Proteins

L-Lactate dehydrogenase 1 (LDH), galactose oxidase (GalOx) and horseradish peroxidase (HRP) were produced as IBs in E. coli cultivations. A monomeric LDH originating from Lactobacillus plantarum is with a size of 34.4 kDa and no disulfide bridges was used in this work. The molecular weight of GalOx from Fusarium graminearum is approximately 68.5 kDa and the enzyme incorporates two disulfide bridges. A thioether-cross-link is formed upon addition of copper as a cofactor [36]. HRP is an oxidoreductase containing a heme cofactor. HRP C1A isoenzyme with a size of 34.5 kDa and four disulfide bridges was used in this work.

Production of IBs

Production of LDH IBs was performed as described in [37]. In brief, E. coli BL21(DE3) cells were cultivated in DeLisa minimal medium [38]. Main cultures were carried out in fed-batch mode in a 3.3-L Labfors bioreactor (Infors AG, Bottmingen, Switzerland) with controlled feeding at a specific glucose uptake rate (q\(_\)) of 0.2 g g\(^\) h\(^\). The culture was induced with 1 mM isopropyl-D-thiogalactopyranoside (IPTG). Induction was carried out for 6 h at 37°C and a q\(_\) of 0.25 g g\(^\) h\(^\). IBs were separated from the harvested biomass by high-pressure homogenization and washing steps as described elsewhere [37].

IBs of GalOx were produced in E. coli BL21 (DE3) cultures using a pET-29b(+) vector and a T7-expression system. Pre-cultures were grown in DeLisa pre-culture medium [38] supplemented with 50 µg mL\(^\) kanamycin and 8.8 g L\(^\) glucose. Baffled shake-flasks with a filling volume of 500 mL were inoculated with 0.5 mL cryo-preserved culture and cultivated (37°C, 16 h, 250 rpm) in an Infors HR Multitronshaker (Infors AG, Bottmingen, Switzerland). Cultivation in a 15-L Biostat® Cplus stainless steel reactor (Sartorius, Göttingen, Germany) was conducted in three phases: first, a batch phase was initiated by the inoculation of 4500 mL of DeLisa batch medium [38] supplemented with 50 µg mL\(^\) kanamycin and 15.5 g L\(^\) glucose with 500 mL of the pre-culture. Then, a fed-batch phase with a specific glucose uptake rate (q\(_\)) of 0.25 g g\(^\) h\(^\) and a feed concentration of 440 g L\(^\) was followed by an induced fed-batch phase with a q\(_\) of 0.20 g g\(^\) h\(^\) that was initiated by the addition of 1 mM IPTG. The temperature was controlled via the heating jacket of the vessel, being 35°C during the first two phases, and 30°C after induction. Adjustment of stirrer speed (500–900 rpm), oxygen mole fraction in the gas flow (20.95–24.11% (v/v)), and pressure (0.5–1.0 bar) were used to maintain the dissolved oxygen tension (DOT) above 40%. A constant pH 6.9 was controlled by the addition of 2 M phosphoric acid and 12.5% ammonium hydroxide (v/v). The biomass was harvested 6 h after induction by centrifugation of the cell suspension (16,000x g, 25 min, 4 °C). The resulting cell pellet was stored at -20 °C until further processing. The cell rupture and IB isolation were conducted as described elsewhere [37].

The production of HRP IBs was previously described in Humer et al [18]. Briefly, the hrp gene coding for HRP variant C1A was codon-optimized for E. coli and obtained from GenScript USA Inc. (Piscataway, NJ, USA). The plasmid pET21d+ was used for HRP IB production in the cytoplasm. A stop codon was introduced to produce HRP without any tags. HRP was produced in E. coli BL21(DE3) in a 10-L Biostat Cplus stainless steel bioreactor (Sartorius, Germany). The pre-culture was grown in 0.5 L DeLisa medium [38] at 37 °C, 230 rpm in a 2.5-L Ultra Yield™ Flask (UYF; Thomson Instrument company, Encinitas, CA, USA) over night. Subsequently, the pre-culture was added to 4.5 L DeLisa medium [38] in the bioreactor vessel and batch fermentation at 35 °C was run for 6 h. The pH was maintained at 7.2 and the DOT was kept above 20%. During the 16 h fed-batch phase q\(_\) was 0.333 g g\(^\) h\(^\), which was set to 0.25 g g\(^\) h\(^\) after induction with 0.5 mM IPTG. After an induction phase of 8 h, the biomass was harvested by centrifugation. IBs were separated from the harvested biomass by high-pressure homogenization and washing steps [18].

