A synthetic enzymatic pathway for extremely thermophilic acetone production based on the unexpectedly thermostable acetoacetate decarboxylase from Clostridium acetobutylicum

Abstract

One potential advantage of an extremely thermophilic metabolic engineering host (T_opt ≥ 70°C) is facilitated recovery of volatile chemicals from the vapor phase of an active fermenting culture. This process would reduce purification costs and concomitantly alleviate toxicity to the cells by continuously removing solvent fermentation products such as acetone or ethanol, a process we are calling “bio-reactive distillation”. While extremely thermophilic heterologous metabolic pathways can be inspired by existing mesophilic versions, they require thermostable homologs of the constituent enzymes if they are to be utilized in extremely thermophilic bacteria or archaea. Production of acetone from acetyl-CoA and acetate in the mesophilic bacterium Clostridium acetobutylicum utilizes three enzymes: thiolase (Thl), acetoacetyl-CoA:acetate CoA transferase (CtfAB), and acetoacetate decarboxylase (Adc). Previously reported biocatalytic pathways for acetone production were demonstrated only as high as 55°C. Here, we demonstrate a synthetic enzymatic pathway for acetone production that functions up to at least 70°C in vitro, made possible by the unusual thermostability of Adc from the mesophile C. acetobutylicum, and heteromultimeric acetoacetyl-CoA:acetate CoA-transferase (CtfAB) complexes from Thermosipho melanesiensis and Caldanaerobacter subterraneus, composed of a highly thermostable α-subunit and a thermally labile β-subunit. The three enzymes produce acetone in vitro at temperatures of at least 70°C, paving the way for bio-reactive distillation of acetone using a metabolically engineered extreme thermophile as production host.

Introduction

Acetone is widely used industrially as a solvent and polymer precursor. An acetone shortage in Great Britain during World War I led to one of the first instances of large-scale industrial fermentation, leveraging the ability of Clostridium acetobutylicum to convert starches and sugars into acetone, butanol, and ethanol (Sauer, 2016). Routes to produce acetone from petroleum eventually eclipsed fermentative processes, and recent research on C. acetobutylicum has predominantly focused on production of bio-butanol as a drop-in replacement for gasoline, in which context acetone is often seen as an undesired byproduct to be minimized (Jang et al., 2012). However, acetone remains a valuable product in its own right (Luo et al., 2016). As a commodity chemical, acetone prices in recent years have exceeded those for ethanol, although it remains less valuable than n-butanol. Acetone is a feedstock in the production of bisphenol A and methyl methacrylate-based polymers, used as a fuel additive, and of course as a solvent (direct solvent use represents ~30% of total demand) (Wu, Wang, Liu, & Huo, 2007). Acetone is also considerably less toxic to cells than n-butanol, which can cause severe toxicity even at low concentrations due to its fluidizing effects on the cell membrane (Peabody & Kao, 2016). Additionally, recombinant expression of enzymes for acetone production in native acetate producers has improved growth (Bermejo, Welker, & Papoutsakis, 1998; Shaw et al., 2015), since acetone appears to be less toxic than acetate, especially at low pH.

The high volatility of acetone (normal boiling point 56°C) makes it a potentially strategic metabolic engineering product for an extremely thermophilic host (T_opt ≥ 70°C), where continuous recovery of product from the bioreactor, termed “bio-reactive distillation”, could be possible even at atmospheric pressure and relatively modest titers (Figure 1). In contrast, implementation of such a concept for mesophilic hosts would require near full vacuum conditions, which then requires energy intensive refrigeration at the distillation condenser (Cysewski & Wilke, 1977). Unfortunately, no microorganism growing at such elevated temperatures is known to produce acetone. In fact, production of solvents is rare among the fermentative extreme thermophiles, which tend to produce organic acids and hydrogen gas instead. The highest reported temperatures for native production of acetone, ethanol, and butanol are 43°C (Weimer, 1984), 72°C (Svetlitchnyi et al., 2013), and 58°C (Freier-Schroder, Wiegel, & Gottschalk, 1989), respectively, while metabolically engineered hosts expressing recombinant enzymes have allowed production as high as 55°C (Shaw et al., 2015), 78°C (Basen et al., 2014), and 60°C (Keller et al., 2015), respectively, as summarized in Table 1.

Figure 1: Vapor-liquid phase envelopes for acetone and ethanol at various pressures of ‘bio-reactive distillation’ operation.

Phase envelopes on T-xy diagram features bubble point (lower line of phase envelope) and dew point (upper line of phase envelope) lines for acetone (solid lines) and ethanol (dashed lines). Biologically reasonable fermentation titers for acetone or ethanol shown in shaded region. The fermentation vessel (reboiler in batch distillation terminology) must operate within the phase envelope for ‘bio-reactive distillation’. The condenser operates at the dew point (top line of phase envelope) corresponding to the relevant product purity. Distillation stages operate at gradually decreasing temperatures from bottom (reboiler) to top (condenser). (Data from http//:VLE-calc.com).

