Design of experiments‐based high‐throughput strategy for development and optimization of efficient cell disruption protocols

Abstract

Efficient and reproducible cell lysis is a crucial step during downstream processing of intracellular products. The composition of an optimal lysis buffer should be chosen depending on the organism, its growth status, the applied detection methods, and even the target molecule. Especially for high‐throughput applications, where sample volumes are limited, the adaptation of a lysis buffer to the specific campaign is an urgent need. Here, we present a general design of experiments‐based strategy suitable for eight constituents and demonstrate the strength of this approach by the development of an efficient lysis buffer for Gram‐negative bacteria, which is applicable in a high‐throughput format in a short time. The concentrations of four lysis‐inducing chemical agents EDTA, lysozyme, Triton X‐100, and polymyxin B were optimized for maximal soluble protein concentration and ß‐galactosidase activity in a 96‐well format on a Microlab Star liquid handling platform under design of experiments methodology. The resulting lysis buffer showed the same performance as a commercially available lysis buffer. The developed protocol resulted in an optimized buffer within only three runs. The established procedure can be easily applied to adapt the lysis buffer to other strains and target molecules.

Abbreviations

ß‐Gal

ß‐galactosidase

BCA

bicinchoninic acid

DoE

design of experiments

HT

high‐throughput

LHS

liquid handling system

1. Introduction

Cell lysis is an important step in industrial production of biomolecules, as well as in miniaturized high‐throughput (HT) screening setups. The composition of an optimal lysis buffer is dependent on the target organism, but also depends on the cultivation conditions, such as medium and cultivation temperature, as well as on the applied cell density. Furthermore, the buffer must be composed in view of the target molecule, of the purification process, and may differ from small molecules to proteins, DNA, and RNAs. For a target protein, all further factors need to be considered, such as the state of oligomerization, redox state, compartmental localization, and cofactors.

A combination of chemical, enzymatic, and physical cell disruption methods is the standard at industrial and laboratory scale and this is normally optimized for a specific situation 1, 2, 3. However, effective methods like such as pressure homogenization (known as French Press) cannot be easily applied in HT screening systems. HT‐compatible methods are ultrasound, bead milling, and the combination of chemical and enzymatic methods.

A variety of newly developed microscale cell disruption methodologies exist 4, for example, the application of microfluidic compact discs, nanoscale barbs, or electric pulses 5, which have the advantage of working without any additional, potentially interfering chemical compound. But the application of such mechanical or physical lysis methods adapted to the microscale applied specialized prototyped apparatuses, which are not easily adaptable to existing liquid handling systems (LHS) or are limited to specific applications. Another option is the application of head‐inducible autolytic vectors 6. In contrast, chemoenzymatic cell disruption methods can be easily applied in microwell plate based experiments without any significant increase in costs and any need to use additional equipment, and they can be adapted to new screening tasks in short time.

A tremendous variety of chemicals and enzymes have been applied for the disruption of microorganisms so far, that is, chaotropic agents (e.g., ethanol, guanidine‐HCl, guanidine‐SCN) 3, anionic (e.g., SDS), nonionic (e.g., Triton, Tween, or Brij family) and zwitterionic (CHAPS) detergents 2, chelating agents (e.g., EDTA) 5, organic solvents (e.g. toluene, butanol) 7, cationic polypeptide antibiotics (e.g., polymyxins), 8 and foreign peptidoglycan digesting enzymes (e.g., lysozyme) 9 are the main classes of lysis buffer components.

