Keywords

1 Introduction

Biomanufacturing is defined as manufacturing the desired products by using living biological organisms (e.g., bacteria, yeasts, plants) or some components from one or several biological organisms. Biomanufacturing has great potential to become a defining technology, especially in the sustainability revolution [1, 2]. Research and development in the fields of biotechnology, bioengineering, and nanomaterials have dramatic impacts on both the products that we are able to create and the ways in which we create them. Potential products that could be produced through biomanufacturing can be listed in an increasing order of selling prices: biocommodities ($0.3–several US dollars per kg), specialties and biomaterials (tens of dollars per kg), fine chemicals (hundreds of dollars per kg), pharmaceuticals (thousands of dollars/kg), to protein drugs (more than tens of thousands of dollars per kg) (Fig. 1).

Fig. 1
figure 1

Typical products produced by cell-free biosystems in terms of their selling prices from biocommodities, specialties and renewable materials, and fine chemicals to pharmaceuticals

Full size image

Biomanufacturing can be classified into two platforms: living organisms and cell-free biosystems [3, 4]. Living organisms, especially microorganisms, have been utilized for several thousand years to produce a number of products that meet mankind’s needs. As a result, living entities dominate as whole-cell biocatalysts in current biomanufacturing systems. With the development of genetic engineering, protein engineering, systems biology, and synthetic biology, we have gained the ability to modify living organisms to produce natural products at high yields or produce non-natural products. However, the potential of cell-free biosystems in biomanufacturing is often ignored. Herein we define cell-free biosystems composed of more than three catalytic components in one reactor. Thus, we intentionally exclude the use of one to three catalytic components as has been widely used to produce fructose from glucose, produce glucose from starch, and produce chiral alcohols, on large scales [5–9]. Therefore, we do not review traditional enzyme-mediated biocatalysis in this chapter.

Cell-free biosystems have many advantages over living-organism biomanufacturing, such as high product yield, fast reaction rate, high product titer, unprecedented level of control and freedom of design, and broad reaction conditions [3, 4, 10–12] (Table 1). Cell-free biomanufacturing usually gives a yield of the desired product that is often close to a theoretical value when the reactions are irreversible or when the products can be removed in situ. Living systems suffer from low product yields [13, 14] because a significant fraction of carbohydrate and/or energy is used for duplication of cells and formation of side-products when using living organism for manufacturing [15]. For example, all hydrogen-producing microorganisms regardless of whether natural or genetically modified cannot produce hydrogen at yields of more than four moles per mole of glucose, called the Thauer limit [16], because complete oxidation of glucose with water as an oxidant cannot generate any ATP supporting living systems. Conversely, cell-free enzyme cocktails are able to produce nearly 12 moles of hydrogen per mole of glucose [17]. Much higher reaction rates are believed to be possible for cell-free systems than living organisms because (i) neither cell membrane nor wall is present to slow down substrate/product transport, (ii) no energy is needed for transport of substrate/product across the membrane, and (iii) much higher concentrations of biocatalysts can be present in the reactors and no side reactions “slow” the production of the desired product [11]. For example, the highest power density of microbial fuel cells is expected to be around 0.79 mW/cm2 [18], while enzymatic fuel cells can generate power densities of up to 8.5–24 mW/cm2 [19, 20]. Sometimes too slow productivities mean low potential in industrial biomanufacturing [14, 21]. A high concentration of fermentation products can greatly decrease product separation costs [15] while often inhibiting the growth of living organisms, resulting in low product titers. Cell-free biosystems can be controlled much more easily than living organism because the latter has complicated feedback control for gene regulation, protein transcription, translation, and metabolite fluxes. Therefore, cell-free protein synthesis (CFPS) has become an alternative choice for fast synthesis of recombinant proteins in academic laboratories. Furthermore, cell-free protein synthesis has reached a 100-L milestone [22]. In principal, enzymes, especially thermostable enzymes, can tolerate higher levels of organic solvents or toxic chemical compounds than living organisms, resulting in high product titers. Recently, Sieber and co-workers demonstrated that the enzyme cocktails can produce isobutanol even at a concentration of more than 4 % [23]. Additionally, cell-free biosystems can be conducted under much broader reaction conditions than living organisms, for example in organic solvents, ionic liquids, or in the presence of compounds that are toxic to microorganisms [24]. Most building blocks of cell-free systems are highly exchangeable [25]. For example, high-yield enzymatic hydrogen can be produced by using enzymes isolated or originated from bacterium, yeast, plant, rabbit, and archaebacterium [26]. Given the advantages of cell-free systems over living organisms, cell-free biosystems are emerging as a powerful biomanufacturing platform to expand the capabilities of natural biological systems without using whole-cell cells.

Table 1 Comparison of cell-free biosystems and a living organism for biomanufacturing
Full size table

Cell-free biosystems may be classified into two distinctive platforms according to their preparation method. The first is based on whole cell extract by breaking the cell membrane (Fig. 2a). The second is based on purified enzymes from different sources that are then mixed together (Fig. 2b). The cell extract can be prepared easily but some cellular components may influence the whole system’s performance and the whole cell extract has a short half-life time [27, 28]. Systems made of purified components require labor-intensive preparation and seem costly but they can be stabilized after each component is carefully optimized and their performance is more precisely controlled [14, 29]. Specific attributes of two cell-free biosystems enable their applicability to different products.

