Fatty Acid and Alcohol Metabolism in Pseudomonas putida: Functional Analysis Using Random Barcode Transposon Sequencing

To characterize the genetic basis of fatty acid metabolism in P. putida, barcoded transposon mutant libraries were grown in minimal medium with straight-chain fatty acids (C₃ to C₁₀, C₁₂, and C₁₄), fatty esters (Tween 20 and butyl stearate), and an unsaturated fatty acid (oleic acid) as sole carbon sources. An overview of sample collection is provided in Fig. 1A. Pearson correlation analyses of global fitness patterns revealed that the metabolisms of straight-chain fatty acids between C₇ and C₁₄ form a clade together, suggesting similar overall catabolic routes (Fig. 1B). Oleic acid, an 18-carbon monounsaturated fatty acid, also grouped within this clade. Shorter-chain fatty acids (shorter than C₇) did not show high correlation with one another based on global fitness analyses, suggesting more independent routes of catabolism (Fig. 1B). Annotations in the BioCyc database, functional assignment from a recent metabolic model of P. putida KT440 (iJN1462), and previous in vitro biochemical work predict the existence of several enzymes encoded in the genome of the bacterium that may be putatively involved in fatty acid catabolism: six acyl-coenzyme A (CoA) ligases, seven acyl-CoA dehydrogenases, seven enoyl-CoA hydratases, four hydroxyacyl-CoA dehydrogenases, and five thiolases (Fig. 2) (17, 24, 42). Our data show discrete fitness patterns for the steps of beta-oxidation that appear to be largely dictated by chain length (Fig. 2).

When grown on fatty acids, many bacteria require the anaplerotic glyoxylate shunt to avoid depleting tricarboxylic acid (TCA) cycle intermediates during essential biosynthetic processes. In P. putida, the two steps of the glyoxylate shunt are encoded by PP_4116 (aceA, encoding isocitrate lyase) and PP_0356 (glcB, encoding malate synthase). Transposon mutants in both of these genes showed serious fitness defects (fitness score, less than −3) when grown on nearly all of the fatty acids tested (Fig. 2). However, the glyoxylate shunt genes appeared to be dispensable for growth on valerate (C₅), and showed a more severe fitness defect when grown on heptanoate (C₇). Complete beta-oxidation of valerate and heptanoate results in ratios of propionyl-CoA to acetyl-CoA of 1:1 and 1:2, respectively. This higher ratio of 3-carbon to 2-carbon production presumably offers an alternate means to replenish TCA cycle intermediates in the absence of a glyoxylate shunt (Fig. 2).

In order to utilize the propionyl-CoA generated by beta-oxidation of odd-chain fatty acids, bacteria often employ the methylcitrate cycle (MCC), producing succinate and pyruvate from oxaloacetate and propionyl-CoA. In P. putida, the MCC is catalyzed via methylcitrate synthase (prpC [PP_2335]), 2-methylcitrate dehydratase (prpD [PP_2338] or acnB [PP_2339]), aconitate hydratase (acnB [PP_2339] or acnA2 [PP_2336]), and 2-methylisocitrate lyase (mmgF [PP_2334]) (see Fig. S1 in the supplemental material). Unsurprisingly, the MCC appeared to be absolutely required for growth on propionate (C₃), valerate (C₅), heptanoate (C₇), and nonanoate (C₉), with PP_2334, PP_2335, and PP_2337 (a putative AcnD-accessory protein) showing severe fitness defects (Fig. 2; see Fig. S1). While PP_2338 (prpD) encodes a 2-methylcitrate dehydratase, transposon mutants showed no fitness defects when grown on odd-chain fatty acids. This reaction is likely carried out by PP_2339 (acnB, encoding a bifunctional 2-methylcitrate dehydratase/aconitase hydratase B); however, there were no mapped transposon insertions for the gene (Fig. 2; see Fig. S1). This suggests that PP_2339 was essential during the construction of the RB–Tn-Seq library. Furthermore, PP_2336 showed relatively modest fitness defects when grown on propionate and other odd-chain fatty acids, suggesting that PP_2339 likely accounts for much of the methylaconitate hydratase activity, as well (Fig. 2; see Fig. S1).

Pearson correlation analysis of fitness data indicated that both long- and medium-chain fatty acids are likely catabolized via similar pathways. Fitness data suggest that FadD1 (PP_4549) catalyzes the initial CoA ligation of C₇ to C₁₈ fatty acids and may potentially act on C₆, as well (Fig. 2). Disruption of fadD2 (PP_4550) did not cause fitness defects as severe as those seen in fadD1 mutants, although it did result in moderate fitness defects when grown on C₈ to C₁₀ fatty acids. These data are consistent with the biochemical characterization of FadD1 from P. putida CA-3, which showed greater activity on longer-chain alkanoic and phenylalkanoic acids than on shorter-chain substrates (43). For fatty acids with chain lengths of C₁₀ and greater, the data suggest that the fadE homolog PP_0368 is the primary acyl-CoA dehydrogenase, while the nearby fadE homolog, PP_0370, appears to be preferred for C₆ to C₈ fatty acids (Fig. 2). Relatively even fitness defects for these two fadE homologs indicates that PP_0368 and PP_0370 may have equal activity on nonanoate (Fig. 2). These data are supported by a previous biochemical characterization of PP_0368, in which it showed greater activity on chain lengths longer than C₉ (44). The fadB homolog PP_2136 showed severe fitness defects when grown on all fatty acids with chain lengths of C₆ and longer, implicating it as the primary enoyl-CoA hydratase/3-hydroxy-CoA dehydrogenase for those substrates (Fig. 2). P. putida was able to grow on the unsaturated substrate oleic acid and is likely able to isomerize the position of the unsaturated bond via the enoyl-CoA isomerase PP_1845, which showed specific fitness defects when grown on oleic acid (Fig. 2). P. putida’s primary long-chain thiolase appears to be the fadA homolog PP_2137, which showed severe to moderate fitness defects when grown on fatty acids with chain lengths of C₈ or longer (Fig. 2). Fitness data for mutant pools grown on heptanoate showed minor fitness defects for both PP_2137 and PP_3754 (bktB), suggesting that both thiolases may work on C₇ substrates (Fig. 2).

