Strain construction and validation
Plasmids containing previously identified malonate transporters MadLM44, MdcF45, and MatC30 were examined for their ability to import malonate into E. coli followed by subsequent conversion to M-CoA by MatB. The success of these importers was quantified using the flaviolin-based M-CoA biosensor discussed further in subsequent sections of this text.46,47 Briefly, all importers proved capable of increasing M-CoA in a dose-dependent manner with increasing concentrations of malonate (SI Fig. 1). However, upon induction of gene expression with tetracycline, both the active transport systems MadLM and MdcF displayed obvious toxicity, resulting in poor growth. No such toxicity was observed for the induction of the MatC-harboring strains. To avoid toxicity issues, strains harboring the MadLM and MdcF transporters were cultured in the absence of an inducer. Low levels of leaky expression from these plasmids resulted in increased flaviolin production, indicative of increased M-CoA concentrations, in response to increased malonate supplementation. Given its lower apparent toxicity upon protein expression, the passive transporter MatC was selected for subsequent experiments.
Genes encoding the malonate importer MatC and malonyl-CoA ligase MatB from Rhizobium trifolii under the control of a lacUV5 promoter were integrated via homologous recombination into the intergenic region downstream of ompW in K207-3, or the essential biotin pathway gene bioH in BAP1 resulting in strains K207-3-MatBC and BAP1-ΔbioH-MatBC respectively. In K207-3-MatBC, two of the most common PKS precursors, M-CoA, and mM-CoA pools, can be increased and controlled by supplying exogenous malonate and propionate to the medium (Fig. 1a). In the BAP1-ΔbioH-MatBC strain, biotin biosynthesis, which plays an essential role as a cofactor in the conversion of Ac-CoA into M-CoA through the ACC complex, the only means for M-CoA production is inhibited. In this strain, M-CoA production and thus growth can be restored either through the exogenous addition of biotin, which is taken up through the YigM transporter48 or through the addition of malonate and subsequent import and CoA ligation by the engineered MatC and MatB pathway (Fig. 1b).
Increasing PKS product flaviolin titer by malonate supplementation
Several M-CoA biosensors have been developed for high-throughput screening of mutations that increase the M-CoA pool. One of these biosensors utilizes the 1,3,6,8-tetrahydroxynaphthalene (THN) synthase, RppA, a type III polyketide synthase (PKS) that converts five molecules of M-CoA to THN, which is then spontaneously converted to flaviolin, a red compound with a unique absorbance peak at 340 nm (Fig. 2a). Flaviolin production can be used as an indirect indicator of the intracellular M-CoA level that can be diverted to PKS biosynthesis by various strains.46,47 Both MatBC-engineered strains and the control strains K207-3 and BAP1-∆bioH-RFP were analyzed for their flaviolin production over three days in LB medium (+ 10 µM IPTG, 0.2% arabinose) with increasing concentrations of malonate. Flaviolin production was quantified by measuring the absorbance at 340 nm (A340) of the culture supernatant. In all cases, flaviolin production was significantly increased for the MatBC strains with malonate supplementation. For both MatBC integrated strains, supplementing with 10 mM and 20 mM malonate resulted in A340 readings of 13–14 (Fig. 2b), which, based on the reported extinction coefficient for flaviolin (ε340 = 3,068 M-1cm-1), equates to approximate flaviolin titers of 0.9 -1 g/L in the supernatant alone, a roughly 1.6-1.7-fold enhancement over the wild-type strains under the same conditions. The increase in titer also correlated with increasing malonate concentrations in the medium, indicating that this enhancement is tunable. Surprisingly, some increase in flaviolin production was observed for the control strains when supplemented with 1 mM malonate; however, no increase in titer was observed as malonate concentrations were further increased. Additionally, both MatBC strains displayed higher titers than the wild-type strains, even when no exogenous malonate was fed to the system. This could be due to the potential for MatB to convert endogenously produced malonate through the natural hydrolysis of M-CoA, thus converting it back into M-CoA (Fig. 2b).
