Purification and basic biochemical characterization of 19 recombinant plant peroxidase isoenzymes produced in Pichia pastoris.

The plant enzyme horseradish peroxidase (HRP) is used in several important industrial and medical applications, of which especially biosensors and diagnostic kits describe an emerging field. Although there is an increasing demand for high amounts of pure enzyme preparations, HRP is still isolated from the plant as a mixture of different isoenzymes with different biochemical properties. Based on a recent next generation sequencing approach of the horseradish transcriptome, we produced 19 individual HRP isoenzymes recombinantly in the yeast Pichia pastoris. After optimizing a previously reported 2-step purification strategy for the recombinant isoenzyme HRP C1A by substituting an unfavorable size exclusion chromatography step with an anion exchange step using a monolithic column, we purified the 19 HRP isoenzymes with varying success. Subsequent basic biochemical characterization revealed differences in catalytic activity, substrate specificity and thermal stability of the purified HRP preparations. The preparations of the isoenzymes HRP A2A and HRP A2B were found to be highly interesting candidates for future applications in diagnostic kits with increased sensitivity.

Introduction

Horseradish peroxidase (HRP2; EC 1.11.1.7) is a class III peroxidase or classical secretory plant peroxidase which oxidizes different substrates (e.g. aromatic phenols, indoles, phenolic acids, amines, sulfonates) using peroxides, commonly H2O2, as initial electron acceptors [1-3]. This enzyme has been studied for more than 200 years. Already in 1810, horseradish roots were observed to cause a color reaction when mixed with the resin of Guaiacum plants [4], probably the oxidation of α-guaiaconic acid to guaiacum blue by HRP [5]. In plants, HRP is involved in numerous reactions, such as the crosslinking of phenolic molecules and the regulation of H2O2 levels, the cell wall network and auxin catabolism [6-8]. Correlating with the large number of different in vivo functions, horseradish was found to contain a multitude of different HRP isoenzymes. Up to 42 isoenzymes were detected by isoelectric focusing of commercial HRP preparations [9]. Jermyn et al. observed multiple proteins in the horseradish plant with peroxidase activity and found seasonal variation in their relative amounts as well as differences in substrate affinity [10,11]. This biochemical versatility of HRP isoenzymes was further demonstrated in several subsequent studies (e.g. [12-16]).

Until now, however, most studies have focused on the isoenzyme C1A [17], which is the only isoenzyme with a solved structure [18]. HRP C1A contains nine potential N-glycosylation sites, defined by the N-X-S/T motif, with X being any amino acid but proline, of which eight are glycosylated when isolated from plant [19]. Plant-derived HRP C1A has a total carbohydrate content of 21.8% [20]. Interestingly, plant-derived HRP isoenzymes with a basic isoelectric point (pI) of >12 were found to be less glycosylated, e.g. only 0.8–4.2% carbohydrate content for isoenzymes E3–E6 [15]. Tams et al. studied the effect of the N-glycans on the biochemical properties of HRP C1A and found that pI, absorption spectrum, peroxidase activity towards o-dianisidine and thermal stability remained the same, whereas the kinetic stability and the solubility in ammonium sulfate were decreased upon deglycosylation [21,22]. Thus, the presence of glycan structures on the enzyme surface has a considerable impact on HRP.

Today, the roots of the horseradish plant are the main source for commercially available HRP preparations. These preparations commonly describe mixtures of isoenzymes whose expression patterns change seasonally and in response to uncontrollable environmental factors [8]. The yields of HRP are rather low with less than 10 mg of total HRP protein, which presents a mixture of different isoenzymes, from 100 g of horseradish roots [23]. Thus, the yield of specific isoenzymes purified from such a mixture is extremely low, e.g. Aibara et al. reported as little as 40 mg of isoenzyme E1 from 200 kg of horseradish roots [15]. Unfortunately, due to intrinsic enzyme properties such as intramolecular disulfide bridges [18], the recombinant production of HRP is challenging. Recombinant production as inclusion bodies in Escherichia coli is possible (e.g. [24,25]), but refolding yields are as low as 10 mg L−1 [25]. Beside the recombinant production of HRP in insect cell cultures (e.g. [26,27]), the currently most promising production systems are yeasts such as Saccharomyces cerevisiae [28-30] and Pichia pastoris [29,31]. However, HRP produced in P. pastoris is heterogeneously hyperglycosylated, causing the enzyme to appear as a smear on a SDS polyacrylamide gel at a size of approximately 65 kDa instead of its unglycosylated size of 35 kDa [29,32,33]. These excessive yeast-type glycans considerably impede classical downstream processing approaches. Whereas plant-derived HRP can be purified either by several consecutive steps of column chromatography (e.g. [12,15,16]) or by affinity chromatography using the lectin concanavalin A (e.g. [34-37]) as an isoenzyme mixture in a quite simple way, yeast-derived HRP cannot be purified by these strategies [33].