Fig. 2
figure 2

Cell-free biosystems can use a cell extract from one host a or purified enzyme from different hosts b for biomanufacturing

Full size image

2 History of Cell-Free Biosystems

The power of cell-free biosystems has been appreciated as a fundamental research tool for more than 100 years. In 1897, Eduard Buchner discovered a yeast extract (not living yeast) that can convert glucose to ethanol [30], leading to his Nobel Prize in Chemistry in 1907. Later, many scientists used cell-free systems to study the basic biological mechanisms and some have received the Nobel Prize for their efforts. An incomplete winners list includes Arthur Harden and Hans von Euler-Chelpin for their investigation of the fermentation of sugar and fermentative enzymes [31], Otto Warburg for his discovery of the nature and mode of action of respiratory enzymes [32], Carl and Gerty Cori for their discovery of the course of the catalytic conversion of glycogen [33, 34], Hans Krebs for his elucidation of the citric acid cycle [35], Melvin Calvin for his investigation of carbon dioxide assimilation in plants [36], and Nirenberg for his elucidation of the genetic codon by using a cell-free system to translate a poly-uracil RNA sequence [37].

Enzyme-based biotransformation became a manufacturing tool approximately 50 years after the discovery of enzymes. The developments of enzyme-based biotransformation can be divided roughly into three phases:

Phase 1 (1960s)—one-enzyme biotransformation [5, 38]. To solve enzyme stability and recycling issues, enzyme immobilization technology was developed. Invertase may be the first immobilized enzyme used commercially for the production of Golden Syrup by Tate and Lyle during World War II. Tanabe Seiyaku Co. (Japan) started the industrial production of l-methionine by using immobilized aminoacylase in a packed bed reactor in 1969. The Clinton Corn Processing Company (USA) produced fructose syrup using glucose isomerase in 1967. Current annual fructose production exceeds 9 million tons [38] and the longest published working lifetime of immobilized glucose isomerase is 687 days at 55 °C and pH 7.5 (Kato Kagaku, Japan) [5]. Semisynthetic beta-lactamase-resistant beta-lactam antibiotics (e.g., cloxacillin, flucloxacillin) are produced by using amidases [5, 39]. Enzymatic acrylamide production was initiated in 1985 and more than 100,000 metric tons of acrylamide per year is produced by using immobilized nitrile hydratases [38]. Progress is accelerating due to fast developments in protein engineering tools including directed evolution, rational design, and their combination [40], high cell-density fermentation for low-cost recombinant protein production [41], the discovery and utilization of thermoenzymes [42], and enzyme immobilization in nanomaterials [3]. As a result, enzyme-mediated biocatalysis is an alternative choice for numerous transformations from commodities to pharmaceuticals [43, 44].

Phase 2 (1990s)—multienzyme one pot for relatively complicated biotransformation. Multienzyme one pot has numerous benefits compared to single-enzyme reactors in cascade: fewer unit operations, smaller reactor volume, higher volumetric and space–time yields, shorter cycle times, and less waste generated. Also, by coupling steps together, unfavorable equilibria can be driven towards the formation of desired products [4, 45, 46]. For instance, enzymatic hydrolysis of crystalline cellulose requires a synergetic action of endoglucanase, cellobiohydrolase, and beta-glucosidase because a single enzyme cannot hydrolyze cellulose efficiently [47, 48]. For cofactor-dependent enzyme reactions that consume reduced NAD(P)H, in situ NAD(P)H-regenerated by another enzyme is becoming more and more accepted, especially for the synthesis of high-value chiral compounds in the pharmaceutical industry [3, 49, 50]. Similarly, NAD(P)H is usually generated by using a pair of a hydrogen-donor substrate and a single enzyme, including formate/formate dehydrogenase [51], glucose/glucose dehydrogenase [52], glucose-6-phosphate/glucose-6-phosphate dehydrogenase [42], dihydrogen/hydrogenase [53], and phosphite/phosphite dehydrogenase [54]. In the organic chemistry field, the synthesis of monosaccharides, activated monosaccharides, oligosaccharides, and glycopeptides by using multienzyme one pot has been intensively investigated [55–58].

Phase 3 (2000s)—the utilization of numerous enzymes (i.e., more than three) for implementing very complicated biotransformations. This cell-free biotransformation scheme has three representative directions: (i) cell-free protein synthesis, which utilizes natural protein synthesis systems in cell lysates for fast synthesis of proteins for research purposes and the production of high-value antibodies or other proteins [59, 60], (ii) in vitro synthetic biology for the production of high-value fine chemicals and pharmaceuticals [27, 61–64], and (iii) synthetic pathway biotransformation (SyPaB) for the production of low-value biocommodities [3, 4, 25]. Herein, we focus on the use of more than three catalytic units in one reactor because of its ability to implement biochemical reactions that living organisms cannot.

3 Examples of Cell-Free Biosystems for Biomanufacturing

We highlight several examples of cell-free biosystems so that readers can easily understand the advantages and applications of cell-free biosystems.

3.1 Cell-Free Biosystems for Biocommodity Engineering

Biocommodities are low-value and large-volume products, including biofuels, biochemicals, bioplatics, feed and food [4, 65]. Among them, transportation biofuels have received wide attention due to their nearly zero net greenhouse gas emissions, enhanced energy security, and new biomanufacturing job creation. Because feedstock costs usually account for more than half of biocommodity selling prices [65, 66], economically viable production of biocommodities requires high product yields and low production costs. Although living microorganisms can duplicate themselves easily and have seemingly low biocatalyst preparation costs, we argue that cell-free biosystems could be an alternative choice due to the unique advantages mentioned earlier: high product yields, fast reaction rates, and potentially very low biocatalyst costs as bulk enzymes become stable and are produced at low cost in the future [3].

