Combined optimization of N‐terminal site‐specific PEGylation of recombinant hirudin using response surface methodology and kinetic analysis

Abstract

In this study, a combined optimization method was developed to optimize the N‐terminal site‐specific PEGylation of recombinant hirudin variant‐2 (HV2) with different molecular weight mPEG‐propionaldehyde (mPEG‐ALD), which is a multifactor‐influencing process. The HV2‐PEGylation with 5 kDa mPEG‐ALD was first chosen to screen significant factors and determine the locally optimized conditions for maximizing the yield of mono‐PEGylated product using combined statistical methods, including the Plackett–Burman design, steepest ascent path analysis, and central composition design for the response surface methodology (RSM). Under the locally optimized conditions, PEGylation kinetics of HV2 with 5, 10, and 20 kDa mPEG‐ALD were further investigated. The molar ratio of polyethylene glycol to HV2 and reaction time (the two most significant factors influencing the PEGylation efficiency) were globally optimized in a wide range using kinetic analysis. The data predicted by the combined optimization method using RSM and kinetic analysis were in good agreement with the corresponding experiment data. PEGylation site analysis revealed that almost 100% of the obtained mono‐PEGylated‐HV2 was modified at the N‐terminus of HV2. This study demonstrated that the developed method is a useful tool for the optimization of the N‐terminal site‐specific PEGylation process to obtain a homogeneous mono‐PEGylated protein with desirable yield.

Abbreviations

CCD

central composition design

Corr

correlation coefficient

HV2

recombinant hirudin variant‐2

Mono‐PEG‐HV2

mono‐PEGylated recombinant hirudin variant‐2

mPEG‐ALD

mPEG‐propionaldehyde

mPEG‐SC

monomethoxy‐PEG‐succinimidyl carbonate

PB

Plackett–Burman

PEG

polyethylene glycol

RE

relative error

RSM

response surface methodology

1. Introduction

Hirudin is a 65‐amino acid polypeptide, which is probably the most potent natural thrombin inhibitor due to its high affinity for thrombin (inhibition constant (K _i), approximately 10⁻¹⁰–10⁻¹⁴ M) 1, 2. Recently, various recombinant hirudin variants have been developed for therapeutic use due to the limited supply of natural hirudin. Although two recombinant hirudin drugs, lepirudin and desirudin, are currently approved as anticoagulants by the FDA, the clinical use of recombinant hirudin is still limited by its short plasma half‐life. Consequently, repeated injections are required to maintain its treatment efficacy, which may also cause other side effects including immunogenicity and bleeding 2, 3. PEGylation has been generally proved to be a successful approach to prolong the plasma half‐life and reduce the immunogenicity of proteins and peptides 4, 5, 6, 7. For recombinant hirudin, PEGylation is also expected to reduce its bleeding risk. Recently, random PEGylation at the amino group of recombinant hirudin has been reported using nonspecific amine reactive PEG reagents, such as monomethoxy‐PEG‐succinimidyl propionic acid ester (mPEG‐SPA) 6, 8 and monomethoxy‐PEG‐succinimidyl carbonate (mPEG‐SC) 7, 9, 10. As recombinant hirudin has one N‐terminal residue and several lysine residues (Supporting information Fig. S1) 11, random amino‐PEGylation usually produces a heterogeneous mono‐PEGylated mixture of different positional isomers. The percentages of these positional isomers are difficult to be constant between different production batches, which is not accepted for drug administration in recent years. Therefore, it is desirable to develop a site‐specific PEGylation method to overcome such drawbacks of random PEGylation and obtain a homogeneous mono‐PEGylated recombinant hirudin.

In terms of the structure of recombinant hirudin, its N‐ and C‐terminus, respectively, bind to the active‐site pocket and the fibrinogen‐binding site of thrombin 11. Therefore, we speculated that site‐specific PEGylation of recombinant hirudin at the N‐terminus may be an alternative method because its C‐terminus can still play an important role in the interaction with thrombin. Recently, N‐terminal site‐specific PEGylation of proteins and peptides has been achieved using aldehyde chemistry under acidic condition due to the different pK _a values between the lysine ε‐amines (10.0–10.2) and the N‐terminal α‐amine (7.6–8.0) 12. Although aldehyde chemistry has high‐selective PEGylation at α‐amine of N‐terminus under acidic condition, minor competitive PEGylation at ε‐amine of lysine residues occur concurrently, which leads to multi‐PEGylated proteins (e.g., di‐PEGylated protein). Therefore, the reaction conditions for aldehyde PEGylation should be carefully optimized to improve the homogeneity and yield of the N‐terminal mono‐PEGylated protein.

