https://orcid.org/0000-0002-9314-7997
Abstract
Engineering enzymes to degrade solid substrates, such as crystalline cellulose from paper sludge or microplastics in sewage sludge, presents challenges for high-throughput screening (HTS), as solid substrates are not readily accessible in cell-based biosensor systems. To address this challenge, we developed a cell-free cellobiose-detectable biosensor (CB-biosensor) for rapid characterization of cellobiohydrolase (CBH) activity, enabling direct detection of hydrolysis products without cellular constraints. The CB-biosensor demonstrates higher sensitivity than conventional assays and distinguishes between CBH subtypes (CBHI and CBHII) based on their modes of action. Integration with the Echo 525 liquid handler enables precise and reproducible sample processing, with fluorescence signals from automated preparations comparable to manual experiments. Furthermore, assay volumes can be reduced to just a few microlitres—impractical with manual methods. This cell-free CB-biosensor with Echo 525 minimizes reagent consumption, accelerates testing, and facilitates reliable large-scale screening. These findings highlight its potential to overcome current HTS limitations, advancing enzyme screening and accelerating the Design-Build-Test-Learn cycle for sustainable biomanufacturing.
1. Introduction
Synthetic biology has garnered significant attention for its potential to enable sustainable biomanufacturing by leveraging engineered biocatalysts that convert waste materials (e.g. paper sludge and discarded plastics) into valuable products [1–5]. Biocatalyst engineering typically relies on an iterative Design-Build-Test-Learn (DBTL) cycle to construct various candidate enzymes and systematically evaluate them [6–8]. Recent advances in AI technologies have fuelled interest in data-driven enzyme engineering, which relies on the large-scale analysis of enzyme sequences and functions (phenotypes) [9, 10]. Traditional methods, such as liquid chromatography or gas chromatography, while highly accurate and specific, are often too slow for large-scale testing. Alternatively, transcription factor (TF)-based biosensors have been offering a sensitive and fast analysis method for high-throughput screening (HTS)-based enzyme engineering [11]. TF-based biosensors translate the target enzyme activities into easily readable output signals such as colorimetry, fluorescence (e.g. GFP), and bioluminescence (e.g. luciferase) [12–16], facilitating the rapid identification of highly active enzymes from large-scale libraries [17, 18].
Whereas TF-based biosensors are commonly employed to measure enzymatic products at the single-cell level [19], practical applications of microbial sensor cells are limited because solid substrates, such as lignocellulose from paper sludge, cannot easily penetrate the cellular membrane [20–22]. Cell-free expression (CFE) systems [23, 24] address these limitations by operating TF-based biosensors in vitro, utilizing transcription and translation processes in open environments, such as test tubes or microplates [25, 26]. In addition to broadening the range of substrates that can be tested, cell-free TF-based biosensors support rapid and efficient identification of target enzymes owing to the simpler setup and reduced dependency on maintaining cell viability, further facilitating their use in high-throughput enzyme screening [25, 27].
Biofoundries—AI-integrated automated facilities—accelerate the DBTL cycle, maximizing research efficiency. Experimental processes in biofoundries typically rely on microwell plate-based protocols, ranging from 96 to 1536 wells, enabling parallel processing of large enzyme libraries. Integrating CFE systems with biofoundries can significantly reduce sample volumes, costs, and analysis time while improving data reproducibility and reliability, which benefits biocatalyst engineering [7, 28–30]. Moreover, CFE systems with automation can further accelerate the DBTL cycle by eliminating time-consuming steps such as plasmid assembly and gene transfer [11, 26], making them particularly advantageous for large-scale enzyme screening, and supporting sustainable biomanufacturing.
Leveraging the capabilities of CFE systems with biofoundries, we explored the use of a cell-free TF-based biosensor for characterizing enzymes involved in the conversion of insoluble substrates, such as waste materials. We focused on cellulases, particularly those that degrade crystalline cellulose. While endo-type cellulases (endoglucanase, EC 3.2.1.4) can hydrolyze internal bonds at amorphous region [31], exo-type cellulases, known as cellobiohydrolases (CBHs, EC 3.2.1.91), can break down crystalline cellulose domain into smaller sugars including cellobiose [32–34]. To streamline this approach, we utilized the Echo 525 liquid handler, a noncontact, high-speed acoustic system that dispenses nanolitre volumes with precise control. We believe this system holds promise beyond CBH engineering, offering rapid enzyme characterization and the potential to significantly increase screening throughput, even when working with solid substrates.