Processing of IBs — solubilization

The solubilization of IBs of LDH and GalOx was performed at a concentration of 100 g IB wet weight L\(^\) at 25°C under slight agitation for 2 h. The resulting suspension was then centrifuged (20,000x g, 4°C) before the supernatant was stored at 4°C until further processing. LDH IBs were solubilized in a buffer containing 150 mM NaH\(_\)PO\(_\), pH 6.0, 4 M GuHCl. GalOx IBs were first mixed with solubilization buffer — 150 mM NaH\(_\)PO\(_\), pH 7.0, 6 M GuHCl, before dithiothreitol (DTT) was added at a concentration of 25 mM to initiate the disruption of disulfide bonds. DTT stocks at a concentration of 1 M were prepared freshly before the adequate volume was added to the solubilization buffer. Concentrations of the solubilized protein were determined before refolding was initiated in a batch dilution approach. IBs of HRP were solubilized at a concentration of 100 g IB wet weight L\(^\) in a buffer containing 6 M urea, 7.11 mM DTT, 50 mM glycine at pH 10 for circa 0.5 h at 4–10°C under slight agitation [18]. The resulting suspension was then centrifuged (20,000x g, 4°C) for 20 min.

Processing of IBs — refolding

Refolding was conducted at 5°C for LDH and GalOx and 7.5°C for HRP with a constant stirring between 500 and 800 rpm. To initiate the refolding process, the solubilized protein was rapidly added into pre-cooled refolding buffer directly inside the fluorescence cuvette (“Online intrinsic fluorescence monitoring”). The samples were incubated at constant temperature and stirring speed for 2.5 h for LDH and GalOx or 22 h for HRP.

Refolding of LDH

The solubilized LDH was refolded in a buffer containing 150 mM NaH\(_\)PO\(_\) at pH 6.0 unless stated otherwise.

Different additives (L-arginine, acetone and glycerol) and the addition of excess GuHCl were tested for their effects on aggregate formation. The chemicals were added to the LDH refolding buffer and refolding was carried out in batch processing mode. The detailed experimental design can be found in Table 1.

Table 1 Experimental design for buffer screening of additives for LDH refolding

To investigate the suitability of different processing modes for LDH refolding, 8 experiments were conducted using the standard refolding buffer. The experiments were set up following a full-factorial design-of-experiment (DoE) approach, where the final dilution factor (10, 30, 50) and the number of pulses (1, 3, 5) were altered. Here, a pulse number of 1 refers to batch dilution and higher numbers to pulsed batch processing. Five different conditions were tested, with the center point being conducted as biological replicates (\(n = 4\)). The complete experimental setup is described in Table 2.

Table 2 Experimental design for pulsed LDH refoldingRefolding of GalOx

Refolding of solubilized GalOx was carried out in a buffer containing 100 mM NaH\(_\)PO\(_\), 5 mM cystamine, 1 M L-arginine at pH 7.4 unless stated otherwise. Refolding was carried out in batch mode at varying dilution factors. Cu\(^\) was added as a cofactor at a concentration of 1 mM. The time of addition was varied according to the experimental setup described in Table 3. The design was based on a full-factorial DoE approach with the center points being carried out as biological replicates (\(n=3\)).

Table 3 Experimental design for GalOx refoldingRefolding of HRP

Refolding of solubilized HRP was performed in a buffer containing 2 M urea, 2 mM CaCl\(_\), 7% (v/v) glycerol, 1.27 mM GSSG — oxidized form of glutathione, at pH 10 [18]. Each measurement was performed in duplicate. The refolding process was initiated by dilution of the solubilized protein in the refolding buffer to the final HRP concentration of 0.5 g L\(^\). The refolding process was monitored over 20 h at 7.5°C with a constant stirring of 800 rpm. After 20 h, hemin cofactor (1 mM stock in 100 mM KOH solution) was added to the refolding samples to reach the hemin concentration of 5 \(\mu \)M or 20 \(\mu \)M in the samples. After the hemin addition, the samples were monitored for additional 2 h at 7.5°C with a constant stirring of 800 rpm.