Cell lysis buffers are commercially available from several manufacturers, for example, BugBuster® (Merck Millipore) or SoluLyse® (Genlantis), which were applied in a comparative study using an expression library by Listwan et al. 10. However, since the formulation is proprietary, not all lysis‐inducing components are known. In the case of protein assays, detergents and EDTA mainly can interfere with chromophore formation 11. For the purification of polyhistidine‐tagged recombinant proteins, lysis buffers should not contain EDTA, which disturbs the immobilized metal ion affinity chromatography 12. Therefore, the ideal solution would be to optimize the lysis buffer composition depending on the specific demands of the expression host's cell wall, the target protein, and the subsequent downstream steps. Hence, a fast and cost‐efficient framework for the development or adaptation of a specific cell disruption mixture will allow a higher degree of specification for each application. In order to maximize the efficiency of the development process, design of experiments (DoE) 1, 13 is a standard way to plan the experimental setup on a liquid handling station. The applications of optimal experimental planning are beneficial for a wide variety of operations in upstream and downstream bioprocess development 2, 3. In formulation development, the design space can get very large, especially in HT screening facilities where a large number of costly and time‐consuming experiments make intuitive design very complex and inefficient. Consequently, liquid handling robotics and laboratory automation have to be coupled to experimental design programs, which select the best combination of experiments for a given task including correct evaluation of datasets.

The aim of this study was to use state of the art methods to design and carry out experiments on a LHS in order to create a framework for fast, cost‐effective, and efficient optimization of cell lysis buffers. The applicability of the method is demonstrated with the development of a lysis buffer for purification of soluble expressed recombinant proteins released from Escherichia coli cells.

The DoE‐based strategy can then be easily applied for reoptimization of the buffer composition suitable for other strains and target proteins within a few days of work.

2. Materials and methods

2.1. Strains

Bacterial cultivations were carried out using E. coli BL21. The strain was transformed with the plasmid pDgPNP for heterologous expression of purine nucleoside phosphorylase from Deinococcus geothermalis 14. Media were supplemented with 100 mg/L ampicillin to maintain plasmid stability. All strains were cryopreserved at –80°C in media containing 20% glycerol.

2.2. Cultivation conditions

All cell lysis experiments were performed using E. coli cells obtained from glucose‐limited fed‐batch cultivations. For this purpose, the EnBase® technology (BioSilta Ltd., Cambridge, UK) was applied. By enzymatic degradation of a polymer, quasi‐constant feeding of cells can be carried out using EnBase media. In this study, a medium for bacterial cultures in ready‐made tablets was used (EnPresso B). The kit contains medium tablets, glucose‐releasing “Reagent A,” and complex “Booster” tablets.

Precultures were carried out at 37°C in 5 mL of Luria–Bertani broth (10 g/L peptone, 5 g/L yeast extract, 10 g/L sodium chloride, pH 7.0) using a 125‐mL UltraYield Flask™ covered with AirOtop Enhance Seal™ (Thomson Instrument Company, Oceanside, USA). The preculture was shaken at 250 rpm with 25 mm amplitude in a Kühner LT‐X incubator (Adolf Kühner AG, Basel, Switzerland) for 6 h. For main cultures, 50 mL of EnPresso B (BioSilta Ltd.) medium was prepared in a 250‐mL UltraYield Flask^TM, according to the manufacturer's instructions. The optical density at 600 nm (OD₆₀₀) of the preculture was measured in an Ultrospec 2100 pro‐spectrophotometer (GE Healthcare Europe GmbH, Freiburg, Germany) in order to inoculate the main culture with an initial OD₆₀₀ of 0.2 AU. Immediately after inoculation, 1.5 U/L of Reagent A (BioSilta Ltd.) was added for controlled glucose release. The culture was incubated at 250 rpm and 30°C overnight. Intracellular recombinant protein production was induced using 100 μM IPTG. At the time of induction, 3 U/L of Reagent A and a tablet of a complex “Booster” mix (BioSilta Ltd.) were added. The culture was continuously incubated for another 24 h, until the shake flask was cooled down on ice for cell harvest. At the end point of the cultivation, OD₆₀₀ was measured and the harvest volume (V _H) was calculated in order to normalize the harvested cells to 1 mL of culture taken at an OD₆₀₀ of 5 AU:

$$V_H=\frac{5\mathrm{mL}}{\mathrm{O}{\mathrm{D}}_{600}}$$ (1)

The culture broth was distributed into 96‐microtube racks (HJ Bioanalytik GmbH, Erkelenz, Germany) and centrifuged at 3000 g and 4°C for 15 min. The supernatant was removed thoroughly and the cell pellets were stored at –20°C.