2. Materials and methods
2.1 Bacterial strains and reagents
Table S1 presents the bacterial strains used in this study. The Escherichia coli DH5α strain was used for cloning, whereas the BL21 (DE3), Rosetta (DE3), and SHuffle T7 strains were used for protein expression. Microcrystalline cellulose was purchased from Daejung (Seoul, South Korea), whereas all other chemical reagents were obtained from Sigma-Aldrich (St. Louis, MO, USA). CBHI from Hypocrea jecorina (HjCel7A), α-amylase from human saliva, and β-amylase from sweet potato were also purchased from Sigma-Aldrich. The Gibson Assembly® master mix, RNase inhibitor, and PURExpress® in vitro protein synthesis kit were sourced from New England Biolabs (Ipswich, MA, USA). Oligonucleotides were synthesized by Macrogen (Seoul, South Korea) and Cosmogenetech (Seoul, South Korea). Polymerase chain reaction (PCR) was performed using a KOD One™ PCR master mix (TOYOBO Life Science, Japan). Plasmids were isolated using the Wizard® Plus SV DNA Purification System (Promega, Madison, WI, USA), and PCR products were purified using the Wizard SV Gel and PCR Clean-Up System (Promega).
2.2 Plasmid constructions
Tables S2 and S3 present the plasmids and primers used in this study. First, the pIVEX-sfGFP plasmid was constructed by PCR amplification of the sfGFP gene using primers designed for the pIVEX2.3d vector, followed by Gibson assembly. Subsequently, to construct the sensory plasmid pIVEX-PT7-CelO-sfGFP, the CelO operator site was inserted into the appropriate location using kinase and ligation methods. The plasmid encoding MtCel6A (GenBank accession no.: MZ826702) was constructed via Gibson assembly utilizing a partial fragment of pET28a as the backbone vector and a synthetic gene insert. The construct was verified via DNA sequencing.
2.3 Determination of fluorescence signal
The fluorescence of sfGFP was measured using either a Tecan Spark plate reader (Tecan, Austria) or a Victor X multilabel plate reader (PerkinElmer, Waltham, MA, USA), using a black 384-well polystyrene cell culture microplate (Greiner Bio-One, Austria). The fluorescence signal of sfGFP was assessed with excitation and emission wavelengths at 485 and 535 nm, respectively.
2.4 Expression and purification of the CelR protein
The plasmid encoding CelR (pET28-CelR) was introduced into E. coli BL21 (DE3) via chemical transformation. A starter culture was initiated by inoculating a single colony into a Luria–Bertani (LB) medium supplemented with 50 µg/ml kanamycin and incubating overnight at 37℃. The starter culture was then transferred to a fresh LB medium with 50 µg/ml kanamycin and cultured further. Isopropyl β-D-1-thiogalactopyranoside (IPTG) was added to induce the T7 promoter when the cell optical density (OD600nm) reached 1.0, to a final concentration of 0.1 mM. The cultivation temperature was set to 20℃. Cells were harvested via centrifugation at 1977 × g for 20 min, and the resultant cell pellet was collected for subsequent purification.
The cell pellet was resuspended in phosphate-buffered saline (PBS, pH 7.4) and subjected to sonication on ice for 12 min at 24% amplitude (2 s on and 2 s off pulses). The cell lysate supernatant, obtained via centrifugation, was applied to affinity chromatography using a Ni-column (Bio-Rad) with an FPLC system (NGC Quest 10, Bio-Rad). CelR was ultimately eluted with PBS (pH 7.4) and concentrated using a Vivaspin® 6 centrifugal concentrator with a molecular weight cutoff of 10 kDa (Satorius AG, Germany). The concentration of the purified CelR was determined using a Bradford assay [35].