N‐terminal site‐specific PEGylation of protein can be affected by diverse reaction conditions (at least six factors, e.g., pH, temperature, protein concentration, NaCNBH₃ concentration, molar ratio of PEG to protein, and reaction time) as well as the properties of the PEG reagent (e.g., PEG size). Traditional method, for example, one‐variable‐at‐a‐time, is not suitable for optimizing this process due to so many factors and their interactions between different variables. Alternatively, statistical methods, including Plackett–Burman (PB) design and response surface methodology (RSM) are simple and effective process optimization approaches, which can not only reduce the experimental numbers, but also analyze the interactions between different variables 13, 14, 15, 16, 17, 18, 19, 20. Recently the reports concerning the application of RSM for optimizing the protein PEGylation process are very limited 21. However, statistical methods are largely dependent on the design of experiments, which would not achieve the global optimum if the experimentally designed ranges of the variables did not include the optimal conditions. For example, in some cases of process optimization by RSM, the obtained response surfaces were slopes, and only local optimum (not the global optimum) was achieved in the boundary conditions of experimental design 15, 22. Beside the statistical methods, a reaction kinetic model‐based approach has been developed for the optimization of protein PEGylation process in our previous reports, which can provide detailed information about the PEGylation mechanism, and also achieve the global optimum of the PEGylation process in a very wide range of the variables 9, 10. However, this approach is difficult to be directly used for the optimization of N‐terminal site‐specific PEGylation of protein affected by so many factors, which requires a series of combinatorial experiments under various reaction conditions to obtain numerous experimental data for kinetic model fitting. To overcome the above mentioned limitations of individual optimization method by statistical methods or reaction kinetic model, the aim of this study is to develop a combined optimization methods based on both the advantages of the two methods. To our knowledge, the combined optimization of protein PEGylation process using statistical methods and reaction kinetic model has not yet been reported to date.

In this study, PEGylation of recombinant hirudin variant‐2 (HV2) using mPEG‐propionaldehyde (mPEG‐ALD) was investigated (Supporting information, Fig. S1). To obtain a desirable yield of mono‐PEGylated‐HV2 (mono‐PEG‐HV2), HV2 PEGylation with 5 kDa mPEG‐ALD was first chosen to screen significantly influencing factors and locally optimize using combined statistical methods, including the PB design, steepest ascent path analysis, and central composition design (CCD) for the RSM. Then, PEGylation kinetics with 5, 10, and 20 kDa mPEG‐ALD were investigated and two significant factors (molar ratio of PEG to HV2 and reaction time) were globally optimized using kinetic analysis. The PEGylation reaction mixture was quantitatively analyzed via RP‐HPLC. Mono‐PEG‐HV2 was purified via anion exchange chromatography and characterized using SDS‐PAGE and RP‐HPLC. The PEGylation site and in vitro anticoagulant activity of the identified mono‐PEG‐HV2 were further characterized.

2. Materials and methods

2.1. Materials

HV2 (HPLC pure >95%) was obtained from Chongqing Kerun Biomedical R&D Co., Ltd. (Chongqing, China). The mPEG‐ALD (MW = 5, 10, and 20 kDa) was purchased from Kaizheng Biotech Development Co., Ltd. (Beijing, China). Thrombin (Sigma‐T4648), fibrinogen (Sigma‐F8630), acetonitrile (HPLC grade), trifluoroacetic acid (HPLC grade), and other chemicals of analytical grade were obtained from Sigma–Aldrich (St. Louis, MO, USA). A LiChrospher 100 RP‐18 column (250  × 4.6 mm, 5 μm) was purchased from Merck (Darmstadt, Germany). A HiTrap Q HP column (5 mL) was purchased from GE Healthcare (Piscataway, NJ, USA).

2.2. Optimization of PEGylation of HV2 with 5 kDa mPEG‐ALD using statistical methods

Different reaction conditions for the PEGylation of HV2 with 5 kDa mPEG‐ALD were optimized by statistical methods including PB design, steepest ascent path analysis, and CCD for RSM. The experimental design and data analysis were performed using the Design‐Expert software version10 trial (State‐Ease Inc., Minneapolis, MN, USA). More details concerning optimization procedures are included in the Supporting information.

2.3. PEGylation kinetics of HV2 with different molecular weight mPEG‐ALD

The molar ratio of PEG to HV2 and reaction time for PEGylation with different molecular weight mPEG‐ALD were further optimized using kinetic analysis referring to our previous report 10. More details concerning optimization procedures were included in the Supporting information.

Statistical analysis was also performed to evaluate the simulation reliability using the parameters of relative error (RE) and correlation coefficient (Corr) between mean experiment value and model‐predicted value. RE was calculated as follow:

$$\mathrm{RE}(\%)=\left|1-\frac{c_{\mathrm{calc},i}}{c_{\exp ,i}}\right|·100\%$$ (1)

where, i is corresponding to HV2, mono‐PEG‐HV2, or di‐PEG‐HV2. c _{calc, i} is model‐predicted value of the concentration of i and c _{exp, i} is experimental value of the concentration of i. Corr was calculated by comparing point‐by‐point the model‐predicted value with the mean experiment value.

2.4. Purification of mono‐PEG‐HV2 via anion exchange chromatography

The PEGylation reaction mixtures were purified via the anion exchange chromatography according to our previous report 9. More details are included in the Supporting information.

