A tryptophan auxotroph of the strain “Micrococcus lysodeikticus” (M. luteus) ISU, trpE16 (Kloos and Schultes, 1969; Angelov et al., 2015), was grown in LB or modified Naylor medium at 30°C and used as host for DNA manipulations and also as a reference strain. The Naylor medium had a pH of 7.2 and contained per liter 10 g glutamate x H₂O, 7 g glucose, 5 g NH₄Cl, 2 g K₂HPO₄, 100 mg MgSO₄ × 7H₂O, 100 mg tryptophan, 10 mg biotin, 4 mg FeSO₄ × 7H₂O, 2 mg MnCl₂ × 4H₂O and was prepared with ultrapure water. For the growth studies with 20 mM amino acids as the only nitrogen source, NH₄Cl and biotin were omitted. The E. coli strain XL1-Blue (Stratagene, La Jolla, CA, United States) was grown in LB medium at 37°C and was used for cloning purposes and propagation of recombinant plasmids. Where appropriate, the growth media were supplemented with kanamycin (60 μg/ml for M. luteus, 20 μg/ml for E. coli), hygromycin (100 μg/ml for M. luteus) and 5-fluorocytosine (500 μg/ml for M. luteus, dissolved in water).

Potential orthologs of branched-chain alpha-keto acid dehydrogenase (BCKD-) complex components, orthologs of BCAA degradation enzymes, and orthologs of short and branched-chain acyl-CoA dehydrogenases were predicted in the genome of M. luteus by blastp using the nr database (Altschul et al., 1997). Amino acid sequence alignments of short- and branched-chain acyl-CoA dehydrogenases were generated by Clustal Omega (Goujon et al., 2010) with a blosum62 matrix and Geneious (version 10.0.2) (Kearse et al., 2012).

Homologous DNA sequence regions upstream and downstream of promoter-insertion sites as well as genes to be deleted, selection markers, and promoter sequences were amplified by PCR, using Q5 polymerase (New England Biolabs). The assembly of different PCR products with overlaps was achieved by Gibson Assembly (New England Biolabs). The in vitro Gibson assembly reactions (∼0.2–0.4 μg DNA) were added directly to M. luteus cells for uptake via natural transformation and plated on LB plates supplemented with the appropriate antibiotic, as described before (Angelov et al., 2015). For construction of markerless deletion mutants, the codAB counterselection method of Kostner et al. (2013) was used. In this approach, amplified flanking regions of the gene to be deleted were fused together and cloned in the plasmid pKOS6b using Gibson Assembly. Chemically competent E. coli XL-1 cells were transformed with the Gibson Assembly reactions. The purified recombinant plasmid, carrying the deletion allele was transformed in naturally competent M. luteus cells and plasmid integrants were selected on LB plates supplemented with kanamycin. Counterselection was performed on LB plates containing 5-fluorocytosine and the correctness of the alterations in the genome of M. luteus were confirmed by PCR and sequencing of the target genomic regions (Angelov et al., 2013). The M. luteus strains used in this study are listed in Table 1.

Micrococcus luteus cultures were grown at 30°C to stationary phase and 10 ml samples were taken. The samples were centrifuged in glass vials with PTFE screw-caps. The cell pellets were re-suspended in residual supernatant and 100 μl 100% acetic acid was added and thoroughly mixed. One ml of methanol and 4 ml of hexane, amended with 10 or 50 μg/ml triacontane (Tokyo Chemical Industry) as an internal standard were added. The whole mixture was shaken overnight at 640 rpm in KS 130B shaker. To facilitate phase separation, the samples were centrifuged and the upper hexane phase was used for analysis.

Cultures were grown to stationary phase and 2.5 ml samples were transferred to glass vials with PTFE screw-caps. The cell suspension was thoroughly mixed with 100 μl 100% acetic acid. Then 500 μl of a 100 μg/ml eicosanoic acid (Larodan AB) solution in ethanol as internal standard and 5 ml of methanol/chloroform (1:1) were added. The whole mixture was shaken overnight at 640 rpm and after centrifugation the lower chloroform phase was transferred to a new glass vial and evaporated completely. Lipid and free fatty acids were methylated by the addition of 500 μl 1.2 M HCl in methanol and incubation at 50°C with shaking overnight. The methylation reaction was quenched by adding 5 ml of 100 mg/ml NaHCO_{3 (aq)}. Finally, the fatty acid methyl esters (FAMEs) were extracted with 1 ml of hexane and the hexane phase was transferred into a GC/MS vial for analysis.