Both long-chain fatty esters tested (Tween 20 and butyl stearate) appeared to utilize the same fad homologs as the long-chain fatty acids. However, before either molecule can be directed toward beta-oxidation, Tween 20 and butyl stearate must be hydrolyzed to generate a C₁₂ or C₁₈ fatty acid, respectively. To date, no such hydrolase has been identified in P. putida KT2440. Comparing the mutant fitness scores between Tween 20 and laurate (C₁₂) carbon source experiments revealed six genes (PP_0765, PP_0766, PP_0767, PP_0914, PP_2018, and PP_2019) that had significant and severe fitness defects specific to Tween 20 (fitness score, less than −2; t-statistic > |4|) in both biological replicates (see Fig. S2 in the supplemental material). The same comparison between butyl stearate and myristate (C₁₄) revealed four genes specific to the fatty ester (PP_0765, PP_0766, PP_2018, and PP_4058) that had significant severe fitness defects (fitness scores, less than −2; t > |4|) in both biological replicates (see Fig. S2). As PP_0765-6 and PP_2018 appear to have specific importance under both of the ester conditions tested, it is possible that they contribute to the hydrolysis of the fatty ester bonds. However, it is also possible that the esterase is secreted or associated with the outer membrane (45), in which case, its enzymatic activity would be shared among the library, and it would not have the associated fitness defect expected (10).

The genes PP_2018 and PP_2019 encode a BNR domain-containing protein and an RND family efflux transporter, respectively, and are likely coexpressed in an operon that also includes PP_2020 and PP_2021. Interestingly, although PP_2021 codes for a putative lactonase, transposon mutants had no apparent fitness defect with either of the fatty esters as the carbon source. PP_0765 and PP_0766 encode a DUF1302 family protein and a DUF1329 family protein, respectively. Given their similar fitness scores, they are likely coexpressed in an operon positively regulated by the LuxR-type regulator PP_0767 (see Fig. S2). Previous work in multiple other Pseudomonas species has shown cofitness of DUF1302/DUF1329 family genes with BNR domain and RND family efflux genes when grown on Tween 20 (41). The authors proposed that the genes may work together in order to export a component of the cell wall. However, an alternative hypothesis could be that PP_0765 and PP_0766 contribute to catalyzing the hydrolysis of fatty esters, accounting for the missing catabolic step of butyl stearate and Tween 20. This hypothesis is bolstered somewhat by the colocalization of PP_0765/PP_0766 near fatty acid catabolic genes in P. putida KT2440 and many other pseudomonads (see Fig. S3 in the supplemental material). Further work will need to be done to biochemically characterize the potential enzymatic activity of these proteins.

In our genome-wide fitness assays, the mutant fitness patterns of C₆ or shorter fatty acid carbon sources had lower Pearson correlation among them than the correlations within long- and medium-chain fatty acids (Fig. 1). These global differences reflect what appear to be discrete preferences in beta-oxidation enzymes for growth on short-chain fatty acids. Fitness data suggest that while CoA ligases PP_0763 and PP_4559 are both required for growth on hexanoate, only PP_0763 is required for growth on valerate (Fig. 2). Furthermore, the putative positive regulator of PP_0763, the LuxR family transcription factor PP_0767, also showed a significant fitness defect (−2.0) when grown on both valerate and hexanoate (Fig. 2). PP_0370 seems to be the acyl-CoA dehydrogenase largely responsible for hexanoate catabolism, though PP_3554 mutants also have minor fitness defects. The dehydrogenation of valeryl-CoA appears to be distributed between the activities of PP_0368, PP_0370, and PP_3554, with no single acyl-CoA dehydrogenase mutant demonstrating a strong fitness defect when grown on valerate (Fig. 2). Interestingly, though previous biochemical analysis had demonstrated that PP_2216 has activity on C₄ to C₈ acyl-CoA substrates with a preference for shorter chain lengths (46), we observed no fitness defects for PP_2216 mutants when grown on any fatty acid carbon source (Fig. 2).

It appears that the role of enoyl-CoA hydratase or hydroxyacyl-CoA dehydrogenase may be distributed across multiple enzymes for both hexanoate and valerate. Growth on hexanoate resulted in moderate fitness defects in mutants with the predicted enoyl-CoA hydratases PP_2136, PP_2217, and PP_3726 disrupted; however, for mutants grown on valerate, there were almost no observable fitness defects for any of the enoyl-CoA hydratase enzymes examined in the study, suggesting that for this chain length, significant functional complementation exists between the fadB homologs (Fig. 2). Fitness data suggest that PP_2136 (encoded by fadB), PP_2214 (a predicted type II 3-hydroxyacyl-CoA dehydrogenase), and PP_3755 (a 3-hydroxybutyryl-CoA dehydrogenase) may all be involved in the dehydrogenation of 3-hydroxyhexanoyl-CoA (Fig. 2), while there appears to be a distribution of fadB-like activity when it comes to the dehydrogenation of 3-hydroxyvaleryl-CoA, with PP_3755 showing only a slight fitness defect on valerate. Intriguingly, mutants with the predicted type 2 acyl-CoA dehydrogenase PP_2214 disrupted showed apparently increased fitness when grown on valerate (Fig. 2). As with heptanoate, fitness data from mutant pools grown on valerate or hexanoate suggest that both PP_2137 and PP_3754 may catalyze the terminal thiolase activity of these substrates. The lack of pronounced fitness phenotypes for the beta-oxidation steps of both valerate and hexanoate underscores the necessity for further in vitro biochemical interrogation of these pathways.