2.5. RP‐HPLC analysis

PEGylation reaction mixtures and the elution fractions were analyzed via RP‐HPLC according to our previous report 10. More details are included in the Supporting information.

2.6. SDS‐PAGE analysis

Elution fractions were analyzed as previously reported 9. More details are included in the Supporting information.

2.7. Identification of PEGylation site of mono‐PEG‐HV2

The elution fractions of mono‐PEG‐HV2 were desalinated, concentrated, freeze dried, and stored at −20°C until subsequent analysis. The N‐terminal PEGylation site was analyzed through the examination of the N‐terminal amino acid sequence of mono‐PEGylated HV2 using Edman degradation method 23, 24. The freeze‐dried solids of unmodified HV2 and mono‐PEG‐HV2 were redissolved in double‐distilled water at a concentration of 2 × 10⁴ pmol/mL. N‐terminal five amino acid residues of each sample (200 pmol) were analyzed using PPSQ‐31A/33A Automated Protein/Peptide Sequencers (Shimadzu, Kyoto, Japan) following the manufacturer's instructions.

2.8. In vitro anticoagulant activity of mono‐PEG‐HV2

The in vitro anticoagulant activities of the unmodified HV2 and mono‐PEG‐HV2 were assessed via the thrombin titration method according to our previous report 9. The retained bioactivity of mono‐PEG‐HV2 was calculated according to the following equation:

$$\begin{array}{c}\hfill \mathrm{Retained}\mathrm{bioactivity}\mathrm{of}\mathrm{mono}-\mathrm{PEG}-\mathrm{HV}2\\ \hfill {\displaystyle =\frac{\mathrm{Bioactivity}\mathrm{of}\mathrm{mono}-\mathrm{PEG}-\mathrm{HV}2}{\mathrm{Bioactivity}\mathrm{of}\mathrm{HV}2}}\end{array}$$ (2)

2.9. Evaluation of the PEGylation efficiency

The parameters of conversion of HV2 (abbreviated as conversion), yield of mono‐PEG‐HV2 (abbreviated as yield), and selectivity of PEGylation (abbreviated as selectivity) were used to evaluate the PEGylation efficiency. They were calculated as follows:

$$\mathrm{Conversion}\mathrm{of}\mathrm{HV}2=\frac{\mathrm{The}\mathrm{total}\mathrm{PEGylated}\mathrm{HV}2\left(\mathrm{mono}-\mathrm{and}\mathrm{multi}-\mathrm{PEG}-\mathrm{HV}2\right)(\mathrm{mol})}{\mathrm{The}\mathrm{initial}\mathrm{HV}2\mathrm{for}\mathrm{PEGylation}(\mathrm{mol})}$$ (3)

$$\begin{array}{ccc}\hfill \mathrm{Yield}\mathrm{of}\mathrm{mono}& -& \mathrm{PEG}-\mathrm{HV}2\hfill \\ & =& {\displaystyle \frac{\mathrm{Mono}-\mathrm{PEG}-\mathrm{HV}2(\mathrm{mol})}{\mathrm{The}\mathrm{initial}\mathrm{HV}2\mathrm{for}\mathrm{PEGylation}(\mathrm{mol})}}\hfill \end{array}$$ (4)

$$\mathrm{Selectivity}\mathrm{of}\mathrm{PEGylation}=\frac{\mathrm{Mono}-\mathrm{PEG}-\mathrm{HV}2(\mathrm{mol})}{\mathrm{The}\mathrm{total}\mathrm{PEGylated}\mathrm{HV}2\left(\mathrm{mono}-\mathrm{and}\mathrm{multi}-\mathrm{PEG}-\mathrm{HV}2\right)(\mathrm{mol})}$$ (5)

2.10. Data analysis

All experimental data were obtained from at least three independently repeated experiments and are represented as the mean ± SD unless particularly outlined.

3. Results and discussion

3.1. Optimization of PEGylation of HV2 with 5 kDa mPEG‐ALD using statistical methods

3.1.1. Plackett–Burman design

PEGylation of HV2 with mPEG‐ALD can be affected by at least six factors. If all six factors were investigated by RSM, it is difficult to perform the optimization process due to too many combination experiments. Therefore, the PB design, a rapid and reliable method requiring few experiments 25, was used first to screen few significant factors from the six factors influencing PEGylation (Supporting information, Table S1). The design matrix contained the six selected factors and three responses (Supporting information, Table S2). The experimental data were statistically analyzed via the first‐order polynomial model (Supporting information, Table S5). The confidence levels > 95% (p < 0.05) were considered to be significant. The fitted model equations for conversion, yield, and selectivity as coded levels of the factors are as follows:

$$\begin{array}{ccc}\hfill \mathrm{Yield}\left(\%\right)& =& 50.74-5.13\mathrm{A}+3.66\mathrm{B}+4.46\mathrm{C}+4.89\mathrm{D}\hfill \\ & & +5.13\mathrm{E}+3.54\mathrm{F}\hfill \end{array}$$ (6)

$$\begin{array}{ccc}\hfill \mathrm{Conversion}\left(\%\right)& =& 57.61-8.02\mathrm{A}+4.76\mathrm{B}+9.01\mathrm{C}+5.16\mathrm{D}\hfill \\ & & +8.81\mathrm{E}+5.16\mathrm{F}\hfill \end{array}$$ (7)

$$\begin{array}{ccc}\hfill \mathrm{Selectivity}\left(\%\right)& =& 90.54+2.88\mathrm{A}-1.24\mathrm{B}-6.04\mathrm{C}-1.37\mathrm{D}\hfill \\ & & -4.52\mathrm{E}-1.63\mathrm{F}\hfill \end{array}$$ (8)