2.3. Experimental design and automated mixture preparation

All experiments were planned and evaluated using MKS Umetrics MODDE® 10 (MKS Umetrics AB, Umeå, Sweden). Experimental plans were imported into a laboratory automation database (iLab‐Bio, infoteam software AG, Bubenreuth, Germany) via the MODDE‐Q interface (Umetrics AB). With the help of a graphical user interface, worklists for the liquid handling robot were generated. The software calculated the individual volumes for each component, based on the concentration of stock solutions and the working volume. A method for mixture design on the LHS (Hamilton Microlab Star, Hamilton Bonaduz AG, Bonaduz, Switzerland) was programmed using the Hamilton VenusOne software. In brief, it enables the operator to import worklists containing volumes from the iLab‐database and distribute the respective compounds into a 96‐deep well plate (Ritter GmbH, Schwabmünchen, Germany) for convenient fully automated preparation of mixtures with up to eight components.

Six compounds were selected for evaluation: EDTA, guanidine hydrochloride (guanidine‐HCl), lysozyme, polymyxin B, Triton X‐100, and Tween 20. For mixture design, all solutions were prepared as 10‐fold concentrates to allow easy mixing of stock solutions with the liquid handler to a final volume of 500 μL. A basic binding buffer (80 mM sodium phosphate buffer, 40 mM imidazole, 0.5 M sodium chloride, pH 7.6) that allows further purification steps was used. Dilutions for cell lysis screening were prepared in binding buffer containing an EDTA‐free blend of protease inhibitors (cOmplete EDTA‐free, Roche Diagnostics GmbH, Mannheim, Germany) on the liquid handling robot. BugBuster® Protein Extraction Reagent (Merck Millipore, Billerica MA, USA) supplemented with 1800 U/mL of lysozyme served as a reference system. All lysis buffers were supplemented with 25 U/mL of Benzonase® Nuclease (Merck Millipore) for viscosity reduction of cell lysates.

2.4. Cell disruption

The frozen cell pellets were incubated at room temperature for approximately 5 min. Then, 300 μL of lysis buffer was added simultaneously to all 96 tubes of the rack, using the 96‐pipettor head of the LHS. The pellets were resuspended immediately afterward, by 30 aspiration and dispense cycles of 300 μL. Incubation was carried out for 20 min at room temperature, followed by centrifugation at 4000 g for 20 min at 4°C. Then, the supernatants were transferred into fresh 96 microtubes for further analysis.

2.5. Bicinchoninic Acid assay

Protein quantification of the soluble fraction was carried out with a bicinchoninic acid (BCA) 15 protein assay kit (BioVision Inc., Milpitas, USA) according to the manufacturer's instructions. In brief, 200 μL of BCA working reagent was distributed into a 96‐well flat bottom plate (Greiner Bio‐One, Frickenhausen, Germany). Then, 25 μL of sample was added and the plate was incubated for 30 min at 37°C. After cooling down for 5 min, absorption was measured at 562 nm in a Biotek Synergy MX plate reader (BioTek Instruments Inc., Winooski, USA), connected to the liquid handling station using the Gen5 software (BioTek Instruments Inc.). Standard curves were prepared using BSA (Sigma‐Aldrich, Munich, Germany) in concentrations ranging from 0.025 to 2 mg/mL.

2.6. β‐Galactosidase assay

β‐Galactosidase (β‐Gal) activity was measured in flat‐bottom 96‐well plates using o‐nitrophenyl‐β‐d‐galactopyranoside as the substrate. Ten microliters of sample was added to 160 μL of Z‐buffer (60 mM NaH₂PO₄, 40 mM Na₂HPO₄, 10 mM KCl, 1 mM MgSO₄, 50 mM β‐mercaptoethanol, pH 7.0) and incubated at 30°C for 3 min. Then, 32 μL of o‐nitrophenyl‐β‐d‐galactopyraonside solution (4 mg/mL in 50 mM Tris‐HCl, pH 8.0) was added simultaneously to all wells using the LHS. The increase in absorbance at 420 nm was measured every 30 s for 10 min. One enzyme unit (U) is defined as the amount of enzyme releasing 1 μmol o‐nitrophenol per minute under the defined reaction conditions.