Both the butanol and butyrate metabolisms of P. putida have been studied in detail through omics level interrogation across multiple strains (28, 29). Previous work showed that during growth on n-butanol, which is later oxidized to butyrate, three CoA ligases are upregulated: PP_0763, PP_3553, and PP_4487 (an acyl-CoA synthase encoded by acsA1) (29). However, our butyrate carbon source experiments revealed strong fitness defects only in PP_3553 mutants (Fig. 2 and 3A). The same work found that PP_3554 was the only upregulated acyl-CoA dehydrogenase, which agrees with the strong fitness defect we observed in mutants of that gene (29). That prior work did not find upregulation of any enoyl-CoA hydratase in P. putida grown on butanol, but that is likely reflective of redundancy in this step; we observed fitness defects in multiple genes, including PP_2136, PP_2217, and PP_3726, with PP_2217 mutants demonstrating the most severe fitness defect (Fig. 2 and 3A). Hydroxyacyl-CoA dehydrogenase PP_2136 and 3-hydroxybutyryl-CoA dehydrogenase PP_3755 (hbd) have both been shown to be upregulated during growth on butanol (29). While our data showed fitness defects in both of the genes, the defect of PP_3755 mutants was much more severe. Three different thiolases (PP_2215, PP_3754, and PP_4636) and the 3-oxoacid CoA-transferase encoded by atoAB were previously observed to be upregulated during growth on butanol, but of the genes, only PP_3754 (bktB) had a strong fitness defect, implying that it is the main thiolase for the terminal step of butyrate catabolism (Fig. 2 and 3A). The inability of the RB–Tn-Seq data to clearly show which enzymes are likely responsible for specific beta-oxidation reactions suggests multiple enzymes may catalyze these steps. In addition to the lack of genotype-to-phenotype clarity in the enzymes responsible for the catabolic steps, we observed additional phenotypes within our fitness data that portray a complex picture of short-chain fatty acid metabolism in P. putida. The TetR family repressor paaX (PP_3286) was shown to have a negative fitness score when mutant pools were grown on fatty acids with chain lengths of C₇ or below (see Fig. S4 in the supplemental material). PaaX negatively regulates the paa gene cluster encoding the catabolic pathway for phenylalanine (47, 48), implying that the presence of phenylalanine catabolism impedes growth on short-chain fatty acids. It is therefore somewhat surprising that no individual mutant within the paa gene cluster showed a fitness increase when grown on short-chain fatty acids, though no robust fitness data exist for paaJ (PP_3275, encoding a 3-oxo-5,6-didehydrosuberyl-CoA thiolase) (see Fig. S4).

MerR family regulator PP_3539 mutants showed very large fitness benefits (fitness scores of 3.8 and 4.7 in two biological replicates) when grown on valerate. PP_3539 likely increases expression of mvaB (PP_3540, encoding hydroxymethyl-glutaryl-CoA lyase), suggesting that decreased levels of MvaB activity may benefit P. putida valerate catabolism. Unfortunately, there are no fitness data available for mvaB, likely because it is essential under the conditions in which the initial transposon library was constructed. The genes hdb and bktB, encoding the two terminal steps of butyrate metabolism, are flanked upstream by an AraC family regulator (PP_3753) and downstream by a TetR family regulator (PP_3756); the latter is likely cotranscribed with the butyrate catabolic genes (Fig. 3B). When grown on butyrate, both PP_3753 and PP_3756 mutants showed decreased fitness; however, previous work to evaluate the global fitness of P. putida grown on levulinic acid showed negative fitness values only for PP_3753, htb, and btkB (Fig. 3B). These results suggest that the TetR repressor may be responding to a butyrate-specific metabolite. Finally, across multiple fitness experiments, the TonB siderophore receptor PP_4994 and the TolQ siderophore transporter PP_1898 showed fitness advantages when grown on fatty acids, especially on hexanoate (see Fig. S5 in the supplemental material). Together, these results suggest that a wide range of environmental signals impact how P. putida is able to metabolize short-chain fatty acids.

In addition to its ability to robustly catabolize a wide range of fatty acid substrates, P. putida is also capable of oxidizing and catabolizing a wide variety of alcohols into central metabolism through distinct pathways. To further our understanding of these pathways, transposon libraries were grown on a number of short n-alcohols (ethanol, butanol, and pentanol), diols (1,2-propanediol, 1,3-butanediol, 1,4-butanediol, and 1,5-pentanediol), and branched-chain alcohols (isopentanol, isoprenol, and 2-methyl-1-butanol). Relative to growth on fatty acids, fitness experiments with P. putida grown on various alcohols showed less correlation with one another, reflecting the more diverse metabolic pathways used for their catabolism (Fig. 4A). The initial step in the catabolism of many primary alcohols is the oxidation of the alcohol to its corresponding carboxylic acid. The BioCyc database features 14 genes annotated as alcohol dehydrogenase genes (PP_1720, PP_1816, PP_2049, PP_2492, PP_2674, PP_2679, PP_2682, PP_2827, PP_2953, PP_2962, PP_2988, PP_3839, PP_4760, and PP_5210) (24). Fitness data showed that the majority of these alcohol dehydrogenases had no fitness defects when grown on the alcohols used in this study (Fig. 4B).

The alcohol dehydrogenases that showed the most consistent fitness defects under multiple conditions were the two PQQ-dependent alcohol dehydrogenases PP_2674 (pedE) and PP_2679 (pedH), as well as the Fe-dependent alcohol dehydrogenase PP_2682 (yiaY) (Fig. 4B). Both pedE and pedH have been extensively studied in P. putida and other related bacteria and are known to have broad substrate specificities for alcohols and aldehydes (25, 26, 49). Their activity is dependent on the activity of pedF (PP_2675), a cytochrome c oxidase that regenerates the PQQ cofactor (25). In Pseudomonas aeruginosa, a homolog of yiaY (ercA) was shown to have a regulatory role in the expression of the ped cluster and was not believed to play a direct catabolic role (50). Recent work has validated that this function is conserved in P. putida (51). Under most conditions tested, disruption of pedF caused more severe fitness defects than disruption of either pedE or pedH individually, suggesting they can functionally complement one another in many cases. However, growth on both 2-methyl-1-butanol and 1,5-pentanediol showed more severe fitness defects in pedE mutants than in pedF (Fig. 4B). In many other alcohols, including ethanol and butanol, even disruption of pedF did not cause extreme fitness defects, suggesting the presence of other dehydrogenases able to catalyze the oxidation (Fig. 4B).