GC/MS analysis was performed using a Trace GC Ultra gas chromatograph in combination with a DSQ mass spectrometer (Thermo Scientific). The gas chromatograph was equipped with a PTV splitless injection system and a Rxi-5 ms (30 m × 0.25 mm i.d. × 0.25 μm) capillary column, and the MS was operated under ionization by electron impact at 70 eV and 200°C. The samples (1 μl) were injected in split mode (split 1:10). Helium flow was maintained at 1 ml/min. The temperature of the column was started at 40°C for 3 min, increased to 250°C at 15°C/min, and held for 19 min. Mass spectra were recorded at m/z 40-600 at a rate of 0.8741/s.

Quantification of olefins was done by direct comparison of their peak areas with the peak area of triacontane (C30 alkane) of known concentration, which had been added to the hexane with which the olefins were extracted. Quantification of FAMEs was done by direct comparison of their peak areas with the peak area of eicosanoic acid (C20:0) of known concentration, which had been added to the sample at the beginning of the extraction procedure.

For RNA extraction, the “ZR Fungal/Bacterial RNA MiniPrep” Kit (Zymo Research) and the corresponding protocol was used, using lysozyme treatment (at 1 μg/ml for 10 min) of the cells instead of bead beating for cell lysis. RNA was converted to cDNA by the “iScript Select cDNA Synthesis” Kit (BIO-RAD Laboratories). qPCR was conducted with “SsoAdvanced Universal SYBR Green Supermix” (BIO-RAD Laboratories). The relative quantification of the transcripts was performed with the 2^-ΔΔC_T method (Livak and Schmittgen, 2001).

The BCKD complex, using thiamine diphosphate, NAD⁺ and CoA-SH as cofactors, converts 3-methyl-2-oxopentanoate, 4-methyl-2-oxopentanoate and 3-methyl-2-oxobutanoate to 2-methylbutyryl-CoA, isovaleryl-CoA and isobutyryl-CoA, respectively, thereby releasing CO₂. This type of reaction is analogous to the activities of the pyruvate and 2-oxoglutarate dehydrogenase complexes, which show also amino acid sequence conservation. The enzyme complex consists of four subunits: two subunits of the branched-chain α-keto acid decarboxylase (E1), a lipoamide acyltransferase (E2), and a lipoamide dehydrogenase unit (E3). In the genome of Micrococcus luteus three gene clusters could be found which potentially code for the pyruvate dehydrogenase-, 2-oxoglutarate dehydrogenase- and BCKD complexes: Mlut_06800-06820, Mlut_13320-13340 and Mlut_17810-17790 (numbering in the order of transcription of the genes of the operons). The Mlut_06800-06820 and Mlut_17810-17790 operons encode E1 and E2 subunits, while the Mlut_13320-13340 operon possesses only genes for E2 and E3 subunits. It is known that different dehydrogenase complexes can share the same E3 monomer (Namba et al., 1969; Wang et al., 1993).

In order to identify if one of the two candidate clusters with genes for the E1 and E2 subunits codes for the BCKD gene cluster we exchanged the promoter regions with potentially strong promoters (promoter region of Mlut_05030 in the case of Mlut_06800, promoter region of Mlut_04850 in the case of Mlut_17810; the promoter regions were chosen based on gene expression data derived from the microarray experiment of Pereira et al., 2012). Changes in the transcription level were checked by qPCR of the target genes after growth to exponential phase in minimal medium. Relative quantification by the 2^-ΔΔC_T method revealed that the genes Mlut_06800 and Mlut_17810 were strongly overexpressed in those mutants in comparison to the parental trpE16 strain (Figure 1A). The GC/MS analysis of olefins produced by the promoter exchange mutants showed that overexpression of the gene cluster Mlut_06800-06820 led to a 3-fold increase of olefin production in our minimal medium. The promoter exchange mutations upstream of Mlut_17810 had no effect on olefin production (Figure 1B). Analysis of the fatty acids by GC/MS (Figure 1C) showed that the overexpression of cluster Mlut_06800-06820 had no effect on the fatty acid amount, while the promoter exchange mutation in Mlut_17810-17790 caused a drop of fatty acid production by about 30%. This is in line with the idea that an increased amount of active pyruvate- and 2-oxoglutarate-dehydrogenase would lead to induction of the citric acid cycle, which is in direct competition with the fatty acid synthesis for acetyl-CoA. The composition of the fatty acids as well as of the olefin isomers was not affected by the overexpression of Mlut_06800-06820 (data not shown).