Online intrinsic fluorescence monitoring

Intrinsic Trp and Tyr fluorescence was measured using an FP-8550 Spectrofluorometer (Jasco, Tokyo, Japan) with a multi-cuvette holder (Jasco, Tokyo, Japan) enabling thermostating and stirring of the cuvettes. Refolding was carried out in 3-mL quartz fluorescence cuvettes (Starna GmbH, Germany) with magnetic stirrers at a volume of either 1.5 mL or 3.0 mL. The temperature of the cuvette holder was set to 5°C (LDH, GalOx) or 7.5°C (HRP) and the stirring speed was set between 500 and 800 rpm. The sample was excited at 280 nm and the emission spectrum was recorded between 310 and 370 nm with a step size of 0.5 nm. Excitation and emission slits of 1 nm and 10 nm, respectively, were used. The scanning speed was set to 200 nm min\(^\), sensitivity to medium, and the response time to 0.5 s. Data processing was conducted using python 3.7. Pre-processing was carried out by calculating the integral of the intensity f from \(\lambda _0\) to \(\lambda _1\) using Eq. 1.

$$\begin F(t) = \int _^ f(t) d\lambda \end$$

(1)

The average emission wavelength (AEW) of the emission spectra was calculated by Eq. 2, where F\(_i\) is the fluorescence emission intensity at wavelength \(\lambda _i\).

$$\begin AEW = \frac \end$$

(2)

For time course measurements, spectra were collected in approximately 1-min intervals between 310 and 370 nm and AEW and fluorescence intensity integrals F were calculated for each spectrum. An exponential decay function as described in Eq. 3, was fit to the development of AEW over time, where y(t) corresponds to the AEW curve fit over process time (t). \(y_0\) describes the y-intercept, k the exponential decay coefficient in min\(^\) and d the final AEW in nm at the equilibrium. Its derivative with respect to time (Eq. 4) was used to assess the reactivity of the process with \(\dot(t)\) describing the change of AEW in nm min\(^\). A refolding reaction was considered to be finished when the derivative was below 5% of the maximum rate of change.

$$\begin y(t) = (y_0-d)\cdot e^+d \end$$

(3)

$$\begin \dot(t) = -(y_0-d)\cdot k \cdot e^ \end$$

(4)

The sensitivity of the exponential decay constant k was assessed by calculation of the signal-to-noise ratio (SNR) as shown in Eq. 5, with k in min\(^\), where \(\sigma _k\) is the standard deviation of k. k was considered to be inconclusive when falling below the threshold of 10\(\sigma \) [39].

$$\begin \text _k = \frac \end$$

(5)

From Eq. 2\(\Delta \)AEW in nm is calculated as a function of process time (t) in min (Eq. 6).

$$\begin \Delta AEW(t) = AEW(t)-AEW(t=0) \end$$

(6)

Offline analytical toolsQuantification of protein concentration

Concentrations of the total protein in the soluble fraction were quantified using reverse-phase high-performance liquid chromatography (RP-HPLC) as described by [40] with a Polyphenyl BioResolve-RP-mAb 2.7 µm 3.0 x 100 mm column (Waters Corporation, Milford, USA) and an UltiMate 3000 HPLC system (Thermo Fisher Scientific, Waltham, MA, USA). Bovine serum albumin was used as a reference standard in a concentration range of 0.05–2.0 g L\(^\). The concentration of insoluble protein was calculated based on the theoretically added total protein subtracted by the amount of soluble protein fraction.

Enzymatic activity assay

Enzymatic activities were all measured using photometric assays conducted in a TECAN Spark® microplate reader (Tecan Trading AG, Männedorf, Switzerland). The temperature was set to 30°C and absorbance was recorded for 2 or 3 min.

To measure the enzymatic activity of LDH, the reaction buffer (100 mM NaH\(_\)PO\(_\), 0.425 mM nicotinamide adenine dinucleotide (NADH), 0.45 mM pyruvate) was mixed with sample at a ratio of 30% (v/v). Absorbance was recorded at 340 nm using the extinction coefficient of NADH, which is 6.22 mM\(^\) cm\(^\) [41]. Here, 1 Unit was defined as the necessary enzyme for the conversion of 1 µmol of NADH per minute. The volumetric enzymatic activity (vAc) was calculated as follows:

$$\begin vAc = \frac} \end$$

(7)

Calculation of the volumetric enzymatic activity (vAc) in U mL\(^\) was based on the change of absorbance over time (\(\Delta A / \Delta t\)). \(V_t\) stands for the total volume of the reaction mixture, \(V_s\) is the volume of the enzyme solution, l is length of the optical path (l = 0.62 cm) and \(\epsilon \) is the extinction coefficient. The specific activity was calculated by dividing the volumetric activity by the target protein concentration determined via RP-HPLC.