The transcriptional-regulatory systems that activate expression of various genes in the ped cluster could also be identified from these data. Mutants of either member of the sensory histidine kinase/response regulator (HK-RR) two-component system, pedS2-pedR2, showed significant fitness defects when 2-methyl-1-butanol was supplied as the sole carbon source. This HK-RR signaling system has been shown to activate the transcription of pedE and repress pedH in the absence of lanthanide ions (52). Since lanthanides were not supplied in the medium, this likely explains the fitness defect observed in pedS2-pedR2. The transcription factor pedR1 (agmR) was also found to affect host fitness when grown on various alcohols (Fig. 5). The gene has been identified in P. putida U as an activator of long-chain (longer than C₆) n-alcohol and phenylethanol catabolism (53). In P. putida KT2440, pedR1 has been associated with the host response to chloramphenicol, and its regulon has been elucidated previously (54). Our data reflect the literature, indicating that pedR1 functions as a transcriptional activator of the ped cluster and pedR2 functions as a specific regulator of pedE and pedH.

Unsurprisingly, the genes required for the biosynthesis of the PQQ cofactor were also among the most cofit (cofitness is defined as the Pearson correlation between the fitness scores of two genes under many independent experimental conditions) with both pedF and yiaY. P. putida synthesizes PQQ via a well-characterized pathway, starting with a peptide encoded by the gene pqqA (PP_0380), which is then processed by pqqE, pqqF, and pqqC to generate the final cofactor (Fig. 4C). The three synthetic genes (pqqEFC) all showed significant fitness defects on the same alcohols as the pedF mutants, while pqqA showed a less severe fitness phenotype (Fig. 4C). However, the small size of pqqA resulted in few transposon insertions, making it difficult to draw conclusions with confidence. Two genes showed similar defective fitness patterns on select alcohols: pqqB, which has been proposed to encode an oxidoreductase involved in PQQ biosynthesis, and pqqD, encoding a putative PQQ carrier protein. Previous work regarding a PqqG homolog from Methylorubrum extorquens suggested that it forms a heterodimeric complex with PqqF that proteolytically processes PqqA peptides, although PqqF was sufficient to degrade PqqA on its own (55). Fitness data from P. putida may support this hypothesis, as there was no observed fitness defect in pqqG mutants when grown on any alcohol, suggesting that the bacterium is still able to process PqqA with PqqF alone (Fig. 4C).

The metabolism of n-alcohols almost certainly proceeds through beta-oxidation, using the same enzymatic complement as their fatty acid counterparts. This relationship is reflected in the high correlation in global fitness data between alcohols and fatty acids of the same chain length (ethanol and acetate, r = 0.72; butanol and butyrate, r = 0.66; pentanol and valerate, r = 0.72). However, given previous work and our fitness data, the initial oxidation of these alcohols appears to be quite complex (Fig. 6). Biochemical characterizations of PedE and PedH have shown that both have activity on ethanol, acetaldehyde, butanol, butyraldehyde, hexanol, and hexaldehyde (25). When grown on n-pentanol, mutants with pedF disrupted show severe fitness defects, suggesting that PedH and PedE are the primary dehydrogenases responsible for pentanol oxidation (Fig. 4B and 5A). However, when grown on either ethanol or n-butanol, both the PQQ-dependent alcohol dehydrogenases (PQQ-ADHs) and pedF show less severe fitness defects than when they are grown on pentanol (Fig. 4B). This implies that other dehydrogenases are also capable of these oxidations. One likely candidate may be PP_3839, which shows a minor fitness defect when grown on n-butanol and has been biochemically shown to oxidize coniferyl alcohol (Fig. 4B) (56). Individual gene deletion mutants of either pedF (PP_2675) or PP_3839 showed only minor growth defects when grown on either ethanol, butanol, or pentanol as a sole carbon source (Fig. 7). However, when both genes were deleted, no growth was observed on these substrates, suggesting that the PQQ-ADHs and PP_3839 are the primary dehydrogenases responsible for the oxidation of short-chain n-alcohols (Fig. 7).

Based on our data and previous work, it is ambiguous which enzymes oxidize the aldehyde to the corresponding carboxylic acid. As mentioned previously, both PQQ-ADHs have been biochemically shown to act on aldehydes and could catalyze the reaction, but the lack of a strong fitness phenotype for both ethanol and n-butanol suggests they are not the only enzymes capable of catalyzing the reaction. The genomically proximal aldehyde dehydrogenase gene pedI (PP_2680) showed minor fitness defects when grown on ethanol and several other alcohols (Fig. 5 and 6A) but showed no fitness defects when libraries were grown on butanol or pentanol. Another aldehyde dehydrogenase gene, aldB1 (PP_0545), showed virtually no fitness defects when grown on any of the short-chain n-alcohols tested here (Fig. 6A). The lack of any one obvious enzyme with a distinct fitness defect supports the notion that multiple enzymes are present and able to catalyze the oxidation of these aldehydes.