One obvious advantage of the recombinant production of single HRP isoenzymes in P. pastoris is the fact that this isoenzyme does not need to be isolated from an isoenzyme mixture, an otherwise time-intensive and tedious purification effort. Consequently, all HRP activity can be ascribed to the produced individual isoenzyme, allowing its specific enzymatic characterization. However, for that purpose, the HRP isoenzyme still has to be purified from yeast proteins. Hyperglycosylated HRP isoenzyme C1A from P. pastoris was previously purified by subsequent steps of hydrophobic interaction chromatography (HIC), size exclusion chromatography (SEC) and anion exchange chromatography (AEC) [29,30]. Recently, we addressed the issue of this cumbersome purification strategy by even making use of the high carbohydrate content of recombinant HRP C1A. We applied hydrophobic charge induction chromatography (HCIC) operated in flowthrough mode to remove contaminating proteins that bound to the resin, whereas the hyperglycosylated HRP eluted in the flowthrough. The glycan coat surrounding HRP C1A seemed to mask the physicochemical properties of the enzyme, allowing this rather unconventional, negative chromatography approach. An additional polishing step by SEC gave a preparation of HRP C1A with a specific activity comparable to the purest commercially available HRP preparation from plant [33].

Recently, we performed a next generation sequencing approach of the horseradish transcriptome which greatly increased the amount of available HRP isoenzyme sequences [38], and thus allowed more detailed studies of single isoenzymes. Considering the numerous applications of HRP as a reporter enzyme in diagnostic assays and histochemical staining as well as in strain engineering studies (e.g. [31,39,40]), biocatalysis (e.g. [41]), wastewater cleanup systems (e.g. [42]) and antibody-directed enzyme-prodrug cancer therapy (e.g. [43]), it is highly interesting to biochemically characterize the different HRP isoenzymes to find the most suitable one for a certain application.

Here, we report the production, purification and basic biochemical characterization of 19 individual HRP isoenzyme preparations. We significantly improved our recently reported 2-step purification procedure [33], replacing the rather inefficient and slow SEC polishing step by using a tube monolithic AEC column. Finally, we performed a basic enzymatic characterization of the final HRP preparations to determine potential differences in their catalytic activities, substrate specificities and thermal stabilities.

Materials and methods

Chemicals

Enzymes were obtained from Thermo Scientific (formerly Fermentas, Germany). 2,2′-Azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) diammonium salt (ABTS), 3,3′,5,5′-tetramethylbenzidine HCl (TMB) and d(+)-biotin were purchased from Sigma–Aldrich (Austria). Difco™ yeast nitrogen base w/o amino acids (YNB), Bacto™ tryptone and Bacto™ yeast extract were obtained from Becton Dickinson (Austria). Zeocin™ was obtained from in vivo Gen (France). Other chemicals were obtained from Carl Roth (Germany).

P. pastoris strains for HRP production

All P. pastoris strains in this study were based on the P. pastoris wildtype strain CBS 7435 (identical to NRRL Y-11430 and ATCC 76273). The MutS (methanol utilization slow) phenotype of P. pastoris was shown to be superior over the Mut+ phenotype in terms of volumetric productivity and production efficiency of HRP [31]. Thus, all HRP production strains in this study were strains with MutS phenotype [44]. A detailed description of the identification of new HRP isoenzyme sequences and the generation of the P. pastoris strains producing the various HRP isoenzymes was given elsewhere [38]. Calculated basic protein parameters and the corresponding database accession codes are shown in Table 1. Synthetic codon-optimized genes encoding mature HRP isoenzymes were N-terminally fused to the prepro signal peptide of the S. cerevisiae mating factor alpha to facilitate efficient secretion of the recombinant HRP to the cultivation supernatant. The expression of the HRP encoding genes was regulated by the methanol-inducible promoter of the P. pastoris AOX1 gene.

Table 1

Calculated basic characteristics of HRP isoenzymes. The isoelectric point (pI) and the molecular weight (MW) were calculated using the Compute pI/Mw tool of the ExPASy server [55,56] and the number of potential N-glycosylation sites (N-X-S/T) was deduced from the NetNGlyc 1.0 Server [57]. All calculations were based on a cleavage of the prepro signal peptide of the S. cerevisiae mating factor alpha between A87 and E88, upstream of the mature HRP peptide.

HRP isoenzyme	pI	MW [kDa]	N-X-S/T	GenBank	UniProt
C1A	5.41	35.82	9	HE963800.1	K7ZWW6
25148.1 (C1C)	6.13	35.86	7	HE963802.1	K7ZWQ1
25148.2 (C1D)	6.50	35.89	7	HE963803.1	K7ZW56
04627 (C2)	8.38	35.67	4	HE963804.1	K7ZW02
C3	7.05	35.48	3	HE963805.1	K7ZWW7
A2A	4.84	32.09	9	HE963806.1	K7ZW28
A2B	4.84	32.12	9	HE963807.1	K7ZWQ2
E5	8.99	33.92	3	HE963808.1	K7ZW57
1805	5.75	35.96	5	HE963809.1	K7ZW05
22684.1	6.39	35.06	4	HE963810.1	K7ZWW8
22684.2	6.00	35.15	4	HE963811.1	K7ZW29
1350	8.47	31.42	3	HE963812.1	K7ZWQ3
5508	8.22	31.35	3	HE963815.1	K7ZWW9
6351	5.99	32.89	2	HE963816.1	K7ZW31
22489.1	8.24	31.37	2	HE963818.1	K7ZW59
22489.2	8.24	31.39	2	HE963819.1	K7ZW11
17517.2	9.30	32.69	4	HE963823.1	K7ZW60
08562.4	8.91	33.26	3	HE963825.1	K7ZWX1
08562.1	8.89	33.81	3	HE963824.1	K7ZW15