Table 1:

Solvent production in native and in metabolically engineered hosts

Solvent	Native	T (°C)	Reference	Engineered	T(°C)	Reference
Acetone	Bacillus macerans	43	(Weimer, 1984)	Thermoanaerobacterium saccharolyticum	55	(Shaw et al., 2015)
Ethanol	Caldicellulosiruptor sp.	72	(Svetlitchnyi et al., 2013)	Pyrococcus furiosus	78	(Basen et al., 2014)
n-Butanol	Clostridium thermosaccharolyticum	58	(Freier-Schroder et al., 1989)	Pyrococcus furiosus	60	(Keller et al., 2015)

The absence of known native acetone producing extremely thermophilic microorganisms suggests that creation of an acetone production pathway in an extremely thermophilic host requires establishing a synthetic production pathway with recruitment of enzymes from multiple thermophilic organisms. Such efforts have been reported previously, an example of which is n-butanol in Pyrococcus furiosus (Keller et al., 2015). The mesophilic three-enzyme pathway in C. acetobutylicum serves as a template for biological acetone production.

Thl: 2 Acetyl-CoA →Acetoacetyl-CoA + CoA

CtfAB: Acetoacetyl-CoA + Acetate → Acetoacetate + Acetyl-CoA

Adc: Acetoacetate → Acetone + CO₂

Net Rxn: Acetyl-CoA + Acetate → Acetone + CO₂ + CoA

First, a thiolase (Thl) extends carbon chains by condensing two acetyl-CoA molecules into an acetoacetyl-CoA. Acetoacetyl-CoA:acetate CoA-transferase (Ctf) then transfers the CoA moiety from acetoacetyl-CoA to acetate, generating acetoacetate while transforming acetate to acetyl-CoA. Finally, acetoacetate decarboxylase (Adc) decarboxylates acetoacetate to acetone and CO₂. The reaction catalyzed by Adc has also been found to occur non-enzymatically, but a mutant strain lacking the gene lags behind wild-type in acetone production (although final titers were found to be comparable) (Han, Gopalan, & Ezeji, 2011), suggesting Adc activity is highly advantageous in acetone production. Therefore, it was necessary to identify thermostable, thermoactive versions of these three enzymes.

The most straightforward method to identify thermostable enzymes is a search for homologs (based upon amino acid sequence) within genomes of thermophiles with optimum growth temperatures at or near the desired working temperature (in this case, at least 70°C). This was the approach previously adopted (Shaw et al., 2015), utilizing Thl from Thermoanaerobacterium thermosaccharolyticum (T_opt 60°C), Ctf from Thermosipho melanesiensis (T_opt 70°C), and Adc from Bacillus amyloliquefaciens (T_opt 50°C). While this collection of enzymes successfully produced acetone as a minor product at 55°C in T. saccharolyticum, the Thl and Adc were from moderately thermophilic organisms, and thus unlikely to function in extreme thermophiles.

Thermophilic homologs can be found for many mesophilic enzymes, but in the case of highly specialized metabolisms, or reactions that are thermodynamically less favorable at higher temperatures, no thermophilic candidates may be available. However, there are cases where enzymes from mesophilic organisms exhibit unusual thermostability, such as the industrially relevant α-amylase from Bacillus licheniformis (Saito, 1973). Another example can be found in the acetone pathway; acetoacetate decarboxylase from C. acetobutylicum was reported to be active at 70°C when first characterized in partially purified forms (Davies, 1943). Subsequent work determined the enzyme retained 50% activity after 30 minutes at 80°C, and activity actually increased following an hour-long incubation at 70°C (Autor & Fridovich, 1970). C. acetobutylicum grows optimally at 35°C, thus it is unclear why the mesophile possesses an enzyme stable above 70°C, although it is worth noting that C. acetobutylicum spores are activated by brief heat shock at temperatures as high as 80°C (Al-Hinai et al, 2014).

The question addressed here is whether a synthetic biochemical pathway for acetone production can be designed and demonstrated to function at 70°C or higher. If available, this would pave the way for the development of metabolic engineering strains that form acetone at temperatures high enough to facilitate its recovery and purification.

Methods

Identification of thermophilic gene candidates

Thermophilic homologs to C. acetobutylicum acetone pathway enzymes Thl (AAK80816.1), CtfAB (NP_149326 & NP_149327), and Adc (NP_149328) were identified by BLASTp (NCBI) searches limited to microbial groups known to be made up primarily of thermophilic organisms. Query results were considered by coverage (the length of the protein that demonstrates a minimum level of homology) and amino acid identify (the number of amino acids that are identical at a given position and the number of amino acids that share the same R-group category of amino acid at a given position). Potential candidate enzymes were narrowed down to the most promising by focusing on organisms with optimum growth temperatures above 70°C.

Protein expression

Genes encoding candidate enzymes were cloned from genomic DNA of C. acetobutylicum ATCC 824 (CA_P0165 = C. acetobutylicum Adc), and C. subterraneus (TTE0549 = C. subterraneus Thl, TTE0720 = C. subterraneus-CtfA, TTE0720 = C. subterraneus-CtfB), or synthesized to match genes from T. melanesiensis (Tmel_1136 = T. melanesiensis CtfA (first base changed from T to A to change start codon from TTG to ATG), Tmel_1135 = T. melanesiensis CtfB), Vulcanisaeta distributa (VDIS_RS01295 = V. distribute Adc) and Sulfolobus sp. (WP_009989587 = Sulfolobus Adc) (Integrated DNA technologies, Skokie, IL), and inserted into PCR amplified plasmid backbones pET-46, pRSF, or pCDF (EMD Millipore, Billerica, MA) with flanking regions appropriate for Gibson assembly, which was performed using NEB Gibson Assembly^® Master Mix (New England Biolabs, Ipswich, MA). All constructs included N-terminal 6-histidine tags. In addition, all three versions of the Adc enzyme candidates were constructed without tags as well. Cell lines used for cloning were chemically competent E. coli cells 5-α (New England Biolabs) for plasmid screening and amplification, and Rosetta 2 (EMD Millipore) for protein expression. Protein expression was carried out in shake flasks containing one liter of ZYM-5052 lactose autoinduction medium (Studier, 2005) and appropriate antibiotics at 37°C for 20–24 hours. Cells were harvested by centrifugation at 10,000 x g for 10 minutes.