While the metabolisms of alcohols and their corresponding fatty acids are similar, their fitness patterns showed distinct differences. When grown on acetate, gacS and gacA (PP_1650 and PP_4099, encoding a two-component system [TCS]), sigS (PP_1623, encoding the stationary-phase sigma factor sigma S), and ptsH (PP_0948, encoding a component of the sugar phosphotransferase system [PTS]) mutants showed large and significant fitness benefits, which were not apparent when they were grown on ethanol (Fig. 5B). The GacS-GacA TCS is widespread across many Gram-negative bacteria and is believed to exert transcriptional control over a wide variety of functions, sometimes in concert with a small RNA binding protein (CsrA) that exerts posttranscriptional control (57). In pseudomonads, the GacA-GacS TCS has been implicated in positively controlling sigS expression in multiple species (58). In P. putida specifically, gacS mutations in strains engineered to produce muconic acid have resulted in higher titers (59), but disruption of the gene was also shown to completely abolish production of medium-chain-length polyhydroxyalkanoates (PHAs) (60). Growth on butyrate also showed that gacS, gacR, sigS, and another component of the PTS (ptsP) had significant fitness benefits if disrupted, which was not observed when the library was grown on butanol (Fig. 5C). Interestingly, gacA and gacS mutants seemed to have fitness benefits when grown on either pentanol or valerate (Fig. 5D). Further work is necessary to precisely characterize the nature of the benefits that occur when these genes are disrupted.

When grown on ethanol compared to acetate, relatively few genes not involved in the oxidation of the short-chain alcohols were found to be specifically and significantly unfit; however, specific phenotypes for acetate catabolism were observed (Fig. 5B). PP_1635 (encoding a two-component system response regulator), PP_1695 (encoding a protein variously annotated as a sodium solute symporter, sensory box histidine kinase, or response regulator), and tal (PP_2168, encoding a transaldolase) mutants all showed fitness defects on acetate that were not observed when libraries were grown on ethanol. PP_1635 and PP_1695 have high cofitness observed across all publicly available fitness data (r = 0.88) and share homology to crbSR systems of other bacteria, where they are known to regulate acetyl-CoA synthetase (61).

Much like ethanol and acetate, there were relatively few genes that showed specific fitness defects when grown on butanol that were not also observed in butyrate. However, the genes glgB (PP_4058, encoding a 1,4-alpha-glucan branching enzyme) and the cotranscribed PP_2354 and PP_2356 (encoding proteins annotated as a histidine kinase-response regulator [HK-RR] and a histidine kinase, respectively) showed specific fitness defects when grown on butyrate relative to growth on butanol. PaperBLAST analysis of PP_2356 and PP_2354 did not reveal any publications that had explored the function of the system, and thus, further work will be needed to better characterize its regulon (62). Mutants of genes encoding three TCSs were found to be specifically unfit when grown on pentanol compared to growth on valerate. PP_2683 (a two-component HK-RR) and pedR1 (PP_2665, an RR) were both specifically unfit and, as previously described, are involved in the regulation of the ped cluster (Fig. 5D). The gene cbrB (PP_4696, a σ⁵⁴-dependent RR) also showed pentanol-specific defects and is known to regulate central carbon metabolism and amino acid uptake in the pseudomonads (63, 64).

Another group of industrially relevant alcohols with potential for biotechnological production is short-chain diols. These compounds have broad utility ranging from plasticizers to food additives (65). As shown in Fig. 5, most of the tested short-chain diols result in significant fitness defects in pedR1, indicating that some of the genes involved in these metabolisms are in the PedR1 regulon. However, only 1,5-pentanediol had a strong fitness defect in pedF, indicating that multiple dehydrogenases may act on the shorter-chain diols. Additionally, both 1,2-propanediol and 1,3-butanediol cause slight defects in mutants of the aldehyde dehydrogenase PP_0545. Although there is some ambiguity as to which enzymes initially oxidize the diols to their corresponding acids, the remaining steps in 1,2-propanediol, 1,3-butanediol, and 1,5-pentanediol catabolism are much more straightforward.

Oxidation of 1,2-propanediol yields lactate, and l-lactate permease PP_4735 (lldP) mutants have a fitness defect of −4.3 when grown on 1,2-propanediol. Furthermore, under these conditions, mutants of the l- and d-lactate dehydrogenases PP_4736 (lldD) and PP_4737 (lldE) have fitness defects of −5.0 and −1.5, respectively. Since we provided a rac-1,2-propanediol as a substrate, this likely explains the fitness defects observed in both dehydrogenases (66, 67). Given these results, it appears that 1,2-propanediol is assimilated into central metabolism via oxidation to pyruvate (see Fig. S6 in the supplemental material).

When grown on 1,3-butanediol, two oxidations of 1,3-butanediol result in 3-hydroxybutyrate, and we observed fitness defects of −2.5 in the d-3-hydroxybutyrate dehydrogenase PP_3073 and −1.8 in the neighboring σ⁵⁴-dependent regulator PP_3075 (68). Dehydrogenation of 3-hydroxybutyrate results in acetoacetate, and we saw fitness defects of −2.9 and −3.0 for the subunits of the predicted 3-oxoacyl-CoA transferase PP_3122-3 (atoAB). This enzyme likely transfers CoA from either succinyl-CoA or acetyl-CoA in order to generate acetoacetyl-CoA. Regarding transport, d-beta-hydroxybutyrate permease gene PP_3074 mutants, located in the same operon as the 3-hydroxybutyrate dehydrogenase gene, have a fitness defect of −0.9, while the RarD permease PP_3776 gene mutants have a fitness defect of −1.2.

Following oxidation by the above-mentioned PQQ-dependent dehydrogenases and aldehyde dehydrogenases in the periplasm, an oxidized intermediate is likely transported into the cell for the next steps in catabolism. This is supported by the observation that mutants of the predicted dicarboxylate major facilitator superfamily (MFS) transporter PP_1400 and its two-component regulator, PP_1401-2, have strong fitness defects on both alpha-ketoglutarate and 1,5-pentanediol. Furthermore, there is a −4.7 fitness defect in mutants of the l-2-hydroxyglutarate oxidase PP_2910, which catalyzes the second step in the glutarate hydroxylation pathway of glutarate catabolism. The glutarate hydroxylase PP_2909, which catalyzes the first step of the pathway, has a much smaller negative fitness of −0.6. This is expected, because glutarate can also be catabolized through a glutaryl-CoA dehydrogenation pathway, so PP_2909 mutants can simply divert flux through the other catabolic route (12). PP_2910 mutants are unable to oxidize l-2-hydroxyglutarate to alpha-ketoglutarate and likely accumulate l-2-hydroxyglutarate as a dead-end metabolite.