There is no question that hydrogen would be the best energy carrier for the transport sector in the future due to the high-energy utilization efficiency through fuel cells and nearly zero pollutants for end-users. Hydrogen can be produced through a number of approaches based on a variety of feedstocks [67]. The production of biohydrogen from low-cost biomass is an excellent solution for producing low-cost hydrogen without net carbon emissions [16, 68, 69]. However, natural and genetically modified hydrogen-producing microbes cannot produce hydrogen in yields of more than four moles hydrogen per mole glucose, called the Thauer limit [14, 16, 70]. Chemical catalysis, such as gasification and aqueous phase reforming, suffers from low product yields due to low chemical selectivity. To solve these problems, a non-natural enzymatic pathway composed of enzymes from bacteria, yeasts, animals, plants, and archae was constructed for generating high-yield hydrogen from starch in 2007 [26] (Fig. 3). The Royal Society (the UK’s National Academy of Science) has praised this breakthrough as “the beginning stage of a cheap, green and high-yield hydrogen production” and as a good example of synthetic biology [71]. This synthetic pathway contains (i) a chain-shortening phosphorylation reaction on glucan for producing glucose-1-phosphate (G1P) catalyzed by glucan phosphorylase (Eq. 1); (ii) the conversion of G1P to glucose-6-phosphate (G6P) catalyzed by phosphoglucomutase (Eq. 2); (iii) a pentose phosphate pathway and gluconeogenesis pathway containing ten enzymes for producing 12 NADPH and 6 CO2 per G6P (Eq. 3); and (iv) the generation of hydrogen from NADPH catalyzed by hydrogenase (Eq. 4).

Fig. 3
figure 3

The synthetic pathway for complete conversion of glucan and water to hydrogen and carbon dioxide. PPP pentose phosphate pathway. The enzymes are: GNP glucan phosphorylase; PGM phosphoglucomutase; G6PDH G-6-P dehydrogenase; 6PGDH 6-phosphogluconate dehydrogenase; R5PI phosphoribose isomerase; Ru5PE ribulose 5-phosphate epimerase; TKL transketolase; TAL transaldolase; TIM triose phosphate isomerase; ALD aldolase; FBP fructose-1,6-bisphosphatase; PGI phosphoglucose isomerase; and H2ase, hydrogenase. The metabolites and chemicals are: g1p glucose-1-phosphate; g6p glucose-6-phosphate; 6pg 6-phosphogluconate; ru5p ribulose-5-phosphate; x5p xylulose-5-phosphate; r5p ribose-5-phosphate; s7p sedoheptulose-7-phosphate; g3p glyceraldehyde-3-phosphate; e4p erythrose-4-phosphate; dhap dihydroxyacetone phosphate; fdp fructose-1,6-diphosphate; f6p fructose-6-phosphate; and Pi inorganic phosphate. Modified from Ref. [26]

Full size image
$$ \left( {{\text{C}}_{ 6} {\text{H}}_{ 10} {\text{O}}_{ 5} } \right)_{n} \, + \,{\text{P}}_{i} \, \rightleftharpoons \,\left( {{\text{C}}_{ 6} {\text{H}}_{ 10} {\text{O}}_{ 5} } \right)_{n - 1} \, + \,{\text{G1P}} $$
(1)
$$ {\text{G1P}}\, \rightleftharpoons \,{\text{G6P}} $$
(2)
$$ {\text{G6P}}\, + \, 1 2 {\text{NADP}}^{ + } \, + \, 7 {\text{H}}_{ 2} {\text{O}}\, \rightleftharpoons \, 1 2 {\text{ NADPH}}\, + \, 1 2 {\text{ H}}^{ + } \, + \, 6 {\text{CO}}_{ 2} \, + \,{\text{P}}_{i} $$
(3)
$$ 1 2 {\text{ NADPH}}\, + \, 1 2 {\text{ H}}^{ + } \, \rightleftharpoons \, 1 2 {\text{H}}_{ 2} \, + \, 1 2 {\text{NADP}}^{ + } $$
(4)

The combination of Eqs. (1)–(4) results in Eq. 5:

$$ \left( {{\text{C}}_{ 6} {\text{H}}_{ 10} {\text{O}}_{ 5} } \right)_{n} \, + \, 7 {\text{H}}_{ 2} {\text{O}}\, \to \,\left( {{\text{C}}_{ 6} {\text{H}}_{ 10} {\text{O}}_{ 5} } \right)_{n - 1} \, + \, 1 2 {\text{H}}_{ 2} \, + \, 6 {\text{CO}}_{ 2} $$
(5)

Thermodynamic analysis of Eq. (5) suggests that the overall reactions from starch or cellulosic materials and water are spontaneous and endothermic (∆G = –50 kJ/mol and ∆Ho = +598 kJ/mol) [17, 26]. These reactions are driven by entropy gains rather than enthalpy losses. The removal of gaseous products, H2 and CO2, from the aqueous phase favors the unidirectional reaction for hydrogen formation. When cellobiose is used as the substrate with a reaction time in the range of one week for a complete reaction, the overall yields of H2 and CO2 are 11.2 moles of H2 and 5.64 moles of CO2 per mole of anhydroglucose unit of cellobiose, corresponding to 93.1 and 94 % of the theoretical yields (Eq. 5). In principle, the theoretical yield of hydrogen (i.e., 12 H2 per glucose equivalent) can be obtained when a continuous reaction is conducted. Over the past few years, intensive efforts have been made pertaining to (i) decreasing enzyme costs and prolonging enzyme lifetime [17, 42, 72–75], (ii) producing all recombinant cytoplasmic enzymes at low costs [42, 76], and (iii) purifying recombinant proteins at low cost using heat precipitation [42, 75], ammonia sulfate precipitation [77, 78], cellulose-binding module-based adsorption and immobilization [79–82].