Malonate growth rescue in biotin auxotrophic strain
To determine if exogenously fed malonate could rescue the growth of the biotin auxotrophic strain BAP1-ΔbioH-MatBC, the strain was grown in a M9 minimal medium (M9-mm) supplemented 2% glucose and increasing concentrations of malonate up to 100 mM. As a control, the strain was also grown in an LB medium, which contains biotin and displays normal growth. The impact of IPTG induction of the lacUV5 promoter was also evaluated to determine if the leaky expression of the MatBC genes was also suitable for growth recovery. With and without IPTG induction, the cells exhibited significant growth with malonate concentrations as low as 1 mM. Cell density comparable to the control grown in LB medium was achieved with concentrations at and exceeding 10 mM malonate after 48 h of cultivation at 37°C, despite displaying decreased growth in high malonate concentrations at the 24h time point (Fig. 2c).
Improved Production of PKS products derived from mM-CoA and M-CoA
Polyketide synthases rely on the acyltransferase (AT) domain to determine the CoA substrate identity incorporated into the growing polyketide chain. To modify the final compound synthesized by a PKS or hybrid PKS, a common approach is to perform an AT exchange using junctions defined by multisequence alignment and predictive tools to introduce an AT with a substrate preference different from the wild-type sequence.
However, when attempting to replace mM-CoA-specific AT domain of first module (M1) of the pikromycin synthase with an M-CoA-specific AT in a hybrid version of the pikromycin synthase consisting of modules 1, 2, and 7, called Pik127 (Fig. 3a), Dickinson et al. observed the formation of both the mM-CoA and M-CoA-derived products with the mM-CoA product being in excess, despite introducing an AT domain that should be specific for M-CoA.49,50 They proposed that the low level of malonate-derived product was due to decreased activity of the AT at this location or gatekeeping by the KS to favor elongation with the native substrate, mM-CoA.50,51 A detailed description of the pikromycin module boundaries, along with native and hybrid domain structures, can be found in SI Fig. 2.
Herein, we also observed that Pik127 favored production of mM-CoA over M-CoA products when attempting the same M-CoA-specific AT exchange, utilizing the AT from pikromycin module 3 (M3) in the second (M2) and third (M7) modules of the hybrid PKS Pik127 (Fig. 3b). Especially interesting was the AT exchange in the third module that re-installs the natural KS-AT junction. This construct (Pik127-3rd AT) should thus be both active towards M-CoA and with its native KS partner, and yet the mM-CoA product still dominated. In addition to the factors of KS-gate keeping and domain activity, we hypothesize that the discrepancy in expected vs. actual product might also be due to an overabundance of mM-CoA when propionate was supplied as a substrate for its production in the K207-3 strain. Even though the AT for pikAT3 naturally prefers the M-CoA substrate, the excess of mM-CoA relative to the tightly regulated M-CoA pool could lead to higher titers of the methylated product.
To determine if increasing the M-CoA substrate pool could increase the production of the desired M-CoA product, we transformed K207-3 and K207-3-MatBC strains with plasmids encoding hybrid Pik127 variants with AT3 exchanges in module 2 – referred to as Pik127-2nd AT – and module 3– referred to as Pik127-3rd AT. The resulting strains harboring these plasmids were designated as KCP9 (Pik127-2nd AT) and KCP3 (Pik127-3rd AT) in the K207-3 strain, and as KCP9-MatBC and KCP3-MatBC in the K207-3-MatBC strain. These newly engineered PKS constructs should utilize both mM-CoA and M-CoA substrates (see Fig. 3a). Strains were cultured in EZ-rich media supplemented with propionate and +/- malonate to generate mM-CoA and M-CoA, respectively. LC-MS analysis was performed to quantify the amount of 6-ethyl-3,5-dimethyldihydro-2H-pyran-2,4(3H)-dione (product 1 derived from three mM-CoAs incorporation), 6-ethyl-5-methyldihydro-2H-pyran-2,4(3H)-dione (product 2 derived from two mM-CoAs and one M-CoA), and 6-ethyl-3-methyldihydro-2H-pyran-2,4(3H)-dione (product 3 derived from two mM-CoAs and one M-CoA). Integration of the peak area in the corresponding extracted ion chromatographs was used to quantify titers of products 1 and 2 while estimating the titer of product 3 using TAL as an internal standard and calibration curve developed for products 1 and 2 (SI Figs. 3–4).
