of the PlsXY pathway for phospholipids biosynthesis and the lack of a FadD homolog suggest that M. buryatense 5G(B1) is not capable of incorporating exogenous fatty acids into its phospholipid membranes.

Neither fabR/desT nor fadR (a known transcriptional regulators of fatty acid biosynthesis/degradation in E.coli) homologs were detected. The lack of any recognizable transcriptional regulatory factors is unique and could imply that FA biosynthesis in M. buryatense 5G(B1) is being regulated by an as-yet-to-be discovered mechanism. It seems unlikely that FA biosynthesis in methanotrophs is not regulated as formation of ICMs is commensurate with an increase of FA content per cell (Gilman et al., 2015).

The overreaching goal of the study was to identify targets to manipulate fatty acid biosynthesis in methanotrophs. The FBA suggests that about 7.5% of methane carbon enters the fatty acid biosynthesis pathway. According to genomic data, M. buryatense 5G(B1) possesses three different metabolic routes for production of acetyl-CoA from single carbon compounds with the main reaction contributing to acetyl-CoA formation predicted to be pyruvate dehydrogenase (Figure 1A). The FBA predictions are consistent with published enzymatic evidence and transcriptomic data (Figure 1; Kalyuzhnaya et al., 1999; De la Torre et al., 2015), as well as with transcriptomic data generated in this study.

ACP is one of the key proteins in FA biosynthesis. In the M. buryatense 5G(B1) genome there are three genes encoding ACP of which only one is highly expressed (Table 4). The differential regulation of ACP expression raised a question about whether M. buryatense 5G(B1) FA metabolism is saturated with ACP or whether overproduction of ACP would stimulate FA synthesis. If ACP is a limiting factor, its overexpression would lead to an overall increase of the FA production. Alternatively, adverse effects on cell physiology might be observed if cells do not tolerate higher ACP levels. To distinguish between these hypotheses, we chose to overexpress the highly expressed acp gene (WP_014148504.1). An ACP-overexpressing plasmid was constructed and introduced into the AP18 strain. Unlike most other strains tested in this study, the ACP-upregulated strain showed growth defects (Figure 2) and no signs of increased FA production (Table 2, Figure 3). Addition of FabB to the overexpressing ACP strain did not enhance its FAME content. Thus, higher levels of ACP production seem to be toxic for the methanotrophic culture, as seen for E. coli (Keating et al., 1995).

To determine whether FA biosynthesis in methanotrophs is limited by the elongation step, ketoacyl-ACP synthases (fabB) overexpression was examined. No changes in either the amount of FAME or composition of FAME were observed (Table 2, Figure 3). In order to define the role of β-oxidation in methanotrophic FA metabolism, the entire cluster of fadA, fadB, and fadE genes was deleted. The Δfad(ABE) mutant did not show significant changes in growth characteristics (Figure 2); furthermore, the Δfad(ABE)-mutant cells had higher levels of FAs (10% increase compared to original AP18 strain; Table 2, Figure 3).

Having established that neither ACP nor elongation of FA synthesis are limiting steps in FA accumulation, we then tested the possibility that directing carbon flux into acyl-CoA and malonyl-CoA pools would have a positive impact on total FA accumulation. M. buryatense 5G(B1) cells excrete acetate (~100 μmol gCDW⁻¹) and formate (~500 μmol gCDW⁻¹) during growth on methane (Gilman et al., 2015). The concentrations of excreted organic acids are not high; however, the leak of C₂-compounds might indicate an additional loss of carbon needed for FA biosynthesis. In order to increase carbon flow into FA biosynthesis two additional modifications were made: deletion of acetate kinase (Δack) and overexpression of acetyl-CoA carboxylase (acc). It should be mentioned that the overexpression of acetyl-CoA carboxylase (accABCD) in the M. buryatense 5G(B1) strain did not show a statistically significant increase (Puri and Lidstrom, unpublished data). Here, we tested acc-overexpression in different genetic backgrounds. The AP18Δack and overproducing Acc strains were found to produce 99 ± 7 and 108 ± 2 mg/gDCW of FAME, respectively (Table 2, Figure 3). Moreover, when combined together in the same strain, the effects of both mutations were additive resulting in the highest FA levels (111 ± 2 mg/gDCW, ~20% more than in the original AP18) achieved. Deletion of fadABE genes in AP18Δack::acc background did not increase FAME levels.