1,4-Butanediol catabolism has been previously studied. Based on the results of expression data and adaptive laboratory evolution, Li et al. proposed three potential catabolic pathways for 1,4-butanediol, including a beta-oxidation pathway (Fig. 8) (30). Their evolved strains had mutations in the LysR activator PP_2046 that resulted in overexpression of the beta-oxidation operon PP_2047-PP_2051 (30). Interestingly, we found that when grown on 1,4-butanediol, transposon mutants of the acyl-CoA dehydrogenase PP_2048 had significant fitness benefits and no CoA ligase mutants showed significant fitness defects. However, a fitness defect of −1.0 in PP_0356 (malate synthase) mutants suggests that there might be flux through the beta-oxidation pathway to glycolic acid and acetyl-CoA. A possible explanation for the positive fitness of PP_2048 mutants is that the beta-oxidation pathway is suboptimal in the wild type, and it may be beneficial to divert flux through the other pathway(s). The same reasoning could also explain the absence of CoA ligases with fitness defects; however, it also could be due to the presence of multiple CoA ligases capable of catalyzing that step. Mutants of the 3-hydroxyacyl-CoA dehydrogenase gene PP_2047, a fadB homolog that likely catalyzes the hydration and dehydrogenation steps to produce 3-oxo-4-hydroxybutyryl-CoA, had a strong fitness defect. When PP_2047 is nonfunctional, 4-hydroxycrotonyl-CoA likely accumulates as a dead-end metabolite, resulting in decreased fitness. Li et al. also showed that deletion mutants of PP_2046 are unable to grow on 1,4-butanediol (30). Our data suggest that this is because PP_2049 appears to be the main alcohol dehydrogenase acting on either 1,4-butanediol or 4-hydroxybutyrate and is encoded in the operon under the control of PP_2046. Although our fitness data suggest that both the oxidation to succinate and beta-oxidation pathways occur, further work is necessary to determine if the pathway to succinyl-CoA is involved in the catabolism.

Due to their superior biofuel properties, branched-chain alcohols have been targets for metabolic engineering as potential alternatives to ethanol and butanol (69). Our fitness data suggest that pedE and/or pedH oxidizes 2-methyl-1-butanol to 2-methylbutyrate, which then undergoes one round of beta-oxidation to produce acetyl-CoA and propionyl-CoA (see Fig. S7 in the supplemental material). Most of the genes involved in 2-methylbutyrate beta-oxidation are located in the operon PP_2213-PP_2217. With mutants having a fitness defect of −3.2, PP_2213 appears to be the main acyl-CoA ligase acting on 2-methylbutyrate. Mutants of two predicted acyl-CoA dehydrogenases, PP_2216 and PP_0358, show fitness defects of −1.1 and −2.6, respectively. The enoyl-CoA hydratase PP_2217 has a fitness defect of −5.7, and the 3-hydroxyacyl-CoA dehydrogenase PP_2214 has a fitness defect of −5.6. Finally, the acetyl-CoA acetyltransferase appears to be PP_2215, with mutants having a fitness defect of −4.8. We also observed fitness defects of −1.8 and −1.6 in mutants of the ABC transporters PP_5538 and PP_2667, respectively. Since 2-methylbutyrate is a known intermediate in the catabolism of isoleucine, we found that the genetic data presented here closely mirror the previous biochemical characterization of the system (70, 71).

P. putida can readily grow on isopentanol and isoprenol, but not prenol (Fig. 9A). All three alcohols have been produced in high titers in Escherichia coli and other bacteria because of their potential to be suitable replacements for gasoline (72, 73). RB–Tn-Seq data for isopentanol and isoprenol showed severe fitness defects in genes of the leucine catabolic pathway (Fig. 10). This is unsurprising, as isopentanol can be generated from the leucine biosynthetic pathway (74). Deletion of the PP_4064-PP_4067 operon, which contains the genes that code for the conversion of isovaleryl-CoA to 3-hydroxy-3-methylglutaryl-CoA, abolished growth on both isopentanol and isoprenol (see Fig. S8 in the supplemental material). Deletion of PP_3122 (acetoacetyl-CoA transferase subunit A) also abolished growth on isopentanol and greatly reduced growth on isoprenol (see Fig. S8). Taken together, these results validate the notion that both of these alcohols are degraded via the leucine catabolic pathway. Transposon insertion mutants of pedF showed strong fitness defects on both isopentanol and isoprenol, suggesting that pedH (PP_2679) and pedE (PP_2674) catalyze the oxidation of the alcohols. pedH deletion mutants showed only a minor delay in growth compared to the wild type when grown on either isopentanol or isoprenol, while pedE mutants showed a more substantial growth defect on both alcohols (Fig. 9A). Deletion of pedF (PP_2675) prevented growth on isopentanol and nearly abolished growth on isoprenol when provided as a sole carbon source in minimal medium (Fig. 9A). When wild-type P. putida was grown in minimal medium with 10 mM glucose and 4 mM either isopentanol, prenol, or isoprenol, each alcohol was shown to be readily degraded, with concurrent observation of increasing levels of the resultant acid (Fig. 9B). Though P. putida was unable to utilize prenol as a sole carbon source, it was still able to readily oxidize prenol to 3-methyl-2-butenoic acid, suggesting there is no CoA ligase present in the cell able to activate this substrate and channel it toward leucine catabolism (Fig. 10). When wild-type P. putida was grown in lysogeny broth (LB) medium supplemented with a 4 mM concentration of each alcohol individually, all the alcohols were completely degraded by 24 h postinoculation (Fig. 9C). In pedF deletion mutants grown under the same conditions, the rate at which the alcohols were degraded was significantly reduced; however, after 48 h, ∼50% of the isopentanol, ∼75% of the isoprenol, and 100% of the prenol were degraded (Fig. 9C). Uninoculated controls showed that no alcohol was lost at greater than 5% on account of evaporation (data not shown). Future efforts to produce any of these alcohols in P. putida will be heavily impacted by this degradation, and greater effort will be needed to identify other enzymes involved in the oxidation of the alcohols or other metabolic pathways that consume them.