Ethanol is the most important gasoline additive because it can decrease air pollutants generated by internal combustion (Otto) engines. It can be produced through microbial anaerobic fermentation by the yeast Saccharomyces cerevisiae or by other microorganisms such as Escherichia coli and Zymomonas mobilis [83]. As early as 1897, Eduard Buchner discovered a yeast extract that converted glucose to ethanol [30]. Much later, Welch and Scopes (1985) started investigating the feasibility of high-yield production of ethanol by a reconstituted yeast glycolytic enzyme system—12 enzymes in total including ten enzymes required for the conversion of glucose to pyruvate and two enzymes required for the conversion of pyruvate to ethanol [84]. In this system, ATP accumulation (i.e., two ATP produced per glucose) prevents complete conversion of glucose to ethanol so that costly ATPase or highly toxic arsenate had to be carefully supplemented in Welch’s systems to achieve the high yield of ethanol. To solve this problem, Volker and co-workers designed a non-natural synthetic ethanol-producing pathway comprising six enzyme-catalyzed reactions only [23] (Fig. 4). In it, only four enzymes are required for the conversion of glucose to pyruvate. This synthetic pathway requires neither ATP nor CoA and has a balanced NAD cofactor. Since this pathway has a much lower Gibbs energy than those in yeast and Z. mobilis, it is anticipated that this cell-free pathway could have very fast production rates after optimization.

Fig. 4
figure 4

Schematic representation of a novel cell-free reaction pathway from glucose to ethanol and isobutanol. In the first part of the reaction (top box), glucose is converted into two molecules of pyruvate by four enzymes, then pyruvate can be either directed to ethanol (lower right box) or isobutanol synthesis (lower left box) in the second part of the reaction cascade. The molecules of CO2 and H2O are not shown for clarity. Enzymes are GDH glucose dehydrogenase; DHAD (gluconate/glycerate) dihydroxy acid dehydratase; ALDH glyceraldehyde dehydrogenase; KDGA 2-keto-3-desoxygluconate aldolase; ALS acetolactate synthase; KARI ketolacid reductoisomerase; KDC 2-ketoacid decarboxylase; PDC pyruvate decarboxylase complex; and ADH alcohol dehydrogenase. Modified from Ref. [23]

Full size image

Isobutanol is a four-carbon liquid alcohol that has several advantages over ethanol, such as, lower water absorption, better blending ability, higher energy density, and compatibility with current internal combustion engines [85]. Volker and co-workers also designed a synthetic pathway that can produce isobutanol from glucose by using a cell-free enzyme mixture (Fig. 4). This pathway is composed of two parts: the production of pyruvate from glucose and the generation of isobutanol through the enzymes in valine biosynthesis pathways along with 2-keto-acid decarboxylases (KDC) and alcohol dehydrogenase (ADH). One of the most beautiful aspects of this cell-free system is that this enzyme system can work well even in the presence of 4 % (v/v) isobutanol, whereas low levels of isobutanol (e.g., 1–2 % v/v) stops microbial isobutanol production [86].

Fructose is the sweetest monosaccharide. High-fructose corn syrup (HFCS) is a mixture of fructose and glucose produced from starch, which has simply replaced sucrose as a sweetener. On average, one person in the USA consumes approximately 20 kg of HFCS per year. HFCS is made from starch through a series of enzymatic conversions: (i) starch liquefaction mediated by amylose, (ii) starch saccharification to glucose mediated by glucoamylase, and (iii) conversion of glucose to fructose mediated by glucose (xylose) isomerase (Fig. 5). In this typical process, fructose yield cannot be very high because the last step is reversible and the process runs at equilibrium. To solve this problem, Benner and his coworker designed a novel enzymatic pathway in one pot containing (i) starch phosphorylase, (ii) phosphoglucomutase, (iii) phosphoglucose isomerase, and (iv) fructose-6-phosphatase (Fig. 5). In the last step, the net hydrolysis of fructose-6-phosphate yields fructose via a transaldolase-catalyzed reaction between fructose-6-phosphate and glyceraldehyde to yield fructose and glyceraldehyde-3-phosphate, which is then hydrolyzed to regenerate glyceraldehyde and inorganic phosphate by using a 3-phosphoglycerate phosphatase. Because the final step is exerogenic and irreversible, this process pulls all of the equilibrium intermediates to the desired product. This pathway design illustrates a general idea that the energetics of a pathway should be considered when designing multistep biocatalytic transformations [87].

Fig. 5
figure 5

Schematic representation of enzymatic fructose production from starch. GNP alpha-glucan phosphorylase; PGM phosphoglucomutase; PGI phosphoglucose isomerase; F6P fructose-6-phosphatase. Arrows in red are reversible reactions. Modified from Ref. [87]

Full size image

3.2 Cell-Free Biosystems for High-Value Product Synthesis

Cell-free systems can be used to produce high-value pharmaceuticals. Organic chemical synthesis plays a dominant role in producing chemicals in the modern pharmaceutical industry, although it is being challenged by enzyme-mediated biocatalysis [44]. For example, fondaparinux (trade name Arixtra) is a synthetic antithrombin III-binding pentasaccharide. Its chemical synthesis involves a large number of chemical synthesis reactions, resulting in very low yields of products (e.g., less than 0.5 %) and the generation of toxic intermediates [88]. As a result, Arixtra is a costly drug. In 2011, Xu et al. reported 10- and 12-step chemoenzymatic synthesis of two structurally homogeneous ultralow molecular weight heparins in 45 and 37 % overall yield, respectively, starting from simple disaccharides [89]. These ultra-low molecular weight heparins display excellent in vitro anticoagulant activity and comparable pharmacokinetic properties to Arixtra in a rabbit model. This enzymatic approach may be scaled up easily and shows great potential for a more cost-efficient way to synthesize this important drug class.