Production of recombinant HRP isoenzymes

Recombinant production of 19 different HRP isoenzymes in P. pastoris was performed in 2.5 L Ultra Yield Flasks from BioSilta (Finland), applying a protocol based on [45] with the following modifications: An overnight culture (ONC) of 30 mL YPD (yeast extract-peptone-dextrose) in 250 mL baffled shake flasks was inoculated with a single colony of a P. pastoris strain producing a specific HRP isoenyzme and incubated at 28 °C, 90 rpm and approximately 50% humidity for at least 12 h. 1.5 mL of this ONC were transferred to 270 mL of iron-supplemented BMD1% (11 g L−1 α-d(+)-glucose monohydrate, 13.4 g L−1 YNB, 0.4 mg L−1 d(+)-biotin, 278 mg L−1 FeSO47H2O, 0.1 M potassium phosphate buffer, pH 6.0) per Ultra Yield Flask and cultivated under the same conditions for approximately 60 h. A first induction pulse was performed by addition of 30 mL BMM10 (5% (v/v) methanol, 13.4 g L−1 YNB, 0.4 mg L−1 d(+)-biotin, 0.1 M potassium phosphate buffer, pH 6.0). 3.0 mL of pure methanol were added approximately 12 h and 36 h after the first induction pulse, 1.5 mL of pure methanol were added approximately 24 h and 48 h after the first induction pulse. 72 h after the first induction pulse, the culture broth was centrifuged (15,000×g, 30 min, 4 °C) and the supernatant was filtered through a 0.2 μm cellulose acetate filter (Sartorius Stedim Biotech, Germany).

Purification of recombinant HRP isoenzymes

In accordance to our previous study, the supernatant was concentrated using the Vivaflow 50 system (Sartorius Stedim Biotech, Germany) with a 10 kDa MWCO membrane prior to hydrophobic charge induction chromatography (HCIC; [33]). The buffer was changed to HCIC-A (500 mM NaCl, 20 mM NaOAc, pH 6.0) and concentrated to a final volume of 10–15 mL. All further steps of concentration and buffer change were performed using Vivaspin 20 tubes (Sartorius Stedim Biotech, Germany) with 10 kDa MWCO. The HCIC resin MEP HyperCel™ was obtained from Pall (Austria), and HCIC was performed in flowthrough mode based on [33]: A column containing approximately 25 mL of MEP HyperCel™ resin was equilibrated with at least 4 column volumes (CV) of buffer HCIC-A. 10–15 mL concentrated HRP solution in HCIC-A were loaded onto the column and washed with at least 220 mL of HCIC-A at a flow rate of approximately 55 cm h−1. Fractions of 10 mL were collected. Fractions containing HRP activity were pooled and concentrated to 500–1000 μL. The column was washed with 5 CV of 800 mM NaOH, then re-equilibrated with HCIC-A for subsequent runs.

Univariate screenings for a potential application of CIM® tube monolithic columns (BIA separations, Slovenia) as a second chromatographic purification step were performed with the partially purified isoenzyme C1A after HCIC. Flowthrough fractions from HCIC purifications were pooled, concentrated and rebuffered in either of the loading buffers: 50 mM Tris–HCl, pH 7.4, 50 mM Tris–HCl, pH 8.0 or 50 mM potassium phosphate, pH 6.0. The respective elution buffers contained 1 M NaCl. The tube monolithic columns tested were 1 mL CIM®-DEAE, 1 mL CIM®-QA and 1 mL CIM®-OH (BIA separations, Slovenia), which were equilibrated in the respective loading buffer at a flow rate of 156 cm h−1. A post-load wash of 4 CV binding buffer was performed before elution was conducted by either increasing the elution buffer in a single step to 100% or in a linear gradient to 100% over 30 CV.

Ultimately, anion exchange chromatography (AEC) with an 8 mL CIM®-DEAE tube monolithic column was performed as a second purification step for all the HRP isoenzymes. The column was equilibrated in loading buffer AEC-A (50 mM Tris–HCl, pH 8.0) at a flow rate of 16.8 cm h−1. Post-HCIC pools of each HRP isoenzyme were subjected to diafiltration in AEC-A and were subsequently loaded onto the AEC column at an average linear flow rate of 16.8 cm h−1. Elution was performed in a single step from 0% to 100% AEC-B (50 mM Tris–HCl, 1 M NaCl, pH 8.0). For column recovery the column was washed with 5 CV of a 1 M NaOH/1 M NaCl solution at an average linear flow rate of 33.6 cm h−1.