One mystery that remains is how isoprenol enters into leucine catabolism. Gas chromatography-mass spectrometry (GC-MS) analysis confirmed oxidation of the alcohol to 3-methyl-3-butenoic acid, but it is unclear what the next step entails. Fitness data suggest that either PP_4063 or PP_4549 may attach CoA to isovalerate, but neither of these genes has a strong phenotype when mutant libraries are grown on isoprenol (Fig. 10). The fact that PP_4064 (isovaleryl-CoA dehydrogenase) shows strong negative fitness values when libraries are grown on isoprenol implies that its degradation goes through an isovaleryl-CoA intermediate; however, this fitness defect may be the result of polar effects that disrupt the downstream steps (Fig. 10). One possibility is that 3-methyl-3-butenoic acid is reduced to isovalerate in the cell; however, this seems unlikely, since no isovalerate was observed via GC-MS when P. putida was fed isoprenol and glucose. Two other possible routes could result from the activation of 3-methyl-3-butenoic acid by an undetermined CoA ligase. If this CoA ligase exists, it is interesting that it would have activity on 3-methyl-3-butenoic acid but not 3-methyl-2-butenoic acid, which accumulates when P. putida is grown in the presence of prenol. Once formed, the 3-methyl-3-butenyl-CoA could be directed into leucine catabolism via either isomerization to 3-methylcrotonyl-CoA or reduction to isovaleryl-CoA. Future work that leverages metabolomics to identify compounds that accumulate in leucine catabolic mutants may reveal the missing steps and help narrow the search for their enzymes.

The large set of global fitness data generated in this study provides an extensive and global overview of the putative pathways of alcohol and fatty acid degradation in P. putida. Overall, our fitness data agree with previously published biochemical data that explored enzymes in both fatty acid and alcohol metabolism. However, there are still many questions that our data leave unanswered. Further investigation will be required to untangle and elucidate which specific enzymes are biologically relevant in the beta-oxidation of short-chain fatty acids. It is likely that biochemical characterization of individual enzymes will be required to determine which of the fad homologs catalyze these reactions. Another intriguing question is the functions of PP_0765 and PP_0766. Biochemical interrogation and mutational analysis of the DUF1302 and DUF1329 family proteins are needed to determine whether these proteins indeed function as an esterase or, as previously predicted, play some other role in outer membrane biogenesis (41). Additional work is also warranted to ascertain which of the proposed 1,4-butanediol catabolic routes the wild-type organism actually uses and to determine whether the beta-oxidation pathway is indeed less preferable than the pathway to succinate.

To our knowledge, our finding that P. putida can consume both isopentanol and isoprenol is the first observation of this metabolism. If metabolic engineers wish to produce these chemicals in P. putida, these pathways will need to be removed. Critically, researchers will need to identify other enzymes that result in the oxidation of these alcohols or other routes of degradation within P. putida. How P. putida is able to utilize isoprenol, but not prenol, as a sole carbon source is metabolically intriguing. One of our proposed pathways of isoprenol catabolism requires the existence of a CoA ligase that shows surprising specificity toward 3-methyl-3-butenoic acid with little to no activity on 3-methyl-2-butenoic acid. More work should be done to leverage other omics level techniques to try to identify this hypothetical enzyme and biochemically verify its activity. Finally, the data set as a whole will likely strengthen the assumptions made by genome scale metabolic models. Previous models of P. putida metabolism have incorporated RB–Tn-Seq data to improve their predictions (17). This work nearly doubles the number of available RB–Tn-Seq data sets in P. putida that are publicly available and will likely contribute greatly to further model refinement. Ultimately, large strides in our understanding of P. putida metabolism leveraging functional genomic approaches will provide the foundation for improved metabolic-engineering efforts in the future.

General E. coli cultures were grown in LB Miller medium (BD Biosciences) at 37°C, while P. putida was grown at 30°C. When indicated, P. putida and E. coli were grown on modified MOPS (morpholinepropanesulfonic acid) minimal medium, which is comprised of 32.5 μM CaCl₂, 0.29 mM K₂SO₄, 1.32 mM K₂HPO₄, 8 μM FeCl₂, 40 mM MOPS, 4 mM tricine, 0.01 mM FeSO₄, 9.52 mM NH₄Cl, 0.52 mM MgCl₂, 50 mM NaCl, 0.03 μM (NH₄)₆Mo₇O₂₄, 4 μM H₃BO₃, 0.3 μM CoCl₂, 0.1 μM CuSO₄, 0.8 μM MnCl₂, and 0.1 μM ZnSO₄ (75). Cultures were supplemented with kanamycin (50 mg/liter; Sigma-Aldrich), gentamicin (30 mg/liter; Fisher Scientific), or carbenicillin (100 mg/liter; Sigma-Aldrich) when indicated. All other compounds were purchased through Sigma-Aldrich. 3-Methyl-3-butenoic acid was not available commercially and required synthesis, which is described below.

All the bacterial strains used in this study are listed in Table 1, and the plasmids used in this work are listed in Table 2. All strains and plasmids created in this work are available through the public instance of the JBEI registry (https://public-registry.jbei.org/folders/456). All plasmids were designed using Device Editor and Vector Editor software, while all primers used for the construction of plasmids were designed using j5 software (76–78). Plasmids were assembled via Gibson assembly using standard protocols (79) or Golden Gate assembly using standard protocols (80). Plasmids were routinely isolated using a Qiaprep Spin Miniprep kit (Qiagen), and all primers were purchased from Integrated DNA Technologies (IDT) (Coralville, IA). Construction of P. putida deletion mutants was performed as described previously (18).