Sheldon and his coworker developed a one-pot procedure including four enzymes for the synthesis of carbohydrates from glycerol and an aldehyde (i.e., butanal) (Fig. 6). The first step is the conversion of glycerol to glycerol-3-phosphate at a cost of pyrophosphate mediated by phytase. The second enzyme (glycerol phosphate oxidase) converts glycerol-3-phosphate to DHAP and generates H2O2 as a by-product, where a third enzyme catalase is responsible for degrading H2O2 to water and oxygen. The fourth enzyme fructose-1,6-bisphosphate aldolase links DHAP and butanal to butanal-DHAP. At the last step, the first enzyme phytase releases phosphate from butanal-DHAP and generates 5-deoxy-5-ethyl-d-xylulose. Although the four enzymes have different optimal pHs, the four enzymes are put in one pot during the whole process. The pH of the reaction vessel is changed from pH 4.0 to pH 7.0 and then to pH 4.0, ensuring the cascading reactions occur in the desired order. The phytase on/off-switch at its respective pH 4.0 and 7.0 was the key to controlling phosphorylation and dephosphorylation [55].

Fig. 6
figure 6

Schematic representation of enzymatic transformation from glycerol to carbohydrates in one pot with pH shifts. GPO glycerol phosphate oxidase; ALD fructose-1,6-bisphosphate aldolase. Modified from Ref. [158]

Full size image

Cell-free biosystems have been developed for the biosynthesis of radio-labeled purines (ATP and GTP) and pyrimidines (UTP and CTP) from glucose and ammonia, which requires 28 and 18 enzymes, respectively [63, 64]. Compared to the chemical synthesis, both purine and pyrimidine biosynthesis pathways can be reconstructed de novo for the incorporation of isotopes into specific sites. This work enables NMR detection for probing structural and dynamic characteristics of nucleic acids.

Enzymatic fuel cells (EFCs) represent a new type of fuel cell devices that can convert chemical energy stored in fuels into electricity mediated by redox enzymes [4, 90]. EFCs are an appealing micro-power source suitable for powering portable electronics because they have a number of features such as high-energy storage density; no flammability or explosion risk; biodegradability; low fuel costs; no costly, rare or heavy metals needed; and rapid “recharge” by injecting a sugar solution [91]. Nearly all EFCs extract only a small fraction of chemical energy in fuels by utilizing only one or a few oxidoreductase enzymes. The complete oxidation of fuels into electricity through engineered cascade pathways would have four benefits: (i) high energy utilization efficiency, (ii) high energy storage density, (iii) low product inhibition, and (iv) high power density [4, 92–94]. EFCs can utilize a large range of chemical compounds as fuels, including methanol, ethanol, glycerol, pyruvate, and glucose (these fuels are listed in increasing order of carbon number in the compound). To increase the fuel utilization efficiency, cascade enzymes can be employed. For example, three cascading redox enzymes have been used in an anode for the complete oxidization of one-carbon methanol to CO2 [95]. Similarly, two-carbon ethanol has been oxidized using an 11-enzyme pathway to generate more electrons [96]. Three-carbon glycerol and pyruvate have been oxidized using two cascading dehydrogenases [97] and enzymes in the Krebs cycle [98, 99], respectively. Furthermore, Minteer et al. have proposed the complete oxidation of glucose using glycolysis and the TCA cycle [92]. Their pathway design is more complicated than the one proposed here and involves ATP/GTP and acetyl-CoA. However, acetyl-CoA and ATP are labile and cannot be utilized for an extended period of time. Recently, Xu and Minteer [100] published another paper on glucose oxidation to CO2 through a new pathway that does not involve ATP and acetyl-CoA (Fig. 7). However, this design suffers from a very low power density because aldolase that breaks the C–C bond has a very low activity on non-phosphorylated carbohydrates [101, 102].

Fig. 7
figure 7

Schematic representation of complete oxidation of glucose for electricity generation through six enzymes. Modified from Ref. [100]

Full size image

Zhang and coworkers utilized four thermophilic enzymes in cascade for deep oxidation of glucose without the use of ATP [94]. Polyphosphate glucokinase converts glucose to glucose-6-phosphate using low-cost, stable polyphosphate rather than costly ATP [81]. Two NAD-dependent dehydrogenases (glucose-6-phosphate dehydrogenase and 6-phosphogluconate dehydrogenase) that were immobilized on the bioanode were responsible for generating two NADH per glucose-6-phosphate (i.e., four electrons were generated per glucose via a diaphorase-vitamin K(3) electron shuttle system at the anode). When the temperature was increased to 50 °C, the maximum power density increased to 0.322 mW/cm−2, which was approximately eight times higher than that based on mesophilic enzymes at the same temperature. These results suggest that the deep oxidation of glucose could be achieved by using multiple dehydrogenases in synthetic cascade pathways and that high power output could be achieved by using thermostable enzymes at elevated temperatures.