Protein purification

Harvested E. coli cells were re-suspended in 5 mL/mg pellet weight of immobilized metal affinity chromatography (IMAC) buffer A (300 mM NaCl, 50 mM sodium phosphate, 1 mM MgCl₂, 20 mM imidazole, 10% glycerol, pH 8 for C. subterraneus Thl, increased to 20% glycerol with 100 mM sodium sulfate added for all Ctf subunit candidates), or lysis buffer (50 mM potassium phosphate, pH 5.9) for Adc candidates, and lysed in a French pressure cell. Lysed cells were heat-treated at 65°C for 10 minutes to denature E. coli proteins, and then centrifuged at 24,000 x g for 20 minutes to generate soluble heat-treated cell extract. No further purification was performed for the untagged Adc candidates. The heat-treated cell extract for the untagged Adc candidate proteins was buffer exchanged into storage buffer (50 mM Tris-HCl, 100 mM sodium chloride, 50% glycerol, pH 7.5) in Vivaspin 20 10,000 Da MWCO filters (Sartorius, Goettingen, Germany). All histidine-tagged proteins were purified by IMAC using 5 mL HisTrap HP columns. Binding was in IMAC buffer A (described above), followed by elution in a gradient up to 500mM Imidazole. Fractions containing the elution peak were pooled, concentrated, and buffer exchanged as described above. Storage buffers consisted of: 50 mM Tris-HCl, 100 mM sodium chloride, 1 mM DTT, 50% glycerol, pH 7.5, for C. subterraneus Thl; 50 mM MOPS, 500 mM ammonium sulfate, 50% glycerol for all Ctf subunits. Purified enzymes were stored at −20°C.

Enzyme assays

Individual enzymes were assayed for activity on a Lambda 25 spectrophotometer with PTP-1 Peltier heaters (Perkin Elmer, Waltham, MA) using 100 μL Quartz Cuvettes (Starna Cells, Atascadero, CA). All assays were performed at 70°C unless otherwise noted.

Thiolase was assayed by coupling to the NADH consuming activity of 3-hydroxybutryl-CoA dehydrogenase (Hbd), as described previously (Loder et al., 2015). The assay mixture consisted of 100 mM MOPS pH 7.9, 0.3 mM NADH, and appropriately diluted enzymes, with Hbd in considerable excess. The reaction was started by adding 0.5 mM acetyl-CoA substrate, which Thl converts to acetoacetyl-CoA and free CoA. The simultaneous consumption of acetoacetyl-CoA and NADH by Hbd was followed by monitoring the decline in NADH absorbance at 340 nm.

Acetoacetyl-CoA:acetate CoA-transferase was monitored by tracking consumption of the Mg-enolate form of acetoacetyl-CoA by decrease in absorbance at 310 nm (Cary, Petersen, Papoutsakis, & Bennett, 1990). The assay mixture consisted of 100 mM Tris, 150 mM potassium acetate, 20 mM MgCl₂ and 5% glycerol at pH 7.5. Acetoacetyl-CoA was added, and absorbance was monitored for 30 seconds to establish the non-enzymatic rate of acetoacetyl-CoA hydrolysis, then an appropriately diluted mix of the Ctf subunits was added (subunit beta was always in slight excess). For inhibition studies substrate concentrations ranged from 15 to 720 mM acetate and 0.03 to 0.4 mM acetoacetyl-CoA.

The acetoacetate decarboxylase assay was adapted from (Ho, Ménétret, Tsuruta, & Allen, 2009). Assay mix consisted of 50 mM potassium phosphate, 300 mM lithium acetoacetate, pH 5.9. The assay was started by adding appropriately diluted enzyme, and consumption of acetoacetate was monitored by the decline in absorbance at 290 nm.

Thermal inactivation studies involved diluting enzymes to appropriate concentration in assay buffer (without substrate), incubating at 70°C for a range of times, which were then assayed for residual activity. Thermal loss of enzyme activity is typically modeled as a simple exponential decay, but in some cases a more complex two-step inactivation model is necessary. This form of thermal inactivation can be conceptualized as initial enzyme E being converted to an intermediate E₁ with fractional activity β, before becoming fully deactivated enzyme E₂ (Epting, Vieille, Zeikus, & Kelly, 2005). The parameter β and inactivation rate constants k₁ and k₂ give the fraction of remaining enzyme activity (y) as a function of time:

$$y\left(t\right)=\left(1+\frac{\beta k_1}{k_2-k_1}\right)e^{-k_{1}t}-\left(\frac{\beta k_1}{k_2-k_1}\right)e^{-k_{2}t}$$

Values of β, k₁, and k₂ were calculated in Microsoft^® Excel using the Solver function to minimize the sum of squared differences between the model and experimental data.