The enzymatic activity of GalOx was determined via a two-stage colorimetric assay, where the reaction of galactose to H\(_2\)O\(_2\) catalyzed by GalOx is coupled to a chromogenic 2,2\('\)-azinobis(3-ethylbenzothiazolinesulfonic acid) (ABTS) assay. In brief, an assay solution consisting of 4% HRP stock solution (0.6 g L\(^\) HRP in 50 mM Tris-HCl, 1 M (NH\(_\))\(_\)SO\(_\), pH 7.5), 10% ABTS stock solution (6.125 g L\(^\) ABTS in 0.1 M NaH\(_\)PO\(_\), pH 7.5) and 40% D-galactose (1 M) was prepared in NaH\(_\)PO\(_\) at pH 7.5. Then, the prediluted sample was mixed with the assay solution at a ratio of 30% (v/v). Absorbance was measured at 420 nm using an \(\epsilon \) of ABTS of 36 mM\(^\) cm\(^\) [42]. The enzymatic activity was calculated according to Eq. 7. Here, 1 Unit was defined as the necessary enzyme for the oxidation of 2 µmol of ABTS per minute.

HRP activity assay also employs ABTS as a peroxidase substrate. The reaction mixture for the assay of a total volume of 200 µL contained: 175 µL of 8 mM ABTS solution in 50 mM phosphate-citrate buffer pH 5, 20 µL of 10 mM H2O2 in ultrapure water and 5 µL of the HRP sample after refolding diluted 1:1,000 in 20 mM BisTris/HCl pH 7. Absorbance at 420 nm was measured and the volumetric enzyme activity was calculated using Eq. 7, using the extinction coefficient at 420 nm (\(\epsilon _\) = 36 mM\(^\) cm\(^\)). The volumetric activity was tested for the samples at the end of monitored refolding process, i.e., after 22 h of refolding. The average volumetric activity and its standard deviation for each process were calculated from 9 independent activity measurements.

Circular dichroism

Circular dichroism (CD) spectra were measured in 0.1 cm pathlength SUPRASIL® quartz cuvettes (HellmaAnalytics, Müllheim, Germany) using a J-815 CD Spectrometer (Jasco, Tokyo, Japan). The temperature during the measurement was set to 5°C. The resulting far-UV CD spectrum was obtained as an average of three scans between 200 and 250 nm and the spectrum of the pure buffer was subtracted. Samples were diluted to a protein concentration of 0.5 g L\(^\). The mean residue ellipticity (MRE) in deg cm\(^\) dmol\(^\) was calculated as described in Eq. 8, where \(\theta _\) is the CD in mdeg, M is the molecular weight of the protein in g dmol\(^\), l is the pathlength in cm, n is the number of amino acid residues, and c is the protein concentration determined via RP-HPLC in g L\(^\).

$$\begin MRE = \frac \cdot M} \end$$

(8)

Process model

The refolding process model used was adapted from Kiefhaber et al. [9]. It is composed of three ordinary differential equations (ODEs) describing the state dynamics of intermediates (I) (Eq. 9), native protein (N) (Eq. 10) and aggregated protein (A) (Eq. 11) in a batch type process with a constant volume.

$$\begin \frac = -k_N \cdot I - k_A \cdot I^n \end$$

(9)

$$\begin \frac = k_N \cdot I \end$$

(10)

$$\begin \frac = k_A \cdot I^n \end$$

(11)

The refolding rate of the native protein \(k_N\) and the aggregation rate \(k_A\) are described by the following algebraic kinetic equations

$$\begin k_N = a_N\cdot (1+D)^ \end$$

(12)

$$\begin k_A = a_A\cdot (1+D)^ \end$$

(13)

with the kinetic model parameters \(a_N\), \(b_N\), \(a_A\) and \(b_A\), that can be experimentally identified and D representing the concentration of the denaturing agent. For the considered LDH refolding process the parameters have been identified elsewhere [43] and were specified with \(a_N=^}}\pm ^}}\), \(a_A=^}}\pm ^}}\), \(b_N=^}}\pm ^}}\) and \(b_A=^}}\pm ^}}\). For the pulsing experiments, discrete events have been introduced where the concentrations of I and D were increased according to the calculated concentration after pulse addition.

The output functions Eqs. (14) and (15) were used to describe the relationship between the model states and intrinsic fluorescence measurements, where \(P = I + N\) and \(\beta _1\) to \(\beta _5\) being the parameters obtained by the experimental data fit. P was considered as the total dissolved protein concentration without considering insoluble aggregates.

$$\begin F(t) = \beta _1\cdot \frac \end$$

(14)

$$\begin \Delta AEW(t) = \beta _3\cdot \frac}} + \beta _5 \cdot \frac \end$$

(15)

Modeling framework

The programming language Julia was used for the model analysis and simulation. The model was defined symbolically using ModelingToolkit.jl [44] and the ODE system was numerically solved using DifferentialEquations.jl [45] from the Julia SciML ecosystem.