To further investigate the metabolic routes for branched-chain fatty acids synthesis in M. luteus, we generated a BCKD complex deletion strain in which the Mlut_06800-06820 gene cluster was replaced with a hygromycin resistance cassette. The deletion of the genes of the BCKD complex led to a strong growth inhibition of the strain in minimal medium, while in complex medium normal growth was observed. In both media, no olefins could be detected in this strain. In our standard minimal medium, a strong shift in the fatty acid composition was observed, with more than 80% of the fatty acids being straight chain, as opposed to the wild type where the fraction of sc fatty acids was less than 16%. This shift was accompanied by a significant decrease in the total amount of fatty acids (Figure 2A). Similar phenotype was also observed in complex medium, where the total amount of fatty acids produced in the BCKD mutant was drastically reduced to less than 20% of that of the parental strain and the fraction of the sc fatty acids was more than 60% (Figure 2). These dramatic changes in the fatty acid composition and the predominance of straight over branched fatty acids may have reached a critical level in minimal medium. The requirement of such a critical level of branched-chain fatty acids for growth has been demonstrated in B. subtilis (Kaneda, 1988). Branched-chain fatty acids in Gram-positive bacteria as well as unsaturated straight chain fatty acids in Gram-negative bacteria, which are incorporated into membrane lipids, contribute to an increased membrane fluidity due to their low phase transition temperatures (Kaneda, 1991). It seems that in the case of the BCKD mutant in minimal medium the fraction of branched-chain fatty acids in the membrane lipids has been reduced to such an extent that cell division is inhibited, probably due to the reduced membrane fluidity.

Supplementation of the minimal medium with valine, leucine or isoleucine (at 4 mM) led to increased production of the corresponding branched fatty acids in the parental strain trpE16 but not in the BCKD deletion mutant (Figure 3A). This result further supports our assignment of the Mlut_06800-06820 genes as encoding the components of the BCKD complex. The observed low level of branched fatty acids in the BCKD deletion mutant in minimal medium could result from unspecific activity of the pyruvate- or the 2-oxoglutarate dehydrogenase complexes.

An alternative route to produce branched fatty acid primer molecules are short- and branched-chain specific acyl-CoA synthetases, which activate externally supplied short branched-chain carboxylic acids directly to the corresponding fatty acid primer molecules. To test if this alternative route is present in M. luteus trpE16 and is still active in the ΔBCKD mutant, the mutant as well as the parental strain were supplemented with 2-methylpropanoic (isobutyric) acid, 3-methylbutyric (isovaleric) acid or 2-methylbutyric acid (at 4 mM) in minimal medium. In both strains the supplementation allowed the synthesis of the corresponding branched fatty acids, and in the case of the deletion mutant the effect of the supplementation was more pronounced because of the missing background production (Figure 3B). In addition, these supplementations restored the olefin production defect in the mutant, leading to olefin isoforms corresponding to the supplied precursor (Figure 3B).

The branched fatty acid primer molecules 2-methylbutyryl-CoA, 3-methylbutyryl-CoA (isovaleryl-CoA) and 2-methylpropanoyl-CoA (isobutyryl-CoA) are also intermediates in the degradation pathway of the BCAAs isoleucine, leucine and valine, respectively. In the supplementation experiment above we observed that the availability of these primers can have a major impact on the isomeric composition of the olefins produced by M. luteus. In B. subtilis, BCAAs degradation follows a common pathway up to the level of the activated acyl-CoA derivatives which are further catabolized by specific acyl-CoA dehydrogenases (Massey et al., 1976). We reasoned that the deletion of specific dehydrogenases of the BCAA degradation pathways could lead to an increased supply of the respective acyl-CoA for fatty acid and olefin biosynthesis and could be used as a means to engineer the branching pattern of the olefins of Micrococcus.