The three enzymes were assayed together as an in vitro pathway in a mixture containing 100 mM Tris, 10 mM MgCl₂, 150 mM potassium acetate, 5 mM acetyl-CoA, pH 7.5, with enzymes added to activities of 5 U/mL for Thl and Adc, 15 U/mL for Ctf. Controls consisted of reaction mixture with each enzyme missing individually, and a no-enzyme control. The resulting mixtures were incubated at 70°C in a thermocycler, and acetone was detected by gas chromatography (GC).

Other methods

Acetone was detected on a GC-2014 gas chromatograph (Shimadzu, Kyoto, Japan) equipped with a ZB-WAXplus 30 m long, 0.53-mm ID capillary column (Phenomenex, Torrance, CA) and flame ionization detector. The GC oven temperature was initially held at 35°C for 3 min, increased to 150°C at 20°C/min, and held for 6 min. The injector was held at 220°C and FID detector at 300°C. Nitrogen was used as the carrier gas at a column flow of 30 cm/s. Samples of 0.1 µL were injected with a 1:10 split ratio using an AOC-20i autosampler.

Protein concentrations were determined by the Bradford method with BSA standards. SDS-PAGE was done using BioRad TGX 4 to 12% gels with standard Tris-glycine buffer. Samples were heated at 95°C for 15 minutes to ensure denaturation of the thermostable enzymes, and gels were stained with GelCode blue. Blue native PAGE used the Novex NativePAGE kit, with a 4–16% Bis-Tris gel. Imaging relied on running the gel with Dark Blue Cathode Buffer, followed by overnight de-straining in 10% acetate, 50% methanol solution.

Results

Extremely Thermophilic Acetone Pathway Candidates

BLAST searches for homologs to the C. acetobutylicum acetone pathway enzymes indicated several promising candidates (Table 2). The thiolase from Caldanaerobacter subterraneus subsp. tengcongensis (formerly Thermoanaeobacter tencongensis) is 68% identical at the amino acid level, and in fact was already characterized and utilized as part of a synthetic pathway for n-butanol production, which starts with the same condensation reaction of two acetyl-CoA molecules to form the four carbon acetoacetyl-CoA (Loder et al., 2015).

Table 2:

Enzyme candidates for thermophilic acetone pathway based on Clostridium acetobutylicum pathway

Source Organism	Topt (°C)	Accession number	% amino acid identity	Evidence
Thiolase (Thl) – C. acetobutylicum: AAK80816.1
C. subterraneus subsp. tengcongensis	70	WP_011024972	68% (99% cov)	Characterized, 70°C (Loder et al., 2015)
Acetate acetoacetyl-CoA transferase (CtfAB) – C. acetobutylicum: NP_149326 & 27
C. subterraneus subsp. tengcongensis	70	WP_009610465 WP_011025123	65% (97% cov) 65% (100% cov)	Homology
Thermosipho melanesiensis	70	WP_012057350 WP_012057349	53% (96% cov) 64% (94% cov)	Activity, 55°C (Shaw et al., 2015)
Acetoacetate decarboxylase (Adc) – C. acetobutylicum: NP_149328
Clostridium acetobutylicum	35	NP_149328	100 (100% cov)	Activity, 70°C (Davies, 1943)
Vulcanisaeta distributa	85	WP_013335399	29% (92% cov)	Homology
Sulfolobus contig	75	WP_009989587	24% (98% cov)	Homology

C. subterraneus also provided a candidate for the acetoacetyl-CoA:acetate CoA-transferase, with 65% identity for both subunits, followed closely by Thermosipho melanesiensis with 53% and 64% identity for the alpha and beta subunits, respectively. The T. melanesiensis enzyme hit was also not surprising, having been identified in the search for a more moderately thermophilic acetone pathway functional at 55°C (Shaw et al., 2015). However, while that report confirmed activity by acetone formation in T. saccharolyticum at 55°C and room temperature activity assay of unpurified cell extracts, no detailed biochemical characterization of the enzyme or its thermostability was reported.

The same study utilized an acetoacetate decarboxylase from Bacillus amyloliquefaciens with significant sequence identity (65%) to C. acebutylicum Adc, but this organism’s optimum growth temperature of 50°C was lower than desired for an extremely thermophilic pathway. Unfortunately, the acetoacetate decarboxylase is a highly specialized enzyme for solvent production, and the only extremely thermophilic candidates identified with a coverage score over 80%, from Vulcanisaeta distributa and a metagenomics sequence annotated as ‘from a Sulfolobus species’, had low sequence homology (< 30% amino acid identity). One surprising finding in the first report characterizing the activity of the C. acetobutylicum Adc (Davies, 1943) indicated that the enzyme retained activity up to 70°C, suggesting that the native enzyme may be viable in an extremely thermophilic pathway despite its mesophilic origins.

Recombinant Production of Acetone Pathway Enzymes

Thiolase:

The thiolase from C. subterraneus had been purified and characterized previously (Loder et al., 2015) and Figure 2 confirms that the IMAC purified preparation utilized in the in vitro experiments forms the expected homotetramer.

Figure 2: SDS-PAGE and Blue-native PAGE of acetone enzymes.