The M. buryatense 5G(B1) genome does not contain proteins homologous to typical bacterial regulators of FA metabolism, such as fabR, fadR, or desT. However, an orf (WP_017839568) with a weak homology to the nucleotide-binding Maf family proteins was identified upstream of the plsXfabHDGacpfabF-cluster. Homologous genes were identified in eight methanotrophic genomes (M. buryatense, Methylomicrobium alcaliphilum, Methylomicrobium marinus, Methylomicrobium luteus, M. sp MK WGS, M. sp.11b, Methylomicrobium methanica, and M. capsulatus), and in each case the gene located upstream of the FA biosynthesis genes. The conserved proximity of the fab-gene suggested it could be linked to FA metabolism. A WP_017839568 deletion mutant was generated and shown to have a severe growth defect (Figure 2). Cellular lipid content was also impaired: not only was the FAME content decreased but also there was an unusually high amount of C18:1 complemented by a decrease in C16:1 (Table 3). Reciprocal up-regulation (overexpression) of the WP_017839568 gene showed no growth defects and approximately the same FA profile as the initial strain (Table 3).

This striking phenotype of the WP_017839568-deleted strain suggested the possibility that this gene might be involved in the regulation FA synthesis, especially elongation of growing FA chains past 16 carbon atoms. In this regard, we proposed the name farE (for Fatty Acid Regulator of Elongation) for the WP_017839568 gene. In order to investigate the mechanism of such regulation, total RNA sequencing was performed (Table S3) for the following strains: (i) original AP18 strain (ii) AP18ΔfarE, (iii) AP18::farE, and (iv) AP18::pAWP78 (empty plasmid). By comparison of AP18 to AP18ΔfarE, ~1600 differentially expressed transcripts were identified. The majority of the differentially expressed genes are related to general metabolism and cellular functions (not shown), which could be explained by significant differences in growth rates between wild type and the farE-mutant. A subset of differentially regulated genes upon farE deletion is involved in FA biosynthesis or maintaining the acyl-CoA pool (Table 4). They include all three copies of the acpP gene present in M. buryatense 5G(B1) genome, two copies of the fabF and fabA genes, and the fabG and fabD genes. Two more insufficiently annotated genes included another copy of 3-ketoacyl-ACP synthase (WP_017841167.1) and a new thioesterase (WP_040575583.1, acyl-CoA thioester hydrolase). Considering their function in elongation of FA biosynthesis, upregulation of either 3-ketoacyl-ACP synthase (two copies of fabF and WP_017841167.1) could be a logical explanation for the elevated levels of the normally uncommon C18 FA in M. buryatense 5G(B1). Upon deletion of farE, the Acp expression pattern changes so that all three acpP genes are slightly upregulated (1.32x, 1.75x, and 2.13x).

In order to further narrow down the list of potential candidates for explaining changes in the ΔfarE mutant, we compiled a table (Table 5) including only those FA biosynthesis-related genes that are regulated in opposite directions in ΔfarE and farE overexpression strains. In addition, genes differentially regulated in AP18 and AP18::pAWP78 (empty plasmid) were also removed. That allowed us to narrow the list to only three entries (FabF, WP_026130034.1; 3-hydroxydecanoyl-ACP dehydratase, WP_017839698.1; 2-isopropylmalate synthase, WP_017840721.1). Effects of up- and down-regulation of these genes will be tested in future studies.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.