For the Pik127-3rd AT (KCP3 and KCP3-MatBC) strains, a significant increase in titer of the desired M-CoA product (2) was achieved in KCP3-MatBC when malonate was supplemented to the medium, reaching 0.96 +/- 0.12 mg/L representing a 14.8-fold increase over that from KCP3 with no supplementation. The ratio of the correct M-CoA product to the incorrect mM-CoA incorporated product (products 2:1) was improved to 5.3:1, an improvement of over 150-fold (Fig. 3b-c). Additionally, in a separate report52, the titers achieved by the K207-3-MatBC strain harboring the Pik127-3rd AT plasmids (KCP3-MatBC), titers increased further to 200–300 mg/L in rich media and 50 mL shake flask conditions and up to 1.1 g/L in a bioreactor under fed-batch conditions.52
For the Pik127-2nd AT (KCP9 and KCP9-MatBC) strains, products 1 and 3 were only detectable in cultures of KCP9-MatBC. Titers of the M-CoA product 3 increased 33-fold to 1.32 +/- 0.85 mg/L with a ratiometric excess of 29:1 when supplemented with malonate in comparison to no supplementation, representing an approximately 600-fold improvement in ratiometric excess (Fig. 3b-c). Exact quantification of product 3 could not be achieved due to the inability to obtain the exact chemical standard. However, given the chemical similarity to product 2, identification and quantification were performed using 6-ethyl-5-methyldihydro-2H-pyran-2,4(3H)-dione as a standard (SI Fig. 4). Larger error bars for the quantification of product 3 may be attributable to the decreased stability of this hybrid protein construct with multiple unnatural junctions resulting in varied protein levels and thus variable product titers.
These results indicate that while KS gatekeeping may play a role in the preferred production of the mM-CoA product, it is evident that by increasing M-CoA levels through matBC integration and malonate supplementation, this substrate preference can be significantly altered to favor the incorporation of the intended AT substrate, M-CoA, into the final PKS product.
Production of uniformly labeled 13C polyketides
In BAP1-ΔbioH-MatBC strain, the engineered pathway is the sole means for generating M-CoA precursors used by various PKSs and endogenous FAS. This design facilitates the production of uniform isotopically labeled polyketides, free from contamination by unlabeled endogenously produced M-CoA without the need to cultivate in [13C]glucose. Here, the flaviolin-producing strain BAP1-ΔbioH-MatBC-RppA is cultivated in M9-mm supplemented with 2% natural abundance glucose and with malonate or [13C]malonate labeled solely at the 2-position. A control without malonate supplemented failed to grow and produce flaviolin, while the strains supplemented with malonate turned turbid and red (SI Fig. 5). LCMS analysis of the cultures revealed a consistent 5-Da mass shift in the peak corresponding to flaviolin when [13C]malonate was used compared to the unlabeled counterpart, in accordance with the incorporation of five M-CoA molecules (Fig. 4a).
To demonstrate the necessity of inhibiting endogenous M-CoA biosynthesis through biotin auxotrophy for achieving uniform 13C labeling, biotin was introduced into the medium along with 13C malonate to enable endogenous M-CoA production. Extracted ion chromatograms (EICs) for masses corresponding to flaviolin labeled with zero to five [13C]carbons were examined for cultures fed solely with [13C]malonate and those also supplemented with biotin. In the case of [13C]malonate alone, a prominent peak at 210 Da was observed, representing five [13C]M-CoA molecules with 99.8% purity over other isotopically labeled species. Conversely, when biotin was included in the medium, detectable peaks for masses corresponding to flaviolin with zero to five [13C]carbons were evident (Fig. 4b). EICs for masses 205–210 Da were similarly analyzed for controls where only malonate or biotin was added to the culture. In both cases, a predominant peak at 205 Da was observed, accompanied by minor peaks at 206 and 207, which is in line with the natural isotopic abundance of carbon (SI Fig. 6). Over the 3-day period, the final OD600 for both strains exceeded 5.0 while the flaviolin absorbance at 340 nm exceeded 13.5, similar to what was observed in the LB medium conditions (Fig. 2b). Notably, when supplementing with both [13C]malonate and biotin over a 72 h period, the ratio of the uniformly labeled 210 Da peak progressively rose from 79.9% on day one to 92.6% by day 3 (SI Fig. 6). This suggests that, over time, the M-CoA pool derived from supplemented malonate dominated the M-CoA derived from Ac-CoA.