3. Results

In order to simplify the planning, execution, and data evaluation for DoE‐based experiments, a workflow of proceeding was established in this study. Independent software solutions for (i) DoE (MODDE 10), (ii) data handling (iLab‐Bio), (iii) liquid handling (VenusOne), and (iv) microplate reading (Gen5) were connected using customized interfaces (Fig. 1). This combination of commercially available software packages is user‐friendly and allows planning and data handling of large experimental setups.

Figure 1.

Integrated system of data transmission and processing. The experiments are planned in the DoE software (MODDE® 10), and translated into worklists for the LHS with the modular database (iLab‐Bio). The LHS manages fully automated multichannel pipetting of chemicals and performs read‐outs in the multiwell plate reader (Gen5) after the cell lysis procedure. The results are manually integrated and saved as spreadsheets, which are then automatically imported back into the iLab‐Bio database via the LHS control software. Finally, the experimenter can use the DoE software to analyze the data statistically and determine the optimal region.

During validation of the integrated automated workflow, preliminary experiments showed that Triton X‐100 outperformed Tween 20, as well as tests using guanidine‐HCl exhibited a significantly negative influence on ß‐Gal activity (data not shown). Therefore, for the main experiments, four components were selected for the screening and the optimization steps (Table 1).

Table 1.

Factors and concentration ranges of ingredients added to cell lysis buffers applied for screening and optimization

Component	First run (screening)	Second run (optimization)
Benzonase (×102 U/mL)	2.5	2.5
EDTA (mM)	0.5–10	–
Lysozyme (×103 U/mL)	0.3–9	4.5–13.5
Polymyxin B (μM)	0.1–50	20–60
Triton X‐100 (%)	0.1–2	0.94–2.82

The experiments were planned as D‐optimal designs, which have the advantage of flexible boundaries and restrictions. For the screening phase, a design with duplicates of 32 experiments and four center points was used. In addition, triplicates of mixtures with three out of four parameters at the highest concentration, triplicates of basic buffer as a negative control and BugBuster as a reference buffer were included. In total, the first round consisted of 91 experiments (72 from MODDE + 19 controls). The experiments were randomized over the plate. The concentrations of lysis‐inducing agents were automatically converted into volumes of stock solutions to be pipetted by the LHS. After incubation and centrifugation, the supernatant was analyzed for protein concentration (BCA assay) and β‐Gal activity. Since each buffer composition gave an individual background signal in the protein assay, blank measurements of all buffers were performed. As a reference cell lysis system, the BugBuster® Protein Extraction Reagent was used.

The data were fitted using the partial least squares (PLS) regression, which gave a good fit for both responses (Table 2). Apart from EDTA, all other factors showed positive and synergistic effects. The response surface was investigated for extreme points, which resulted in a recommendation for high concentrations of lysozyme, polymyxin B, and Triton X‐100. The predicted activity of released β‐Gal for the optimium is 58% of the activity obtained with the commercial cell lysis kit considered as reference.

Table 2.

Comparison of the statistics parameters summarizing the fit of the models used for screening (first run) and optimization (second run)

Statistical parameters	Responses
	Screening	β‐Gal activity	Optimization	β‐Gal activity
	Soluble protein concentration		Soluble protein concentration
R2 a	0.76	0.81	0.55	0.74
Q2 b	0.62	0.71	0.50	0.69
Reproducibilityc	0.77	0.84	0.75	0.92

^aCoefficient of determination.

^bFuture prediction precision.

^cVariation of the replicates compared to overall variability.

In order to get a better picture of the optimal region and improve the lysis efficiency, a second D‐optimal experiment was performed in the optimal region. The design space was shifted to higher concentrations of the three buffer's ingredients with positive influence, that is, lysozyme, polymyxin B, and Triton‐X100. The concentration ranges of lysis‐inducing agents were defined around the optimal region of the first experiment with 50% surpluses and insufficiencies. Again, duplicates of 32 runs with four center points were selected. The results were combined with the data from the previous experiment, resulting in a model for the complete design space, which is depicted in Fig. 2. The final model showed high reproducibility but with a reduced goodness of fit and prediction precision. The maximum predicted ß‐Gal activity of the model is 0.35 U/mL, which is 92% of the activity obtained using BugBuster.