SDS-PAGE indicates that all enzymes have the expected M_r and are at least 80% pure by determined densitometry (except for C. subterraneus CtfB). Blue-Native PAGE also confirms that enzymes exhibit the expected multimolecular arrangement, with the exception of C. subterraneus Ctf (possibly a result of the low purity of the beta subunit). Neither Ctf beta subunit is visible on Native PAGE, despite all lanes loaded at equal mass. The smear in both Ctf beta lanes suggests these subunits may have denatured.

Acetoacetyl-CoA:Acetate CoA-Transferase:

The CtfAB complex (α₂β₂) has been purified from E. coli (Sramek & Frerman, 1975b) and C. acetobutylicum (Wiesenborn, Rudolph, & Papoutsakis, 1989), and in both cases is described as a fastidious enzyme requiring a buffer with at least 20 wt% glycerol and 500 mM ammonium sulfate to remain stable. In both previous studies, the alpha and beta subunits were co-purified as a complex, as did a subsequent study expressing the C. acetobutylicum CtfAB enzyme in E. coli (Cary et al., 1990). The salting-out purification described by these earlier reports was utilized here for mixtures of the T. melanesiensis and C. subterraneus subunits expressed individually in E. coli. This resulted in partially purified cell extracts with Ctf activity at 70°C, which was the first confirmation that the C. subterraneus enzyme functions as a thermophilic Ctf, but purified enzymes were needed for subsequent assays.

Histidine-tagged fusion proteins are readily purified (as was the case for Thl), but the tag can interfere with enzyme activity if the tagged N or C terminus is near the enzyme active site or dimerization interfaces. Fortunately, examination of a crystal structure of the E. coli CtfAB complex (PDB ID: 5DBN) indicated that all subunits N and C termini are at the outer surface of the heterotetramer, well separated from the subunit interfaces and active sites. Given the structural similarities between family-I transferases, even those from vastly different lineages (Coros, Swenson, Wolodko, & Fraser, 2004), the extremely thermophilic candidates here seemed likely to share this general structure. Therefore, N-terminal histidine-tagged alpha and beta subunits of both CtfAB complexes were expressed in E. coli for purification. An initial lysis in standard IMAC buffer (lacking glycerol and sulfate) resulted in cell extracts with a strong band for both T. melanesiensis and C. subterraneus alpha subunits on SDS-PAGE, but bands for the beta subunits were only visible in the insoluble cell fraction. Subsequent lysis in IMAC A buffer reformulated for Ctf purification was able to recover both subunits in soluble form, but yields of the beta subunits were low, since most was still lost to the insoluble fraction (particularly for C. subterraneus). Purification using IMAC buffers containing glycerol and sulfate resulted in highly purified alpha subunits, as well as purified T. melanensiensis CtfB, but the C. subterraneus CtfB subunit was only partially purified (78%), as determined by SDS-PAGE densitometry (Figure 2).

In contrast to previous efforts, individually purified Cft alpha and beta subunits from separate expressions were obtained for biochemical characterization; the E. coli alpha subunit had been previously purified alone for structural but not biochemical analysis (Korolev et al., 2002). The results shed light on the difficulty of recovering active Ctf in previous studies. It appears that the cause of Ctf’s low stability is specifically attributable to the beta subunit, since alpha subunits from each thermophile purified easily even without ammonium sulfate and glycerol as stabilizers. It is also possible that the beta subunit could only be recovered alone in this case due to the greater stability inherent in thermophilic proteins. Blue Native PAGE analysis confirms that the T. melanensiensis CtfAB complex exists as the expected heterotetramer (α₂β₂), although bands for the heterodimer and heterooctamer are also visible (Figure 2). The Ctf subunits were also loaded individually at equal levels, but both beta subunits are only faintly visible as broad smears – likely reflecting denaturation or aggregation in the gel, and further evidence of their low stability. No oligomers are visible for the C. subterraneus CtfAB complex, although this could be a result of either the poor stability or low purity of the beta subunit.

Acetoacetate Decarboxylase:

The acetoacetate decarboxylase candidate from Sulfolobus sp. was expressed in both His-tagged and untagged forms and its presence as a soluble protein was confirmed via SDS-PAGE gel. Here, neither form demonstrated any acetoacetate decarboxylase activity at temperatures ranging from 40°C to 80°C. The V. distributa enzyme was also expressed in E. coli in tagged and un-tagged forms, but neither could be solubilized under any conditions tested. Inclusion bodies containing the protein of interest were solubilized by urea extraction and SDS-PAGE gel confirmed the presence of the protein in the inclusion bodies but the protein could not be refolded to soluble form via dialysis after many trials. In contrast, the C. acetobutylicum Adc was functionally expressed in both his-tagged and untagged forms, but the tagged enzyme exhibited significantly reduced activity (data not shown). The active enzyme complex is a homododecamer (Ho et al., 2009), and it is plausible that the tag interferes with subunit assembly or interferes with substrate access to the active site. Therefore, untagged C. acetobutylicum Adc was expressed and the cell extract was heat-treated, resulting in a surprisingly pure protein (84% according to densitometry), which Native PAGE confirmed formed the expected 12-subunit complex (Figure 2).