In Eqs. (6)–(8), the positive or negative regression coefficient of each factor shows its positive or negative effect on the response. The effect size of each factor on the response can be evaluated by the p‐value (Supporting information, Table S5). The smaller the p‐value, the greater the effect is. As shown in Supporting information, Table S5, all R ² of the models for conversion, yield, and selectivity were greater than 0.9, indicating that these models were well‐fitted to the experimental data. All six investigated factors have significant effects on the conversion and yield, while pH, molar ratio of PEG to HV2, and reaction time have significant effects on the selectivity. After an overall analysis, pH, molar ratio of PEG to HV2, and reaction time were screened as the three most significant factors for further optimization via RSM, while HV2 concentration, NaCNBH₃ concentration, or temperature were fixed at 1.5 mg/mL, 30 mM, and 30°C for subsequent experiments.

3.1.2. Steepest ascent path analysis

The PB design was unable to obtain the optimum values of the most significant factors. Therefore, steepest ascent path analysis was used to find the values closer to the optimum values as the center points for CCD 26. As shown in Supporting information, Table S3, the highest yield (65.9%) was achieved at the point of experiment run 3.Thus, pH 5.0, molar ratio of PEG to HV2 4:1, and reaction time 8 h were chosen as the center points of CCD in further optimization step.

3.1.3. Central composition design

According to the results of PB design and steepest ascent path analysis, CCD for RSM was used to determine the optimal values of the three screened factors for obtaining a maximum yield and a high selectivity. The designed actual levels of the screened factors and the experimental values of the two responses are shown in Table 1. The experimental data were statistically analyzed using the second‐order polynomial model (Supporting information, Table S6). As shown in Supporting information, Table S6, the model for yield or selectivity was significant (p < 0.05), while the corresponding lack of fit was not significant (p > 0.05). These results indicated that the two model equations were reliable. The fitted model equations for yield and selectivity as coded levels of three screened factors are as follows:

$$\begin{array}{ccc}\hfill \mathrm{Yield}\left(\%\right)& =& 66.45-1.01\mathrm{A}+0.59\mathrm{C}+0.90\mathrm{E}-0.49\mathrm{AC}\hfill \\ & & +0.96\mathrm{AE}-0.51\mathrm{CE}-0.61{\mathrm{A}}^2-1.07{\mathrm{C}}^2\hfill \\ & & -1.24{\mathrm{E}}^2\hfill \end{array}$$ (9)

$$\begin{array}{ccc}\hfill \mathrm{Selectivity}\left(\%\right)& =& 81.81-0.67\mathrm{A}-1.53\mathrm{C}-1.06\mathrm{E}-0.61\mathrm{AC}\hfill \\ & & +1.75\mathrm{AE}-0.030\mathrm{CE}+0.38{\mathrm{A}}^2-0.41{\mathrm{C}}^2\hfill \\ & & -0.95{\mathrm{E}}^2\hfill \end{array}$$ (10)

Table 1.

Experimental design and results of the central composition design

Run	Factors (actual levels)			Response
	A (pH)	C (Molar ratio of PEG to HV2 )	E (Reaction time [h])	Yield (%)	Selectivity (%)
1	4.5	3	6	62.2	84.5
2	5.5	3	6	60.0	81.3
3	4.5	3	10	64.3	79.9
4	5.5	3	10	64.4	82.4
5	4.5	5	6	66.5	83.4
6	5.5	5	6	60.8	76.5
7	4.5	5	10	65.0	77.4
8	5.5	5	10	64.7	78.7
9	4.16	4	8	66.5	84.2
10	5.84	4	8	63.1	82.5
11	5.0	4	4.64	62.0	81.7
12	5.0	4	11.36	64.0	77.5
13	5.0	2.32	8	62.9	83.8
14	5.0	5.68	8	64.1	78.5
15	5.0	4	8	66.4	81.8
16	5.0	4	8	66.6	82.5
17	5.0	4	8	66.3	82.0
18	5.0	4	8	67.3	82.2
19	5.0	4	8	66.2	81.2
20	5.0	4	8	65.9	81.0

Other reaction conditions: HV2 1.5 mg/mL (2.14 × 10⁻⁴ mol/L), NaBH₃CN 30 mM, and temperature 30°C.