Quantification of intracellular CoA levels, fatty acid content, and protein expression
While flaviolin production provides easy indirect measurement of M-CoA levels, LC-MS analysis was used for the quantification of the M-CoA pool directly in the absence of a PKS. BAP1 and BAP1-ΔbioH-MatBC strains were cultured in M9-mm with 2% glucose and increasing concentrations of malonate for 36 h at 37°C. At this time point, the cultures had entered the early stationary phase where M-CoA and Ac-CoA levels have been reported to be lower than during exponential growth53. For the BAP1-ΔbioH-MatBC, both OD600 and M-CoA increased with increasing malonate concentrations. The highest M-CoA level, 436.9 +/- 57.5 nM, was observed for BAP1-ΔbioH-MatBC supplemented with 100 mM malonate. When normalized for cell density, this equated to 0.206 +/- 0.025 nmol/mg dry cell weight (DCW), reflecting an 18.11-fold improvement compared to the 0.011 +/- 0.09 nmol/mg DCW M-CoA levels in BAP1 (Fig. 5a).
The biosynthesis of fatty acids (FA) is the major route through which M-CoA is utilized in E. coli. To assess the impact of malonate supplementation and increased M-CoA levels on fatty acid biosynthesis, the fatty acid profiles of BAP1-ΔbioH-MatBC and control BAP1-∆bioH-RFP strains were analyzed using GC-MS. With 20 mM malonate supplementation, a significant increase in C14 and C16 fatty acids was observed in the BAP1-ΔbioH-MatBC strain over all other strains and conditions. During this cultivation, malonate was almost completely consumed in the BAP1-ΔbioH-MatBC cultures with 20 mM malonate (SI Fig. 6). Overall, the total FA content, including unsaturated and cyclopropanated FAs observed in smaller quantities, increased significantly for the BAP1-ΔbioH-MatBC strain when supplemented with malonate over the control BAP1-∆bioH-RFP strain in the same conditions (Fig. 5b).
Similarly, to assess the impact of the engineered MatBC system and subsequent malonate supplementation on the proteome, BAP1-ΔbioH-MatBC and BAP1-∆bioH-RFP was grown in M9 medium with 20 mM malonate, biotin, or both malonate and biotin and analyzed via shotgun proteomics to quantify expression levels of their endogenous proteins. Expression levels of proteins significantly altered by media conditions or strain identity were identified using Welch’s T-test parameters (Fig. 5c).
When comparing the BAP1-ΔbioH-MatBC strain cultured with biotin versus that cultured in malonate, increased expression of proteins in the Rut pathway (RutB, RutC, RutE) was observed in the malonate-only condition. Notably, the Rut pathway is involved in pyrimidine degradation, allowing E. coli to utilize pyrimidine nucleosides as the sole nitrogen source, generating malonate semialdehyde as an intermediate that is then reduced to 3-hydropropanoate (3-HP) by one of two malonic semialdehyde reductases, RutE or YdfG.54 We had contemplated the possibility that instead of reducing 3-HP, the oxidation of malonate semialdehyde to malonate could present an alternative endogenous route for M-CoA production in our strain, potentially allowing it to evade selective growth on malonate. Despite not observing this escape route, the upregulation of these proteins is noteworthy, particularly in the context of subsequent adaptive laboratory evolution performed on this strain.