Biochemical Characterization of Extremely Thermophilic Acetone Pathway Enzymes

Thiolase:

As the C. subterraneus Thl had been previously characterized (Loder et al., 2015), only thermal stability studies were conducted on this enzyme before use in the three enzyme pathway assays. The functional enzyme complex is a homotetramer displaying remarkable thermostability. High temperature incubation actually increases the enzyme activity initially, before starting a slow, linear decline, such that there is a small amount of residual activity even after 20 hours at 70°C (Figure 3d).

Figure 3: Thermal stability of thermophilic acetone production enzymes.

Residual enzyme activity (y-axis - log scale) following incubation at 70°C for various times (x-axis – minutes), relative to initial enzyme activity (normalized to 1). (a) C. subterraneus Thl exhibits brief heat activation followed by a long, approximately linear decay. (b) T. melanesiensis Ctf in complex exhibits the common two-step thermal inactivation. (c) C. acetobutylicum Adc shows significant activation following 1 h incubation, after which activity slowly declines. (d) Individual subunits of T. melanesiensis Ctf incubated separately have very different thermostabilities: alpha shows no decline after one h (linear trendline), while beta loses 90% of activity within 5 min (exponential decay).

Acetoacetyl-CoA:Acetate CoA-Transferase:

Purification of the individual alpha and beta subunits of Ctf from T. melanesiensis and C. subterraneus allowed assay of the individual subunits for activity: neither subunit from either species was capable of catalyzing the CoA transfer reaction alone.

While we observed no activity from either the alpha or beta subunits alone, activity was observed using hybrid assemblies (C. subterraneus CtfA with T. melanesiensis CtfB and vice-versa), but the hybrid complexes exhibited specific activities roughly 1% of T. melanesiensis CtfAB. Interestingly, the individual subunits also displayed dramatically different thermostabilities; incubation of the beta subunit of either species at 70°C for a few minutes was enough to eliminate activity, while the alpha subunit showed no reduction in activity after an hour (Figure 3b). The combined subunits of T. melanesiensis exhibited a stability intermediate between the two pure subunits, with a half-life at 70°C of 95 minutes. Thermal degradation of T. melansiensis CtfAB complex followed a two-step inactivation, where activity was rapidly reduced to approximately 60% in the first 15 minutes of high temperature incubation, but then declined much more slowly (Figure 3a).

Inactivation parameters (equation in Methods) for T. melanesiensis CtfAB at 70°C were: β = 0.64, k₁ = 0.324 min⁻¹, and k₂ = 2.68*10⁻³ min⁻¹, which implies a half-life of 96 minutes. The low purity and poor stability of C. subterraneus CtfB complicated activity assays, but based on limited data the C. subterraneus CtfAB enzyme complex appears to have thermostability comparable to T. melanesiensis, with residual activity evident even after 12 h at 70 °C.

Observed specific activities for the purified T. melanesiensis CtfAB were similar to those reported for other versions of the enzyme. With substrate concentrations of 25 mM acetate and 0.2 mM acetoacetyl-CoA we observed 80 µmol/min/mg protein (or U/mg), while under comparable conditions E. coli acetoacetyl-CoA:acetate CoA-transferase has specific activity of approximately 150 U/mg (Sramek & Frerman, 1975a). The value obtained for purified C. acetobutylicum Ctf under similar assay conditions was 29.1 U/mg, compared to 0.36 U/mg in raw cell extracts (Wiesenborn et al., 1989). A specific activity of 3.57 U/mg was reported in cell extracts of recombinant T. saccharolyticum over-expressing T. melanesiensis Ctf, although that study assayed the thermophilic enzyme at room temperature utilizing a significantly different method (Shaw et al., 2015).

Acetoacetate Decarboxylase:

The results here confirm previous evidence of heat activation (Autor & Fridovich, 1970; Neece & Fridovich, 1967) with incubation at 70°C leading to an increase in activity over the first hour (Figure 3c), resulting in two-step inactivation parameters β = 2.28, k₁ = 0.077 min⁻¹, and k₂ = 0.5 ×10⁻³ min⁻¹. Since the activity of the intermediate enzyme E1 is greater than the initial, β has a value above 1. Reported k_cat values for C. acetobutylicum Adc at room temperature range from 165 s⁻¹ (Ho et al., 2009) to 1560 s⁻¹ (Highbarger, Gerlt, & Kenyon, 1996) with acetoacetate. Given that these values differ by roughly a factor of 10, and calculation of k_cat depends on molar concentration of enzyme, it seems likely that the larger value was calculated using the molecular weight of the dodecameric holozyme (330 kDa), while the smaller used the subunit weight of 27.5 kDa. If this is the case, converting the k_cat values to specific activity gives V_max values of 360 and 280 U/mg, respectively. Here, C. acetobutylicum Adc exhibited maximal activities in excess of 1,000 U/mg at 70°C, an increase which is probably attributable to the higher assay temperature, or to heat-activation during high-temperature incubations (not mentioned in either reference above).

Acetone Production in vitro by the Extremely Thermophilic Enzyme Pathway

A mixture of all three enzymes converted acetyl-CoA and acetic acid to acetone at 70°C; omitting any one enzyme eliminated acetone production (Figure 5). Sample chromatograms are shown alongside a 2.5 mM acetone standard in assay buffer, which seems to distort the acetone peak, since it displays a broad right shoulder in both the standard and in vitro reaction. However, when acetone standards were analyzed in water, no such shoulder was present. The relative size of the peaks suggests evaporation or incomplete conversion, as the reaction stoichiometry would predict that 5mM acetyl-CoA with excess acetate would result in 5 mM acetone. The peak at 4.1 min matches the acetoacetic acid standard, which unsurprisingly is most prominent in the no-Adc control, but some acetoacetate is also visible as an intermediate in the full reaction.