3.3 Cell-Free Protein Synthesis

While cell-free protein synthesis (CFPS) has been used for decades as a foundational research tool for understanding transcription and translation, recent advances have made possible cost-effective micro-scale to manufacturing scale synthesis of complex proteins [22]. CFPS is becoming a more acceptable tool for the fast synthesis of recombinant proteins, especially when the proteins are in vivo cytotoxic, regulatory, or unstable proteins that are difficult to express in living cells and/or the proteins contain unnatural amino acids [10, 59, 103]. CFPS is usually conducted by using a crude cell lysate from any given organism (e.g., bacterial, plant, or animal cells) supplemented with the DNA template encoding the desired protein, NTPs, a highly processive RNA polymerase, amino acids, and an energy supply while the cell lysate provides the translational machinery, accessory enzymes, tRNA, and cofactors [10]. The use of cell lysate greatly simplifies CFPS but it may have some negative impacts due to other cellular components in cell extracts, for example, a rapid depletion of energy charge [59] and the degradation of protein products or template nucleic acids by proteases or nucleases [29]. In 2001, Shimizu et al. reported a protein-synthesizing system reconstituted from recombinant tagged protein factors purified to homogeneity [29]. The system – termed the “protein synthesis using recombinant elements” (PURE) system—contains all necessary translation factors, purified with high specific activity, and allows efficient protein production. This system has more than 100 well-defined molecules for implementing this complicated protein biosynthesis. The PURE system exhibits high translational efficiency with the added advantage of simple manipulation of reaction conditions and easy purification of untagged protein product. As a result, New England Biolabs Inc. sells the PURE fast protein synthesis kit.

Currently CFPS can produce protein in yields exceeding grams of target protein per liter of reaction volume, in batch reactions lasting multiple hours, with cost decreases to several orders of magnitude, and at scales reaching the 100-L milestone by SutroBio [11]. These advances have inspired new applications in the synthesis of protein libraries for functional genomics and structural biology, the production of protein therapeutics [104], personalized medicines [105], vaccines [106], short antimicrobial peptides [107], and the expression of virus-like particles, among others.

4 Challenges

The potential of cell-free biosystems for biomanufacturing is often ignored or underappreciated by many bioengineers and scientists because this paradigm shift is an out-of-the-box solution so that most do not realize that technological breakthroughs in other fields may be game changing in their field. A good example is no-till farming replacing tilling or discing farmland. Mankind has tilled land for thousands of years so it was easy to forget that the primary goal of tilling was to kill weeds. Beveridge described it as conditional thinking in his famous book entitled “The art of Scientific Investigation” [108]. Even long after the invention of selective herbicides and complementary herbicide-resistant seeds, farmers continued to till their land due to the ruling paradigm. The situation changed when several agricultural engineering pioneers in the nonpoint water pollution field realized that tilling was not necessary with proper herbicide and improved seed use. In reality, these no-tilling pioneers did not invent any new technologies but proposed a new concept to solve a key problem in their field. Now no-till farming is widely adopted because of the environmental and economic benefits [109]. Similarly, doubts regarding cell-free biosystems may include (i) enzymes cannot be produced and purified at low cost, (ii) enzymes are not stable enough, (iii) coenzymes are expensive and labile, and (iv) optimal conditions for numerous enzymes are different (Table 2). To address the above challenges, the respective solutions are listed in Table 2.

Table 2 Challenges of cell-free systems and respective solutions
Full size table

4.1 Low-Cost Enzyme Production and Purification

Cell-free protein synthesis usually uses cell extracts to decrease enzyme costs, while most cell-free biosystems prefer using partially purified enzymes that avoid unnecessary side-reactions and may prolong reaction time up to weeks, months, or even years. Low-cost production of bulk enzymes has been achieved but most academic researchers do not know this because most purchase costly enzymes from Sigma or other enzyme vendors. For example, bulk enzymes, such as protease and amylase produced by Bacillus sp., cellulase produced by Trichoderma and Aspergillus sp., have selling prices of ca. 5–10 US dollars per kg of dry protein [4, 110]. The cost of protein production is highly related to its expression level. The higher the expression level, the lower the cost of the purified protein. Codon usage optimization is a common way to enhance recombinant protein expression levels. For example, more than 500-fold improvement in the expression of soluble Thermotoga maritima 6-phosphogluconate dehydrogenase has been achieved in E. coli by codon usage optimization, accounting for >30 % of the total cellular protein [42]. Some recombinant formate dehydrogenase expression levels in E. coli are as high as 50 % cellular proteins [111]. It is estimated that current costs of recombinant proteins produced by E. coli BL 21 are approximately 100 US dollars per kg of dry protein including materials, labor, and capital depreciation [43]. As enzyme production is scaled up, production costs would decrease further. Dr. Tao at EnzymeWorks (China) pointed out that current enzyme costs in his company were approximately 70 US dollars per kg of enzyme because they can grow the E. coli cell densities of more than 100 g dry cell weight per liter without the use of costly pure oxygen (personal communication). It is anticipated that the cost of bulk recombinant enzymes will decrease greatly when their markets are ready.

The impression of costly enzyme purification is often gained from high-purity enzymes produced by academic labs and for pharmaceutical protein drugs, which are purified through a series of chromatographic methods. However, cell-free biosystems can utilize relatively low-quality enzymes as building blocks. Please bear in mind that cell extracts without purification could work well. In addition, several low-cost scalable protein purifications are available and have been developed recently. For example, ammonia sulfate precipitation can be used as the first step of most protein purifications [77]. Instead of using commercial costly protein purification resins like Ni–NTA resin, and glutathione Sepharose 4B, low-cost cellulosic materials can be used for protein adsorption/desorption, purification, and immobilization [79–82]. For highly thermophilic enzymes expressed by E. coli, heat precipitation could be the simplest way to obtain purified proteins from E. coli cell extracts [42, 75]. In our laboratory, we have systematically investigated heat precipitation for the purification of recombinant enzymes produced from E. coli. We found that six recombinant enzymes cloned from T. maritima and produced in E. coli, for example, 6-phosphogluconate dehydrogenase [42], ribose-5-phosphate isomerase [75], aldolase [112], transaldolase [113], transketolase, and xylulokinase can be purified to more than 80–90 % homogeneity by simple heat treatment at 80 °C for 20 min.