In order to identify the genes encoding the enzymes of the downstream part of the BCAA degradation routes in M. luteus, we searched the genome for putative short-chain acyl-CoA dehydrogenases. We identified 7 candidate ORFs and inspected their genomic surroundings. Of the 7 candidate short-chain acyl-CoA dehydrogenases, 3 were found in gene clusters potentially involved in the degradation of BCAAs, i.e., ORFs Mlut_02820, Mlut_02900 and Mlut_06870. Mlut_02820 and Mlut_02900 were located near a set of genes which seem to encode almost all steps of the isoleucine/valine degradation pathways, and Mlut_06870 was found in a gene cluster which appears to encode all steps of the leucine degradation pathway (Figure 4). Amino acid sequence alignments of these dehydrogenases with homologs with known substrate specificities supported the annotation of Mlut_02820 as isobutyryl-CoA dehydrogenase (valine degradation), while Mlut_06870 was most similar to an enzyme from Streptomyces avermitilis with broad substrate specificity (Djordjevic et al., 1995; Tiffany et al., 1997; Zhang et al., 1999; Battaile et al., 2002; Battaile et al., 2004; Förster-Fromme and Jendrossek, 2008) (Figure 5).

Next, we generated deletion strains for all seven predicted short-chain acyl-CoA dehydrogenase ORFs by allelic exchange with a kanamycin marker. Some of these deletions were introduced also in a M. luteus trpE16 strain with elevated expression of the ole genes, named M. luteus ope (for olefin promoter exchange). The effect of each deletion on the isomeric composition of fatty acids and olefins in M. luteus allowed us to further specify the substrate specificities of the dehydrogenases. Additional experiments were performed, i.e., addition of BCAA to the growth medium and deletion of putative transcriptional regulators in the BCAA degradation gene clusters, in order to reveal the phenotypes of the generated mutants and to aid in deducing the substrate specificity of the acyl-CoA dehydrogenases. Due to their localization in putative BCAA degradation gene clusters, we focused on the deletion mutants of ORFs Mlut_02820, Mlut_02900 and Mlut_06870.

Under the growth conditions used by us, notable changes in the amounts and/or the isomeric composition of fatty acids and olefins could be observed only for the deletion mutant Δ06870:kan. Compared to the parental strain, the Mlut_06870 deletion mutant displayed a shift toward iso-branched odd-numbered fatty acids (from leucine-derived acyl-CoA primer) and in consequence a shift toward isoiso-branched olefins. These shifts in the fatty acids and olefin profiles may be explained by the increased supply of the leucine-derived acyl-CoA primer (isovaleryl-CoA) which is expected to accumulate if its further degradation is blocked by the deletion of the isovaleryl-CoA dehydrogenase. This was further supported by leucine supplementation experiment, where in the mutant the leucine-caused shift to isoiso-olefins was more pronounced (Figure 6).

Under our standard growth conditions, the deletions of Mlut_02900 and of Mlut_02820 did not have a measurable effect on the olefin or on the fatty acid profile of M. luteus (Figure 6). However, valine supplementation revealed marked differences in the branching pattern of the olefins of both Δ02900:kan and Δ02820:kan strains in comparison to the parental strain. In the parental strain, valine addition led to a shift to almost exclusively isoiso-olefins (and iso-even fatty acids), while in the Δ02900:kan and Δ02820:kan strains this shift was less pronounced and the mutants had a higher proportion of isoai- and aiai-olefins (Figure 6B). The observed olefin profiles of Δ02900:kan and of Δ02820:kan after valine supplementation are in agreement with a specificity of these dehydrogenases toward ai-acyl-CoA derivatives (2-methyl-butyryl-CoA, isoleucine degradation).

Interestingly, supplementation of the minimal medium with 4 mM isoleucine resulted in both cases, the reference strain as well as in the Mlut_02900 deletion strain, in a threefold higher total olefin production (from ∼8 mg per gram dry cell weight without supplementation to ∼25 mg per gram cells with isoleucine supplementation) of almost exclusively the anteisoanteiso- (aiai-) configuration type (Figure 7), apparently due to the overall metabolic preference of M. luteus for isoleucine derivatives. Similar increase in olefin amounts could be observed also after supplementation with isoleucine of wild type and Mlut_02820 deletion mutant cultures.