Electrophoresis

To check the electrophoretic purity of HRP isoenzyme preparations SDS–PAGE was performed using a 5% stacking gel and a 10% separating gel in 1× Tris–glycine buffer. Unless otherwise stated, samples were diluted to a protein concentration between 0.1 and 0.5 mg mL−1 before loading. Gels were run in a vertical electrophoresis Mini-PROTEAN Tetra Cell apparatus (Biorad, Austria) and stained with Coomassie blue. The protein mass standard used was the PageRuler Prestained Ladder (Fermentas, Austria).

Data analysis and basic enzymatic characterization of purified recombinant HRP isoenzymes

Protein concentrations were determined at 595 nm by the Bradford assay using the Sigma–Aldrich (Austria) Protein Assay Kit with bovine serum albumin as standard in the range of 0.2–1.2 mg mL−1.

The enzymatic activity of HRP was measured using an ABTS assay in a CuBiAn XC enzymatic robot (Innovatis, Germany). 10 μL of sample were mixed with 140 μL 1 mM ABTS solution (50 mM potassium phosphate buffer, pH 6.5). The reaction mixture was incubated at 37 °C for 5 min before the reaction was started by the addition of 20 μL 0.078% (w/w) H2O2. Changes in absorption at 415 nm were measured for 80 s and rates were calculated. The standard curve was prepared using a commercially available HRP preparation (Type VI-A, Sigma–Aldrich, Austria) in the range from 0.02 to 2.0 U mL−1.

The efficiency of the applied purification approach was evaluated by determining the purification factor (PF) and the recovery yield of HRP activity in percentage (R%). PF and R% were calculated by Eqs. (1) and (2):

The suffixes “pre” and “post” indicate the respective values before and after a purification step. To obtain an overall PF and R% for the 2-step purification approach, we combined the values we determined for the single purification steps (Table 2). In case one purification step did not work or could not be evaluated, e.g. due to too low HRP activity, we only presented the successful purification step. The pooled active fractions after AEC were concentrated using Amicon Ultra-15 Centrifugal Filter Units with 10 kDa MWCO (Merck-Millipore, Austria) to the final enzyme preparation of a volume of approximately 1.5 mL.

Table 2

Summary of the 2-step purification approach of 19 recombinant HRP isoenyzmes. The purification factor (PF) and the recovery of HRP activity in percentage (R%) of the applied HCIC and AEC flowthrough steps, as well as the combined purification results are shown. Some isoenzymes did not show any detectable peroxidase activity. In these cases, no values for PF or R% are available (n/a). The combined PF value is a product of the PFs of the HCIC and AEC step; in case no values were available for one purification step (n/a), the combined value only describes the working step.

HRP isoenzyme	HCIC		AEC		Combined
	PF	R%	PF	R%	PF	R%
C1A	7.02	95.4	10.92	58.5	76.66	55.8
25148.1 (C1C)	6.80	95.5	2.28	66.4	15.50	63.4
25148.2 (C1D)	8.96	32.9	2.42	66.5	21.68	21.9
04627 (C2)	10.08	100.0	6.17	45.7	62.19	45.7
C3	2.17	17.6	2.50	100.0	5.43	17.6
A2A	15.85	100.0	43.18	15.6	684.40	15.6
A2B	5.82	92.0	6.64	80.0	38.64	73.6
E5	3.89	33.1	0.54	81.0	2.10	26.8
1805	n/a	n/a	15.68	38.4	15.68	38.4
22684.1	0.11	1.4	n/a	n/a	n/a	n/a
22684.2	n/a	n/a	n/a	n/a	n/a	n/a
1350	n/a	n/a	n/a	n/a	n/a	n/a
5508	2.18	33.2	66.25	31.5	144.43	10.5
6351	3.01	32.6	n/a	n/a	n/a	n/a
22489.1	0.42	3.8	n/a	n/a	n/a	n/a
22489.2	0.02	0.3	2.09	74.6	0.04	0.2
17517.2	3.48	38.9	0.98	75.0	3.41	29.2
08562.4	n/a	n/a	0.89	32.6	0.89	32.6
08562.1	1.18	16.5	0.27	45.5	0.32	7.5

Characterization of the purified HRP isoenzyme preparations included the determination of the basic kinetic parameters, Vmax and KM, for the electron acceptors ABTS and TMB in a spectrophotometer UV-1601 from Shimadzu (Austria). These peroxidase substrates are commonly used in enzyme-linked immunosorbent assays (ELISA), The reaction mixture with a final volume of 1.0 mL contained a concentration of 1 mM H2O2, 10 μL of HRP isoenzyme preparation and varying concentrations of ABTS (0.05–10 mM) or TMB (0.005–0.5 mM) in 50 mM potassium phosphate buffer, pH 6.5. The increase in absorption was followed at 420 nm for ABTS and at 653 nm for TMB at 30 °C for 180 s, respectively. Absorption curves were recorded with an adapted software program (UVPC Optional Kinetics software, Shimadzu). The maximum reaction rate (Vmax) and the Michaelis constant (Km) were calculated with the Sigma Plot software (Version 11.0, Systat Software Inc., USA).

The thermal stability of individual HRP isoenzyme preparations was tested at 60 °C. The residual activity towards ABTS was measured after 5, 10, 15, 20, 30 and 60 min of incubation at 60 °C. The residual activities were plotted versus the incubation time and the half life times of thermal inactivation at 60 °C () were calculated using Eq. (3) [46]:kin rate of inactivation (slope of the logarithmic residual activity).