Inspired by integrated circuits and natural metabolons [114], we developed a new protein purification method which can enrich, purify, and immobilize three cascade enzymes in one step (Fig. 8a). This method was based on the highly species-specific interaction between cohesins and dockerins from natural cellulosomes [112, 115–117]. A synthetic protein scaffold called scaffoldin was constructed containing a CBM3 module and three cohesins. The CBM3 module can tightly bind on cellulosic materials and cohesins are responsible for binding with the specific dockerin-containing enzymes. Dockerins can be located in either the N- or C-terminal of the target protein [118]. After mixing four cell extracts containing the synthetic scaffoldin and dockerin-containing target proteins with cellulosic materials (e.g., regenerated amorphous cellulose, RAC) and centrifugation, the synthetic three-enzyme metabolon can be purified and immobilized on RAC (Fig. 8a). Triosephosphate isomerase (TIM), aldolase (ALD), and fructose 1,6-bisphosphatase (FBP) were chosen for demonstration purposes. The protein expression levels of dockerin-containing TIM, ALD, and FBP were low (Fig. 8b, Lane 1, 2, 3), but they can be easily enriched and purified by mixing with the synthetic scaffoldin (Lane 5, 6, and 7). The self-assembled three-enzyme metabolon can be obtained in one step (Fig. 8b, Lane 9 and Fig. 8c). Such an enzyme complex also has the unique feature of substrate channeling because of the proximity of the cascade enzymes [114]. This enzyme complex showed more than one order of magnitude enhancements on reaction rates compared to the noncomplexed TIM, ALD, and FBP mixture [112] (Fig. 9). In conclusion, numerous methods can decrease enzyme purification costs.

Fig. 8
figure 8

a Schematic representation for the purification and co-immobilization of the synthetic three-enzyme complex, where the mini-scaffoldin contained three different types of cohesins and one family 3 carbohydrate-binding module, and three enzymes contained respective dockerin. b SDS-PAGE analysis of the E. coli cell extracts containing the recombinant proteins and RAC pull-down proteins (a). Lane M, protein marker; Lane 1–4, cell extract containing TIM-Doc1, ALD-Doc2, FBP-Doc3, and CBM-Scaf3, respectively; Lane 5, RAC adsorbed mini-scaffoldin, Lane 6–8, RAC adsorbed CBM-Scaf3 and TIM-Doc1, ALD-Doc2, and FBP-Doc3, respectively; and Lane 9, RAC adsorbed CBM-Scaf3, TIM-Doc1, ALD-DoC2, and FBP-Doc3. c Schematic representation of the self-assembled three-enzyme complex containing TIM-Doc1, ALD-Doc2, FBP-Doc3, and CBM-Scaf3 containing three different types of cohesins and one family 3 carbohydrate-binding module. Modified from Ref. [159]

Full size image
Fig. 9
figure 9

Profiles of fructose-6-phosphate production catalyzed by 2-μM synthetic enzyme complex (■) and 2-μM noncomplexed enzyme mixture (●) in 2.5-mM glycerealdehyde-3-phosphate at 60 °C, where the slope of the dashed lines was defined as the initial reaction rate. Modified from Ref. [112]

Full size image

4.2 Prolonging Enzyme Lifetime

Most enzymes in academic laboratories deactivate rapidly, giving researchers an impression that enzymes are not stable. In reality, a few enzymes are very stable with a shelf lifetime of years and have been widely used in daily life, for example, protease used in detergents, or enzymes used in diabetic test strips. Another famous example is immobilized glucose isomerase lasting for more than 2 years [5].

Enzyme deactivation can be addressed by using thermoenzymes, enzyme immobilization, protein engineering through directed evolution and rational design, and their combination. Our economic analyses pertaining to enzymatic hydrogen production suggest that enzyme costs would be minimal when total turnover numbers (TTN) of all enzymes are larger than 107–108 mol of product per mol of enzyme [4, 77]. When cell-free biosystems are used to produce high value products, acceptable TTN values could be lower. On the basis of our experiences, it is very easy to obtain enzymes with such high TTN values by using natural thermostable enzymes (Table 3). Also, Bommarius and his coworker suggest a very simple way to calculate the TTN value of an enzyme: TTN = k cat/k d, where k cat and k d are the turn-over number and degradation constant of the enzyme, respectively [119].

Table 3 Examples of stable enzyme building blocks suitable for biocommodity production
Full size table

The discovery and utilization of thermoenzymes may be the simplest strategy. During the past few years, we have produced a number of thermophilic enzymes in E. coli, such as Clostridium thermocellum cellodextrin phosphorylase [17], C. thermocellum cellobiose phosphorylase [17], C. thermocellum glucan phosphorylase [72], C. thermocellum phosphoglucomutase [73], T. maritima 6-phophogluconate dehydrogenase [42], T. maritima fructose bisphosphatase [74], and T. maritima pentose phosphate isomerase [75]. Most of them have TTN values of more than 107 as shown in Table 3.