In Eqs. (9) and (10), the positive or negative regression coefficients of individual factor terms (A, C, and E), interaction terms (AC, AE, and CE), and quadratic terms (A², C², and E²) show their positive or negative effects on the responses. The greater the absolute value of the regression coefficient, the greater the effect is. The effect sizes of these terms on the responses can also be evaluated by the p‐value (Supporting information, Table S6). The smaller the p‐value, the greater the effect is. As shown in Supporting information, Table S6, all individual factor terms (A, C, and E), interaction terms (AC, AE, and CE), and quadratic terms (A2, C2, and E2) were significant for the response of yield. For the response of selectivity, the terms (A, C, E, AC, AE, and E2) were significant. To visually understand the interactions of the factors and to analyze their optimized values for obtaining the maximum yield and high selectivity, the response surface and contour plots were also constructed (Fig. 1 and Supporting information, Fig. S3). Regression results for the response surface of the quadratic polynomial model are listed in Supporting information, Table S7. To better understand these regression results, comparison of actual and predicted yields of mono‐PEGylated HV2 and selectivity of PEGylation are shown in Supporting information, Fig. S4. As shown in Supporting information, Table S7, all confidences of determination (R ²) and adjusted R ² of the models for yield and selectivity were greater than 0.9, indicating that the models were well‐fitted to the corresponding experimental data. However, predicted R ² for yield and selectivity were not as close to the adjusted R ², probably indicating a large block effect. The adequate precision for yield and selectivity were greater than 4, indicating that the model can be used to navigate the design space. Based on the model equations (9) and (10), a series of optimized combinations of pH, molar ratio of PEG to HV2, and reaction time for obtaining a maximum yield and a high selectivity were obtained using the Design‐Expert software, and verified by the corresponding experiments (Supporting information, Table S8). The experiment values are in good agreement with the model‐predicted values (Supporting information, Table S8), indicating that this optimization method is usable.

Figure 1.

Response surface plot (A, C, and E) and the corresponding contour plot (B, D, and F) of the effects of two different factors and their interactions on the yield of mono‐PEG‐HV2 predicted by RSM.

3.2. Kinetics analysis of PEGylation of HV2 with different molecular weight mPEG‐ALD

Among the factors influencing the PEGylation of HV2 with mPEG‐ALD, the molar ratio of PEG to HV2 and reaction time have the most effects on the selectivity of PEGylation and yield of mono‐PEG‐HV2. According to our previous study concerning PEGylation of HV2 with mPEG‐SC 10, we demonstrated that the concentration change of mono‐PEG‐HV2 as a function of the molar ratio of PEG to HV2 and reaction time is very complex, which cannot be described using the second‐order polynomial model derived from RSM. Thus, optimization the PEGylation reaction conditions using RSM can only achieve a local optimum in the experimentally designed ranges of the variables. Alternatively, we have also demonstrated that PEGylation kinetics can be used to describe real PEGylation reaction process and optimize efficiently the PEGylation reaction conditions to achieve the global optimum 10. Therefore, PEGylation kinetics of HV2 were investigated and the two factors were future optimized using kinetic analysis. On the other hand, PEG size also has significant effect on PEGylation of HV2 with mPEG‐ALD, which has little interactions with other variables 10. To reduce the number of experiments, the molar ratio of PEG to HV2 and reaction time for PEGylation of HV2 with 10 or 20 kDa mPEG‐ALD were directly optimized using kinetic analysis under the locally optimized reaction conditions derived from PEGylation of HV2 with 5 kDa mPEG‐ALD.

As shown in Fig. 2, the established kinetic model fit well with the experimental data in most cases. To quantificationally understand the reliability of kinetic modeling, statistical analysis was also performed using the parameters of RE and Corr, and the results are listed in Supporting information, Tables S9–S11 and Table 2. On one hand, the Corr values in all cases were more than 0.93, indicating that this model was reasonable. On the other hand, the RE is a little high in some cases. As shown in Supporting information, Tables S9–S11, the main source of error was from di‐PEG‐HV2. One possible reason is that di‐PEG‐HV2 concentration is too low to be effectively quantitated using RP‐HPLC, resulting in a measurement error of the corresponding experiment data. The second source of error was from mono‐PEG‐HV2 at the initial stage of the reaction. One possible explanation is as follow: the sample taken at different reaction time intervals was added an excess glycine solution to stop the reaction. Because the amine of glycine can competitively react with mPEG‐ALD, minor PEGylation reaction can still occur concurrently. The sample taken at the initial stage could continue to generate a certain amount of mono‐PEG‐HV2, leading to a higher measured experiment value than the real value (as shown in Supporting information, Tables S9–S11).

Figure 2.

PEGylation kinetics of HV2 with mPEG‐ALD. Other reaction conditions: HV2 1.5 mg/mL (2.14 × 10⁻⁴ mol/L), NaBH₃CN 30 mM, pH 4.5, and temperature 30°C. Dots are experimental points and solid lines are theoretical prediction.

Table 2.