Results and discussion

HRP is a well-studied enzyme which is used in numerous industrial and medical applications (e.g. [31,39-43]). Due to certain intrinsic enzyme features, the recombinant production and purification of this important enzyme is quite cumbersome, and HRP is still isolated from horseradish roots as a mixture of isoenzymes at low yields. A recombinant production process in combination with an efficient purification strategy would allow the reliable production of individual HRP isoenzymes for the various applications in large amounts.

HRP production

In this study, 19 HRP isoenzymes were recombinantly produced in the methylotrophic yeast P. pastoris in shake flask cultivations. By fusing the genes to the prepro signal sequence of the S. cerevisiae mating factor alpha, the HRP isoenzymes were secreted to the cultivation broth. Due to the fact that P. pastoris actively secretes only few endogenous proteins [47,48], the subsequent downstream process was thereby considerably facilitated already. After centrifugation and diafiltration, the HRP isoenzymes could already be subjected to chromatography. Typical total protein concentrations in the cultivation broth at the time of harvesting were in the range of 200–500 mg L−1. The amount of obtainable purified HRP isoenzyme preparation differed vastly; for the isoenzymes which could be purified best using the here presented 2-step strategy the following protein contents per liter cultivation broth were obtained: 1.0 mg C1A, 0.6 mg C2, 0.1 mg A2A and 0.15 mg 5508. Also volumetric HRP activities with ABTS as reducing substrate varied considerably from isoenzyme to isoenzyme. For instance, the HRP isoenzymes 8562.1, 22489.1, A2A, C1A and E5 gave approximately 3, 70, 220, 440 and 670 U L−1, respectively.

Hydrophobic charge induction chromatography (HCIC)

In a recent multivariate Design of Experiments screening study, we found HCIC operated in the flowthrough mode to be very effective for the purification of the hyperglycosylated recombinant HRP C1A, allowing a 5-fold purification at almost 100% recovery [33]. However, in the present study the application of this flowthrough purification step to the 19 different HRP isoenzymes produced in P. pastoris led to quite diverse results in terms of purification factor (PF) and recovery yield (R%; Table 2), indicating significant differences in the physicochemical properties, to some extent probably caused by the different degrees of glycosylation of the individual HRP isoenzymes. However, the HCIC elution profile of the HRP isoenzyme C1A that was shown previously [33], could be reproduced under the conditions applied in the present study as the whole HRP activity was found in the flowthrough (Fig. 1). The higher PF for HRP C1A in this study compared to our previous results [33] (i.e. 7-fold versus 5-fold, respectively) might be explained by the different cultivation approaches. In the present study, we produced HRP C1A in shake flask cultivations, whereas previous cultivations were done in the controlled environment of a bioreactor [33]. The latter allowed cultivation under optimized conditions, thus limiting cell lysis and contamination of the cultivation broth by intracellular proteins. Presumably, the amount of contaminating proteins in the starting solution was therefore lower, causing an overall lower PF for the C1A preparation from the bioreactor.

Fig. 1

HCIC chromatogram of the recombinant HRP isoenzyme C1A. The HRP activity (dashed line) was determined by using ABTS as reducing substrate. Protein content was followed throughout the run by recording the absorption at 280 nm (solid line).

Elution profiles similar to the one shown in Fig. 1 were found for ten other HRP isoenzymes (graphs not shown), indicating the applicability of the HCIC flowthrough purification for these isoenzymes (Table 2). Remarkably, a 16-fold purification at 100% recovery was achieved for isoenzyme HRP A2A (Table 2). However, for some isoenzymes the flowthrough based HCIC step could not be applied successfully as no purification was achieved (e.g. HRP 22489.1; Table 2). To find an explanation for that phenomenon, we looked at the single isoenzymes in more detail. Predictions of potential N-glycosylation sites, based on the identification of the conserved N-X-S/T motif, were performed using the NetNGlyc 1.0 Server (Table 1). Interestingly, the number of predicted potential N-glycosylation sites correlated well with both the PF and the recovery yield (R%; Fig. 2). This observation strongly underlines our previous hypothesis that extensive glycosylation prevents the interaction of recombinant HRP from P. pastoris with the HCIC material, hence allowing the negative chromatography purification step [33]. For example, HRP C1A and HRP A2A each contain nine N-X-S/T motifs and could be purified 7.0- and 15.9-fold at 95.4 and 100.0% recovery, respectively. HRP 6351 and HRP C3, on the other hand, contain only 2 and 3 N-X-S/T motifs and could only be purified 3.0- and 2.2-fold at 32.6 and 17.6% recovery, respectively (Tables 1 and 2). Outliers from that correlation, e.g. HRP 22684.1, which could not be purified via HCIC (PF of 0.1; Table 2) despite containing four N-X-S/T motifs, might be explained by varying degrees in glycosylation due to steric hindrance at certain N-glycosylation sites. However, the generally high correlation between the number of glycosylation sites and both the PF and the R%, as evident in Fig. 2, allows the design of an appropriate purification strategy for extensively glycosylated enzymes produced in P. pastoris.