Enzyme immobilization technology has been used to prolong the lifetime of enzymes for a long time [3, 120–122]. For instance, a one-step protein purification and immobilization method has been developed by using low-cost, ultra-high adsorption capacity RAC to adsorb CBM-tagged thermophilic C. thermocellum phosphoglucose isomerase (PGI) [82]. The resulting immobilized PGI is highly active and ultra-stable compared to the nonimmobilized PGI, with a TTN of more than 109 mol of product per mol of enzyme at 60 °C (Table 3). In the food industry, immobilized thermophilic glucose isomerase exhibits TTN values of ~5 × 108 mol of product per mol of enzyme. As a result, no company has a motivation to further prolong the lifetime of this enzyme.

Directed evolution and rational design are powerful approaches to enhancing the thermostability of enzymes [123, 124]. Among different desired properties of engineered enzymes, improving enzyme stability is the easiest. A number of companies and academic laboratories have developed tools for enhancing enzyme lifetime. For example, some enzyme companies and start-ups, such as Codexis, Biomethodes, EnzymeWorks, and Arzeda have demonstrated a number of successful examples for enhancing enzyme stability meeting their customers’ needs. For example, Codexis is developing a stable carbonic andrase used for capturing high-temperature and low-pH CO2 from the waste gas of power stations. Also, a combination of rational and random design also have a significant effect on enzyme stability [40].

4.3 Redox Enzyme Engineering

Since both NADP and NAD are not stable in vitro and are costly, it is important to replace them with low-cost biomimetic cofactors especially when cell-free systems are used to produce low-value biofuels and biochemicals (Fig. 10). The bulk prices of NADP, NAD, and nicotinate mononucleotide (NMN) are $4,500, $1,500, and $250 per kg, respectively (personal communication from Alex Tao). Biomimetic cofactors, such as NMN and 1-benzyl-1,4-dihydronicotinamide (BDN), not only have much lower selling prices but also have much better stability.

Fig. 10
figure 10

Structures of natural cofactors and biomimetic cofactors

Full size image

Redox enzyme engineering was initiated by Perham’s group over 20 years ago [125]. By using molecular modeling and comparing amino acid sequences responsible for cofactor binding sites, they changed NADP-preferred glutathione dihydrogen to NAD-preferred by site directed mutagenesis [125]. After this, by using rational design, a number of studies were conducted by swapping cofactor preferences from NADP to NAD [126–132], from NAD to NADP [133–137], and relaxing or broadening cofactor specificity [138–142].

In the 1990s, Lowe and coworkers developed a series of biomimetic analogues of NAD(P) based on triazine dyes [143–147]. Some natural dehydrogenases, such as horse liver alcohol dehydrogenase, can utilize such biomimetic cofactors for implementing redox reactions [146]. Later, Fish et al. determined that the pyrophosphate and adenosine groups associated with NAD are not essential in the hydride transfer and proposed the use of BDN or its analogues to replace natural cofactors [148]. Later, they also showed that wild-type horse liver alcohol dehydrogenase [149] and monooxygenase [150] can work on such biomimics. However, most wild-type redox enzymes cannot work on such biomimics. Clark and Fish demonstrated that a P450 mutant with two amino acid changes can work on these biomimics [151]. Also, another group demonstrated that engineered P450 can utilize Zn dust as an electron source rather than natural cofactors [152, 153]. In 2011, Zhao and coworkers presented a bio-orthogonal system that catalyzed the oxidative decarboxylation of l-malate with a dedicated biomimetic cofactor, nicotinamide flucytosine dinucleotide, where the redox enzymes were engineered by using saturation mutagenesis of the key amino acid sites [154].

NMN is a precursor of NAD(P) with a much smaller size as compared to NAD(P) (Fig. 6). A few wild-type redox enzymes function using NMN, including liver alcohol dehydrogenase [155] and glutamic dehydrogenase [156]. Recently, Scott et al. demonstrated that engineered Pyrococcus furiosus alcohol dehydrogenase has an ability to work on NMN [157].

Although the importance of redox enzyme engineering is gaining recognition for future biomanufacturing [1, 4, 14, 28], redox enzyme engineering remains at an early stage because no framework or general rules exist for engineering redox enzymes on non-natural cofactors [157]. This direction may become one of the top R&D priorities of cell-free biosystems, especially for the production of biocommodities but not for the production of pharmaceuticals that can use more costly natural cofactors.

4.4 Compromised Reaction Conditions

Although numerous enzymes used for the proof-of-concept cell-free biotransformation experiments have different optimal conditions [17, 26], they can work together under compromised conditions (e.g., 30–32 °C) where hyperthermophilic hydrogenase exhibited a very low activity. According to our experiences and literature data, most enzymes from different sources are highly exchangeable in cell-free biosystems [25]. In future applications, we will discover and utilize nearly all enzymes from one source and/or engineer some unmatched enzymes to make their optimal condition match most of other enzymes.

5 Conclusion

The significant advantages provided by cell-free biosystems (Table 1) are motivating their transition from basic research tools to a future biomanufacturing platform. Although a few obstacles to cell-free biosystems remain, all of them can be addressed by using well-known technologies (Table 2). The development of cell-free biosystems may have a similar trend to modern computers. The first prototypes were extremely costly, with low performance, and few applications. After the performance of each part (e.g., CPU, RAM) was improved greatly and standard parts were produced on a large scale, it was simple and less costly to assemble a customized high-performance computer at low prices by using the available standardized parts. Now it is time to discover and develop more stable enzymes as standardized building blocks and engineer redox enzymes that can work with less costly and more stable biomimic cofactors. Cell-free biosystems will eventually become a new biotechnology platform for biomanufacturing numerous products, especially for biocommodities because they are highly cost-sensitive to both product yields (or energy efficiencies) and production costs.