T. fusca (strain BAA-629) was obtained from the American Type Culture Collection (ATCC) (Manassas, Virginia). The frozen stock was revived as directed by ATCC using their standard TYG 741 media, and cultures were incubated at 50°C. Experimental cultures were grown in 100 ml of Hagerdahl media (ATCC medium 2382) supplemented with 0.5% (w/v) of a carbon source: glucose, cellobiose, cellulose, or avicel. Each culture was initiated with 3 ml of inoculum and cultures were grown at either 37°C or 50°C for up to 2 days.

Cells were collected from each culture by centrifugation at 5000 × g for 5 min, and the wet weight of the cell pellet was determined. Cultures grown at the suboptimal temperature (37°C) did not consume all of the insoluble solid carbon source (i.e., cellulose or avicel). These cell pellets were washed with sterile water and the final, washed cell pellet was weighed. Cell pellets were flash frozen in liquid nitrogen, lyophilized for 48 h, and the dry weight was recorded prior to fatty acid extraction.

Fatty acids were extracted from three aliquots of cells taken from a T. fusca culture. Cells were pelleted and lyophilized. Lyophilized cell pellet aliquots (10 mg each) were transferred to a glass tube and spiked with 10 µg of nonadecanoic acid as an internal standard. The pellets were then suspended by vortexing for 15 min in a solution of 5% (v/v) sulfuric acid in methanol and the suspension was incubated at 80°C for 1 h. After cooling to room temperature, 1 ml of hexane: chloroform (4:1 v/v) solution and 1 ml of 0.9% (w/v) NaCl were added to each tube, and the mixture was vortexed for 5 min. The organic and aqueous phases were separated by centrifugation, and the organic phase containing the resulting fatty acid methyl esters was transferred to a separate test tube. The aqueous phase was extracted an additional time with hexane: chloroform (4:1 v/v) solution, and the organic phases were collected and pooled. The extracts containing fatty acid methyl esters were concentrated by evaporation under a stream of nitrogen gas. GC-MS analysis was performed with an Agilent 6890 GC equipped with a DB-1 MS capillary column (Agilent 122–0112). Chromatography was performed using helium gas at a flow-rate of 1.2 ml/min, using an inlet temperature set at 280°C. Individual fatty acid methyl esters were identified by GC-MS fragmentation patterns in tandem with NIST AMDIS software (Stein, 1999), as well as by comparing their retention times to known fatty acid methyl ester standards obtained from Supelco Inc. (Bellefonte, PA) and Metraya LLC (State College, PA). Peak areas of individual fatty acid methyl esters were integrated with AMDIS software, and these were converted to fatty acid concentrations relative to the peak area of the known amount of nonadecanoic acid internal standard that was added to each sample.

Principal component analysis was conducted using Metaboanalyst software (Pang et al., 2021) with data from all replicates of T. fusca cultures grown in each of the eight growing conditions. Quantitative fatty acid abundance data (µmoles/g dry cell weight) were uploaded from Microsoft Excel to Metaboanalyst statistical software and autoscaling was done as enabled by Metaboanalyst to create a PCA plot and a PCA biplot.

The sequenced T. fusca genome was queried with the BLASTP algorithm (Altschul et al., 1990) using query sequences of experimentally confirmed acetyl-CoA carboxylase (ACCase) and Type II fatty acid synthase (FAS) components, either originating from Actinomycetes or from Escherichia coli (Cronan and Thomas, 2009; Shivaiah et al., 2021). Additionally, genes were identified using the KEGG genome browser (Kanehisa et al., 2002) based on sequence homology, key processes, and operon organization.

Fatty acid analyses show that when T. fusca was grown at 50°C it primarily synthesizes fatty acids that are between 14 and 18 carbons in length, although trace amounts of 13-carbon fatty acids were also detected. The majority of the fatty acids produced were BCFAs, particularly iso- or anteiso-BCFAs (i.e., the methyl branch is present at the ω-1 or ω-2 position of the acyl chain, respectively) (Figure 1). Small amounts of mid-chain BCFAs were also detected, these being 10-methyl BCFAs ranging between 16 and 18 carbons in length. Linear fatty acids were also present, accounting for approximately 10% of total fatty acids produced. Additionally, we determined that T. fusca can produce minor amounts (<5%) of mono-unsaturated C16 and C18 fatty acids with a single double bond at the 9th position.