Rate constants process parameters of PEGylation of HV2 with mPEG‐ALD

PEG Size (kDa)	k 1 (L·mol−1·min−1)	k 2 (L·mol−1·min−1)	k d (min−1)	m crit	y max (%)	Apparent RE (%)	Apparent Corr
5	26.42	5.18	0.0098	4.57	67.21	8.3	0.9968
10	17.00	2.45	0.0123	8.71	72.17	19.5	0.9757
20	9.28	1.24	0.0143	17.79	73.31	16.9	0.9747

Rate constants (k ₁, k ₂, and k _d) were estimated by fitting the reaction kinetics model to the experimental data in Fig. 2. The m _crit and y _max could be found in Fig. 4. Apparent RE (%) is defined as: apparent RE (%) = (average RE (%) of HV2 + average RE (%) of mono‐PEG‐HV2 + average RE (%) of di‐PEG‐HV2)/3. Apparent Corr is defined as: apparent Corr = (average Corr of HV2 + average Corr of mono‐PEG‐HV2 + average Corr of di‐PEG‐HV2)/3. The two parameters were calculated using the data in Supporting information, Tables S9–S11.

Based on the calculated rate constants (k ₁, k ₂, and k _d) (Table 2), the concentration change of mono‐PEG‐HV2, that is, the yield as molar ratio of PEG to HV2 and reaction time can be predicted by the established kinetic model in a very wide range of the two variables (theoretically, from 0 to + ∞) and consequently global optimum can be achieved (Fig. 3). As shown in Figs. 1E, F and Figs. 3A, B, the surface plot and the corresponding contour plot derived from reaction kinetic model were different from those derived from RSM based on the second‐order polynomial model. Relatively, reaction kinetic model, which can describe real PEGylation reaction process, is more efficient than RSM in the optimization of PEGylation conditions in a very wide range of the two variables. In our previous report 10, we found several important process parameters (the theoretical yield of mono‐PEG‐HV2 [y _theo], the maximum theoretical yield of mono‐PEG‐HV2 [y _max], the critical molar ratio of PEG to HV2 to achieve y _max [m _crit], and the reaction time to achieve y _max [t _max]) and their mathematical equations to determine the optimal conditions. These process parameters were also found in the process of PEGylation of HV2 with mPEG‐ALD (Fig. 4), which had the similar trends with those of mPEG‐SC 10. As shown in Fig. 4, the y _max can be achieved at any molar ratio of PEG to HV2 (m) greater than m _crit and the corresponding t _max, which was demonstrated in our previous report 10. This interesting and useful result cannot be obtained using RSM. The mathematical equations of y _max and m _crit and an empirical equation to describe the relationship between t _max and molar ratio of PEG to HV2, which have been described in details in our previous report 10, were also applicable for this study. Furthermore, the optimal conditions were further verified by the corresponding experiments at a molar ratio of PEG to HV2 slightly greater than m _crit and the reactions were stopped at t _max (Table 3). For 5 kDa mPEG‐ALD, the optimized PEGylation reaction conditions and yield of mono‐PEG‐HV2 obtained from kinetic analysis are identical to those obtained from RSM (Supporting information, Table S8 and Table 3), which indicated the two optimization methods can supply and verify each other to some extent. Compared with random PEGylation of HV2 with mPEG‐SC, PEGylation of HV2 with identical molecular weight mPEG‐ALD had higher yield and selectivity (Table 3), indicating that N‐terminal site‐specific PEGylation improved the homogeneity and yield of mono‐PEG‐HV2. Under the optimal conditions (Table 3), N‐terminal site‐specific PEGylation improved both the yields and selectivity of different molecular weight mono‐PEG‐HV2 more than 20% over those obtained from random PEGylation reported in our previous study 10, indicating that N‐terminal site‐specific PEGylation is superior to random PEGylation from the view of industrial production.

Figure 3.

Surface plot (A, C, and E) and the corresponding contour plot (B, D, and F) of the effects of molar ratio of PEG to HV2 and reaction time on the theoretical yield of mono‐PEG‐HV2 predicted by kinetic analysis. (A, B) PEG 5 kDa; (C, D) PEG 10 kDa; and (E, F) PEG 20 kDa. Kinetic analyses were performed based on rate constants listed in Table 2.

Figure 4.

Effects of molar ratio of PEG to HV2 and reaction time on the theoretical yield of mono‐PEG‐HV2 (A) and several important process parameters (B, C, and D) predicted by kinetic analysis. Kinetic analyses were performed based on rate constants listed in Table 2. y _theo = theoretical yield of mono‐PEG‐HV2; y _max = maximum theoretical yield of mono‐PEG‐HV2; m _crit = critical molar ratio of PEG to HV2 to achieve y _max; and t _max = reaction time to achieve y _max. The fitted constants of the equation to describe the relationship between t _max and m are listed in Supporting information, Table S12.

Table 3.