Figure 5: In vitro function of full acetone pathway.

In vitro function of the three enzymes (Rxn) at 70°C is confirmed by production of acetone. Omitting any one enzyme eliminated acetone production, although a peak consistent with the acetoacetate intermediate is visible in the no-Adc control.

Discussion

Acetone is a promising candidate for production via bio-reactive distillation in an extreme thermophile using a pathway requiring three enzymes which function up to at least 70°C (Figure 6). The previously characterized C. subterraneus thiolase meets this requirement, as does the surprisingly thermostable C. acetobutylicum acetoacetate decarboxylase, which is fortunate since the putative thermophilic acetoacetate decarboxylase homologs from this study exhibited no relevant activity. The ease with which untagged C. acetobutylicum acetoacetate decarboxylase was separated from contaminating E. coli proteins and brought to greater than 80% purity by simple heat treatment serves as a reminder as to why recombinant thermophilic proteins can be strategic for biochemical studies. As such, other unusually thermostable proteins from mesophiles could be identified simply by heat-treating mesophilic cell extracts to remove the most labile proteins. The proteins could be separated by various means including liquid chromatography and the identity of the given thermostable proteins could be identified by mass spectrometry. Additionally, screening the heat-treated supernatant for residual activities may provide insights into the types of thermostable enzymes present.

Figure 6: Three enzyme pathway to acetone production.

The enzymes catalyzing the three steps of the acetone pathway (along with equilibrium constants K_eq calculated at 1 mM concentration, pH =7, T = 25°C for the reactions in the direction shown). High acetate concentrations drive the reaction forward to favor products by improving the Ctf enzyme kinetics and thermodynamics. The pathway can be used with a native acetate producer, where it simultaneously detoxifies acetate while generating acetone. Equilibrium constants from eQuilibrator (Flamholz, Noor, Bar-Even, & Milo, 2012).

Thiolase:

The C. subterraneus Thl was characterized in detail previously (Loder et al., 2015). The functional enzyme complex is a homotetramer displaying remarkable thermostability with high temperature incubation increasing initial activity. The C. subterraneus Thl also functions well in the desired direction (formation of acetoacetyl-CoA), with V_max = 74 U/mg and a strong affinity for the acetyl-CoA substrate (K_M = 271 µM). The strong affinity for acetyl-CoA is desirable in order for the thiolase to operate in the thermodynamically unfavorable acetoacetyl-CoA forming direction. Many thermophilic thiolases are known to catalyze the reverse reaction, cleavage of acetoacetyl-CoA to two acetyl-CoAs, and are prevalent in the carbon fixation cycle of the extremely thermoacidophilic Sulfolobales (Berg, Kockelkorn, Buckel, & Fuchs, 2007). Thus all enzyme parameters must be considered when seeking a thermophilic analog of a known mesophilic enzyme.

Acetoacetyl-CoA:Acetate CoA-Transferase:

With a previously characterized extremely thermophilic thiolase and the surprising stable acetoacetate decarboxylase from C. acetobutylicum, the missing link for acetone production was an extremely thermophilic acetoacetyl-CoA:acetate CoA-transferase, two of which have been reported here. Either T. melanesiensis or C. subterraneus could serve as the basis for a thermophilic acetone pathway, since both complexes display sufficient activity and thermostability. In addition to reporting the first purified thermostable acetoacetyl-CoA:acetate CoA-transferase subunits alpha and beta and their heterotetramer complex (α₂β₂), the results here shed light on the properties of acetoacetyl-CoA:acetate CoA-transferases. The function of the hybrid T. melanesiensis / C. subterraneus Ctf complexes is likely due to the strong structural and sequence similarities among acetoacetyl-CoA:acetate CoA-transferases, particularly around the active site, and warrants further investigation to determine if this cross-functionality is evident among homologs from other species as well. Meanwhile, the dramatic difference in thermostability between the Ctf alpha and beta subunits, and poor stability of the beta subunits in general, helps to explain why this class of enzymes has been so challenging to purify in the past. The dramatically improved stability of the full CtfAB complex compared to the beta subunit highlights the role that the alpha subunit plays in stabilizing its partner, and serves as example of how important quaternary structure interactions can be for protein stability (the surprisingly thermostable dodecameric C. acetobutylicum acetoacetate decarboxylase is yet another example).

While Ctf complexes from E. coli and C. acetobutylicum have been purified and characterized for kinetics and substrate preferences, subunits from Ctf enzymes have not been purified separately for analysis. There is one instance of the co-purified E. coli subunits being separated by subsequent urea denaturation, which claimed the pure beta subunit had approximately 2% of the activity of the intact complex (Frerman & Duncombe, 1979). However, since the two subunit complex was used initially, the possibility of a small amount of residual contamination of active complex in the individual subunit fractions cannot be ruled out. The same report indicated that the alpha subunit was involved in structural support or maturation, which agrees with the findings here that it dramatically increases the stability of the beta subunit. At the same time, this rules out the possibility of the alpha subunit playing a catalytic role because it does not contain the nucleophilic glutamate involved in catalysis. More recent reports indicate the alpha subunit plays an important role in binding the CoA group (Korolev et al., 2002), in addition to the structural support it provides the beta subunit. The presence of highly conserved residues neighboring the active site in each subunit also suggests that both contribute to catalysis (Figure 4).