Overall fatty acid yield was 2- to 3- fold higher when T. fusca was grown at the optimal growth temperature (50°C) as compared to growth at the suboptimal temperature (37°C) (Figure 2), and this phenomenon was observed independent of the carbon sources that were evaluated. Principal component analysis of the fatty acid compositional data visualized the factors contributing to the different fatty acid profiles of T. fusca through generation of a PCA plot (Figure 3). Principal component 1 accounts for ∼52% of the sample variation, indicating that the samples are primarily separated by growth temperature, whereas carbon source contributed to ∼26% of the sample variation (represented by principal component 2). Indeed, the data points in the PCA plot cluster distinctly by growth temperature, and to a lesser extent by carbon source, with soluble and insoluble carbon sources typically clustering together within each growth temperature cluster. The conclusion that growth temperature is the primary driver of fatty acid composition is indicated by the PCA biplot (Supplementary Figure S1), as most of the vectors representing the different fatty acids are pointed horizontally, indicating that they contributed more to principal component 1. Additionally, some of the longer-chain fatty acids (18-carbons and longer) and the 10-methyl BCFAs are more associated with changes in the carbon source, as their vectors point more vertically.

The fatty acid profile of T. fusca shifted when it was cultured at its suboptimal temperature; the relative abundance of BCFAs was increased when the bacterium was cultured at 37°C (Supplementary Figure S2). Growth temperature also affected the acyl-chain lengths of the fatty acids produced by T. fusca. While C13, C14, and C15 fatty acid species account for approximately 10–15% of the fatty acids present at 50°C, they are almost completely absent from the cultures grown at 37°C (Supplementary Figure S3). Although unsaturated fatty acids were only detected at trace amounts at the optimum growth temperature, they make up a more significant proportion of the total fatty acid content when T. fusca was cultured at 37°C (∼5%).

While growth temperature was the primary factor determining the types of fatty acids produced, carbon source also contributed to a shift in fatty acid profiles. Changes were primarily observed when comparing cultures grown on soluble versus insoluble carbon sources. Specifically, the fatty acid profiles of T. fusca grown on cellulose were very similar to those obtained when T. fusca was grown on Avicel, while fatty acid profiles of cultures grown on cellobiose resembled those grown on glucose. The main shift between the cultures grown on the soluble versus insoluble carbon sources can be attributed to the types of BCFAs present. While iso-BCFAs are the main species present in all growth conditions, when T. fusca was grown on soluble carbon sources we observed an even higher proportion of iso-BCFAs at 37°C. This relationship is reversed when using insoluble carbon sources; namely there is a higher proportion of iso-BCFAs at 50°C. In both cases, the change in abundance of iso-BCFAs is complemented with a corresponding change in the abundance of anteiso-BCFAs.

Although many of the enzymes involved in fatty acid biosynthesis have not been specifically characterized from T. fusca, they are identifiable by their sequence homology to enzymes from other bacteria, and by the operon organization in the T. fusca genome (Table 1). Querying the sequence of the T. fusca genome indicates that like many other bacteria, T. fusca utilizes a Type II fatty acid synthase (FAS) system to assemble fatty acids (Cronan and Thomas, 2009; Gago et al., 2011; Gago et al., 2018). In most organisms, FAS utilizes acetyl-CoA and malonyl-CoA as substrates, and the latter substrate is generated by the carboxylation of the former, a reaction catalyzed by acetyl-CoA carboxylase (Waldrop et al., 2012). Multiple iterations of the FAS cycle using these two substrates generates linear, saturated fatty acids, but these are minor components in T. fusca. The iso-and anteiso-BCFAs that account for a large portion of the fatty acids of T. fusca are produced by this FAS system by using branched-chain acyl-CoA substrates, rather than acetyl-CoA, which can be generated by the deamination of branched chain amino acids (i.e., valine, leucine, or isoleucine) (Kaneda, 1991; Beck et al., 2004; Zhu et al., 2005). Thus, the fatty acid biosynthetic machinery of T. fusca can be considered as consisting of at least four modules (Figure 4): 1) the module that generates the acyl-CoA starting substrate for FAS; 2) the module that generates the malonyl-CoA elongating substrate for FAS; 3) the FAS system itself; and 4) a fatty acid modifying module which generates the unsaturated and the internally BCFAs.