Growth studies of bacterial strains were conducted using microplate reader kinetic assays as described previously (81). Overnight cultures were inoculated into 10 ml of LB medium from single colonies and grown at 30°C. These cultures were then washed twice with MOPS minimal medium without any added carbon and diluted 1:100 into 500 μl of MOPS medium with 10 mM carbon source in 48-well plates (Falcon; 353072). The plates were sealed with a gas-permeable microplate-adhesive film (VWR), and then, the optical density and fluorescence were monitored for 48 h in a BioTek Synergy 4 plate reader at 30°C with fast continuous shaking. The optical density was measured at 600 nm (OD₆₀₀).

RB–Tn-Seq experiments utilized the P. putida library JBEI-1, which has been described previously, with slight modification (18). Libraries of JBEI-1 were thawed on ice, diluted in 25 ml of LB medium with kanamycin, and then grown to an OD₆₀₀ of 0.5 at 30°C, at which point three 1-ml aliquots were removed, pelleted, and stored at –80°C. The libraries were then washed once in MOPS minimal medium with no carbon source and then diluted 1:50 in MOPS minimal medium with 10 mM each carbon source tested. Cells were grown in 10 ml of medium in test tubes at 30°C with shaking at 200 rpm. One 500-μl aliquot was pelleted and stored at –80°C until BarSeq analysis was performed as previously described (19, 40). The fitness of a strain is defined here as the normalized log₂ ratio of barcode reads in the experimental sample to barcode reads in the time zero sample. The fitness of a gene is defined here as the weighted average of the strain fitness for insertions in the central 10% to 90% of the gene. The gene fitness values are normalized so that the typical gene has a fitness of zero. The primary statistical t value represents the form of fitness divided by the estimated variance across different mutants of the same gene. Statistical t values of >|4| were considered significant. A more detailed explanation of calculating fitness scores can be found in a previous study by Wetmore et al. (40). All the experiments described here passed testing using the quality metrics described previously unless otherwise noted (40). All experiments were conducted in biological duplicate, and all fitness data are publicly available at http://fit.genomics.lbl.gov.

To examine the oxidation of isopentanol, prenol, and isoprenol to their corresponding acids, 10 ml of MOPS minimal medium supplemented with 10 mM glucose and a 4 mM concentration of the tested alcohol added were inoculated with a 1:100 dilution of overnight P. putida culture and incubated at 30°C with 200-rpm shaking. At 0, 12, 24, and 48 h postinoculation, 200 μl of medium was sampled and stored at –80°C. Alcohols and fatty acids were extracted by acidifying the medium with 10 μl of 10 N HCl, followed by addition of 200 μl of ethyl acetate. To detect alcohols and their corresponding carboxylic acids via GC-MS, an Agilent 6890 system equipped with a DB 5-ms column (30 m by 0.25 mm by 0.25 μm) and an Agilent 5973 MS were used. Helium (constant flow at 1 ml/min) was used as the carrier gas. The temperature of the injector was 250°C, and the following temperature program was applied: 40°C for 2 min; increases of 10°C/min to 100°C and then increases of 35°C/min to 300°C; the temperature was then held at 300°C for 1 min. Authentic standards were used to quantify analytes. Determination of isopentanol, prenol, and isoprenol consumption was conducted in 10 ml LB medium with 4 mM either alcohol added. The cultures were inoculated with a 1:100 dilution of overnight P. putida culture and incubated at 30°C with 200-rpm shaking. At 0, 24, and 48 h postinoculation, 200 μl of medium was sampled and stored at –80°C. The remaining concentration of each alcohol was determined by GC-flame ionization detection (FID) as previously described (82).

To a 25-ml round-bottom flask were added chromium(VI) oxide (0.69 g; 6.9 mmol) and distilled water (1 ml). The reaction mixture was then cooled to 0°C before concentrated sulfuric acid (0.6 ml; 10.5 mmol) was added dropwise, forming Jones reagent. The solution of Jones reagent was then diluted to a total volume of 5 ml with distilled water. To a stirred solution of 3-methyl-3-buten-1-ol (0.59 g; 6.9 mmol) in acetone (7 ml) was added dropwise the Jones reagent at 0°C. After being stirred for 8 h at room temperature, the mixture was quenched with ethanol. The mixture was then diluted with water, and the acetone was evaporated in vacuo. The residue was extracted with dichloromethane (DCM), and the organic layers were combined and washed three times with saturated aqueous NaHCO₃ solution. The aqueous phase was acidified with a 2 M aqueous HCl solution to pH 2 to 3 and was then extracted again with DCM. The extract was successively washed with water and brine, dried over MgSO₄, and concentrated in vacuo. The residue was distilled (90°C; 100 mtorr) to yield 3-methyl-3-butenoic acid as a clear oil. ¹H nuclear magnetic resonance (NMR) spectroscopy was performed for validation of the product (300 MHz; chloroform d): δ 4.92 (d, J = 19.1 Hz, 2H), 3.08 (s, 2H), 1.84 (s, 3H) (Fig. 11).

PaperBLAST was routinely used to search for literature on proteins of interest and related homologs (62). All statistical analyses were carried out using either the Python Scipy or Numpy library (84). For the phylogenetic reconstructions, the best amino acid substitution model was selected using ModelFinder as implemented on IQ-tree (85); phylogenetic trees were constructed using IQ-tree, and nodes were supported with 10,000 bootstrap replicates. The final tree figures were edited using FigTree v1.4.3 (http://tree.bio.ed.ac.uk/software/figtree/). Orthologous syntenic regions were identified with CORASON-BGC (86) and manually colored and annotated.