Figure 2.

Response surface and coefficient plots for soluble protein concentration (A–C) and for β‐Gal activity (D–F). The influence of Triton X‐100 and lysozyme (for 30 μM polymyxin B and 0 mM EDTA) on soluble protein concentration (A) and β‐Gal activity (D), as well as the influence of polymyxin B and EDTA (for 2.82% Triton X‐100 and 13.5 × 10³ U/mL lysozyme) on soluble protein concentration (D) and β‐Gal activity (E) are presented. Coefficient plots for soluble protein concentration (C) and β‐Gal activity (F), indicating positive or negative influence of the factors on the responses. The error bars show the confidence interval. 163 experiments were performed.

Based on the model obtained, it can be concluded that high lysozyme (>9000 U/mL) and high Triton X‐100 (>2%) levels at moderate polymyxin B concentrations (35 μM) are necessary to accomplish efficient E. coli cell disruption. Interestingly, at a detergent concentration of around 2%, the concentration of soluble protein reaches a saturation point. At detergent levels greater than 2%, only the enzyme activity increased, but not the soluble protein concentration. At high lysozyme concentrations, less soluble protein is detected with increasing detergent concentration compared to low lysozyme levels. Polymyxin B appears to have a negative effect on protein content and enzyme activity at the given concentration range of used cationic polypeptide antibiotic. Finally, β‐Gal activity is negatively influenced by EDTA, especially at moderate polymyxin B concentrations.

In summary, it can be concluded that the integration of independent software units for DoE, data handling, liquid handling, and plate reading enhances the efficiency of method development in HT‐miniaturized systems. The presented workflow has been applied for screening cell lysis agents and choosing those with a positive effect on chemoenzymatic disruption of E. coli cells. Performing the optimization step and combining both datasets gave more detailed information on the influence of the additives. The detergent (i.e., Triton X‐100) and muramidase (lysozyme B) were found to be crucial for efficient cell lysis, while the cationic polypeptide antibiotic (i.e., polymyxin) assisted the breakdown of the cells.

4. Discussion

The appearance of commercially available laboratory automation systems for liquid and plate handling gave rise to the introduction of HT technologies in the field of bioprocess engineering and applied biotechnology. However, increased experimental throughput requires more sophisticated methods for experimental planning and data evaluation. Therefore, a fully integrated platform of interacting units for DoE, data management, experimentation, and analysis will be a milestone in increasing productivity of laboratories. Up to date, some attempts in the matter of integrated automation of bioprocess development have been achieved and recently reviewed 1, 13, 16, 17.

In this study, we demonstrate the effective interaction of different modules that minimizes the human interaction and especially enables complex experimentation schemes in an automated way. However, human involvement was still needed in most steps, for example, in designing the experimental plan, in programming and initiating the liquid handling unit, in data treatment obtained from the plate reader, as well as in modification the plan of experiments for the next round. Such a strategy is in accordance with previously published automated HT platforms for bioprocess development 18, 19. Further extension of such an automated flow can be envisioned by a full closed‐loop system as recently published by Wu and Zhou 18. Such an “intelligent” system could perform the modification of experimental plans for the next round automatically (e.g., optimization step) based on the results obtained from a previous round (e.g., screening step).

In this report, an automated workflow functionally combined the above‐mentioned units into one HT screening and optimization platform, which has been used for the development of a chemoenzymatical cell lysis buffer for E. coli cells. The developed lysis buffer emerged from six components in only three experimental runs, and finally resulted in a similar disruption efficiency compared to a commercial system.