2.5 Expression and purification of cellulases
The pET28a-MtCel6A was introduced into the E. coli SHuffle T7 strain by electroporation. The pET22b-CelEdx16 plasmid, encoding a bifunctional endo-/exocellulase (CelEdx16, GenBank accession no.: HM475108) constructed in a previous study [36], was introduced into the E. coli Rosetta (DE3) strain by electroporation. A single colony was cultivated in LB medium supplemented with 30 µg/ml kanamycin (pET28a-MtCel6A) or 100 µg/ml ampicillin (pET22b-CelEdx16) and incubated overnight at 30°C (pET28a-MtCel6A) or 37°C (pET22b-CelEdx16). The starter cultures were then transferred to fresh LB media with the same antibiotic concentrations for further growth.
To induce the T7 promoter, IPTG was added when the OD600nm reached 0.5, to a final concentration of 0.4 mM. Thereafter, the cultures were maintained at 16°C overnight. Cells were then collected via centrifugation at 2691 × g for 20 min. For purification, the obtained cell pellets were resuspended in a buffer containing 150 mM KCl, 25 mM KH2PO4, and 2.5 mM imidazole at pH 8, and then lysed via sonication on ice for 10 min at 26% amplitude pulsing (5 s on and 5 s off pulses). The lysates were then centrifuged to separate the supernatants, which were then processed using affinity chromatography on a Ni-column (Bio-Rad) in an FPLC system (NGC Quest 10, Bio-Rad). The cellulases were eluted using a PBS (pH 7.4), concentrated using a Vivaspin® 6 centrifugal concentrator with a 10-kDa cutoff (Sartorius AG), and quantified using a Bradford assay.
2.6 Conventional cellulase activity assay
The activity of the cellulases with microcrystalline cellulose was measured using a well-known method involving dynamic light scattering to detect reducing sugars [37]. The reaction mixture included 50 μl of 2% microcrystalline cellulose in PBS (pH 7.4) and 50 μl of cellulase (0.02 mg/ml) and was incubated at 40°C for 12 h. To stop the reaction, 100 μl of 3,5-dinitrosalicylic acid (DNS) reagent was added, the mixture was boiled for 5 min, and then cooled on ice for 5 min. Following centrifugation at 10 000 rpm for 10 min, the absorbance was measured at 540 nm using a Tecan Spark plate reader (Tecan, Austria GmbH, Austria).
Another method for detecting cellulase activity involved measuring the presence of p-nitrophenol (pNP) produced by cellulase [38]. The absorbance was measured at 405 nm using a Victor X multilabel plate reader (PerkinElmer). The assays were conducted in a 100 μl reaction volume of PBS (pH 7.4) containing 2 mM p-nitrophenyl β-D-cellobioside (pNPG2) and cellulase (0.01 mg/ml), incubated for 12 h at 40°C. The reaction was stopped by adding 100 μl of 500 mM glycine buffer (pH 10.4).
The α-amylase and β-amylase were employed under identical experimental conditions as those used for cellulases.
2.7 Cellulase activity assay using a CB-biosensor with CFE systems
CFE was evaluated using the PURExpress system and comprised the following components: 5 μl PURExpress solution A, 3.75 μl solution B, 0.5 μl plasmid DNA (2 ng/μl), 0.1 μl RNase inhibitor (0.32 U/μl), 0.1 μl purified CelR (0.504 ng/µL), and 3.05 μl reactant in a 12.5 µl reaction volume. The reactant for the hydrolysis was prepared by incubating 2 mM pNPG2 or 1% microcrystalline cellulose with cellulase in PBS (pH 7.4) for 12 h at 40°C. The 1% microcrystalline cellulose was vortexed to minimize variations, and a well-mixed sample was manually transferred into 96-well plates. The α-amylase and β-amylase were used under the same experimental conditions as the cellulases. The reactions were prepared both manually and robotically.
Utilizing a robotic system (Echo 525), all reaction components, excluding the hydrolysis reactant, were dispensed from a 384-well PP 2.0 microplate (Labcyte, PPL-0200) into the destination plate. This transfer was performed using three distinct calibrations: 384PP_Plus_AQ_BP, 384PP_Plus_AQ_SP, and 384PP_Plus_AQ_GP (Table S5). Then, the destination plate was sealed with a Microseal B PCR Plate Sealing Film and incubated for 2 h at 37°C in an Eppendorf ThermoMixer® C (Eppendorf, Germany).