The acyl-CoA substrates required by this organism’s FAS system are products of primary metabolism from sugars, generating acetyl-CoA, as well as products of branched chain amino acid metabolism, generating isobutyryl-CoA and 2-methylbutrylyl-CoA. Sequence homology identified multiple candidates for both the branched chain aminotransferase (Tfu_0616, Tfu_2112) and branched chain α-keto acid dehydrogenase (Tfu_0180, Tfu_0181, Tfu_0182) enzymes required to generate isobutyryl-CoA and 2-methylbutyryl-CoA from valine, leucine, or isoleucine. Alternatively, these branched chain acyl-CoAs may be generated from the α-keto acids that are intermediates of branched chain amino acid biosynthesis. Indeed, the branched chain aminotransferase encoded by Tfu_0616 is located in the genome adjoining genes that encode enzymes involved in branched chain amino acid biosynthesis, including genes with high sequence homology to acetolactate synthase (Tfu_0611, Tfu_0612), keto-acid isomeroreductase (Tfu_0613), 3-isopropylmalate dehydrogenase (Tfu_0615), 2-isopropylmalate synthase (Tfu_0617), and 3-isopropylmalate dehydratase (Tfu_0626, Tfu_0627). (Franco and Blanchard, 2017). Acetyl-CoA can be produced through several biological processes (Krivoruchko et al., 2015), including the oxidative decarboxylation of pyruvate catalyzed by the pyruvate dehydrogenase complex (PDH) (Tfu_3049, Tfu_3050, Tfu_3051). Alternatively, acetyl-CoA can be generated through the activation of acetate catalyzed by: 1) an acetate kinase (Tfu_2971); 2) an AMP-forming acetyl-CoA synthetase (Tfu_1546, Tfu_2808, Tfu_2856); or 3) an ADP-forming acetyl-CoA synthetase (Tfu_1302, Tfu_1473). T. fusca also possesses three genes that encode for proteins that resemble citryl-CoA lyase (Tfu_0341, Tfu_1285, Tfu_1313), a component of the reductive TCA cycle capable of converting citryl-CoA to acetyl-CoA and oxaloacetate (Aoshima et al., 2004; Hügler et al., 2005; Hügler et al., 2007; Katiyar et al., 2018).

The biotin-containing enzyme, acetyl-CoA carboxylase (ACCase) converts acetyl-CoA to malonyl-CoA, a reaction that is classically considered the first and rate-limiting reaction of fatty acid biosynthesis. As with all biotin enzymes, sequences of these proteins can be recognized by sequence homology among three different functional domains: the biotin carboxylase (BC), biotin carboxyl carrier protein (BCCP), and the carboxyl transferase (CT) domains (Cronan and Waldrop, 2002; Tong, 2013). The tertiary and quaternary organization of these domains varies considerably, depending on the phylogeny of the organism. E. coli, for examples, has ACCase components that are organized as individual proteins that come together to form the enzyme complex. In contrast, several ACCases from Actinomycetes consist of two subunits: the A subunit that encompasses both the BC and BCCP functional domains, and the B subunit that encompasses the CT domain (Gago et al., 2011; Gago et al., 2018). Additionally, some Actinomycete ACCases have a third non-catalytic subunit, E, that is needed for proper assembly of the holoenzyme complex (Shivaiah et al., 2021). Moreover, the A and B subunit quaternary organization of biotin enzymes is also common to propionyl-CoA carboxylase (Tong, 2013) and methylcrotonyl-CoA carboxylase (Song et al., 1994; McKean et al., 2000; Wurtele and Nikolau, 2000), which complicates the sequence-based identification of ACCase in the T. fusca genome.