Figure 4: Alignment of conserved active site residues from Family I CoA transferases.

Residues neighboring the active site glutamate (red arrow) are highly conserved in Family I CoA transferase alpha and beta subunits. Boxes indicate Prosite entries PS01273 and 01274, the alpha and beta subunit sequence motifs. Pig succinyl-CoA transferase consists of a single peptide, but shares sequence and structural similarities with the heteromeric bacterial enzymes. Sequence alignments made in Geneious 8.1 (Kearse et al., 2012).

Family-I CoA transferases are known to exhibit Ping-Pong enzyme kinetics (also called “double displacement” or “substituted-enzyme” kinetics). In this case, one substrate binds the enzyme, is modified, and then dissociates as the first product, leaving the enzyme in a modified intermediate state, followed by binding of the second substrate, which is modified and dissociates as the second product, recovering the original enzyme. In the case of acetoacetyl-CoA:acetate CoA-transferase, the net reaction can be broken up into the two component steps:

Net Rxn: Acetoacetyl-CoA + Acetate → Acetoacetate + Acetyl-CoA

Step #1: Acetoacetyl-CoA + Enzyme → Enzyme-CoA + Acetoacetate

Step #2: Acetate + Enzyme-CoA → Enzyme + Acetyl-CoA

The two substrates and sequential nature of the reaction leads to unusual kinetics, where substrate inhibition appears at relatively low concentrations, but can be overcome by increasing the concentration of the other substrate (Wiesenborn et al., 1989). This unusual substituted-enzyme substrate inhibition is easily apparent on single and double-reciprocal plots (Cornish-Bowden, 1995), which matched kinetic data from this study for T. melanesiensis, confirming that catalysis with this enzyme proceeds through the same Ping-Pong mechanism observed in mesophilic versions. The relative thermal stabilities of the enzymes in the synthetic pathway should be taken into account when designing cloning constructs for recombinant expression in extremely thermophilic hosts. While the discovery that T. melanesiensis acetoacetyl-CoA:acetate CoA-transferase has a half-life of over 1 hour at 70°C indicates that the proposed extremely thermophilic acetone pathway is viable, it remains the least thermostable of the three pathway enzymes, inactivating more rapidly than the C. subterraneus thiolase.

Acetoacetate Decarboxylase:

C. acetobutylicum Adc has been thoroughly characterized including the precise nature of its active site and catalytic mechanism which have been of interest to biochemists (Highbarger et al., 1996; Ho et al., 2009), and historically its surprising thermostability was a topic of study (Autor & Fridovich, 1970; Neece & Fridovich, 1967).

Unexpectedly, the mesophilic C. acetobutylicum Adc is the most stable enzyme in this studied extremely thermophilic acetone pathway. The dramatic increase in activity at the beginning of high temperature incubation meant that even after 12 h, activity was still greater than prior to heat-treatment.

Acetone Pathway:

Given that Adc also appears to be the most active enzyme, it could be expressed at the end of an operon, or separately under the control of a moderate transcription level promoter. Thl and Ctf have comparable stabilities and activities, so equivalent expression would be appropriate. Given the instability of Ctf beta subunit, and the fact that all Ctf genes appear to exist in a ctfAB operon (often with overlapping start/stop codons), co-expression of the two subunits seems to be essential, especially at high temperatures.

As shown in Figure 6, only the reaction catalyzed by Adc is strongly favored thermodynamically (k_eq significantly above 1). The thiolase reaction dramatically favors the reverse direction, as indicated by the very low equilibrium constant, such that intracellular acetoacetyl-CoA concentrations will be two to three orders of magnitude lower than acetyl-CoA. One way to drive the CoA transferase reaction forward is to increase the concentration of acetate, which is a known factor in driving the switch to solventogenesis in C. acetobutylicum (Wiesenborn et al., 1989). Using a host that is an efficient natural acetate producer and tolerates relatively high concentrations of acetate may facilitate acetone production. Fortunately, there are a number of fermentative extreme thermophiles that produce acetate as a major fermentation product and for which genetic tools are available (Loder et al., 2017; Zeldes et al., 2015).

Additionally, this acetate consuming pathway may be useful in converting the acetate present in the lignocellulosic biomass in the form of acetyl groups. Lignocellulosic biomass contains an appreciable amount of acetate (2–5 dry wt%) in the form of acetylated compounds, primarily found in the hemi-cellulose component (Kong et al, 1992), and is thus a potential carbon source for conversion to biofuels and biochemicals. During biomass processing, whether chemical or biological, the acetyl groups are cleaved from carbohydrates via hydrolysis resulting in free acetate (Pawar et al, 2013). Acetate is typically not metabolized or consumed by microorganisms targeted for conversion of lignocellulosic materials to biofuels via carbohydrate fermentation, yet there is potential for a carbon efficiency improvement if this fraction can be converted to a useful product.

The next steps are to demonstrate that the extremely thermophilic acetone pathway can indeed be utilized in an extremely thermophilic host and to achieve titers necessary for efficient bio-reactive distillation. Efforts targeting this objective are underway.