EDTA, lysozyme, polymyxin B, and Triton X‐100 have been taken into consideration for the optimization step, which were ordinarily used in previously reported studies focused on chemoenzymatic disruption of bacterial cells 2, 3, 7, 8. The optimization of the considered lysis buffer yielded more detailed answers on the influence of lysing factors mentioned above on E. coli disruption efficiency and on stability of the released intracellular enzyme. As an outcome, the lysis buffer containing high amount of nonionic detergent (i.e., Triton X‐100) and lytic enzyme (i.e., lysozyme), with low‐mid amount of cationic polypeptide antibiotic (i.e., polymyxin B) at low concentrations of chelating agents (i.e., EDTA) has been found to release the highest level of the intracellular target enzyme.

We considered that the developed lysis buffer, as an EDTA‐free variant, extends the applicability of the lysing system also in the case of methodologies that require cell lysates to be free of chelating agent for further enzymatic assays, that is, if binding of divalent metal ions (e.g., Ca²⁺, Co²⁺, Mg²⁺, or Mn²⁺) influences on results of the enzymatic assays as essential cofactors, or by displacing the intrinsic factors 20, 21. Alternatively, sonication could be used as applied as the reference method in a study by Listwan et al. 10, which also points out the importance of testing a larger variety of proteins for validation of a lysis method.

The cells pellets, which have been used as biomass for lysis, were taken from glucose‐limited fed‐batch cultures using EnBase® technology 22, 23. Such culture conditions prevent overfeeding of the cells, which is relevant for scale‐up to production scale 24, 25, however fed‐batch grown cells are also more difficult to lyse. Many of previously published data on methods for lysis of bacterial cells 4 miss details on the cultivation conditions, or the cells used in experiments came from batch cultures with complex media, which makes it hard to compare the outcome of different studies. In the context of bioprocess development, the cell wall composition of E. coli grown under glucose limitation should be recognized as a reference point for further comparative studies on the topic of the efficient disruption of Gram‐negative bacteria.

In general, the overall DoE‐aided methodology, which was applied for the development of a lysis buffer designed for E. coli cells, shows a great potential for application flexibility. This gives a robust possibility for very fast, cost effective, no time, and no labor‐consuming reoptimization of the end user specified solution for chemoenzymatically induced lysis of cells. Such a strategy makes it possible to easily improve the efficiency of other downstream processes of the target intracellular molecules purification, and it is not limited to only proteins but also for wide range of other intracellular products, as small molecules, DNA, plasmids, RNAs, and others 26.

5. Concluding remarks

A functional integration of software units for DoE, data handling, liquid handling, and multiwell plate reader supplied by various producers has been developed. Based on this, a HT miniaturized‐format platform for cell lysis buffer screening and its further optimization has been developed.

The chemoenzymatic lysis buffer containing nonionic surfactant, muramidase, cationic polypeptide antibiotic, and low concentrated chelating agent has been recognized as supporting the most effective conditions for releasing soluble proteins from E. coli, as well as for retaining the active structure of the released intracellular enzyme. The developed lysis buffer showed the same performance than a commercially available product.

Summarizing, the idea of the DoE‐aided rapid optimization of a lysis buffer suitable for maximization of recombinant protein production, and to provide maximal activity of intracellular enzyme released from Gram‐negative bacteria has been fulfilled. Moreover, the presented methodology of the HT automated lysis of cells can be easily adapted to change lysis‐inducing constituents, as well as various strains of E. coli or other species of Gram‐negative bacteria.

Practical application

The efficiency of a cell lysis procedure is influenced by the organism and growth status, and affects the consecutive downstream processing. For high‐throughput applications, the adaptation to the specific campaign is an urgent need since the low sample volume requires efficient chemical or enzymatic lysis. Here, we present a computer‐aided design of experiments (DoE) procedure for the optimization of a multiagent buffer for the lysis of Gram‐negative bacteria with up to eight components. The optimization is exemplarily shown at the release of cytoplasmic β‐galactosidase from Escherichia coli cells. The power of the approach lies in the functional integration into an automated high‐throughput robot‐based screening platform of independent software packages for (i) DoE, (ii) data processing, (iii) liquid handling, and (iv) spectrophotometric read‐out. The presented protocol may be applied for any lysis buffer optimization for bacterial, plant, or animal cell cultures or for quantitative assay development.

The authors have declared no conflict of interest.