Comparing the BAP1-ΔbioH-MatBC strain cultured with malonate versus the BAP1-ΔbioH-RFP strain cultured with biotin and malonate, the MatB and MatC expression increased significantly in the MatBC strain, while RFP decreased significantly, consistent with the strain genotypes. Additionally, an upregulation in biotin pathway proteins (BioA, BioD, BioF) was observed, along with increased expression of glycolysis-related proteins PfkB and GlpX and downregulation of heat-shock proteins (IbpA, HspQ). Furthermore, propionate-CoA ligase PrpE and pantothenate kinase CoaA expression decreased significantly, both being related to the utilization and generation of the CoA pool. CoaA, in particular, plays a crucial role as the initial step in CoA biosynthesis and is a key regulator of this biosynthetic process.55 The full results of the T-test comparisons of the proteomic analysis are provided (SI proteomic analysis).
Adaptive evolution experiments utilizing biotin auxotrophic strain
Another advantageous aspect of the biotin auxotrophic strain is that growth recovery by malonate enables adaptive laboratory evolution (ALE) experiments to enhance the production of M-CoA through novel mutations not previously identified through rational engineering and design. To perform ALE, BAP1-ΔbioH-MatBC was transformed with mutagenesis plasmid-6 (MP6), which harbors dnaQ926, dam, seqA, emrR, ugi, and cda1. When induced, these genes enable a 322,000-fold increase in the mutation rate of E. coli chromosomal DNA with a wide mutational spectrum.56 The strain was then cultivated in M9-mm with 2% glucose and 1 mM malonate, as this was the lowest concentration where significant but not full growth recovery was observed (Fig. 2c). The culture was maintained at an OD600 of 0.75 by dilution with medium using a Chi.Bio turbidostat.57 After 7 days, the evolved culture was streaked out onto M9 minimal solid agar plates containing 3 mM malonate and grown at 37°C. Twenty four individual colonies (C1-C24) were inoculated into LB, M9-mm, and M9-mm with 1 mM malonate and cultured at 37°C for 24 h. Mutant strains with beneficial adaptations were identified by their inability to grow on M9 medium but increased growth when supplemented with malonate. Six colonies (C11, C12, C13, C15, C16, C18) from the original 24 plus the unevolved strain were diluted 1:100 in fresh M9-mm with 1 mM malonate and grown for 24 h at 37°C in 24-well plate format. Significantly increased growth was observed for four out of six strains, indicating positive growth adaptations (Fig. 6a). To identify the genetic basis for this enhanced growth, the genomes of the six evolved strains were sequenced.
Many mutations were observed in all strains relative to the non-mutagenized control and are fully reported herein (SI adaptive laboratory evolution whole genome sequencing). One obvious mutation of note that arose in all six colonies was a single point mutation (C > A) to the lac operator upstream of the matB and matC genes, which is known to disrupt repressor affinity.58 Due to the negative effect of expressing the MadLM and MdcF in our initial screen, it was unclear if a high expression level of MatB, MatC, or both would lead to negative growth effects, especially during low resource conditions as in our mutagenesis conditions. To determine the abundance of this mutation in the total population, the total evolved population was PCR amplified using primers flanking the bioH site and sequenced. In 99.6% of the sequencing reads, the lac operator had acquired the disruptive mutation of (C > A), thus indicating that increasing the expression of MatC and MatB is favorable for cell growth in this context.
Another mutation observed in all strains was the mutation from a GTG to ATG start codon for the tolA gene. In E. coli, transcripts starting with AUG are, on average, translated at significantly higher levels than transcripts that start with GUG.59 TolA consists of three domains and spans the inner membrane and periplasm, where it acts as part of the Tol-Pal system, which is involved in the regulation and maintenance of the outer membrane. Deletions to tolA result in the leakage of periplasmic proteins and altered morphology, while overexpression had no noticeable effects on cell morphology.60–63 One possible explanation for the change in the start codon for TolA could be due to the presence of MatC in the membrane, where overexpression of TolA results in a beneficial phenotype.