Experimental verification of the optimal conditions derived from kinetic analysis

PEG reagent	PEG size (kDa)	Molar ratio of PEG to HV2	Reaction time (min)	Yield (%)		Selectivity (%)		Ref.
				Calc.	Exp.	Calc.	Exp.
mPEG‐ALD	5	4.6	435	67.21	67.1±1.6	83.6	84.2±1.5	This work
mPEG‐ALD	10	9.0	255	72.17	71.4±1.4	87.4	88.1 ±1.2	This work
mPEG‐ALD	20	18.0	295	73.31	72.6±1.3	88.2	88.7±1.6	This work
mPEG‐SC	5	1.8	94	51.5	51.6±0.1	67.6	68.2±1.5	10
mPEG‐SC	10	2.1	90	57.5	56.4±1.9	71.3	72.1±1.8	10
mPEG‐SC	20	2.5	106	59.5	59.8±0.5	72.6	73.8±2.2	10

Exp, experimental value; calc, model‐predicted value. Experimental verification was carried out under the optimal condition at a molar ratio of PEG to HV2 (m) slightly greater m _crit (in Table 2) and the corresponding t _max (in Fig. 4) to achieve the y _max. Other reaction conditions for mPEG‐ALD: HV2 1.5 mg/mL (2.14 × 10⁻⁴ mmol/L), NaBH₃CN 30 mM, pH 4.5, and temperature 30°C. Other reaction conditions for mPEG‐SC: HV2 1 mg/mL (1.43 × 10⁻⁴ mol/L), pH 8.0, and temperature 25°C. The yield and selectivity were calculated according to Eqs. (4) and (5) in Section 2.9, respectively

It is worth noting that we gave priority to obtain high yield rather than high selectivity in this study. In fact, high selectivity is significant for industrial production, which could simplify downstreaming purification. As shown in Supporting information, Table S2, 100% selectivity could be achieved in some cases, although the yields were not high. In further studies, an integrated reaction and separation process will be developed to isolate and recycle the unreacted HV2 under these conditions of 100% selectivity (Supporting information, Table S2), and finally achieve a high selectivity and high yield after multibatch recycling of unreacted HV2.

3.3. Purification of mono‐PEG‐HV2 via anion exchange chromatography

The reaction mixtures were effectively separated via anion exchange chromatography (AEC) (Supporting information, Fig. S5). The elution peaks were collected and analyzed via SDS‐PAGE and RP‐HPLC (Supporting information, Fig. S6). The results indicated that peaks 1, 2 and 3 in Supporting information, Fig. S5 corresponded to unmodified HV2, mono‐PEG‐HV2, and di‐PEG‐HV2, respectively. Thus, these peaks were labeled in Supporting information, Fig. S5. The N‐terminal mono‐PEG‐HV2 (mono‐PEG‐HV2 derived from mPEG‐ALD) eluted at smaller elution volumes compared with the mono‐PEG‐HV2 derived from mPEG‐SC, but the elution volumes of unmodified HV2 were identical 7. The results indicated that N‐terminal mono‐PEG‐HV2 achieved a better separation resolution than the mono‐PEG‐HV2 derived from mPEG‐SC. SDS‐PAGE and RP‐HPLC analyses indicated that the collected mono‐PEG‐HV2 from AEC had both a high electrophoresis and HPLC purity (both >95%) (Supporting information, Fig. S6) and can be used for the characterization of its PEGylation site and in vitro anticoagulant activity.

3.4. Identification of PEGylation site of N‐terminal mono‐PEG‐HV2

Common method to identify the PEGylation sites is tryptic peptide mapping (combined tryptic digestion with LC–MS analysis). Alternatively, the N‐terminal mono‐PEGylated protein could also be identified through examining its N‐terminal amino acid sequence by Edman degradation method, because the PEGylated α‐amine can resist peptide cleavage by fluorodinitrobenzene (FDNB) 24. As shown in Supporting information, Fig. S7, the first amino acid residues (Ile or I in Supporting information, Fig. S1) of mono‐PEG5k‐HV2 (Supporting information, Fig. S7C(a)), mono‐PEG10k‐HV2 (Supporting information, Fig. S7D(a)), and mono‐PEG20k‐HV2 (Supporting information, Fig. S7E(a)) had no signal peak at the corresponding position compared with unmodified HV2 (Supporting information, Fig. S7B(a)). The result indicated that almost 100% of the obtained mono‐PEG5k‐HV2, mono‐PEG10k‐HV2, or mono‐PEG20k‐HV2 were modified at the N‐terminus of HV2, that is, it is a homogeneous product, whereas random PEGylation process reported in our previous study could only obtain a heterogeneous mono‐PEGylated HV2 mixture of different positional isomers 10.