3. Results and discussion
3.1 Development and optimization of cell-free CB-biosensor
We sought to apply a cellobiose-detectable TF-based biosensor (CB-biosensor), originally developed for a microbial cell platform [39], to a CFE system using the PURExpress® in vitro protein synthesis kit. The sensory plasmid, pIVEX-PT7-CelO-sfGFP, was introduced into the CFE system along with the purified repressor TF (CelR), which was expressed in E. coli BL21 (DE3). In this system, CelR binds to the CelO operator site on the sensory plasmid, suppressing downstream expression of the reporter gene (sfGFP). Upon binding to cellobiose, CelR dissociates from the CelO operator site, allowing for sfGFP expression. Initial optimization of the cell-free reaction conditions was performed using the sensory plasmid pIVEX-PT7-CelO-sfGFP without considering regulation by CelR.
To optimize the CB-biosensor, cell-free reactions were conducted with DNA amount ranging from 10 ng to 200 ng in a total reaction volume of 12.5 μl. We established a reaction time of 2 h, as fluorescence significantly increased and reached saturation at this point, while background signal (without plasmid DNA) continued to slightly increase over the entire 6 h reaction period (Fig. 1a). Furthermore, no significant differences in fluorescence signals were observed at various tested RNase inhibitor concentrations (Fig. 1b). However, to ensure batch-to-batch reproducibility and to account for potential contamination during plasmid preparation, we decided to include 4 U of RNase inhibitor. Additionally, DNA input optimization revealed that 25 ng produced the highest fluorescence signal, which was 500 times greater than the background signal (Fig. 1c). This large difference between the background and fluorescence signals verifies the effective CelR regulation at the CelO operator site, demonstrating its suitability for further development of a cell-free TF-based biosensor with an anticipated broad dynamic range.
Optimization of the CFE system using the sensory plasmid only for (a) reaction time, (b) RNase inhibitor, and (c) DNA input. All measurements are based on triplicates, with error bars representing the standard deviation (SD).
3.2 Utilizing the Echo 525 liquid handler with the cell-free CB biosensor
Using the optimized conditions, we evaluated the cell-free CB biosensor with the Echo 525, which offers precise, tipless transfer of small-volume samples, even with high-viscosity solutions such as 50% glycerol. We optimized reagent transfer conditions using a drop test to assess droplet morphology. Notably, standard polypropylene microplates (384PP) were inadequate for transferring RNase inhibitors in 50% glycerol. However, we achieved the desired droplet morphology using a specialized plate for high-viscosity fluids (384PP_Plus) and further calibration (384PP_Plus_GP) (Table S4). Additionally, transfer conditions for all necessary reagents were validated (Table S5).
We compared manual (using a pipette) and robotic (using Echo 525) operations of the cell-free CB-biosensors. Without CelR, fluorescence signals were approximately ‘176120 a.u. (CV = 1.02%)’ and ‘182356 a.u. (CV = 4.56%)’ for manual and Echo 525 operations, respectively (Fig. 2a and d). With CelR, fluorescence signals decreased with increasing purified CelR input, likely due to sfGFP expression inhibition via binding to the CelO site (Fig. 2b and e). sfGFP expression was strongly repressed by CelR amounts exceeding 50 ng, with fluorescence signals similar to the background signal (without DNA). Moreover, we observed dose-dependent fluorescence signals in response to cellobiose, which decreased CelR binding to CelO site, thus allowing for sfGFP expression (Fig. 2c and f). Therefore, fluorescence signals in automated reactions were comparable to those in manual operations, verifying the feasibility and optimized conditions for automated cell-free CB-biosensor operations using the Echo 525 while significantly improving throughput. This demonstrates that our Echo 525-based sample transfer is properly calibrated and optimized, ensuring negligible deviation from manual operations and validating its reliability for high-throughput automated workflows.
Comparison of CB-biosensor operating conditions using manual pipetting and the Echo 525. The fluorescence signal depends on the addition of the sensory plasmid and an RNase inhibitor, examined (a) manually and (d) using the Echo 525. Fluorescence signals obtained with different CelR concentrations were examined (b) manually and (e) using the Echo 525. Fluorescence signals in response to different cellobiose concentrations were examined (c) manually and (f) using the Echo 525. All measurements are based on triplicates, with error bars representing the standard deviation (SD).