Further evidence about the specificity of the acyl-CoA dehydrogenases of the Mlut_02820-2900 gene cluster was obtained by deleting the potential transcriptional regulator Mlut_02880 which acts as a repressor of all adjoining genes of this cluster. This regulation scheme could be revealed by qPCR and was confirmed by RNAseq data (unpublished results). A single deletion of the regulator (Δ02880:hyg) as well as double deletion mutants of the regulator and each of the two dehydrogenases of the cluster were generated in a M. luteus ope background and were further investigated by GC/MS. Compared with the control strain, deletion of the regulator gene Mlut_02880 led to an increase of leucine-derived (iso-branched odd-numbered) and to a drop of valine-derived (iso-branched even-numbered) and isoleucine-derived (ai-branched odd-numbered) fatty acids (Figure 6A). With respect to olefin production this resulted in an almost uniform distribution of isoiso- and aiai-branched olefins (Figure 6B). These changes in the isomer composition of the fatty acids and olefins can be explained by higher activities of acyl-CoA dehydrogenases with specificities toward the valine- and isoleucine-derived acyl-CoAs in the regulator deletion strain. In agreement with this interpretation, the additional deletion of either one of the two nearby dehydrogenase genes (Mlut_02900, Mlut_02820) in the Δ02880:hyg background fully restored the fatty acid/olefin isomer profile of the wild-type.

We hypothesized that, if an externally supplied BCAA is the only source of nitrogen, the ability of M. luteus to grow in such medium would be indicative of its ability to catabolize the supplied amino acid (Dawes, 1952). The M. luteus strain used by us was unable to grow in glucose-containing medium with isoleucine or valine as the only source of nitrogen, while good growth could be observed with leucine. We assumed that the lack of growth on isoleucine and valine may be due to the continuous repression of the degradation gene cluster Mlut_02820-2900 by the regulator Mlut_02880. However, an additional growth study using the regulator deletion strain revealed that also under conditions of induction of the isoleucine and valine degradation cluster growth was still not possible on these amino acids (data not shown).

Growth with leucine was no longer possible for the Mlut_06800-06820 and for the Mlut_06870 deletion mutants. Deletion of the BCKD complex (Mlut_06800-06820) on one hand prevents the formation of branched-chain fatty acid primer molecules from any supplemented amino acid and on the other hand leads to the accumulation of metabolic intermediates (α-keto acids, substrates of the BCKD reaction) which presumably inhibits further deamination of amino acids and formation of glutamate, which is crucial for the biosynthesis of other amino acids. Therefore, growth of this mutant in minimal medium with supplementation of any of the tested amino acids did not occur. Interestingly, glutamate supplementation also did not allow growth of the BCKD mutant (Figure 8). As shown previously (see above), deletion of Mlut_06870 did not prevent free fatty acid synthesis, but this block in leucine catabolism may lead to the accumulation of the upstream intermediates and thereby inhibit a sufficient (i.e., growth-supporting) extent of deamination of leucine and formation of the central N-metabolite glutamate. In accordance, the supplementation with glutamate allowed growth of M. luteus trpE16 Δ06870:kan.

The ORF Mlut_06840 codes for a putative transcriptional regulator and is located near the predicted leucine degradation genes Mlut_06850-06890 and simultaneously in the vicinity of the BCKD complex-encoding genes Mlut_06800-06820 (see Figure 4). This putative regulator belongs to the TetR family whose members are known to control catabolic pathways.

To identify possible targets of the regulator encoded by Mlut_06840 among the surrounding branched amino acid degradation genes, the levels of expression of Mlut_06800 and Mlut_06850 in the deletion mutant trpE16 Δ06840 (markerless) and in the parental strain trpE16 were analyzed by qPCR. Relative quantification by the 2^-ΔΔC_T method revealed that the product of Mlut_06840 acts as a repressor on the genes of the cluster Mlut_06850-90, but not on Mlut_06800-06820, since deletion of Mlut_06840 led to an induction of the leucine degradative pathway (Figure 9). On the level of fatty acids, the induction of the Mlut_06850-06890 gene cluster during growth in minimal medium led to a shift from branched fatty acid production, which derive from any kind of BCAA, to sc fatty acid production. In complex medium the deletion of Mlut_06840 negatively affected only the odd-numbered leucine-derived fatty acids (Figure 6A). Also, regarding olefin production, the induction of Mlut_06850-06890 in the Mlut_06840 mutant led to less branched-chain and to more straight chain olefin isomers in minimal medium, which reflects the shift from branched to more straight chain fatty acid production (Figure 6B).