3.5. In vitro anticoagulant activity of mono‐PEG‐HV2

The mono‐PEG5k‐HV2, mono‐PEG10k‐HV2, and mono‐PEG20k‐HV2, respectively, retained 25.3, 22.7, and 21.3% in vitro anticoagulant activity compared with unmodified HV2 (Supporting information, Fig. S8). Compared with an identical molecular weight mono‐PEG‐HV2 derived from random PEGylation with mPEG‐SC 7, the in vitro anticoagulant activities of N‐terminal mono‐PEG‐HV2 decreased significantly. This is because PEGylation at the N‐terminus of HV2 prevents it from binding to the active‐site pocket of thrombin. As our expectation, the in vitro anticoagulant activities of N‐terminal mono‐PEG‐HV2 were not completely lost because its C‐terminus can still bind to the fibrinogen‐binding site of thrombin 11. For the development of mono‐PEGylated HV2 as a candidate antithrombotic drug, there exists a trade‐off between the bioactivity and bleeding risk. From the view of retaining HV2's bioactivity, N‐terminal PEGylation may not be appropriate because N‐terminus is its activity site, resulting in a remarkably decreased in vitro anticoagulant activity. Consequently, the loss of in vitro anticoagulant activity should be offset by its significantly prolonged plasma half‐life in vivo, and thus improving its in vivo therapeutic efficacy. A successful case has been proved by a marketed PEG‐protein drug, namely Pegasys^® (peginterferon‐α‐2a; Hoffmann‐La Roche, Inc., Basel, Switzerland), which retained 7% of in vitro activity of interferon‐α‐2a but prolonged the plasma half‐life of interferon‐α‐2a from 0.7 to 51 h (73‐fold) 27. From the view of decreasing HV2's bleeding risk to improve the medical safety, the decreased in vitro anticoagulant activity of N‐terminal mono‐PEG‐HV2 may be beneficial because many studies indicated that bleeding risk of recombinant hirudin was positively correlated with its strong binding affinity for thrombin 2, 28. To balance the trade‐off between the bioactivity and bleeding risk of N‐terminal mono‐PEG‐HV2, the in vivo pharmacological efficacy and safety (e.g., bleeding and immunogenicity) should be evaluated in future studies.

4. Concluding remarks

Site‐specific PEGylation at the N‐terminus of HV2 using mPEG‐ALD were investigated. First, the combined statistical methods including PB design, steepest ascent path analysis, and CCD for RSM were successfully applied for the optimization of the reaction conditions for PEGylation of HV2 with 5 kDa mPEG‐ALD. The results indicated that PB design is useful for quick screening of the significant factors influencing the PEGylation. Additionally, RSM is an effective tool for analyzing the interactions between these factors and determining the locally optimized reaction conditions in the experimentally designed ranges of the variables. Furthermore, kinetic analysis was used to further optimize the molar ratio of PEG to HV2 and reaction time for PEGylation with different molecular weight mPEG‐ALD, which have the most effects on the selectivity of PEGylation and yield of mono‐PEG‐HV2. More accurate and globally optimized conditions were achieved in a very wide range of the variables and verified by the corresponding experiments. The results demonstrated that the combined optimization using RSM and kinetic analysis is a useful tool for the optimization of PEGylation conditions to obtain a desirable yield of mono‐PEGylated product. Compared with random PEGylation of HV2 with mPEG‐SC, PEGylation of HV2 with identical molecular weight mPEG‐ALD had a higher yield of mono‐PEG‐HV2 and a higher selectivity of PEGylation, indicating that N‐terminal site‐specific PEGylation improved the homogeneity and yield of mono‐PEG‐HV2. PEGylation site analysis revealed that almost 100% of the obtained mono‐PEG5k‐HV2, mono‐PEG10k‐HV2, or mono‐PEG20k‐HV2 were modified at the N‐terminus of HV2, indicating that N‐terminal site‐specific PEGylation can produce a homogeneous mono‐PEGylated product. The N‐terminal mono‐PEG5k‐HV2, mono‐PEG10k‐HV2, and mono‐PEG20k‐HV2, respectively, retained 25.3, 22.7, and 21.3% in vitro anticoagulant activity of unmodified HV2, indicating that N‐terminal mono‐PEG‐HV2 can be used as an alternative to being developed as a candidate antithrombotic drug if its loss of binding affinity for thrombin can be offset by its significantly prolonged plasma half‐life in vivo.

Practical application

N‐terminal site‐specific PEGylation of recombinant hirudin variant‐2 (HV2) with different molecular weight mPEG‐propionaldehyde (mPEG‐ALD) was investigated. A combined method using statistical methods and kinetic analysis was successfully developed to optimize this multifactor‐influencing process. Plackett–Burman design is useful for quick screening of the significant factors influencing the PEGylation. Additionally, response surface methodology (RSM) is an effective tool for analyzing the interactions between these factors and determining the locally optimized conditions in the experimentally designed ranges of the variables. Kinetic analysis is more efficient than RSM in the optimization of PEGylation conditions in a very wide range of the variables to achieve globally optimized conditions. The developed combined optimization method using RSM and kinetic analysis might be useful for scale‐up the manufacturing of mono‐PEGylated product on commercial scale more economically. This combined optimization method might be also have wide application in other biotechnological processes.

The authors have declared no conflict of interest.

Supporting information

Supporting Information