Importantly, the total reaction volume can be reduced to as low as 1.56 μl, with a linear correlation observed between fluorescence signals and reaction volume (Fig. S1). This reduction in reaction volume can significantly decrease processing time; for example, the time required for 100 reactions could be reduced by at least 50-fold based on rough estimation (Table S6). Generally, tip-based liquid handlers such as Opentrons OT-2, Gilson PIPETMAX, and Perkin Elmer Zephyr are useful when preparing and dispensing reaction components at microlitre scales, especially when distributing identical reagents across well plates. However, as this reaction involves multiple components requiring sub-microlitre scale dispensing, few liquid handlers, including the Echo system, can achieve this level of precision. Additionally, Echo is cost-effective as it operates without disposable tips and significantly reduces the risk of cross-contamination owing to its acoustic, noncontact liquid handling mechanism. These advantages highlight the Echo 525-assisted cell-free CB-biosensor as a promising tool for time- and cost-efficient large-scale screening in future applications.
3.3 Prototyping of exo-type cellulase activity using the cell-free CB-biosensor
We evaluated the applicability of the cell-free CB-biosensor to measure cellulase activity, specifically exo-type cellulase, CBH, using pNPG2 and microcrystalline cellulose. The enzyme activity measurements with our cell-free CB-biosensor were compared with two conventional methods: the reducing sugar assay using DNS and the chromogenic assay using pNPG2. CBH enzymes are classified into CBHI and CBHII based on their mode of action at the reducing or nonreducing ends of cellulose chains, respectively [34, 40, 41]. While pNPG2 is commonly used for CBHI screening due to its sensitivity and simplicity, it cannot detect CBHII activity, presenting a challenge for CBHII characterization. The CB-biosensor, utilizing microcrystalline cellulose, offers a promising solution for measuring both CBHI and CBHII activities.
To evaluate its versatility, we tested MtCel6A (CBHII from Myceliophthora thermophila) [42], HjCel7A (CBHI from H. jecorina) [43, 44], and CelEdx16 (bifunctional endo-/exo-cellulase from ruminal bacterium AN-C16-KBRB) [36], along with α- and β-amylases as negative controls. The α-amylase cleaves α-(1→4) glucosidic bonds in starch, while β-amylase hydrolyzes α-(1→4) glycosidic bonds from the nonreducing ends of starch [45]. The characterization involved a two-step process: an enzymatic reaction with appropriate substrates followed by activity measurements using the cell-free CB-biosensor. Purified recombinant enzymes were used for MtCel6A and CelEdx16 (Figure S2), while HjCel7A and two amylases were tested using diluted commercial enzymes.
Figure 3a shows that the pNP assay detected significant signals for CelEdx16 and HjCel7A (CBHI), with fold changes of 35.857 and 4.922, respectively. In contrast, MtCel6A (CBHII) and the negative controls (α- and β-amylases) produced negligible signals. The DNS assay, on the other hand, detected a measurable but weaker signal for CelEdx16 (fold change of 1.442) with a net signal-to-noise ratio (SNR) of 3.138. HjCel7A and MtCel6A exhibited only weak signals (fold changes of 1.261 and 1.141, respectively), both with SNRs < 3, indicating insignificance. The net SNR was calculated by dividing the net signal (obtained by subtracting the control signal from the target signal) by the standard deviation of the control, with an SNR of 3 typically considered the limit of detection. Therefore, the difference between HjCel7A and MtCel6A in the DNS assay is not statistically meaningful. As expected, the control enzymes showed no detectable activity in the DNS assay (Fig. 3b). These results verify that the pNP assay is more sensitive for detecting CelEdx16 and HjCel7A (CBHI) activities but fails to detect MtCel6A (CBHII) activity. Conversely, while the microcrystalline cellulose-based DNS assay can be used for CBHII prototyping, its sensitivity remains insufficient.
Prototyping of cellobiohydrolase activity with the conventional and cell-free CB-biosensor methods. Five enzymes, CelEdx16, HjCel7A, MtCel6A, α-amylase, and β-amylase, were used. (a) Conventional methods of pNP assays. (b) Conventional methods of DNS assays. (c) Manual assays using the cell-free CB-biosensor manually. (d) Robotic assays using the cell-free CB-biosensor with Echo 525. All measurements are based on triplicates, with error bars representing the standard deviation (SD). p-nitrophenyl β-D-cellobioside (pNPG2) and microcrystalline cellulose were used as substrates. All enzyme reactions were conducted for 12 h. For (c) and (d), an additional 2 h cell-free reaction was performed following the enzyme reaction.