Fig. 2

Correlation of the number of N-glycosylation sites and HCIC purification parameters. A, purification factor (PF); B, recovery yield of HRP activity in percentage (R%). The average PF and R% for HRPs with n N-X-S/T sites are shown with the corresponding calculated standard deviations; n2 = 3, n3 = 4, n4 = 3, n7 = 2, n9 = 3.

Anion exchange chromatography (AEC)

Recently, monolithic columns were discovered as a powerful tool for both analytical purposes and preparative protein purification [49-51]. The solid support, a uniform monolithic porous material (e.g. glycidyl methacrylate-based materials), is simple to handle and to scale up, allows elevated operating flow rates and pressures (e.g. flow rates of up to 336 cm h−1 and a back pressure of up to 20 bar for an 8 mL tube monolithic column from BIA separations), and provides high binding capacity (>20 mg mL−1). These beneficial features are mainly enabled by the convective mass transfer of the target molecule through the highly interconnected channel structure of the porous polymer block. In convective processes, both resolution and binding capacity are not affected by the flow rate, an effect that is emphasized when large biomolecules such as proteins are separated due to their high diffusion coefficients [49]. On the other hand, porous particles which are applied in conventional chromatographic media require diffusive transport of the molecules which have to enter the pores to get in contact with the active surface. This diffusive transport results in increased separation times and void volumes.

In our previous study, we polished partially purified recombinant HRP C1A after HCIC by SEC [33]. Although this strategy gave a good PF of >2.0 and a recovery yield of 100%, SEC has several uneconomical disadvantages such as low flow rates, sample dilution, temperature effects due to long process times, limited sample volumes and limited scalability. Thus, we tested monolithic columns as an alternative to SEC. A wide range of monolithic formats and ligands is available today [52]. In this study, we tested two AEC resins and a HIC resin, since these two purification principles had shown promising results using particle-based resins for recombinant HRP C1A before [33]. We used different buffer systems and elution profiles for the potential application of CIM® tube monolithic columns as polishing step for recombinant HRP C1A, partially purified after HCIC. Active flowthrough fractions from HCIC purifications were pooled, concentrated via diafiltration and loaded on the different CIM® tube columns. The HIC resin CIM®-OH was not able to purify recombinant HRP C1A after HCIC any further, regardless of the buffers applied. We believe that this was due to the fact that the vast majority of hydrophobic proteins had already been retained on the HCIC resin. Using the strong AEC resin CIM®-QA, the enzyme preparation could not be purified more than 1.3-fold regardless of the buffer, a phenomenon which we also observed with particle-based strong AEC materials before [33]. However, the weak AEC tube monolithic column CIM®-DEAE gave satisfactory results. Using Tris–HCl (50 mM, pH 8.0) as loading buffer, a PF of nearly 11.0 was obtained for recombinant HRP C1A when the negative chromatography approach was applied.

Interestingly, similar to the HCIC flowthrough step, operation of AEC in the flowthrough mode also gave diverse results in terms of PF and R% for the different recombinant HRP isoenzymes (Table 2). Some isoenzymes could not be purified by this strategy, whereas other isoenzymes were purified up to 66-fold. Remarkably, for isoenzymes 1805, 5508 and 22489.2 only the second purification step via the CIM®-DEAE monolithic column worked, whereas HCIC could not improve the preparations in terms of enzyme purity; in fact, PFs of more than 30 showed that for some HRP isoenzymes the AEC step alone already described an efficient purification strategy (Table 2). The HRP isoenzymes 22684.1, 6351 and 22489.1 did not show any detectable peroxidase activity after diafiltration indicating instability under the conditions applied. Furthermore, no HRP activity was detected prior to AEC for 22684.2 and 1350 due to the high dilution of the enzyme at that stage (Table 2). Therefore, we cannot comment on the applicability of the AEC strategy for these five isoenzymes.

Summarizing, the CIM®-DEAE tube monolith describes a highly interesting alternative to SEC as a polishing step for partially purified recombinant HRP isoenzymes. Compared to the PF of around 2.0 which we achieved for the recombinant HRP isoenzyme C1A using SEC before [33], the flowthrough step applying an anion exchange monolith presented here is not only advantageous in terms of flow rates, sample volumes and thus process time, but also gave a 5-fold higher PF of nearly 11.0. In Fig. 3 we exemplarily show a SDS gel of the different steps during AEC purification of the HRP isoenzyme C1A. Although there are no striking bands indicating contaminant proteins in the flowthrough fraction of AEC, the specific activity of the purified HRP C1A preparation in this study was remarkably lower than in our previous study [33]: The preparation of HRP C1A in this study yielded a specific activity of only approximately 100 U mg−1, whereas the C1A preparation in our previous study yielded a specific activity of approximately 1000 U mg−1 [33]. We ascribe this phenomenon to the different cultivation procedures. Whereas HRP C1A was produced in the controlled environment of a bioreactor constantly providing optimal conditions for P. pastoris in our previous work [33], in the present study the HRP isoenzymes, including HRP C1A, were produced in shake flasks where conditions were not controlled. Limitations in oxygen and nutrients as well as gradients, which can occur in shake flasks, apparently influence the physiology of the cells and hence their ability to produce catalytically active enzyme. In fact, this is a very good example how the upstream process might influence the downstream process and the final product quality.