Previous studies have experimentally characterized the T. fusca operon (Tfu_2555, Tfu_2556, Tfu_2557) that encodes the B, E, and A subunits of an acyl-CoA carboxylase (AcCCase) (Shivaiah et al., 2021). This enzyme is promiscuous and can carboxylate acetyl-CoA, propionyl-CoA, and butyryl-CoA. Other Actinobacteria (e.g., Streptomyces coelicolor) also express such a promiscuous carboxylase, in addition to a highly specific propionyl-CoA carboxylase (Gago et al., 2018). The more promiscuous AcCCase enzymes can thereby generate not only malonyl-CoA, but also methylmalonyl-CoA and ethylmalonyl-CoA, and may therefore have multiple metabolic functions. For example, the catabolism of valine, isoleucine and odd-numbered fatty acids generates propionyl-CoA, which is further metabolized via the TCA cycle after the sequential conversion to methylmalonyl-CoA and succinyl-CoA (Wongkittichote et al., 2017). Alternatively, methylmalonyl-CoA and ethylmalonyl-CoA can be used as substrates by polyketide synthases, generating polyketides with methyl- or ethyl-branches in the final structure (Khosla and Keasling, 2003; Risdian et al., 2019).

Our BLAST-based search of the T. fusca genome identified additional genes that encode homologs of biotin-containing carboxylase proteins, namely Tfu_0947, Tfu_0948, Tfu_1215, Tfu_1228, Tfu_1513, and Tfu_1530. The adjacent Tfu_0947 and Tfu_0948 genes suggest that they are on a single operon, with Tfu_0947 encoding a subunit with the BC and BCCP domains, and Tfu_0948 encoding a subunit carrying the CT domain. This subunit/domain organization suggests that this operon may encode either a propionyl-CoA carboxylase (Gago et al., 2018) or methylcrotonyl-CoA carboxylase (Tomassetti et al., 2018), although ACCases with such quaternary subunit organizations also occur in many Actinomycetes, including Streptomyces coelicolor and Mycobacterium tuberculosis (Gago et al., 2011; Tong, 2013).

The Tfu_1228 and Tfu_1215 genes encode proteins that resemble CT subunits of biotin-carboxylases, but neither gene lies within an operon that houses other functional subunits necessary to form the holoenzyme complex. It is a common feature among Actinomycetes to mix and match different CT-subunits with a common BC/BCCP subunit, and thus generate different enzymatic capability with a single BC/BCCP subunit (Gago et al., 2011; Gago et al., 2018). The Tfu_1228 and Tfu_1215 genes may instill such a mechanism, and thus these genes could also provide a means for generating malonyl-CoA for FAS. A similar mechanism may occur in T. fusca, as Tfu_1228 and Tfu_1215 could be part of additional ACCase complexes that use an A subunit from another operon (such as Tfu_2557 or Tfu_0947). An additional gene with sequence homology to known ACCase components is Tfu_1513, which encodes a protein with high sequence homology to a BCCP that is not located near a BC or CT domain. Such a genome organization is similar to that found in E. coli, where the BC, BCCP, and CT components are separated into four individual proteins encoded by 3 separate operons (Cronan and Waldrop, 2002; Gago et al., 2018).

The Tfu_1530 gene encodes a large protein of 1849 residues, and it appears to encompass all three catalytic domains required for the carboxylation reaction (i.e., BC, BCCP and CT domains). This homomeric domain organization is common to such biotin carboxylases as ACCases from eukaryotes (i.e., plants, fungi, and animals) (Nikolau et al., 2003; Sasaki and Nagano, 2004), pyruvate carboxylases (Jitrapakdee and Wallace, 1999) and a long chain acyl-CoA carboxylase from Mycobacterium species (Tran et al., 2015; Lyonnet et al., 2017). The latter enzyme is involved in the biosynthesis of mycolic acid, a fatty acid specifically associated with the Mycobacterium genus (Marrakchi et al., 2014). Thus, the specific enzymatic function encoded by the Tfu_1530 gene is not recognizable by sequence homology but may include the ability to generate malonyl-CoA for FAS.