Figure 3c and d illustrate cellulase activities measured using the CB-biosensor under both manual and automated conditions with pNPG2 and microcrystalline cellulose as substrates. The results demonstrated minimal differences between manual and automated operations. The pNPG2-based assay produced fluorescence signals comparable to the conventional pNP assay, detecting significant activity only for CelEdx16 and HjCel7A (CBHI). In the microcrystalline cellulose-based CB-biosensor assay, strong fluorescence signals were observed not only for CelEdx16 and HjCel7A (CBHI) but also for MtCel6A (CBHII). While MtCel6A showed high activity on microcrystalline cellulose but not on pNPG2, this highlights the CB-biosensor’s ability to distinguish different enzymatic modes of action, specifically measuring MtCel6A’s nonreducing-end activity.
Compared to traditional methods, the CB-biosensor, which utilizes fluorescence-based detection, offers a high sensitivity even with a crystalline cellulose substrate. The control enzymes (α- and β-amylases) did not produce significant fluorescence, reaffirming the sensor’s specificity for cellulases. More importantly, no significant differences in fluorescence between manual and automated operations demonstrate the CB-biosensor’s suitability for HTS applications.
4. Conclusion
In this study, we validated the operation of a CB-biosensor in a CFE system for detecting cellobiose produced by CBHs via fluorescence signals. CBHs, which degrade solid substrates such as paper sludge, have presented challenges for HTS. By implementing this system in a CFE setup, we demonstrated its potential for HTS while also achieving higher sensitivity compared to conventional assays. Additionally, the CB-biosensor successfully distinguished the activities of CBHI and CBHII subtypes, leveraging substrate-specific discrimination to elucidate their distinct modes of action. To enable large-scale experiments and facilitate efficient data acquisition, automated equipment is essential. The Echo 525 liquid handler supports precise, small-scale experiments within a condensed timeframe, reducing reagent consumption and increasing throughput. However, successful integration of manual workflows into automated setups requires continued efforts towards standardizing protocols. The optimized liquid handling conditions for various cell-free resources presented in this study provide a valuable framework for researchers aiming to screen enzyme candidates from large-scale libraries using CFE systems. While this study primarily demonstrated the CB-biosensor with the commercial PURE system, we also verified its functionality in an E. coli lysate-based CFE assay. Using E. coli lysates can lower costs and enhance system versatility, making it more accessible for broader applications. Overall, this work highlights the promise of integrating cell-free biosensors with automated technologies to overcome current limitations in enzyme screening and accelerate biocatalyst engineering for sustainable biomanufacturing.
Acknowledgments
The authors are grateful to Dong-Myung Kim (Chungnam National University) for providing the plasmid, pIVEX 2.3d EGFP.
Author contributions
T.K. and E.J.J.: writing—original draft and editing, visualization, validation, formal analysis, and investigation.; K.K.K.: conceptualization, investigation, and methodology; M.K. and H.N.K.: investigation and formal analysis; S.K.K., E.R.,: investigation and resource acquisition; J.S., H.K and D.H.L.: writing—review and resource acquisition; B.H.S: writing—review and resource acquisition, funding acquisition.; S.J.K: supervision, writing—review and editing; H.L. and S.G.L.: conceptualization, supervision, funding acquisition, and writing—review and editing.
Supplementary data
Supplementary data is available at SYNBIO online.
Conflict of interest:
The authors declare no competing financial interest.
Funding
This work was supported by the Bio & Medical Technology Development Program (grant number 2021M3A9I4022731), the Next Generation Biorefinery Research Program (NRF-2022M3J5A1056169), and the Basic Science Research Program (RS-2024-00336627) of the National Research Foundation funded by the Ministry of Science and ICT of the Republic of Korea, and the Korea Research Institute of Bioscience and Biotechnology (KRIBB) Research Initiative Program (grant number KGM1302511).
Data availability
Data from this article will be shared on reasonable request to the corresponding author.
References
Author notes
contributed equally to this work.