Fig. 3

SDS–PAGE of fractions from AEC with HRP C1A. Lane 1, molecular mass standard; lane 2, cell-free cultivation supernatant (5 μg); lane 3, flowthrough (5 μg); lane 4, flowthrough (10 μg); lanes 5 and 6, fractions eluted with buffer AEC-B (5 μg).

Basic biochemical characterization of HRP isoenzyme preparations

After the chromatographic 2-step purification procedure, the flowthrough fractions of the single HRP isoenzymes were pooled and concentrated by ultrafiltration before basic biochemical characterization was done. Especially, for HRP isoenzymes 22684.2 and 1350, where the concentration of HRP in the collected fractions was very low, this step was essential to be able to obtain reliable kinetic data.

As anticipated from our preliminary data [38], the preparations of the recombinant HRP isoenzymes featured significantly different biochemical properties. Not only physicochemical parameters, such as the predicted pI (Table 1), covered a broad range, but also the enzymatic activity towards the two tested electron donors ABTS and TMB were found to be highly versatile (Table 3; examples for Michaelis–Menten plots shown in Fig. 4).

Table 3

Kinetic parameters of recombinant HRP isoenzyme preparations after 2-step flowthrough purification. Kinetic data of the purified HRP isoenzyme preparations were recorded for the electron donors ABTS and TMB at a concentration of 1.0 mM H2O2. In some preparations, no peroxidase activity could be detected. In these cases, no values for Vmax or KM are available (n/a).

HRP isoenzyme	ABTS		TMB
	Vmax [U mg−1]	KM [mM]	Vmax [U mg−1]	KM [mM]
C1A	105.54	1.01	2031.50	0.11
25148.1 (C1C)	8.13	4.02	243.14	0.16
25148.2 (C1D)	9.94	3.55	139.77	0.13
04627 (C2)	5.52	4.49	82.34	0.15
C3	0.31	12.5	2.60	0.08
A2A	483.02	1.95	397.99	0.12
A2B	538.40	1.73	1049.11	0.18
E5	33.15	3.51	14.38	0.06
1805	2.95	2.36	39.02	0.11
22684.1	n/a	n/a	n/a	n/a
22684.2	1.08	3.36	11.02	0.07
1350	2.64	2.62	23.62	0.06
5508	42.40	0.46	9.89	0.10
6351	n/a	n/a	n/a	n/a
22489.1	0.17	3.03	n/a	n/a
22489.2	2.42	2.70	1.89	0.16
17517.2	0.11	0.39	0.24	0.32
08562.4	0.05	0.31	0.10	0.20
08562.1	0.10	0.18	0.06	0.11

Fig. 4

Michaelis Menten plots for preparations of recombinant HRP C1A and A2B. (A) HRP C1A with ABTS; (B) HRP A2B with ABTS; (C) HRP C1A with TMB; (D) HRP A2B with TMB. The Michaelis–Menten plots for all HRP isoenzyme preparations of this study are shown in Supplementary Figs. 1–6.

For the oxidation of ABTS, the highest Vmax values were obtained for the preparations of the isoenzymes A2A and A2B. These two isoenzyme preparations were able to oxidize ABTS 4- to 5-fold better than the preparation of the well-studied isoenzyme C1A (Table 3), rendering our preparations of HRP A2A and A2B particularly interesting for diagnostic bioassays with increased sensitivity. In a previous study on commercial preparations of acidic HRP isoenzymes from the plant, a comparatively high KM value of 4.0 mM was reported for ABTS as the reducing substrate [53]. In contrast, the here reported KM values of recombinant preparations of the acidic isoenzymes A2A and A2B were significantly lower with 1.95 and 1.73 mM, respectively. An explanation for this difference in substrate affinity remains speculative, but might be ascribed to the slightly different amino acid sequences of isoenzymes A2A and A2B used in this study compared to the commercial isoenzymes.

The here presented apparent KM of 1.01 mM for ABTS for the HRP C1A preparation was higher than the previously published KM values of 0.27 mM and 0.18 mM for C1A preparations from plant and E. coli, respectively [54]. In a previous study on recombinant HRP C1A from P. pastoris a KM of 0.68 mM was reported [30]. Apparently, yeast-derived HRP C1A preparations generally have a tendency for a lowered affinity for ABTS, probably related to the yeast-type hyperglycosylation compared to preparations from plant and E. coli.

Interestingly, some HRP isoenzyme preparations did not show any (e.g. 22684.1) or only very low (e.g. 08562.4) catalytic activity with H2O2 and ABTS. Nevertheless, bearing the biochemical diversity of HRP isoenyzmes in mind, these isoenzymes might be more active towards other substrates that were not tested in this study.