When examining the genome sequences of the six evolved strains, three of the six had a similar set of loss of function mutations via frame shifts (f.s.) or stop codon insertion (ΔfadI, ΔmcrB, ΔstpA, ΔyafS, ΔyjhC, and ΔECD_03713) in addition to the lac operator and TolA mutations. Of these mutations, we deemed ΔfadI of high interest and determined the abundance of the Trp329* mutation in the total evolved pool to be 52%. FadI is known to catalyze the final step of FA β-oxidation by utilizing one CoA and hydrolyzing acyl-CoA to form one Ac-CoA molecule.64 FadI’s exact role in the β-oxidation cycle is debated as either primarily acting on long-chain FAs, while its homolog FadA performs this activity on shorter acyl-CoAs during aerobic conditions65 or as primarily functioning during anaerobic conditions in the presence of nitrate as an electron acceptor.66 It has also been proposed to perform as “back up” to the primary machinery of FadA during β-oxidation where knockout mutations to both fadA and fadI are required to abolish fatty acid catabolism fully.64,67 Notably, in the strains without the ΔfadI mutation, C11 possessed a frameshift mutation in the β-oxidation pathway gene fadL, while C15 possessed an amino-acid mutation of Arg538Trp in the β-oxidation gene fadK of unknown significance. (Fig. 6b).
Based on the genome sequence data, we developed the following hypothesis to explain the observed mutations. First mutations to the lac operator increase the expression of MatC and MatB, thus increasing flux towards M-CoA. The membrane-stabilizing protein TolA may have evolved to increase expression to accommodate the presence of MatC in the cell membrane. This pool of M-CoA is then directed toward fatty acid biosynthesis to support cell growth. Loss of function mutations to β-oxidation genes, such as fadI, prevent the degradation of the newly synthesized fatty acids back to Ac-CoA, which becomes a dead end for our strain regarding M-CoA generation (Fig. 6c).
To test this hypothesis, the fadI gene was knocked out in the BAP1-ΔbioH-MatBC strain, generating the BAP1-ΔbioH-MatBC-ΔfadI strain. The BAP1-ΔbioH-MatBC, BAP1-ΔbioH-MatBC-ΔfadI, and evolved-C12 strain, which also has the ΔfadI mutation, were then grown in M9-mm supplemented with 2% glucose and 50 mM malonate; additionally, samples with and without 10 µM IPTG were analyzed to investigate the impact of lac operator mutation in the evolved-C12 strain. The strains were grown for 48 h at 37°C and M-CoA and Ac-CoA concentrations were measured. Comparing conditions with and without IPTG, it is apparent that induction with IPTG does indeed lead to increased M-CoA levels and increased final OD600 across all strains, including the evolved strain C12, which has the lac operator mutation. This suggests incomplete inhibition of the lac repressor binding ability in the C12 strain. For the condition without IPTG, the evolved C12 strain had a significant 1.6-fold increase in M-CoA concentration over BAP1-ΔbioH-MatBC MatBC. Also interesting was a significant increase in the Ac-CoA concentration in the C12 strain, which was observed with and without IPTG and must be attributed to other mutations acquired during the evolution process, increasing the conversion of glucose to Ac-CoA.
When IPTG was present, there was an average increase in M-CoA levels from BAP1-ΔbioH-MatBC to BAP1-ΔbioH-MatBC-ΔfadI to the C12 strain. However, only the 1.82-fold increase in M-CoA from the MatBC to C12 strain was significant, with the BAP1-ΔbioH-MatBC-ΔfadI strain displaying an intermediary concentration and not significantly different from either strain (Fig. 6d). Finally, to determine if the evolved strain C12 with its enhanced M-CoA pool resulted in increased PKS production, the RppA plasmid was transformed into the C12 strain. The C12-RppA and the BAP1-ΔbioH-MatBC-RppA strains were cultured in M9-mm with 5 or 20 mM malonate and ± IPTG at 30°C for four days. Aliquots of culture were removed at time points from 42–96 h and analyzed for flaviolin production by HPLC. Across almost all conditions and time points, flaviolin production was significantly increased for the C12 strain compared to the MatBC base strain, with greater differences observed in the media conditions lacking IPTG, which displayed over a two-fold improvement in titer for both the 5- and 20-mM conditions. For the condition without IPTG, significant increases in titer were observed at the earlier time points, but those differences diminished at later time points. (Fig. 6e).