Type II FAS systems use cyclic iterations of four reactions that are each catalyzed by distinct enzymes (Cronan and Rock, 2008). Each cycle of the process adds two carbon atoms to the growing acyl-chain, with the donor of these two carbon subunits being the malonyl moiety of malonyl-CoA. The malonyl-moiety is first loaded onto the acyl-carrier protein (ACP) subunit of the FAS system, a reaction catalyzed by malonyl-CoA: ACP transacylase (MCAT). Subsequently, each FAS cycle begins with a Claisen condensation reaction between a preexisting “starting” acyl-CoA or acyl-ACP and malonyl-ACP to generate a 3-ketoacyl-ACP intermediate, which is 2-carbons longer than the initial acyl moiety. The first of these Claisen condensation reactions is between an acyl-CoA and malonyl-ACP, catalyzed by a 3-ketoacyl-ACP synthase III (encoded by the FabH gene). The chemical nature of the acyl-CoA substrate used in this condensation reaction determines the nature of the ω-end of the resulting fatty acid product; namely, utilizing acetyl-CoA, isobutyryl-CoA or methylbutyryl-CoA as the substrate leads to the generation of linear, iso-BCFA or anteiso-BCFA, respectively. The subsequent three reactions of each FAS cycle involve sequential reduction, dehydration and further reduction, catalyzed by 3-ketoacyl-ACP reductase (FabG), 3-hydroxyacyl-ACP dehydratase (FabA), and enoyl-ACP reductase (FabI), respectively. The product of each FAS cycle results in the generation of an acyl-ACP product that is two carbons longer than the pre-loaded acyl chain, and it serves as the substrate for the Claisen condensation reaction of the next round of the FAS cycle; these subsequent Claisen condensation reactions with a malonyl-ACP substrate are catalyzed by a 3-ketoacyl-ACP synthase II isozyme (FabF). These catalytic processes generate saturated fatty acids, and typically the process is terminated by transfer of the acyl moiety from acyl-ACP to glycerol-3-phosphate (glycerol-3-P), a reaction catalyzed by glycerol-3-P acyltransferases (GPATs) (Yao and Rock, 2013). The substrate specificity of GPATs determine the chain-lengths of the fatty acids produced by FAS, and in most bacteria they are typically of between 14 and 18 carbon atoms.

The T. fusca genome contains multiple operons that could encode for FAS genes, although the similarity to enzymatic functions associated with polyketide biosynthesis, catalyzed by Type II polyketide synthase or a nonribosomal peptide biosynthetic system (Corre and Challis, 2009), may confound these identifications. Specifically, Tfu_1973 (MCAT) is part of a large operon that includes Tfu_1974 (FabH), Tfu_1975 (ACP), and Tfu_1976 (FabF). Another copy of MCAT is encoded by Tfu_1231, which is also positioned near an additional copy of FabH (Tfu_1229). We found three FabG genes in the T. fusca genome that appear to encode for the 3-ketoacyl-ACP reductase. Two of these FabG genes, Tfu_1841 and Tfu_1843, are located in the same operon that also contains Tfu_1842 (FabI). A third copy of FabG (Tfu_2308), is found separately and does not appear to be part of a larger operon.

The fatty acid modification module includes genes that are required to generate unsaturated and 10-methyl BCFAs. Recent studies have indicated that Actinomycetes express a three-reaction pathway that metabolically links these two fatty acids. Specifically, a Δ-9 acyl-CoA desaturase (Tfu_0413) (Lykidis et al., 2007) can generate monounsaturated fatty acids, which are substrates for BfaB (Tfu_2160) and BfaA (Tfu_2161) enzymes, which transform the monounsaturated acyl-CoA to 10-methylene BCFA and to 10-methyl BCFA, respectively (Blitzblau et al., 2021).