The oxidation of TMB was catalyzed best by the HRP C1A preparation, followed by HRP A2A and A2B (Table 3). Interestingly, HRP A2A oxidized TMB slower than ABTS, whereas most other isoenzymes – including C1A and A2B – oxidized TMB faster than ABTS (Table 3). Once more, these kinetic differences demonstrate the diverse substrate affinities and biochemical properties of the individual HRP isoenzymes. Keeping this variance in mind, it is of considerable importance to choose the most suitable isoenzyme for a certain application in the future, e.g. to use a HRP A2B preparation for a diagnostic kit with ABTS as substrate, but a HRP C1A preparation for diagnostics with TMB as substrate, to achieve optimal assay sensitivity. On that note, the here described efficient purification strategy is a prerequisite for the application of specific HRP isoenzyme preparations. Also, the possibility for the recombinant production of a certain HRP isoenzyme with favorable characteristics for a given application in P. pastoris is superior to the currently applied, but unpredictable and irreproducible isolation of a mixture of HRP isoenzymes from horseradish roots.

Thermostability of HRP isoenzyme preparations

The HRP isoenzyme preparations did not only differ in terms of enzymatic activity and substrate specificity, but also in thermal stability (Table 4). The preparations of HRP C1A and HRP 1805 did not show a detectable decrease in catalytic activity after 60 min of incubation at 60 °C, whereas the activity of HRP 22489.2 was already below 20% of the initial activity after 30 min. A summary of all the calculated thermal half-life times () is given in Table 4. The thermal stability profiles of the stable HRP C1A preparation, the moderately stable HRP A2A and the quite unstable HRP 22489.2 at 60 °C over time are exemplarily shown in Fig. 5. As shown in Table 4, the most thermostable HRP preparations of this study were HRP C1A and HRP 1805, which both did not show any detectable loss in catalytic activity at 60 °C after 60 min. HRP A2A and A2B, which are highly interesting in terms of catalytic activity with ABTS and TMB (Table 3), showed a significantly lower thermal stability (Table 4, Fig. 5). However, for possible future applications of these isoenzymes in sensitive bioassays, their stability is supposedly sufficient. In addition, no significant loss of peroxidase activity of these two isoenzymes could be detected over weeks when stored at 4 °C.

Table 4

Calculated half life times at 60 °C () of recombinant HRP isoenzyme preparations. Some isoenzymes did not show any detectable loss in HRP activity after 60 min. The thermal stability of recombinant HRP preparations with an initial Vmax lower than 0.5 U mg−1 for ABTS were not determined (n.d.). Due to a limited amount of purified enzyme, HRP 04627 (C2) and HRP E5 were not included in this study (n.i.).

HRP isoenzyme	τ_1/2 [min]
C1A	Stable for 60
25148.1 (C1C)	159
25148.2 (C1D)	21
04627 (C2)	n.i.
C3	n.d.
A2A	64
A2B	55
E5	n.i.
1805	Stable for 60
22684.1	n.d.
22684.2	46
1350	62
5508	17
6351	n.d.
22489.1	n.d.
22489.2	11
17517.2	n.d.
08562.4	n.d.
08562.1	n.d.

Fig. 5

Thermal stability profiles of selected recombinant HRP preparations. Filled circles, HRP C1A; open triangles, HRP A2A; filled squares, HRP 22489.2. Residual HRP activity was determined over 60 min of incubation at 60 °C.

Conclusions

In the present study, we recombinantly produced 19 single HRP isoenzymes in P. pastoris in shake flask cultivations. We optimized our recently reported 2-step purification approach for recombinant hyperglycosylated HRP replacing the tedious SEC step with an AEC step using a tube monolithic column. After purification, we biochemically characterized the individual HRP isoenzyme preparations with different substrates and evaluated their thermal stability. The main outcomes of this study can be summarized as:

The novel 2-step flowthrough purification strategy gave a recovery yield of 55% and a PF of approximately 77 for the recombinant HRP isoenzyme C1A. Although the recovery yield was lower, the PF was more than 7-fold higher compared to our previous study, where we achieved a recovery yield of 93% but only a PF of 10. Despite the lower recovery, the here presented strategy is superior, since the second purification step can be run in flowthrough mode, thus allowing both high sample volumes and flow rates.

Regarding the other isoenzymes especially HRP 04627 (C2), A2A and 5508 could be purified very efficiently with PFs of 62, 684 and 144, respectively. HRP 25148.1 (C1C), 25148.2 (C1D), 04627 (C2), A2B and 1805 were purified 15- to 38-fold.

The correlation between the amount of potential N-glycosylation sites and the success in flowthrough purification can be used to design an efficient purification strategy for glycosylated proteins expressed in P. pastoris in general.

Basic biochemical characterization using ABTS and TMB revealed significant differences of the individual isoenzyme preparations. The preparations of HRP A2A and HRP A2B turned out to be highly active with H2O2 and ABTS and hence are especially interesting for applications in diagnostic assays with high sensitivity.

The data provided in this study pave the way for cost-effective recombinant production of HRP isoenzymes in P. pastoris. Current efforts are made in our lab to provide detailed information on the identification of new HRP isoenzymes from a next generation sequencing of the horseradish transcriptome and to show classifying data on the new HRP isoenzyme sequences (Näätsaari et al., in preparation). Future in-depth studies will provide information on the molecular mechanisms underlying the differences in activity and stability of the various interesting HRP isoenzymes.